Final Research Report

PATIENT-CENTERED OUTCOMES RESEARCH INSTITUTE FINAL RESEARCH REPORT

Comparing Statistical Models That Predict if Patients Will Take a New Medicine as Directed

Joshua J. Gagne, PharmD, ScD1; Moa P. Lee, PharmD, MPH1; Ajinkya Pawar, PhD, MS1; Yaa-Hui Dong, PhD1

AFFILIATIONS: 1Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts

Institution Receiving the PCORI Award: Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital Original Project Title: Adherence Prediction Algorithms to Explain Treatment Heterogeneity and Guide Adherence Improvement PCORI ID: ME-1309-06274 HSRProj ID: 20152264

______To cite this document, please use: Gagne JJ, Lee MP, Pawar A, Dong Y-H. (2019). Comparing Statistical Models That Predict if Patients Will Take a New Medicine as Directed. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/05.2020.ME.130906274

TABLE OF CONTENTS

ABSTRACT ...... 5 BACKGROUND ...... 7 Aim 1: Develop and Compare Algorithms for Predicting Adherence to Various Interventions, Based on Data Collected Prior to the Start of Those Interventions ...... 8 Aim 2: Assess Whether Predicted Adherence Explains Heterogeneity of Treatment Effects in PCOR Studies ...... 9 Aim 3: Develop and Compare Different Algorithms for Predicting Adherence During Treatment ...... 9 PARTICIPATION OF PATIENTS IN THE DESIGN AND CONDUCT OF RESEARCH ...... 10 METHODS ...... 11 Data Source ...... 11 Aim 1: Develop and Compare Algorithms for Predicting Adherence to Various Interventions, Based on Data Collected Prior to the Start of Those Interventions ...... 11 Figure 1. Overview of aim 1 ...... 12 Figure 2. Aim 1, part I: Study design ...... 13 Table 1. Definitions of Measures of Prior Adherence ...... 15 Aim 2: Assess Whether Predicted Adherence Explains Heterogeneity of Treatment Effects in PCOR Studies ...... 20 Aim 3: Develop and Compare Different Algorithms for Predicting Adherence During Treatment ...... 25 RESULTS ...... 29 Aim 1, Part I: Developing a Statin Adherence Prediction Algorithm Enriched With Measures of Prior Adherence Patterns ...... 29 Table 2. Baseline Characteristics of Statin Users in Aim 1, Part I (n = 89 490) ...... 30 Table 3. C Statistics From Models Predicting High Statin Adherence (PDC ≥ 80%) Including Different Sets of Patient Baseline Information in the Testing Cohort, in Aim 1, Part I ...... 33 Table 4. C Statistics From Models Predicting High Statin Adherence (PDC ≥ 80%) Including Only Measures of Prior Adherence in the Testing Cohort in Aim 1, Part I ...... 34 Table 5. C Statistics From Models Predicting High Statin Adherence (PDC ≥ 80%) Including Combinations of Patient Baseline Information and Prior Adherence Measures in the Testing Cohort, Aim 1, Part I ...... 35 Figure 3. Calibration plots for fixed- and random-effects unified models ...... 36

Table 6. Coefficients and Odds Ratios From the Final Statin Adherence Prediction Model in Aim 1, Part I ...... 38 Aim 1, Part II: Assessing Transportability of the Statin Adherence Prediction Algorithm to Other Drug Cohorts ...... 40 Table 7. Baseline Characteristics of the 3 Drug Cohorts in Aim 1, Part II ...... 41 Table 8. Logistic Regression Coefficients of the Selected Predictors Estimated in Each of the 3 Study Cohorts in Aim 1, Part II ...... 44 Table 9. Comparison of Adherence Prediction Model Performance in Statin and Bisphosphonate Cohorts in Aim 1, Part II ...... 45 Table 10. Model Performance in the Developing (Statin Initiators) and Validating Cohort (ACEI/ARB Initiators) ...... 45 Aim 1, Part III: Identifying and Evaluating a Single, Unified Adherence Prediction Model for Application Across Different Drug Classes ...... 47 Aim 2: Assess Whether Predicted Adherence Explains Heterogeneity of Treatment Effects in PCOR Studies ...... 48 Table 11. Descriptive Statistics for Adherence in Baseline and Follow-up Before Propensity Score Matching in the Statin Cohort ...... 49 Table 12. Mortality Incidence Rates (IR) and IR Differences Between Low- and High-intensity Statin Initiators After Propensity Score Matching in As-treated Analysis ...... 52 Table 13. Fracture Incidence Rates (IR) and IR Differences Between Bisphosphonate and Calcitonin Initiators After Propensity Score Fine Stratification in As-treated Analysis ...... 54 Table 14. Fracture Incidence Rates (IR) and IR Differences Between Bisphosphonate and Raloxifene Initiators After Propensity Score Fine Stratification in As-treated Analysis ...... 55 Table 15. Fracture Incidence Rates (IR) and IR Differences Between Raloxifene and Calcitonin Initiators After Propensity Score Fine Stratification in As-treated Analysis ...... 55 Aim 3: Develop and Compare Different Algorithms for Predicting Adherence During Treatment ...... 56 Figure 4. Distribution of patients with and without subsequent dispensation at each dispensation during follow-up ...... 57 Table 16. Distribution of Patient Characteristics, Adherence Measures for Statins, and Adherence to Other Medications During Follow-up ...... 58 Table 17. C Statistics of Models Including Baseline Patient Characteristics and Prior Adherence Measures in Predicting Discontinuation After the First Statin Dispensation in the Testing Cohort ...... 62

Table 18. C Statistics for Models Including Only Baseline Information, Baseline Information Plus Each Follow-up Variable Separately, and Baseline Model Plus All Follow-up Variables in Predicting Statin Discontinuation at the Second Through 12th Dispensation ...... 64 DISCUSSION ...... 65 Context for Study Results ...... 65 Uptake of Study Results ...... 67 Study Limitations ...... 68 Future Research ...... 69 CONCLUSIONS ...... 71 REFERENCES ...... 72 APPENDICES ...... 76

ABSTRACT

Background: Medication nonadherence is a major public health problem. It can lead to poor patient outcomes and can also complicate the interpretation of patient-centered outcomes research (PCOR), including comparative effectiveness research (CER) studies.

Objectives: The objectives of this project were to (1) develop and compare algorithms for predicting adherence to various interventions, based on data collected prior to the start of those interventions; (2) assess whether predicted adherence explains heterogeneity of treatment effects in PCOR studies; and (3) develop and compare different algorithms for predicting adherence during treatment.

Methods: To predict adherence to a newly started drug treatment, we used large health care databases to compare many different approaches to measuring patients’ prior medication adherence. We assessed whether models developed among patients starting one type of drug (eg, statins) are transportable to cohorts of patients starting other drugs (eg, bisphosphonates, antihypertensives) and examined the feasibility of a unified adherence prediction score. In 4 CER studies conducted in electronic health care databases (high- vs low-intensity statins on inpatient mortality and cardiovascular events; and 3 comparisons of osteoporosis medications on fracture: bisphosphonates vs calcitonin, bisphosphonates vs raloxifene, and calcitonin vs raloxifene), we stratified patients into tertiles based on their adherence prediction score and evaluated heterogeneity of treatment effects across tertiles. Finally, we examined whether the addition of posttreatment initiation information improves ability to predict whether patients will receive subsequent medication dispensation.

Results: To measure prior adherence, we identified 89 490 eligible patients who initiated a statin and had at least one previous medication dispensation. Median baseline adherence was 58%, and the 25th and 75th percentiles were 25% and 90%, respectively. Measures of prior adherence were strongly predictive of adherence to the newly started statin and improved the discriminative ability of a model containing usual adherence predictors from a C statistic of 0.665 (95% CI, 0.665-0.670) to 0.696 (95% CI, 0.691-0.701). This model did not perform as well in other drug cohorts (C-statistic in bisphosphonate initiators, 0.638 and C statistic in antihypertensive initiators, 0.654), but discrimination was improved via reestimating model parameters (0.665 and 0.672, respectively). A unified model containing a common set of predictors performed relatively well (0.684; 95% CI, 0.682-0.686). In the high- vs low-intensity statin comparison, we found some indication vs that stratification by predicted adherence score may be useful for explaining heterogeneity of treatment effects for the outcome of mortality (ie, incidence rate differences of –1.90, –2.59, and –4.43 per 1000 person-years in tertiles 1, 2, and 3 of predicted adherence, respectively), but not in the other outcomes we examined or in any of the 3 osteoporosis medication comparisons. As compared with models containing only baseline information, individual variables assessed after statin initiation provided little additional discriminative ability in predicting treatment discontinuation.

Conclusions: Incorporating measures of prior adherence can improve claims-based adherence prediction models, but the overall ability of claims data to discriminate who will and will not adhere to a new medication remains modest. While future activities involving adherence prediction should consider measures of prior adherence, we found little evidence that claims- based adherence scores can explain heterogeneity of treatment effects in CER studies and little evidence that incorporating post-initiation data meaningfully improves prediction of treatment discontinuation.

Limitations: Many potential predictors of medication adherence may not be available in administrative health care data. The modest discriminative ability of the adherence prediction scores might have precluded our ability to observe differences in treatments across strata of predicted adherence. It is also possible that there are no or only very small true differences in effects between compared treatments in the populations studied, which would have precluded us from observing differences in treatment effects across strata.

BACKGROUND

Medication nonadherence is a major public health problem.1 Less than half of patients persist with cardiovascular drugs for a year following heart attack,2 despite compelling evidence of clinical benefits of these life-saving treatments.3 Poor adherence has substantial clinical and economic consequences.4 In the United States, suboptimal adherence accounts for between 33% and 69% of medication-related hospital admissions and US$100 billion of potentially avoidable health care spending each year.1

Various effective medication adherence interventions exist.5-7 Even small improvements in adherence to evidence-based treatment can improve clinical outcomes.8,9 A key challenge in maximizing the benefit and value of these interventions is identifying populations of patients that are expected to have low adherence at the time of treatment initiation, as these patients may stand to benefit the most from adherence interventions.

In addition to contributing to poor clinical and economic outcomes for patients, medication nonadherence also complicates the interpretation of patient-centered outcomes research (PCOR), including comparative effectiveness research (CER) studies. Many randomized clinical studies use a pretrial run-in phase to exclude patients who do not adhere to the intervention(s) under study.10 However, no analogue to the run-in period exists for identifying patients who may not adhere to medications in clinical practice settings or in observational PCOR studies. As a result, patients enrolled in trials tend to adhere better than do patients who use the treatments outside of the clinical trials and in “real-world” care settings, which may be a driver of apparent discrepancies between the effects of treatments observed in observational studies vs randomized trials. Moreover, it is unclear to what extent observational study results apply to patients who adhere more or less than the “average” patient in the study population.

In this project, we sought to develop, validate, and assess adherence prediction scores, based on electronic health care data. An adherence prediction score would enable researchers to stratify study results by predicted adherence (rather than by observed adherence, which can induce bias in CER studies); identify which patients are most likely to adhere to a given

treatment; and potentially even quantify differences between randomized trial results and observational study results that are likely due to adherence. We focused on whether stratifying by predicted treatment adherence explains heterogeneity of treatment effects in CER studies. CER studies generally report only average results across all patients, regardless of their likelihood to adhere. We hypothesized that incorporating an adherence prediction score into CER studies would enable researchers to determine whether certain treatments work better than others for patients with different levels of predicted adherence.

In addition to developing an adherence prediction score that can predict whether patients will adhere to a newly initiated treatment, we also explored innovative adaptations of adherence prediction scores that incorporate information collected during treatment, to predict whether and when patients will stop adhering to the treatment. The specific aims of this project are outlined below.

Aim 1: Develop and Compare Algorithms for Predicting Adherence to Various Interventions, Based on Data Collected Prior to the Start of Those Interventions Many attempts have been made to predict medication adherence using routinely collected health care data.11-16 Although previously developed medication adherence algorithms have had limited predictive performance, preliminary evidence suggests that using measures of prior adherence to other chronically used medications may be a strong predictor of future adherence to a newly initiated medication.17,18 However, it is not known how best to measure and use prior medication adherence to predict adherence to a new drug. No prior study has systematically compared alternative approaches for defining prior medication adherence to optimize prediction, and no study has developed and tested prediction scores that can be applied across a wide range of drugs, patient populations, and clinical conditions. The first aim was to determine the best approaches for defining and including prior adherence in adherence prediction algorithms and to determine if a single unified prediction score applies across different patient populations and drug classes or if more specific algorithms are needed for different types of drugs.

Aim 2: Assess Whether Predicted Adherence Explains Heterogeneity of Treatment Effects in PCOR Studies In addition to identifying patients at the time of treatment initiation who may benefit from adherence improvement interventions, we hypothesized that adherence prediction scores may also be useful for explaining heterogeneity in CER studies. In this aim, we assessed whether stratifying by predicted adherence yielded evidence of heterogeneity of treatment effects in 3 CER use cases. We hypothesized that treatment effects would be greatest among patients with high predicted adherence and smallest among patients with lowest predicted adherence.

Aim 3: Develop and Compare Different Algorithms for Predicting Adherence During Treatment In addition to predicting whether patients will adhere to their medication, using information available at the time of treatment initiation, an algorithm to identify when patients might discontinue their medication would also help target adherence intervention strategies. During treatment, many changing factors can affect whether a patient will continue to adhere or persist. In this aim, we examined the use of post-initiation information to develop a dynamic adherence prediction score for predicting treatment discontinuation.

PARTICIPATION OF PATIENTS IN THE DESIGN AND CONDUCT OF RESEARCH

A 10-person patient advisory board (PAB) provided input into key aspects of project design and conduct and reviewed study results. The PAB comprised patients with some prior experience or education in health care who could offer meaningful input on this methods project. PAB members were recruited from Consumer Reports’ advocacy members, a group of ~7,000 educated consumers who have expressed interest in reviewing and participating in health care and consumer advocacy efforts. Consumer Reports’ advocacy members were recruited if they indicated interest in discussing issues that might be important to themselves or other patients and had some familiarity with the conduct of clinical research. The PAB gave important input into (1) patient factors that may be potential determinants or predictors of medication adherence; and (2) the design and evaluation of the aim 2 CER studies, including outcomes of interest. For the latter, PAB members reviewed and gave feedback on key decision points in the design and conduct of the aim 2 studies, such as whether each aim 2 outcome was relevant to patients. The PAB also provided insights into additional patient factors that might affect patient adherence but that are not well measured in administrative claims data, such as patients’ social support systems and the use of pill boxes, reminders, and other adherence improvement tools.

METHODS

As the 3 aims are distinct but aims 2 and 3 build directly upon aim 1, we first describe the data used in all aims and the role of the PAB and then describe the approaches used for each aim separately and in turn.

The Brigham and Women’s Hospital Institutional Review Board approved this study. Data Source All 3 aims used deidentified data from the Optum Clinformatics Data Mart. The database comprises a large, geographically and clinically diverse population of health insurance beneficiaries enrolled in commercial UnitedHealth Group–affiliated health plans as well as patients with a Medicare supplement plan. Approximately 10% of patients in the database are Medicare beneficiaries. Overall, the database includes information relating to approximately 14 million patients on a yearly basis and 60 million patients in total. The data include medical claims from health care providers and facilities, outpatient pharmacy dispensation records, and enrollment information that provides demographic data and dates of insurance eligibility for persons in the database.

Aim 1: Develop and Compare Algorithms for Predicting Adherence to Various Interventions, Based on Data Collected Prior to the Start of Those Interventions Aim 1 involved 3 main parts: (1) developing a statin adherence prediction algorithm enriched with measures of prior adherence patterns; (2) assessing the transportability of the statin adherence prediction algorithm to other drug cohorts; and (3) identifying and evaluating a single, unified adherence prediction model for application across different drug classes (Figure 1).

Figure 1. Overview of aim 1

Focusing initially on statins, a class of cholesterol-lowering drugs and one of the most frequently prescribed drug classes in the United States, part I sought to systematically evaluate different metrics and chronic medication classes for measuring prior adherence and, using administrative claims data, to assess the added value of these measures of prior adherence for predicting future medication adherence. In part II, we used the top-performing prediction algorithm developed in part I and applied it to cohorts of patients initiating bisphosphonates and angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers (ACEIs/ARBs). This enabled us to evaluate the transportability of the model for predicting adherence to different drugs and different populations from part I. We also compared the performance of the model fit in these groups with the performance of cohort-specific models developed and validated specifically to predict adherence to bisphosphonates and ACEI/ARB in these groups. Finally, in part III of aim 1 and based on the learnings from parts I and II, we developed and examined the performance of a unified prediction model that would more broadly apply across different drug classes and patient populations.

Aim 1, Part I: Developing a Statin Adherence Prediction Algorithm Enriched With Measures of Prior Adherence Patterns

Study cohort. We identified patients aged 18 years or older who initiated a statin between July 1, 2010, and December 31, 2011. Initiation of a statin was defined as a new statin dispensation following a 365-day baseline period in which patients were required to have continuous enrollment in the health plan and no statin dispensation. The day of the first statin dispensation was defined as the cohort entry date. For patients with multiple eligible cohort entries, only the first was included. The primary analysis focused on patients who were continuously enrolled in the plan for at least 365 days following statin initiation and who had at least one dispensation for a medication used for measuring prior adherence in the 365-day baseline period (Figure 2).

Figure 2. Aim 1, part I: Study design

Study outcomes. The outcome of interest was adherence to statins in the 365 days following the first dispensation. We measured adherence using proportion of days covered (PDC), defined as19:

= × 100 365.25 ∑ 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑜𝑜𝑜𝑜 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑃𝑃𝑃𝑃𝑃𝑃 � � PDC was then dichotomized using a threshold of ≥ 80%, indicating “full adherence.”20

Predictors of adherence. We measured many potential adherence predictors that are often included in claims-based studies of medication adherence. Demographic variables

included age and sex at cohort entry. We used claims with service dates occurring during each patient’s 365-day baseline period to define an extensive list of comorbidities, medication use, and health services use measures (eg, number of physician visits, hospitalized days, use of colonoscopy, mammography, and vaccinations; see Appendix 1 for a full list of covariates). We assessed measures of patients’ medication burden by counting the number of dispensations of all drugs and the number of unique drug types dispensed during the baseline period; evidence for medication refill synchronization, which indicates patients filling multiple prescriptions on the same day; and the number of concurrently used medications on the cohort entry date.21 We also assessed proxies of patients’ medication cost burden, including plan benefit type, total copayment for all drugs during the baseline period, and index statin copayment. These variables formed the set of basic potential predictors.

Medications for prior adherence measurement. We measured patients’ adherence to the following types of medications, using dispensations in the 365-day baseline period: ACEI/ARB, renin inhibitors, beta-blockers, calcium channel blockers, thiazide diuretics, loop diuretics, potassium-sparing diuretics, other antihypertensives, oral anticoagulants, digoxin, antianginal agents, serotonin reuptake inhibitors, typical antipsychotics, atypical antipsychotics, lipid-lowering agents (other than statins), antidiabetics, osteoporosis drugs, thyroid hormone, nonbiologic disease-modifying antirheumatic drugs, antiparkinson agents, anticonvulsants, and antiglaucoma agents. We included a broad range of medications for balance, including as many patients as possible in the analysis cohort, while focusing on drugs intended to treat chronic conditions.

Prior adherence. Measurement. We assessed prior adherence using 7 different measures of prescription coverage, drug discontinuation, and dispensation counts (see Table 1 for measure definitions).

Table 1. Definitions of Measures of Prior Adherence

Measurement Type Measurement Name Definition Measures based on Continuous measure Numerator: total days’ supply dispensed prescription of medication during denominator coverage days acquisition (CMA) Denominator: time between the first dispensation in the baseline period and the end of the baseline period (date of statin initiation) Continuous measure Numerator: total gap days (number of days of medication gaps without medication days’ supply) (CMG) Denominator: time between the first dispensation in the baseline period and end of the baseline period Proportion of days Numerator: total days’ supply capped at covered (PDC) maximum number of days in denominators Denominator: time between the first dispensation in the baseline period and the end of the baseline period Measures based on Discontinuation Presence (yes = 1 or no = 0) of one or more discontinuation (dichotomous; DD) periods of at least 30 days without medication supply following the days’ supply of a dispensation in baseline period Time to Days between the date of the first discontinuation (TTD) dispensation and discontinuation or the end of the baseline period Measures based on Refill counts (counts) Number of dispensations in the baseline dispensation counts period beyond the first fill for each drug Lack of second fills Not having a second dispensation of a drug (dichotomous) in the baseline period when the first dispensation occurred more than 30 days + days’ supply before the statin initiation

Measurement was performed over the 365-day baseline period prior to statin initiation. Measurement for each drug began at the first observed dispensation of the particular drug or drug class in the baseline period and ended on the statin initiation date. Prescriptions filled

before the start of the baseline period, but with days’ supply that elapse the start of the baseline period, were not considered. Medications that were first dispensed in the 90 days immediately preceding statin initiation were excluded to ensure at least a 90-day assessment period for each drug or drug class. Switching between medications within the same pharmacologic class was allowed. For patients with multiple medications used for prior adherence measurement, we summarized across drugs by taking the mean, median, maximum, and minimum for each adherence metric.

Statistical analysis. We divided the overall statin cohort into a 50% training cohort and a 50% testing cohort. We first selected potential predictors of future statin PDC in the training cohort from among the list of covariates other than measures of prior adherence, using the least absolute shrinkage and selection operator (lasso). We fit lasso logistic regression models with a binary outcome of good adherence defined as PDC ≥ 80%, with a shrinkage parameter lambda chosen to minimize the models’ Bayesian information criterion. We selected variables with nonzero beta coefficients as predictors for subsequent models.

We then developed multivariable logistic regression models predicting PDC ≥ 80% in the training cohort by including various groups of the lasso-selected variables. For example, we built a model that included only demographic characteristics and another model that included only comorbidities. We then built a model that included all variables selected by lasso from the set of basic predictors. We also built separate univariable logistic regression models in which each of the prior medication measures was included as the only explanatory variable. We then developed models that included all the variables selected by lasso from the basic set of predictors plus the best-performing measures of prior adherence. Finally, we developed a separate model using only data available at the pharmacy at the time of prescription dispensation (ie, demographics, baseline medication use, index statin type, medication burden, and prior medication adherence) to examine the performance of adherence prediction if medical data were not available.

Each model created in the training cohort was applied to the testing cohort to predict each patient’s probability of having a 1-year PDC ≥ 80% for the newly initiated statin. The

predicted probabilities were used to calculate C statistics to compare discrimination across models. The C statistic (also known as the concordance statistic) is a measure of discrimination that can be interpreted as the probability that a randomly selected patient with high adherence (ie, PDC ≥ 80%) has a higher predicted probability of adherence than a randomly selected patient with low adherence. We assessed the calibration of the models using calibration plots. The model with the best performance was used in aim 1, parts II and III, and in aims 2 and 3.

To assess the strength of association between prior adherence and subsequent statin adherence, we fit a modified Poisson regression model using sandwich variance estimators,22 with a binary-dependent variable of high adherence (PDC ≥ 80%). We categorized patients into 3 prior adherence groups based on their mean PDC during the baseline period: high prior adherence (mean PDC ≥ 80%), moderate prior adherence (25% ≤ mean PDC < 80%), and low prior adherence (mean PDC < 25%). We compared the probability of having high adherence to statins across prior adherence groups, adjusting for all previously selected potential predictors from lasso regression.

Aim 1, Part II: Assessing the Transportability of the Statin Adherence Prediction Algorithm to Other Drug Cohorts

Bisphosphonate cohort. Based on the same sampling scheme applied to identify the cohort of statin initiators in part I, we identified a cohort of patients who initiated an oral bisphosphonate between January 1, 2006, and December 31, 2013. Initiation was defined as a first dispensation of an oral bisphosphonate (alendronate, risedronate, or ibandronate) with no prior dispensation of any oral bisphosphonates during a 365-day baseline period in which patients were required to have continuous enrollment in the health plan. We applied the same eligibility criteria as above to identify bisphosphonate initiators—with the additional requirements of having no diagnosis for Paget’s disease (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 731.0) and no claims for parenteral administration of osteoclast inhibitors during the baseline period.

ACEI or ARB cohort. Similarly, we identified a cohort of ACEI/ARB initiators from January 1, 2013, to December 31, 2013. Initiation was defined as a first dispensation of an ACEI or ARB with no prior dispensation of any ACEI or ARB during the baseline.

Study outcomes. We used the same PDC outcome from part I in part II.

Predictors of adherence. In addition to the predictors included in part I, we assessed other factors that may be associated with bisphosphonate adherence, including prior bone fracture, corticosteroid use, and bone mineral density (BMD) tests. For the ACE/ARB cohort, we also included additional variables related to the severity of hypertension and cardiovascular comorbidities .

Statistical analysis. To assess the transportability (ie, external validation) of a model developed in statin initiators to other drug classes and populations, we compared its performance in the cohorts of bisphosphonate initiators and ACEI/ARB initiators with the performance of models developed specifically in these cohorts. To facilitate comparisons across the 3 cohorts, we first refit the part I model in the statin cohort using a 10-fold cross-validation process (in contrast to the split-sample approach used in part I). We used the C statistic to evaluate the model’s discriminative ability. We estimated the calibration slope by regressing the observed adherence outcome on the predicted probability from the prediction model: logit (full adherence) = a + b1 logit (predicted probability of full adherence); the estimated b1 represents the calibration slope. A model with a perfect calibration will have a calibration slope of 1.23 We then refit the model in the bisphosphonate and ACEI/ARB cohorts by reestimating the coefficients for each potential predictor and augmented these models by adding the cohort-specific variables described previously. Measures of performance (ie, C statistics and calibration slope) were averaged over the 10 values estimated during cross-validation for each model.

We examined the distribution of baseline potential predictors across the cohorts of statin, bisphosphonate, and ACEI/ARB initiators. We also assessed the association between

each potential predictor and the outcome across the 3 cohorts by comparing regression coefficients of the predictors estimated in each cohort. To further understand the importance of each variable in predicting adherence to different drug classes, we used lasso and performed 10-fold cross-validation to determine the frequency of selection of each variable across the 10 cross-validation steps. The frequency of variables selected for each fold was reported as a percentage. We also used a generalized boosted model (GBM) to assess the influence of each variable in predicting adherence across the drug classes. For each variable, the GBM computes the “relative influence,” which estimates the proportion of the model prediction ability due to each variable.24,25 The sum of the relative influence of each variable adds to 100, with higher numbers indicating a stronger contribution to predicting the response. The initial specifications of the GBM were 10 000 iterations, a learning rate parameter of 0.01, and an interaction depth of 1 for the additive approximations.

Aim 1, Part III: Identifying and Evaluating a Single, Unified Adherence Prediction Model for Application Across Different Drug Classes The purpose of part III of aim 1 was to develop a single, uniform prediction score to predict adherence across different drug classes and populations and to compare its performance with the cohort-specific adherence prediction scores.

Study cohort. For part III, we pooled the statin, bisphosphonate, and ACEI/ARB initiator cohorts developed in parts I and II to form a single larger cohort of patients initiating different medications to which we sought to predict adherence.

Predictors of adherence. The candidate set of potential predictors for the unified adherence score included those selected with high frequency during the 10-fold cross- validation performed in part II. Among the basic set of potential predictors and the measures of prior adherence assessed in each study cohort in parts I and II, we identified the variables that were selected in at least 50% of the lasso models from the cross-validation folds. Categorical variables with more than 2 levels were included if at least 1 level was selected in ≥ 50% of the lasso models.

Statistical analysis. We considered 2 approaches for modeling the unified adherence prediction score in the pooled cohort. First, we fit a fixed-effects logistic regression model to estimate the probability of full adherence with the variables selected as described previously. This model ignores variability in adherence to the drugs or classes being initiated and evaluated for the adherence outcome. The second model included drug class–specific random intercepts. Including random intercepts permits variation in the baseline probability of the outcome by allowing the logit of the outcome probabilities to vary across the drug classes.26 We assessed and compared the performance of these models with each other and with the cohort-specific models using C statistics and calibration plots.

Aim 2: Assess Whether Predicted Adherence Explains Heterogeneity of Treatment Effects in PCOR Studies In aim 2, we used the cohort-specific adherence prediction scores developed in aim 1 in 4 CER use cases in 2 clinical contexts. In particular, we designed and conducted CER studies comparing major cardiovascular events between patients who received low-intensity vs high- intensity statins and comparing fracture events in patients who received bisphosphonates, calcitonin, or raloxifene. We used the adherence prediction scores to estimate predicted adherence to the study drugs, to stratify patients into groups based on this predicted adherence, and to compare the outcomes between treatment groups within each stratum, in accordance with the “descriptive heterogeneity of treatment effects analysis” and the “confirmatory heterogeneity of treatment effects analysis” approaches described in the PCORI Methodology Report (Heterogeneity of Treatment Effect Standard 7.3.1). We hypothesized that the observed treatment effects would increase across strata of increasing predicted adherence score (Heterogeneity of Treatment Effect Standards 7.3.2 and 7.3.4). To determine the interaction between the treatment effect and the adherence prediction score, we tested this hypothesis using statistical modeling.

Study Design For each of the 3 CER use cases, we used a new-user, parallel group, propensity score– matched cohort study design. We identified new users of the drugs of interest (high-intensity

statins vs low-intensity statins; bisphosphonates vs calcitonin; bisphosphonates vs raloxifene; and calcitonin vs raloxifene). For each use case, new use was defined as having a prescription dispensation (the date of which was defined as the date of cohort entry) for either the drug of interest or the comparator following a period of at least 1 year with continuous health plan enrollment and no use of either drug or drug class. The eligible cohorts were defined using information available at the time of treatment initiation (ie, the date of cohort entry; PCORI Methodology Report Causal Inference Standard 7.2.1). All covariates (described below) were measured during the 1-year baseline period preceding the cohort entry date (Causal Inference Standard 7.2.4). We also used the 1-year baseline period to estimate each patient’s predicted adherence using the adherence prediction scores developed in aim 1.

Study Cohorts and Exposures

High-intensity vs low-intensity statins on major cardiovascular events. We identified all eligible episodes of new use of a high-intensity or low-intensity statin between January 1, 2004, and September 30, 2015. High-intensity statins included high-potency statins used at high doses (ie, atorvastatin > 10 mg, lovastatin > 40 mg, rosuvastatin > 5 mg, simvastatin > 40 mg) as well as any statin used at any dose in combination with ezetimibe. Statins in the high-intensity group were those that would be expected to lower low-density lipoprotein cholesterol (LDL-C) by 40% or more.27 To identify new users, we excluded episodes with any prescription dispensations for statins or ezetimibe within the year prior to the potential date of cohort entry, requiring and required continuous health plan enrollment during this baseline year. For patients with multiple eligible new-user episodes, only the first episode was included and the cohort entry date was defined as the dispensation date of the first prescription in the episode. We excluded patients with more than one prescription for a statin or a statin-containing product on the cohort entry date and those younger than 18 years on the cohort entry date.

Bisphosphonates vs calcitonin vs raloxifene on fracture events. For the bisphosphonates vs calcitonin, bisphosphonates vs raloxifene, and calcitonin vs raloxifene use

cases, we identified all eligible episodes of new use of these drugs between January 1, 2004, and September 30, 2015. To define new use, we excluded episodes with any prior prescription dispensations for bisphosphonates, calcitonin, or raloxifene within the year prior to the potential cohort entry date and required continuous health plan enrollment during this baseline year. We excluded episodes with a prescription dispensation of more than one drug of interest on the potential cohort entry date as well those in which the patient was younger than 18 years on the cohort entry date. To focus on patients using these drugs for osteoporosis, we excluded patients with Paget’s disease, defined as ICD-9-CM code 731.0, in the 365 days before the cohort entry date as well as those with treatment with any of the following in the 365 days before the cohort entry date: risedronate 30 mg if quantity dispensed was more than 12 tablets, alendronate 40 mg if quantity dispensed was more than 12 tablets, and calcitonin injection (nonspray forms). For patients with multiple eligible new-user episodes, only the first episode was included and the cohort entry date was defined as the dispensation date of the first prescription in the episode.

Outcomes and Follow-up In accordance with Causal Inference Standard 7.2.3, we identified all outcomes following the start of exposure.

High-intensity vs low-intensity statins on major cardiovascular events. In the statin-use case, the outcomes of interest were hospitalization for cerebrovascular event or myocardial infarction (MI),28 as well as in-hospital mortality.

Bisphosphonates vs calcitonin vs raloxifene on fracture events. For the osteoporosis drug–use cases, outcomes of interest were hospitalization for fracture of the hip, humerus, pelvis, or wrist, as fractures at these 4 sites are typically considered fragility fractures.29

In both use cases, outcomes were assessed over a 1-year period, starting the day after each patient’s cohort entry date. Patients were permitted to have more than one outcome of

interest but no more than one of each (ie, we considered only the first event for each outcome). To achieve high specificity, all outcomes were defined based on hospitalizations and using validated algorithms that have been shown to have high performance in claims data.

The primary analysis used an intention-to-treat (ITT) model for follow-up. Follow-up began on the day after the cohort entry date and ended at the first occurrence of one of the following events: outcome occurrence, inpatient death, health plan disenrollment, end of available data, or 365 days from the cohort entry date. In an “as-treated” sensitivity analysis, patient follow-up was further censored at the time of index therapy change (eg, switching from low-dose to high-dose statin, switching from bisphosphonate to calcitonin) or discontinuation of index therapy.

Covariates Covariates included all variables needed for the respective adherence prediction scores as well as other predefined variables that are known or suspected risk factors for the outcomes of interest. These included demographics, measures of health services use and preventive care, the combined comorbidity score,30 clinical comorbidities, and use of other medications. For all 3 use cases, we assessed covariates based on enrollment information and claims during the 1 year prior to (and including) each eligible patient’s new-user cohort entry date. Age, sex, and state/region were measured on the cohort entry date only. Unless specified otherwise, covariates were identified based on one inpatient or one outpatient diagnosis in any position.

Stratifying by adherence prediction scores. We estimated each patient’s predicted adherence using the adherence prediction scores developed in aim 1—that is, we determined each patient’s values for the variables in the adherence prediction score and multiplied these values by the corresponding coefficients from the prediction models. We then summed these products to estimate each patient’s log predicted adherence.

We stratified the cohorts of eligible new users in each use case into tertiles, based on patients’ estimated predicted adherence scores. We estimated the association between the drug of interest vs comparator and each outcome of interest separately within each stratum of

predicted adherence and used statistical modeling to determine the interaction between the treatment effect and the adherence prediction score. Because not all patients used medications that assessed adherence before cohort entry, some patients had missing prior adherence values. The primary analysis included only patients with prior adherence values. We also conducted a sensitivity analysis in which we imputed missing values. Because results were similar to those of the main analyses, only the main analyses are presented.

Achieving balance in patient covariates. We used propensity scores to reduce confounding by balancing measured baseline covariates between the new users of the drugs of interest and comparison groups.31 All covariates described previously were included in logistic regression models to predict patients’ probabilities (ie, propensity scores) of being in the drug of interest group vs the comparison group. Because including only the adherence prediction score in the propensity score models would achieve balance on predicted adherence but not necessarily on all variables in the adherence prediction scores, we included all covariates that went into adherence score calculation in the propensity score model as well as the mean baseline PDC.

For the high-intensity statin vs low-intensity statin comparison, we matched patients by propensity score in a 1:1 ratio within a caliper of 0.01 on the probability scale using a nearest- neighbor matching algorithm. For the osteoporosis use cases, we used fine stratification on the propensity score to achieve covariate balance, because the relative sizes of the raloxifene and calcitonin groups were small compared with those of the bisphosphonate groups. Propensity score fine stratification has been found to have superior performance to 1:1 matching in this setting of disproportionate group sizes.32

Covariate balance between the treatment groups was assessed within each tertile as well as in each of the full cohorts.

Statistical Analysis We report number of events, person-time, and incidence rates for each outcome, stratified by tertile of predicted adherence and treatment group. We used incidence-rate

differences and Cox proportional hazards regression models to compare rates of each outcome between treatment groups. Hazard ratios and 95% CIs were estimated overall for each use case as well as within each tertile. We used the Schoenfeld test to assess for violations of proportional hazards. We tested for differences in treatment effects by predicted adherence using an interaction term between the treatment effect and the adherence prediction score. All analyses were performed using SAS statistical software, version 9.3 (SAS Institute, Inc, Cary, North Carolina).

Aim 3: Develop and Compare Different Algorithms for Predicting Adherence During Treatment Aim 3 built on aim 1 by incorporating post-initiation information into the adherence prediction model to predict, at each prescription dispensation, the likelihood that a patient would subsequently refill the prescription. We focused on a cohort of statin initiators.

Study Cohort We identified patients aged 18 years or older who initiated a statin between January 1, 2010, and September 30, 2014. Statin new use was defined the same way as in aim 1. We further restricted the cohort to patients with at least one dispensation for a medication used for measuring prior adherence in the 365 days before the cohort entry date, as in the main analysis in aim 1.

Outcomes and Follow-up The outcome of interest was statin discontinuation, which we defined as a lack of subsequent statin dispensations within 14 days following the end of a previous dispensation (ie, after the days’ supply of the previous dispensation ran out). Specifically, at each observed statin dispensation, we sought to predict the likelihood that that would be the last statin prescription that the patient would fill. For example, at the initial statin prescription, there could be 1 of 3 possible outcomes: (1) the patient filled the prescription again within the 14-day grace period, which we defined as not discontinuing provided that the date of the second dispensation occurred before a competing event (defined below); (2) the patient had no refill or competing

event within 14 days after the end of the first statin dispensation, which we defined as discontinuation; or (3) the patient was censored due to a competing event that precluded us from observing whether the patient received the second dispensation. At each observed prescription, we repeated the same steps and created longitudinal diaries to determine whether patients experienced discontinuation following the first through up to the 11th statin dispensation.

We followed patients from cohort entry to the earliest statin treatment discontinuation or occurrence of any competing risk event, including change in statin treatment, death, disenrollment from the health plan, end of the study period (September 30, 2015), 365 days from cohort entry, or date of dispensation of the 12th statin prescription. Statin treatment change (switch or addition) was defined as a dispensation of any other lipid-lowering agent (see Appendix Table 1 for a list of statins and other lipid-lowering agents).

Covariates We assessed the same baseline covariates as in the aim 1 statin cohort, including age at cohort entry, sex, information on index statins (specific statin and days’ supply for index statins), proxies of patients’ medical cost burden (copayment for index statins, benefit plan type, and total copayment for any drugs within 365 days preceding the cohort entry date), resource use, comorbidities, and other medication use within 365 days preceding the cohort entry date. We also estimated measures of prior medication adherence as described in aim 1.

We also assessed statin-specific covariates at each dispensation during follow-up, including specific statin as well as the days’ supply and copayment of the statin prescription. In addition, we measured covariates occurring between the current statin dispensation and the previous statin dispensation. These included all-cause hospitalization and use of 13 specific classes of cardiovascular medications (ACEIs, ARBs, renin inhibitors, beta-blockers, calcium channel blockers, thiazide diuretics, loop diuretics, potassium-sparing diuretics, other antihypertensives, oral anticoagulants, heparin and low-molecular-weight heparin, antiplatelets, and antianginal agents).

We hypothesized that if patients came back to fill their prescriptions before the end of the days’ supply, they would be more likely to adhere to their statin treatment. Therefore, we estimated several measures of adherence to statins up until each prescription dispensation of interest. These included the lag time for filling each prescription, which was calculated as the date of current statin dispensation minus the end date of the previous statin dispensation. The value of the lag time ranged from 1 (days’ supply of the current prescriptions) to 14. We also counted the number of chronically used medication classes that a patient discontinued in the interval between the date of current statin dispensation and the date of previous statin dispensation.

Statistical Analyses We report the number and proportion of patients who did and did not discontinue statin treatment at each of the 11 dispensations during follow-up. We also tabulated patient characteristics, adherence to other chronically used medications, and adherence measures for statins at baseline prior to cohort entry and during each interval between observed statin dispensations.

To examine whether information at or prior to statin initiation can predict whether patients will receive a second statin dispensation, we first randomly divided the eligible cohort into a training cohort and a testing cohort of equal size. In the training cohort, we fit separate logistic regression models with discontinuation (ie, lack of a second statin dispensation) as the dependent variable and each domain of baseline patient characteristics and prior adherence measures as independent variables. We also fit models that included both baseline patient characteristics and prior adherence measures as independent variables, which served as a final baseline model, as in aim 1. Each model created in the training cohort was applied to the testing cohort to predict each patient’s probability of experiencing continuation after the first statin dispensation. The predicted probabilities were used to calculate C statistics to compare discrimination across models.

To continue predicting patients’ likelihood of discontinuing statin treatment during follow-up, we created separate logistic regression models predicting discontinuation after each of the second through 11th dispensations. Each model included both baseline patient characteristics and prior adherence measures as independent variables. As patient characteristics and measures of statin adherence can change during treatment in ways that may predict whether patients are likely to stop adhering to their medication, we also included these updated variables assessed in the interval between the current dispensation and the previous dispensation. For example, to predict if patients discontinue after the second dispensation, in addition to including baseline patient characteristics and prior adherence measures in the model, we also included information related to the second statin dispensation, indications of whether patients were hospitalized, and indicators for whether patients used 13 predefined cardiovascular medications between the dates of the first and second statin dispensations. We also included the lag time between the first and second dispensations and number of maintenance medication classes that were discontinued between the first and second dispensations. We applied the same approach described previously to assess the predictive performance of the models over time.

RESULTS Aim 1, Part I: Developing a Statin Adherence Prediction Algorithm Enriched With Measures of Prior Adherence Patterns

Patient Characteristics and Prior Adherence for the Statin Cohort We identified 243 051 statin initiators between July 1, 2010, and December 31, 2011; 165 620 had complete follow-up during the 365 days after cohort entry. The primary analysis cohort comprised 89 490 eligible patients who had at least one dispensation for a medication used for prior adherence measurement (please see patient flow chart in Appendix 2). The average age of patients in the analysis cohort was 55 years, and 54% were female. The most common comorbidities were hypertension (56%), diabetes (26%), depression (15%), and cancer (13%) (Table 2). The median baseline PDC was 58%, and the 25th and 75th percentiles were 25% and 90%, respectively.

Table 2. Baseline Characteristics of Statin Users in Aim 1, Part I (n = 89 490)

Characteristic Result Age, mean (SD) 54.5 (10.2) Female, n (%) 48 713 (54.4) Regions, n (%) Midwest 20 893 (23.3) Northeast 6997 (7.8) South 50 067 (55.9) West 11 533 (12.9)

Use of preventive services Fecal occult blood tests, n (%) 8344 (9.3) Colonoscopy, n (%) 7888 (8.8) Mammography, n (% of females) 17 125 (35.2) Number of hospitalizations in prior year, mean (SD) 0.17 (0.6) Total days in hospital in prior year, mean (SD) 0.89 (4.6)

Comorbidities, n (%) Peripheral vascular disease 3031 (3.4) Liver disease 3895 (4.4) Renal disease 4833 (5.4) Recent MI 1181 (1.3) Prior MI 1524 (1.7)

Recent stroke Prior stroke 809 (0.9) Ischemic heart disease 10 398 (11.6) Transient ischemic attack 290 (0.3) Hypertension 50 207 (56.1) Diabetes 23 232 (26.0) Depression 13 061 (14.6)

Characteristic Result Cancer 11 860 (13.3) Combined comorbidity score, mean (SD) 0.17 (1.5)

Baseline medication use, n (%) ACEIs 31 028 (34.7) ARBs 17 156 (19.2) Beta-blockers 22 627 (25.3) Calcium channel blockers 18 519 (20.7) Thiazides 29 711 (33.2) Oral anticoagulants 3021 (3.4) Antiplatelets 3421 (3.8) Antidiabetics 22 515 (25.2) NSAIDs 20 305 (22.7) SSRIs 19 350 (21.6) Other lipid-lowering agents 4392 (4.9)

Index statin, n (%) Atorvastatin 15 067 (16.8) Fluvastatin 104 (0.1) Lovastatin 4382 (4.9) Pitavastatin 1046 (1.2) Pravastatin 16 384 (18.3) Rosuvastatin 12 711 (14.2) Simvastatin 39 796 (44.5) High-intensity dose 8021 (9.0) Refill synchronization measure, n (%) 0.26 (0.2)

Medication burden, mean (SD) Number of drug dispensations in prior year 26.5 (21.9) Number of unique drugs dispensed in prior year 7.6 (5.2)

Characteristic Result Number of concurrently used medications 2.4 (2.3)

Health plan related: Plan types, n (%) EPO 14 146 (15.8) HMO 7439 (8.3) IND 3609 (4.0) OTH 103 (0.1) POS 61 500 (68.7) PPO 2693 (3.0) Total deductibles in US dollars, mean (SD) 558.7 (554.1) Copay of index statin in US dollars, mean (SD) 17.0 (19.4) Abbreviations: ACEIs, angiotensin-converting enzyme inhibitor; ARBs, angiotensin II receptor blocker; EPO, exclusive provider organization; HMO, health maintenance organization; IND, individual health plan; MI, myocardial infarction; NSAIDs, nonsteroidal anti-inflammatory drugs; OTH, other; POS, point-of-service; PPO, preferred provider organization; SD, standard deviation; SSRIs, selective serotonin reuptake inhibitors.

Statin Adherence Prediction Models Using the basic set of variables selected by lasso models—including variables related to medication burden (eg, number of any drug dispensations, number of concurrently used medications) as the only explanatory variables—yielded the highest C statistic for discriminating between those with and without full statin adherence in the testing cohort (0.614; 95% CI, 0.609-0.619; Table 3), followed by demographics (0.578; 95% CI, 0.573-0.584) and proxies of drug cost burden (0.575; 95% CI, 0.570-0.581). The model with only comorbidities had the lowest C statistics (0.540; 95% CI, 0.535-0.545). Combining all such baseline data components yielded a C statistic of 0.665 (95% CI, 0.659-0.670).

Table 3. C Statistics From Models Predicting High Statin Adherence (PDC ≥ 80%) Including Different Sets of Patient Baseline Information in the Testing Cohort, in Aim 1, Part I

C Statistic Variables Included in Each Model (95% CI) Demographics (age, sex) 0.578 (0.573-0.584) Health services use (number of physician visits, 0.549 (0.543-0.554) colonoscopy, mammography, vaccinations, and hospitalized days) Comorbidities (PVD, liver disease, renal disease, recent 0.540 (0.535-0.545) MI, prior stroke, ischemic heart disease, hypertension, diabetes, depression, cancer) Baseline medication usea 0.545 (0.539-0.551) Index statin information (high-/low-intensity dose, 0.543 (0.538-0.547) dispensed days’ supply) Medication burden (number of any drug dispensations, of 0.614 (0.609-0.619) unique drugs, and of concurrent medication dispensed) Financial burden (plan benefit type, total copayment 0.575 (0.570-0.581) during baseline, copay of the index statin) All components combined 0.665 (0.659-0.670) Abbreviations: MI, myocardial infarction; PVD, peripheral vascular disease. a Baseline medication use is specified in Table 2.

Among models including only prior adherence measures, the continuous coverage measures achieved the highest C statistics (Table 4; range: 0.614-0.666). Among them, PDC yielded the highest C statistic -up (C statistic for the model with max PDC: 0.666; 95% CI, 0.661- 0.671). Minimum PDC or minimum continuous measure of medication acquisition (CMA) yielded slightly lower C statistics compared with the mean, median, and maximum of the 2, which were all similar. Count measures including refill counts and lack of second fill measures achieved the poorest discrimination (C statistics: 0.533-0.575).

Table 4. C Statistics From Models Predicting High Statin Adherence (PDC ≥ 80%) Including Only Measures of Prior Adherence in the Testing Cohort in Aim 1, Part I

C Statistics Variables Included in Each Model (95% CI)

Univariable models

Refill count measures Any “no second fill” 0.533 (0.529-0.537) Median refills 0.574 (0.569-0.580) Mean refills 0.575 (0.569-0.580)

Discontinuation measures Any discontinuation 0.587 (0.583-0.592) Mean TTD 0.625 (0.620-0.630)

Coverage measures Min CMA 0.614 (0.608-0.619) Max CMA 0.640 (0.635-0.646) Median CMA 0.639 (0.634-0.644) Mean CMA 0.640 (0.635-0.646) Min CMG 0.661 (0.656-0.666) Max CMG 0.629 (0.623-0.634) Median CMG 0.659 (0.654-0.665) Mean CMG 0.657 (0.651-0.662) Min PDC 0.633 (0.628-0.639) Max PDC 0.666 (0.661-0.671) Median PDC 0.665 (0.659-0.670) Mean PDC 0.663 (0.658-0.668) Abbreviations: CMA, continuous measure of medication acquisition; CMG, continuous measure of medication gaps; PDC, proportion of days covered; TTD, time to discontinuation.

Combining prior adherence measures with the baseline data components increased C statistics (Table 5, models [a] through [d]), with addition of mean PDC achieving the highest C statistic (0.696; 95% CI, 0.691-0.701) and good calibration as portrayed in the calibration plot (Figure 3). Adding combinations of different prior adherence measures to the basic set of predictors selected by lasso did not result in material improvement over the model with a single prior adherence measure (Table 5, models [e] through [g]). Excluding variables defined using medical data yielded a C statistic of 0.687 (95% CI, 0.682-0.692; Table 5, model [h]).

Table 5. C Statistics From Models Predicting High Statin Adherence (PDC ≥ 80%) Including Combinations of Patient Baseline Information and Prior Adherence Measures in the Testing Cohort, Aim 1, Part I

Variables Included in Each Model C Statistic (95% CI)

Clinical variables + prior adherence measures (a) All baseline informationa + median CMA 0.684 (0.679-0.689) (b) All baseline informationa + mean CMA 0.685 (0.680-0.690) (c) All baseline informationa + median PDC 0.694 (0.689-0.699) (d) All baseline informationa + mean PDC 0.695 (0.690-0.700) (e) All baseline informationa + mean PDC + mean refills 0.696 (0.691-0.701) (f) All baseline informationa + mean PDC + DD 0.695 (0.690-0.700) (g) All baseline informationa + mean PDC + DD + no second 0.696 (0.691-0.701) fill + mean TTD + mean CMA

Pharmacy-based data components (h) Demographics + medication use + index statin 0.687 (0.682-0.692) information + medication burden + mean PDC Abbreviations: CMA, continuous measure of medication acquisition; DD, discontinuation (dichotomous); PDC, proportion of days covered; TTD, time to discontinuation. a “All baseline information” includes all variables listed in Table 2: patient demographics, health services use, comorbidities, medication use, index statin information, medication burden, and financial burden.

Figure 3. Calibration plots for fixed- and random-effects unified models

Table 6 presents the beta-coefficients for the variables and the intercept from the model that included basic predictors and mean prior PDC; this was the final model used in subsequent analyses for aims 2 and 3. In addition to prior adherence, female sex (odds ratio [OR], 0.73; 95% CI, 0.70-0.77), history of mammography exams (OR, 1.21; 95% CI, 1.14-1.28), past medical events including MI and stroke, past or current use of oral anticoagulants and nonstatin lipid-lowering drugs, and the use of high-dose statins were strong predictors of full adherence. Appendix 3 presents the information for the model without medical data.

Table 6. Coefficients and Odds Ratios From the Final Statin Adherence Prediction Model in Aim 1, Part I

Variables Coefficients Odds Ratios (95% CI) Intercept –2.976 Demographics Age 0.015 1.015 1.012 1.017 Female sex –0.315 0.730 0.695 0.767 Health services Physician visits 0.005 1.005 1.003 1.007 use Hospitalizations 0.065 1.067 1.017 1.119 Colonoscopy 0.071 1.073 0.998 1.154 Mammography 0.188 1.207 1.138 1.281 Vaccinations 0.099 1.104 1.062 1.147 Comorbidities PVD –0.123 0.884 0.786 0.995 Prior liver –0.140 0.869 0.783 0.965 Renal –0.094 0.911 0.825 1.005 Recent MI 0.847 2.332 1.929 2.819 Prior stroke 0.604 1.83 1.468 2.281 Ischemic heart 0.151 1.163 1.081 1.251 Diabetes –0.101 0.904 0.835 0.98 Cancer 0.062 1.064 1.000 1.132 Drug use ACEIs –0.003 0.997 0.950 1.047 Antidiabetics –0.081 0.922 0.848 1.003 Antiparkinson agents –0.195 0.823 0.694 0.975 Antiplatelets –0.122 0.886 0.790 0.992 ARBs –0.140 0.869 0.818 0.923 Calcium channel blockers –0.162 0.851 0.805 0.899 Digoxin –0.324 0.724 0.568 0.921 Loop diuretics 0.038 1.039 0.938 1.15 NSAIDs –0.058 0.944 0.895 0.995 Oral anticoagulants 0.263 1.301 1.154 1.468 SSRIs –0.001 0.999 0.946 1.056 TCAs 0.187 1.205 1.065 1.364 Thiazide diuretics –0.013 0.987 0.937 1.039

Variables Coefficients Odds Ratios (95% CI) Nonstatin lipid-lowering 0.399 1.49 1.353 1.64 drugs Number of any drug 0.008 1.008 1.005 1.01 Prescription Number of unique drug –0.053 0.949 0.941 0.956 burden type Number of concurrently 0.085 1.089 1.075 1.104 prescribed drugs as the index statin Number of drugs among –0.001 0.999 0.996 1.002 those listed above Refill synchronization –0.144 0.866 0.778 0.964 Sum of copayment for 0.0001 1.000 1.000 1.000 drugs Baseline Plan type EPO –0.3259 0.722 0.635 0.82 financial burden Plan type HMO –0.1097 0.896 0.783 1.026 Plan type IND 0.0695 1.072 0.917 1.253 Plan type OTH 0.1437 1.155 0.662 2.012 Plan type POS –0.0983 0.906 0.805 1.02 Index statin Copay amount –0.005 0.996 0.994 0.997 information Days’ supply 0.007 1.007 1.006 1.008 High-dose statin (No vs 0.2539 1.289 1.192 1.394 Yes) Prior medication Mean PDC (PDC = 1 vs 1.788 5.985 5.410 6.621 adherence PDC = 0) Abbreviations: ACEIs, angiotensin-converting enzyme inhibitors; ARB, angiotensin II receptor blockers; EPO, exclusive provider organization; HMO, health maintenance organization; IND, individual health plan; MI, myocardial infarction; NSAIDs, nonsteroidal anti-inflammatory drugs; OTH, other; PDC, proportion of days covered; POS, point-of-service; PVD, peripheral vascular disease; SSRIs, selective serotonin reuptake inhibitors, TCAs, tricyclic antidepressants.

We observed a strong association between prior adherence and future adherence after adjusting for all predefined clinical variables. For patients with mean prior PDC < 25%, and for those with mean prior PDC f 25% to 79%, the likelihood of having high adherence to newly initiated statins was 51% (risk ratio, 0.49; 95% CI, 0.46-0.50) and 36% (risk ratio, 0.64; 95% CI, 0.62-0.65) lower, respectively, compared with those with mean prior PDC 80% and above.

Aim 1, Part II: Assessing Transportability of the Statin Adherence Prediction Algorithm to Other Drug Cohorts

Comparison of the Study Cohorts In addition to statin initiators identified in aim 1, part I, 67 607 bisphosphonate and 109 059 ACEI/ARB initiators were included in the part II analyses. Several patient characteristics differed across drug cohorts, including age, sex, and prior adherence patterns (Table 7). The proportions of patients classified as fully adherent (ie, PDC ≥ 80%) were similar between statin (35%) and bisphosphonate (34%) cohorts but were higher for ACEI/ARB initiators (46%).

Table 7. Baseline Characteristics of the 3 Drug Cohorts in Aim 1, Part II

Bisphosphonate Statin Cohort Cohort (n = 67 ACEI/ARB Cohort (n = 89 490) 607) (n = 109 059) Full adherence, n (%)a 31 001 (34.6) 22 704 (33.6) 50 340 (46.2) Age, mean (SD) 54.54 (10.2) 60.37 (9.9) 64.53 (13.7) Female, n (%) 48 713 (54.4) 61 614 (91.1) 58 063 (53.2) Regions, n (%) Midwest 20 893 (23.3) 15 513 (22.9) 23 136 (21.2) Northeast 6997 (7.8) 6247 (9.2) 10 389 (9.5) South 50 067 (55.9) 34 516 (51.1) 48 428 (44.4) West 11 533 (12.9) 11 331 (16.8) 27 106 (24.9) Number of physician office 13.61 (14.2) 18.52 (18.0) 7.83 (6.2) visits, mean (SD) Total days in hospital in 0.89 (4.6) 1.10 (5.4) 1.47 (6.1) prior year, mean (SD) Use of preventive services, n (%) Fecal occult blood test 8344 (9.3) 10 571 (15.6) 9501 (8.7) Colonoscopy 7888 (8.8) 8920 (13.2) 9834 (9.0) Mammography, 17 125 (35.2) 25 308 (41.1) 19 956 (34.4) Comorbidities, n (%) Myocardial infarction 1524 (1.7) 222 (0.3) 2170 (2.0) Stroke 809 (0.9) 157 (0.2) 1044 (1.0) Ischemic heart disease 10 398 (11.6) 6151 (9.1) 20 987 (19.2) Hypertension 50 207 (56.1) 28 982 (42.9) 87 293 (80.0) Diabetes 23 232 (26.0) 8832 (13.1) 40 424 (37.1) Hyperlipidemia 65 477 (73.2) 39 230 (58.0) 75 160 (68.9) CHF 3159 (3.5) 2291 (3.4) 10 660 (9.8) Parkinson’s disease 221 (0.2) 336 (0.5) 835 (0.8)

Bisphosphonate Statin Cohort Cohort (n = 67 ACEI/ARB Cohort (n = 89 490) 607) (n = 109 059) Alzheimer’s disease or 954 (1.1) 1404 (2.1) 5198 (4.8) dementia Depression 13 061 (14.6) 10 613 (15.7) 16 170 (14.8) Rheumatoid arthritis 1361 (1.5) 3104 (4.6) 2485 (2.3) Cancer 11 860 (13.3) 12 958 (19.2) 38 322 (35.1) Combined comorbidity 0.17 (1.5) 0.48 (1.6) 0.82 (2.2) score, mean (SD) Baseline medication use Antidiabetics 22 515 (25.2) 8018 (11.9) 33 228 (30.5) ACEIs 31 028 (34.7) 15 303 (22.6) − ARBs 17 156 (19.2) 11 127 (16.5) − Beta-blockers 22 627 (25.3) 16 093 (23.8) 43 353 (39.8) Calcium channel blockers 18 519 (20.7) 11 919 (17.6) 30 718 (28.2) Thiazides 29 711 (33.2) 18 315 (27.1) 38 227 (35.1) Other antihypertensives 3433 (3.8) 1909 (2.8) 7902 (7.2) Oral anticoagulants 3021 (3.4) 2994 (4.4) 7789 (7.1) NSAIDs 20 305 (22.7) 15 523 (23.0) 24 990 (22.9) COX-2 inhibitor 2031 (2.3) 3363 (5.0) 2635 (2.4) DMARDs-biologic 483 (0.5) 832 (1.2) 466 (0.4) DMARDs-nonbiologic 1449 (1.6) 3793 (5.6) 2297 (2.1) Antiparkinson agents 1349 (1.5) 1535 (2.3) 2697 (2.5) SSNRIs 4934 (5.5) 4725 (7.0) 5511 (5.1) SSRIs 19 350 (21.6) 17 270 (25.5) 20 279 (18.6) TCAs 2518 (2.8) 3107 (4.6) 4035 (3.7) Other antidepressants 6272 (7.0) 6161 (9.1) 8549 (7.8) Number of any medication 26.46 (21.9) 33.77 (28.2) 30.86 (27.1) fills, mean (SD)

Bisphosphonate Statin Cohort Cohort (n = 67 ACEI/ARB Cohort (n = 89 490) 607) (n = 109 059) Number of unique 7.58 (5.2) 9.25 (6.2) 8.38 (5.6) medications filled, mean (SD) Number of concurrently 2.41 (2.3) 1.80 (2.1) 1.75 (2.0) used medications, mean (SD) Number of chronic 14.15 (11.8) 16.32 (14.6) 14.87 (13.4) medications, mean (SD) Refill synchronization 0.26 (0.2) 0.41 (0.3) 0.33 (0.2) metric, mean (SD) Total copayment, 558.70 (554.1) 716.83 (681.2) 507.90 (609.9) mean (SD) Primary nonadherence to 3694 (4.1) 1806 (2.7) − antihypertensives Adherence to prior 0.69 (0.3) 0.73 (0.2) 0.70 (0.3) medications, mean (SD) Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; CHF, congestive heart failure; DMARDs, disease-modifying antirheumatic drugs; NSAIDs, nonsteroidal anti-inflammatory drugs; SD, standard deviation; SSNRIs, selective serotonin-norepinephrine reuptake inhibitors; SSRIs, selective serotonin reuptake inhibitors; TCAs, tricyclic antidepressants. a Study outcome: full adherence was defined as PDC ≥ 80%.

The coefficients describing the strength of the associations between potential predictors and subsequent adherence varied across cohorts (Table 8). For example, a history of MI had a coefficient (standard error) of 0.696 (0.061) for adherence to statins and coefficients of –0.012 (0.161) and 0.270 (0.050) for adherence to ACEI/ARBs and bisphosphonates, respectively.

Table 8. Logistic Regression Coefficients of the Selected Predictors Estimated in Each of the 3 Study Cohorts in Aim 1, Part II

Bisphosphonate ACEI/ARB Statin Cohort Cohort Cohort ß SE ß SE ß SE Age 0.013 0.001 0.008 0.001 0.006 0.001 Sex (female) –0.320 0.018 –0.213 0.031 –0.099 0.015 Fecal occult blood test 0.071 0.026 0.159 0.023 0.060 0.023 Colonoscopy 0.090 0.026 0.134 0.025 0.065 0.023 Myocardial infarction 0.696 0.061 –0.012 0.161 0.270 0.050 Stroke 0.570 0.079 –0.210 0.190 0.339 0.069 Ischemic heart disease 0.131 0.028 –0.117 0.035 –0.079 0.020 Hypertension 0.023 0.020 –0.080 0.024 0.031 0.015 Hyperlipidemia 0.003 0.017 –0.046 0.018 0.031 0.015 Depression –0.077 0.025 –0.124 0.029 –0.124 0.029 Rheumatoid arthritis –0.081 0.072 –0.167 0.054 –0.123 0.051 Cancer 0.077 0.023 0.114 0.023 –0.009 0.014 Antidiabetics –0.112 0.030 –0.240 0.049 –0.153 0.023 Beta-blockers –0.031 0.019 –0.166 0.023 0.055 0.015 Oral anticoagulants 0.200 0.052 –0.009 0.052 0.082 0.033 Other antihypertensives 0.015 0.040 –0.117 0.054 –0.060 0.026 Other antidepressants 0.030 0.031 –0.125 0.033 –0.021 0.026 COX-2 inhibitor 0.051 0.050 –0.069 0.041 –0.022 0.042 SSNRIs 0.096 0.035 –0.339 0.039 0.047 0.032 SSRIs 0.003 0.022 –0.281 0.023 –0.072 0.019 TCAs 0.166 0.045 –0.082 0.043 –0.053 0.035 Number of concurrent medications 0.081 0.005 0.123 0.006 0.080 0.004 Primary nonadherence to –0.013 0.043 –0.175 0.059 − − antihypertensives Mail synchronization metric –0.158 0.039 0.263 0.038 –0.011 0.029 Average PDC to prior medications 1.782 0.037 0.865 0.044 1.545 0.032 Abbreviations: PDC, proportion of days covered; SSNRIs, selective serotonin-norepinephrine reuptake inhibitors; SSRIs, selective serotonin reuptake inhibitors; TCAs, tricyclic antidepressants.

Model Development and Performance The model developed in the statin cohort performed less well when applied to the bisphosphonate cohort, with a C statistic of 0.638 (Table 9) compared with 0.695 from the statin cohort.

Table 9. Comparison of Adherence Prediction Model Performance in Statin and Bisphosphonate Cohorts in Aim 1, Part II

Reestimation With Reestimation in Additional Variables Statin Bisphosphonate Bisphosphonate in Bisphosphonate Cohorta Cohortb Cohorta Cohorta,c C statistic 0.695 0.638 0.665 0.679 Calibration slope 0.990 0.695 0.982 0.981 a Evaluated using 10-fold cross-validation and by taking the average across folds. b Estimated by applying the model developed from the statin initiator cohort in the bisphosphonate initiator cohort. c Added variables are specified in Appendix 4.

The same model performed slightly better in predicting adherence to ACEI/ARBs (C statistic, 0.654), but still considerably lower than among statin initiators (Table 10).

Table 10. Model Performance in the Developing (Statin Initiators) and Validating Cohort (ACEI/ARB Initiators)

Reestimation With Reestimation in Additional Variables Statin ACEI/ARB ACEI/ARB in ACEI/ARB Cohorta Cohortb Cohorta Cohorta,c C statistic 0.694 0.654 0.672 0.674 Calibration slope 0.991 0.718 0.991 0.989 Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker. Note: Variables excluded: ACEI, ARB, and primary nonadherence to hypertension medications. a Evaluated using 10-fold cross-validation and by taking the average across folds. b Estimated by applying the model developed from the statin initiator cohort in the ACEI/ARB initiator cohort. c Added variables are specified in Appendix 5.

Reestimating the coefficients in the bisphosphonate and ACEI/ARB cohorts improved the discriminative ability substantially, with C statistics from the refitted models of 0.665 and 0.672 in the bisphosphonate cohort and ACEI/ARB cohort, respectively. Adding variables specific to the clinical scenario further improved model discrimination in the bisphosphonate cohort (0.679) but made little improvement in the ACEI/ARB cohort (0.674). Transporting the statin model to the bisphosphonate and ACEI/ARB cohorts also resulted in substantial miscalibration, as evidenced by decreases in the calibration slope from 0.990 to 0.695 and 0.718 for the bisphosphonate cohort and ACEI/ARB cohort, respectively (Tables 9 and 10).

Predictor Importance To assess the relative importance of each variable in separately predicting adherence to statins, bisphosphonates, and ACEI/ARBs, we calculated the frequency that each was selected by lasso in the 10-fold cross-validation. A total of 68, 70, and 66 variables were selected in at least 50% of the cross-validations for the statin, bisphosphonate, and ACEI/ARB cohorts, respectively (Appendix 6). Frequently selected variables among the common set of predictors available for all cohorts included age, sex, region, and receipt of colonoscopy, mammography, and vaccinations. Prior mean PDC for maintenance medications dispensed during baseline was selected in 100% of the models. Selection of certain variables varied across cohorts. Some of the more frequently selected variables for the statin cohort included history of cardiovascular events, such as MI and stroke, while bisphosphonate-specific variables included rheumatoid arthritis, number of BMD tests, and baseline use of other osteoporosis medications. Appendix 6presents the frequency with which each variable was selected among the 10-fold cross- validation.

Variable influence also varied across the 3 drug cohorts in the GBMs. Prior mean PDC was the strongest predictor of outcome adherence for all 3 cohorts, though the extent of its contribution to the models varied across the cohorts (Appendix 7). Its relative influence was more than 45% in predicting adherence to statins and ACEI/ARBs and approximately 25% in the model predicting adherence to bisphosphonates. In addition to variables reflecting medication burden (eg, number of chronic medications, concurrent medications), which tended to be

important in predicting adherence in all 3 drug cohorts, demographic variables, such as age and region, were more important in the statin cohort; however, proxies for osteoporosis severity, such as BMD tests and use of other osteoporosis medications, contributed more in predicting adherence to bisphosphonates.

Aim 1, Part III: Identifying and Evaluating a Single, Unified Adherence Prediction Model for Application Across Different Drug Classes Keeping in mind the differences in relative importance and influence identified in part II, we sought to evaluate in part III whether a universal model with a common set of variables could predict adherence to each of the 3 drug classes of interest. The unified model included those variables in part II that were selected in at least 50% of the cross-validations. The C statistic for the universal model was 0.675 (95% CI, 0.673-0.677) for the fixed-effects model and 0.684 (0.682-0.686) for the random-effects model. These values were higher than the C statistics for the models refit to the bisphosphonate (0.665) and ACEI/ARB (0.672) cohorts in part II but lower than the C statistic from the statin model in aim 1 (0.696; 95% CI, 0.691-0.701).

The ORs relating the predictors to the adherence outcome were similar for the fixed- effects and random-effects models (Appendix 8). In the random-effects model, the estimated intercept was −2.057, and the estimated variance of the random effects was 0.054; this was significantly different from 0 (p = 0.005), indicating substantial variation in baseline risk of the outcome across the drug classes. The estimated ORs reported in Appendix 8 can be interpreted as the associations between the variables and the outcome based on within-drug-class comparisons. In addition to a higher C statistic, including the random intercept to allow the baseline probability of outcome adherence to vary by drug class also improved calibration, as evidenced by the calibration plots from the 2 models (Figure 3).

Aim 2: Assess Whether Predicted Adherence Explains Heterogeneity of Treatment Effects in PCOR Studies

High-intensity vs Low-intensity Statin Cohorts For the aim 2 statins cohort, we identified 2 090 347 eligible statin initiators who were at least 18 years of age and enrolled in a health care plan for a minimum of 12 months before treatment initiation (Appendix 9). Of them, 1 523 949 were low-intensity and 566 398 were high-intensity statin users. The prior mean (SD) PDC to other maintenance medications was 0.76 (0.26); a total of 481 479 (23%) statin initiators were excluded from the primary analysis because they did not have the PDC information.

We used the final statin adherence prediction model from aim 1 to estimate each patient’s adherence prediction score. We then stratified patients into tertiles of predicted adherence based on these scores. As expected, the prior mean PDC as well as the average PDC to statins in follow-up increased monotonically from the first to third tertile (Table 11). Prior mean (SD) PDC was 0.53 (0.26) in the first tertile, 0.84 (0.18) in the second tertile, and 0.91 (0.13) in the third tertile. Predicted adherence was similar for low- and high-intensity statin initiators within each tertile even before balancing by propensity score.

Table 11. Descriptive Statistics for Adherence in Baseline and Follow-up Before Propensity Score Matching in the Statin Cohort

Exposure Lower Upper Group Variable Mean Median Quartile Quartile Min Max Tertile 1 Low-intensity Adherence score 1.81 1.92 1.54 2.17 –2.22 2.36 statins Mean PDC (baseline) 0.53 0.54 0.33 0.72 0.00 1.00 Mean PDC (follow-up) 0.40 0.39 0.16 0.61 0.00 1.00 High-intensity Adherence score 1.79 1.89 1.50 2.16 –1.13 2.36 statins Mean PDC (baseline) 0.53 0.54 0.33 0.73 0.00 1.00 Mean PDC (follow-up) 0.40 0.38 0.16 0.61 0.00 1.00 Tertile 2 Low-intensity Adherence score 2.66 2.67 2.52 2.80 2.36 2.95 statins Mean PDC (baseline) 0.84 0.89 0.74 1.00 0.00 1.00 Mean PDC (follow-up) 0.58 0.61 0.35 0.84 0.00 1.00 High-intensity Adherence score 2.66 2.66 2.52 2.80 2.36 2.95 statins Mean PDC (baseline) 0.85 0.91 0.75 1.00 0.00 1.00 Mean PDC (follow-up) 0.57 0.60 0.33 0.83 0.00 1.00 Tertile 3 Low-intensity Adherence score 3.40 3.31 3.12 3.59 2.95 7.26 statins Mean PDC (baseline) 0.91 0.96 0.86 1.00 0.01 1.00 Mean PDC (follow-up) 0.66 0.72 0.48 0.91 0.00 1.00 High-intensity Adherence score 3.41 3.32 3.11 3.62 2.95 7.53 statins Mean PDC (baseline) 0.91 0.97 0.86 1.00 0.01 1.00 Mean PDC (follow-up) 0.63 0.67 0.42 0.87 0.00 1.00

Exposure Lower Upper Group Variable Mean Median Quartile Quartile Min Max Overall data Low-intensity Adherence score 2.64 2.68 2.19 3.13 –2.22 7.26 statins Mean PDC (baseline) 0.76 0.84 0.61 0.98 0.00 1.00 Mean PDC (follow-up) 0.55 0.57 0.31 0.82 0.00 1.00 High-intensity Adherence score 2.58 2.61 2.10 3.08 –1.13 7.53 statins Mean PDC (baseline) 0.75 0.84 0.59 0.99 0.00 1.00 Mean PDC (follow-up) 0.53 0.54 0.28 0.78 0.00 1.00 Abbreviation: PDC, proportion of days covered.

Despite increases in mean age across tertiles (56, 60, and 69 years across the first through third tertiles, respectively), patients in the second tertile tended to be the healthiest, with a mean combined comorbidity score of 0.36 (1.77) compared with 0.52 (1.92) and 0.80 (2.20) for patients in the first and third tertile, respectively (Appendices 10-13). Between treatment groups, patients had similar mean age (60 [13.4] vs 59 [13] years); however, fewer low-intensity statin initiators were female, and low-intensity initiators had lower combined comorbidity scores (0.53 [1.89] vs 0.63 [1.96]). Low-intensity statin initiators also had a lower prevalence of heart failure, acute coronary syndrome, hyperlipidemia, hypertension, and ischemic heart disease. High-intensity statin users tended to use more antiplatelets, ARBs, beta- blockers, nitrates, and other lipid-lowering drugs. Differences in covariates between treatment groups tended to increase from the first to third tertile.

Across the tertiles, 94% (n = 147 442 pairs), 95% (n = 135 118 pairs), and 91% (n = 122 870 pairs) of high-intensity statin initiators were matched by propensity score, respectively. Within each stratum, patient characteristics between treatment groups were well balanced after matching (Appendices 14-17). In the matched cohorts, the prior mean PDC for both low- intensity and high-intensity statin groups were 0.53, 0.85, and 0.91 across tertiles (Appendix 18). The follow-up statin PDC was 0.40, 0.57, and 0.63 for both groups across tertiles (Appendix 18).

High-intensity vs Low-intensity Statins on Major Cardiovascular Events Within the propensity score–matched cohorts, the incidence rates for inpatient mortality were lower for high-intensity statin initiators compared with low-intensity statin initiators in the overall data and within tertiles in both the as-treated and ITT analyses (Table 12). In the overall cohort, the as-treated incidence rate per 1000 person-years (IR per 1000py) was 19.6 (2165 events) for low-intensity and 15.6 (1643 events) for high-intensity statin initiators. IR differences increased from –1.90 in the first tertile to –2.59 in the second tertile and –4.43 in the third tertile.

Table 12. Mortality Incidence Rates (IR) and IR Differences Between Low- and High-intensity Statin Initiators After Propensity Score Matching in As-treated Analysis

Low-intensity Statins High-intensity Statins No. of IR/1000 No. of IR/1000 events person-years events person-years IR difference Tertile 1 1106 12.90 910 11.01 –1.90 Tertile 2 1128 13.01 865 10.42 –2.59 Tertile 3 1842 21.43 1392 17.00 –4.43 Overall data 4165 16.06 3168 12.74 –3.32

The hazard ratios (HRs) for inpatient mortality were 17% (HR, 0.83; 95% CI, 0.75-0.93), 20% (HR, 0.80; 95% CI, 0.72-0.90), and 23% (HR, 0.77; 95% CI, 0.71-0.84) lower in high-intensity statins in the first, second, and third tertiles, respectively. Compared with that of low-intensity statins, the rate of inpatient mortality overall was 20% lower among high-intensity statin initiators (HR, 0.80; 95% CI, 0.76-0.85) in the as-treated analysis and 13% lower in the ITT analyses (HR, 0.80; 95% CI, 0.77-0.84). HRs in the ITT analyses were 0.86 (95% CI, 0.79-0.94), 0.83 (95% CI, 0.76-0.90), and 0.81 (95% CI, 0.76-0.86) in the first, second, and third tertile, respectively.

The incidence rates for MI and cerebrovascular event hospitalizations were similar for high-intensity and low-intensity statin initiators overall and within tertiles in both as-treated and ITT analyses (Appendices 19 and 20). Overall, the as-treated IR per 1000py for MI was 8.8 (2267 events) for low-intensity and 8.9 (2203 events) for high-intensity statin initiators. IRs per 1000py were 6.8 vs 6.5 (558 vs 556 events) in the first tertile, 6.8 vs 6.7 (564 vs 578 events) in the second tertile, and 13.4 vs 13.4 (1090 vs 1150 events) in the third tertile. Corresponding HRs comparing high- vs low-intensity statin initiators were 1.06 (95% CI, 0.92-1.21), 1.01 (95% CI, 0.89-1.16), and 1.00 (95% CI, 0.91-1.09), respectively, in as-treated analyses. HRs were 1.02 (95% CI, 0.92-1.12), 1.00 (0.91-1.11), and 0.94 (0.87-1.01), respectively, in the ITT analyses.

Overall, the as-treated IR per 1000py for cerebrovascular event hospitalizations was 12.8 (3294 events) for low-intensity and 13.3 (3285 events) for high-intensity statin initiators.

IRs per 1000py were 10.2 vs 10.1 (839 vs 861 events) in the first tertile, 10.9 vs 10.5 (898 vs 906 events) in the second tertile, and 18.8 vs 17.5 (1532 vs 1495 events) in the third tertile. Corresponding HRs comparing high- vs low-intensity statin initiators were 0.99 (95% CI, 0.89- 1.11), 0.98 (95% CI, 0.89-1.09), and 1.05 (95% CI, 0.97-1.14), respectively. HRs were 1.07 (95% CI, 0.98-1.16), 1.01 (0.93-1.09), and 1.08 (1.01-1.16), respectively, in the ITT analyses.

Bisphosphonates vs Calcitonin vs Raloxifene Cohorts We identified 589 788 eligible patients who were at least 18 years of age; were new users of bisphosphonates, calcitonin, or raloxifene (with no use of the index therapy or teriparatide in the prior year); and had no ICD-9-CM diagnosis code for Paget’s disease in the prior year and no fracture in the prior year. The overall cohort comprised 510 962 patients who initiated bisphosphonates, 26 851 who initiated calcitonin, and 51 975 who initiated raloxifene (Appendix 21).

The prior mean (SD) PDC for cohort members was 0.76 (0.23); 226 131 (50.4%) did not have prior PDC information. As expected, prior mean PDC and PDC for the index drug during follow-up increased monotonically across tertiles for all comparisons (Appendices 22-24).

Raloxifene initiators (62 years) were the youngest, followed by bisphosphonate initiators (65 years) and calcitonin initiators (68 years; Appendices 25-48). Before application of propensity score methods, the calcitonin group, compared with the bisphosphonate group, had a higher prevalence of several comorbid conditions, including coronary atherosclerosis (11% vs 7%), cancer, heart failure (8% vs 4%), chest pain (17% vs 13%), chronic obstructive pulmonary disease (COPD; 14% vs 6%), depression (11% vs 8%), history of falls (3% vs 1%), hypertension (40% vs 27%), and ischemic heart disease (11% vs 4%). Bisphosphonate initiators tended to have higher prevalence of comorbidities than did raloxifene initiators, including COPD (10% vs 6%), hyperlipidemia (45% vs 35%), hypertension (39% vs 27%), and renal dysfunction (6% vs 3%),

Bisphosphonates vs Calcitonin on Fracture Events After application of the propensity score fine stratification technique in the primary as- treated analyses (Table 13), the IR per 1000py for fractures in the overall cohort PDC values was 14.16 for bisphosphonates and 13.66 for calcitonin. IRs per 1000py across the first through third tertiles were 15.47, 11.38, and 15.44, respectively, for bisphosphonate initiators and 17.24, 9.58, and 13.91, respectively, for calcitonin initiators. The overall HR comparing bisphosphonate vs calcitonin initiators was 0.97 (95% CI, 0.96-1.35) in the as-treated analysis and 1.01 (95% CI, 0.83-1.24) in the ITT analysis.

Table 13. Fracture Incidence Rates (IR) and IR Differences Between Bisphosphonate and Calcitonin Initiators After Propensity Score Fine Stratification in As-treated Analysis

Bisphosphonates Calcitonin No. of IR/1000 No. of IR/1000 events person- years events person-years IR difference Tertile 1 719 15.47 29 17.24 –1.77 Tertile 2 558 11.38 15 9.58 1.79 Tertile 3 780 15.44 21 13.91 1.53 Overall data 2074 14.16 65 13.66 0.49

Bisphosphonates vs Raloxifene on Fracture Events After application of the propensity score fine stratification technique in the primary as- treated analyses, the IR per 1000py was 9.03 for bisphosphonate initiators and 6.57 for raloxifene initiators (Table 14). IRs per 1000py across the first through third tertiles were 8.43, 7.28, and 10.28, respectively, for the bisphosphonate initiators and 7.24, 5.38, and 7.07, respectively, for raloxifene initiators. The overall HR comparing bisphosphonate vs raloxifene initiators was 1.37 (95% CI, 1.06-1.79) in the as-treated analysis and 1.38 (95% CI, 1.12-1.75) in the ITT analysis.

Table 14. Fracture Incidence Rates (IR) and IR Differences Between Bisphosphonate and Raloxifene Initiators After Propensity Score Fine Stratification in As-treated Analysis

Bisphosphonates Raloxifene No. of IR/1000 No. of IR/1000 events person- years events person-years IR difference Tertile 1 389 8.43 21 7.24 1.19 Tertile 2 356 7.28 16 5.38 1.90 Tertile 3 524 10.28 23 7.07 3.21 Overall data 1324 9.03 60 6.57 2.46

Calcitonin vs Raloxifene on Fracture Events After application of the propensity score fine stratification technique in the primary as- treated analyses, the IRs per 1000py across the first through third tertiles were 17.28, 9.03, and 13.20, respectively, for the calcitonin initiators and 13.70, 7.85, and 9.73, respectively, for raloxifene initiators (Table 15). The overall HR comparing calcitonin vs raloxifene initiators was 1.35 (95% CI, 0.98-1.86) in the as-treated analysis and 1.39 (95% CI, 1.07-1.80) in the ITT analysis.

Table 15. Fracture Incidence Rates (IR) and IR Differences Between Raloxifene and Calcitonin Initiators After Propensity Score Fine Stratification in As-treated Analysis

Calcitonin Raloxifene No. of IR/1000 No. of IR/1000 events person-years events person-years IR difference Tertile 1 28 17.28 40 13.70 3.58 Tertile 2 14 9.03 24 7.85 1.18 Tertile 3 19 13.20 32 9.73 3.47 Overall data 65 13.71 92 9.98 3.74

Aim 3: Develop and Compare Different Algorithms for Predicting Adherence During Treatment

Patient Characteristics For aim 3, we identified 1 126 274 adult statin initiators between January 1, 2010, and September 30, 2014; 695 253 patients had at least one dispensation for a medication used for prior adherence measures and were eligible for the analysis. The mean (SD) age of the cohort was 63 (13) years, and 45% were male. Simvastatin was the most commonly used statin (36%), followed by atorvastatin (27%), and pravastatin (20%). Most patients had an ICD-9-CM diagnosis code for hyperlipidemia (80%) in the baseline period; 74% had codes for hypertension and 36% had codes for diabetes. Frequently used medications during the baseline period included ACEIs (41%), opioids (38%), beta-blockers (36%), thiazide diuretics (34%), antidiabetics (29%), and calcium channel blockers (27%). On average, patients received 25 prescriptions and 10 unique drug dispensations during the baseline period; on average, 5 unique drugs were concurrently dispensed when patients initiated statin treatment. The prior mean PDC to maintenance medications was 0.76 (Appendix 49).

Sequential Statin Dispensations During Follow-up Figure 4 depicts the number and proportion of patients who did and did not discontinue statin therapy after each prescription during follow-up. Among the 695 253 eligible patients, only 377 658 (54%) filled a second prescription; 265 160 (38%) did not fill their second prescription and 52 435 (8%) were censored, precluding us from observing whether they filled a second prescription. Less than 8% (n = 50 177 patients) filled 12 prescriptions during the follow- up period; however, among those patients who adhered to treatment at each sequential dispensation in follow-up, the proportion who continued adhering to statin treatment increased: 74% to 79% of patients filled their third through sixth prescriptions, and 83% to 88% filled their seventh through 12th prescriptions.

Figure 4. Distribution of patients with and without subsequent dispensation at each dispensation during follow-up

700000

600000

500000

400000

300000 No of patients with Rx 200000 No of patients without Rx 100000 No of patients with missing Rx 0

100% 90% 80% 70% 60% 50% 40% Proportion of patients with Rx among the eligible cohort 30% 20% Proportion of patients with Rx among those on previous 10% Rx 0%

Among those who continued to fill prescriptions, the proportion with hospitalization for any cause decreased from 3% before filling the second prescription to 1% before filling the 12th prescription. In addition, the proportion of patients receiving specific cardiovascular medications increased slightly over time, and we observed a decreased trend in discontinuation of other maintenance medications among those who continued to adhere to their statin therapy (Table 16).

Table 16. Distribution of Patient Characteristics, Adherence Measures for Statins, and Adherence to Other Medications During Follow-up

Prescription dispensation number 2 3 4 5 6 7 8 9 10 11 12 No. of patients with 377 658 278 956 220 790 73 296 132 212 109 602 92 635 80 200 69 442 60 802 50 177 dispensation Patient characteristics Days’ supply of each 30 30 30 30 30 30 30 30 30 30 30 statin Rx, median (30-30) (30-30) (30-30) (30-30) (30-30) (30-30) (30-30) (30-30) (30-30) (30-30) (30-30) (IQR) Copayment of each 9 8 8 7 7 7 6 6 6 6 6 statin Rx, median (3-15) (3-15) (3-15) (3-13) (3-10) (3-10) (3-10) (3-10) (3-10) (3-10) (2-10) (IQR) Benefit plan type, % POS 37.9 37.7 37.3 38.1 40.1 40.1 40.3 40.4 40.8 40.7 40.0 EPO 6.9 6.7 6.5 6.5 6.8 6.8 6.8 6.8 6.7 6.7 6.5 HMO 25.8 25.8 25.8 25.4 24.4 24.3 24.1 24.0 23.9 23.7 24.2 IND 2.0 2.0 2.1 2.2 1.9 1.9 1.8 1.8 1.9 1.9 2.0 PPO 6.4 6.3 6.3 6.1 5.7 5.5 5.4 5.3 5.3 5.2 5.3 OTH 21.1 21.4 21.9 21.7 21.2 21.4 21.5 21.6 21.5 21.7 22.1 All-cause 2.6 2.0 1.8 1.6 1.4 1.3 1.3 1.2 1.1 1.1 1.1 hospitalization, %

Prescription dispensation number 2 3 4 5 6 7 8 9 10 11 12 Use of specific classes of cardiovascular medications, % ACEIs 27.1 28.0 28.6 28.6 28.2 28.6 28.9 28.8 29.3 29.2 29.5 ARBs 13.6 14.2 14.7 14.8 14.6 14.9 15.0 15.1 15.6 15.5 15.4 Renin inhibitor 3.9 4.0 4.2 4.2 4.2 4.2 4.1 4.1 4.2 4.2 4.1 Beta-blockers 25.6 26.9 28.2 28.5 28.0 28.5 28.8 29.1 29.7 29.7 30.4 Calcium channel 18.1 18.7 19.4 19.4 19.0 19.2 19.6 19.7 20.0 20.1 20.1 blockers Thiazide diuretics 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 Loop diuretics 6.6 7.0 7.2 7.3 7.3 7.4 7.5 7.7 7.6 7.7 8.1 Potassium-sparing 3.9 4.1 4.2 4.3 4.1 4.2 4.2 4.3 4.3 4.3 4.6 diuretics Other 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.00 0.01 0.01 antihypertensives Antianginal agents 3.0 2.7 2.7 2.7 2.7 2.7 2.7 2.7 2.8 2.7 2.8 Oral anticoagulants 4.7 5.0 5.3 5.3 5.1 5.2 5.3 5.3 5.4 5.4 5.6 Heparin and LMWH 0.3 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.1 0.1 0.2 Antiplatelets 8.2 8.8 9.0 9.4 9.6 9.6 9.6 9.7 9.7 9.7 9.7 Time lag between –1 0 0 0 0 0 0 0 0 0 0 dispensations, days, (–4 to 3) (–3 to 4) (–2 to 4) (–2 to 4) (–2 to 3) (–2 to 3) (–2 to 3) (–2 to 3) (–2 to 3) (–2 to 3) (–2 to 2) median (IQR)

Prescription dispensation number 2 3 4 5 6 7 8 9 10 11 12 No. of medication 0.13 0.10 0.08 0.07 0.06 0.06 0.05 0.05 0.04 0.04 0.00 classes discontinued, (0.42) (0.37) (0.34) (0.31) (0.29) (0.27) (0.26) (0.25) (0.23) (0.22) (0.04) mean (SD) Abbreviations: ACEIs, angiotensin-converting enzyme inhibitors; ARBs, angiotensin II receptor blockers; EPO, exclusive provider organization; HMO, health maintenance organization; IND, individual health plan; IQR, interquartile range; LMWH, low-molecular-weight heparin; OTH, other; PPO, preferred provider organization; SD, standard deviation.

Performance of Baseline Information in Predicting Statin Discontinuation Similar to predicting full adherence in aim 1, using the baseline patient characteristics as independent variables for predicting discontinuation after the first statin dispensation, the model including medication burden showed the best performance in the validation cohort (n = 321 400) among those including only a single domain of variables (C statistic, 0.583), followed by medical cost burden (C statistic, 0.567), and information on index statins and other medications (C statistic, 0.557). The model including only demographic variables had the lowest C statistic (0.526). Combining all such data components into a single model yielded a C statistic of 0.624 (Table 17).

Table 17. C Statistics of Models Including Baseline Patient Characteristics and Prior Adherence Measures in Predicting Discontinuation After the First Statin Dispensation in the Testing Cohort

Variables Included in Each Model C Statistics Baseline patient characteristics Demographicsa 0.525 Information on index statinsb 0.557 Medical cost burdenc 0.566 Resource use 0.528 Comorbidities 0.546 Other medication use 0.549 Medication burdend 0.582 All baseline patient characteristics combined 0.624 Prior adherence measures Mean PDC 0.588 Median PDC 0.588 Maximum PDC 0.579 Minimum PDC 0.570 Medication classes discontinued 0.532 Mean PDC + number of medication classes discontinued 0.589 Median PDC + number of medication classes discontinued 0.590 Maximum PDC + number of medication classes discontinued 0.588 Minimum PDC + number of medication classes discontinued 0.569 Combination of baseline patient characteristics and prior adherence measures All baseline patient characteristics + mean PDC 0.643 All baseline patient characteristics + median PDC 0.642 All baseline patient characteristics + maximum PDC 0.639 All baseline patient characteristics + minimum PDC 0.638 All baseline patient characteristics + mean PDC + number of medication 0.645 classes discontinued All baseline patient characteristics + median PDC + number of medication 0.645 classes discontinued All baseline patient characteristics + maximum PDC + number of medication 0.644 classes discontinued All baseline patient characteristics + minimum PDC + number of medication 0.640 classes discontinued Abbreviation: PDC, proportion of days covered. a Age and sex. b Type and days’ supply for index statin. c Copayment for index statin, benefit plan type, and copayment for all drugs. d Number of dispensations, number of unique drugs, and number of concurrent unique drugs dispensed.

With respect to prior adherence measures, as in aim 1, the model including the prior mean or median PDC showed the best ability in predicting discontinuation after the initial dispensations (C statistic, 0.588), which was better than the performance of any individual baseline patient characteristics. The model including the number of medication classes that were discontinued in baseline yielded the lowest C statistic (0.532). Combining all baseline characteristics, prior mean PDC, and number of medication classes discontinued during the baseline period yielded a C statistic up to 0.645. The discriminative ability of this model, using only baseline information, decreased at each sequential dispensation during follow-up, with C statistics of 0.629 for predicting discontinuation after the second dispensation and 0.598 for predicting discontinuation after the 11th dispensation (Table 18).

Incorporating Dynamic Post-initiation Information for Predicting Statin Discontinuation Overall, in comparison with the models with only baseline information, the individual variables assessed after statin initiation provided little additional discriminative ability in predicting discontinuation after the second through 11th dispensations. Days’ supply of each statin prescription and the lag time between dispensations provided slightly better predictive performance than did other post-initiation information. Combining all baseline and follow-up variables measured up to the most recent dispensation offered modest improvement in C statistics, although the C statistics still declined over time (Table 18).

Prescription dispensing number 2 3 4 5 6 7 8 9 10 11 12 No. of patients with 321 400 180 59 134 647 102 239 75 050 61 613 51 285 43 715 37 475 32 483 26 431 dispensing Baseline model 0.645 0.629 0.613 0.606 0.596 0.593 0.586 0.583 0.588 0.578 0.598 + specific statin NAa 0.630 0.614 0.607 0.596 0.594 0.586 0.583 0.589 0.580 0.603 + statin days’ supply NAa 0.633 0.620 0.616 0.617 0.607 0.601 0.589 0.598 0.578 0.598 + statin copayment NAa 0.631 0.615 0.609 0.602 0.596 0.590 0.586 0.592 0.580 0.601 + plan type NAa 0.629 0.613 0.606 0.596 0.593 0.586 0.583 0.588 0.578 0.598 + all-cause hospitalization NAa 0.629 0.614 0.607 0.597 0.595 0.590 0.585 0.590 0.579 0.598 + use of specific NAa 0.631 0.614 0.608 0.598 0.595 0.589 0.585 0.592 0.585 0.607 cardiovascular medications + time lag between statin NAa 0.629 0.616 0.615 0.605 0.597 0.588 0.586 0.597 0.585 0.611 dispensations + no. of medication classes NAa 0.630 0.613 0.605 0.594 0.591 0.582 0.578 0.586 0.574 0.592 discontinued Baseline model + all NAa 0.636 0.624 0.615 0.624 0.613 0.608 0.596 0.609 0.596 0.624 variables above Abbreviation: NA, not applicable. a No additional variables after statin initiation are included in the model.

DISCUSSION Context for Study Results This project was designed to advance methods for predicting medication adherence using administrative claims data. Claims data are frequently used for CER of medications and are becoming increasingly available to clinicians at the point of service. In particular, we conducted a systematic evaluation of approaches to using measures of prior adherence to predict patients’ future adherence to a newly initiated medication, and we explored whether a universal adherence prediction score could be useful for predicting adherence to multiple drug classes. We also examined whether stratifying patients in CER studies by predicted adherence could explain apparent heterogeneity of treatment effects. Finally, we attempted to further advance adherence prediction in claims data by updating models with post-initiation information.

Aim 1: Develop and Compare Algorithms for Predicting Adherence to Various Interventions, Based on Data Collected Prior to the Start of Those Interventions Aim 1 focused on developing drug-specific as well as universal adherence prediction models for use in administrative claims data. Across 3 drug cohorts, we found that among claims-based predictors of adherence, measures of prior adherence are relatively strong predictors of future adherence and that adding measures of prior adherence to models containing usual claims-based adherence predictors modestly improved model discriminative ability; however, even the top-performing models achieved C statistics less than 0.70.

In applying a model developed specifically among statin initiators to other drug cohorts, we observed reduced model performance. This lack of calibration when transporting a prediction model from one population to another is often underappreciated.33 We found that performance could be improved by reestimating the model coefficients in the drug cohorts of interest. We also found that different variables may have different influences on adherence prediction for different drugs. Despite these variances, we did find that a model that included a

common set of variables that were predictors of adherence to each of the 3 drug classes performed relatively well in the pooled cohort.

Aim 2: Assess Whether Predicted Adherence Explains Heterogeneity of Treatment Effects in PCOR Studies We had hypothesized that stratifying patients in a CER study by their predicted adherence would reveal heterogeneity of treatment effects—that is, we expected to observe larger differences in treatment effects among patients who were predicted to adhere to the treatments than among patients who were less likely to adhere, as drugs work only if patients take them. As expected, we found that patients’ adherence to the study drugs in follow-up increased monotonically across strata of predicted adherence. We did find some indication in the high- vs low-intensity statin mortality comparison that such stratification might be useful for identifying patients who may be more likely to benefit from high-intensity statins, but not in the other outcomes we examined or in any of the 3 osteoporosis comparisons.

Aim 3: Develop and Compare Different Algorithms for Predicting Adherence During Treatment Aim 3 sought to build on aim 1 by assessing whether incorporating post-initiation information could further enhance claims-based medication adherence prediction algorithms. We observed several interesting findings relating to patterns of treatment discontinuation over time. For example, only about half of patients filled an initial statin prescription more than once. As one might expect, as long as patients continued receiving subsequent dispensations, their probability of continuing to adhere increased over time because those who discontinued treatment were selectively removed from further consideration. Also, as one might expect, the ability of baseline information to predict subsequent discontinuation decreased the further out from treatment initiation. We also observed a decline in C statistics over each of the 12 dispensations when variables measured in follow-up were added to the baseline variables, though this trend was not monotonic. With each sequential prescription dispensation, the population of patients who persisted with the medication decreased in size. It is possible that as the cohort dwindled, the distributions of measured variables that predict discontinuation

narrowed among those who remained, making it difficult to compare C statistics over time in these slightly different populations.34 Nevertheless, despite the interesting patterns that we observed, we found little incremental improvement from the addition of post-initiation information.

Uptake of Study Results Findings from aim 1 suggest that researchers or practitioners seeking to predict medication adherence using administrative claims data should incorporate measures of prior adherence to chronic medications when patients have been prescribed such treatments. Mean PDC for a wide range of maintenance medications used prior to the start of a new drug tended to perform relatively well. Findings from aim 1 also suggest that drug-specific adherence prediction models perform better than universal adherence prediction scores that are developed for one drug class and applied to another. These findings are consistent with prior research on other types of clinical prediction models.35,36 However, when it is infeasible to estimate or use a drug-specific model, a universal adherence prediction score may be useful. We expect that this finding would apply to other populations and settings as well—that is, it would be preferable to develop and validate an adherence prediction model in the specific population or setting in which it is to be applied. However, when this is not feasible, transporting a drug-specific model developed in a different population or setting, or even transporting a universal adherence prediction score, may be a viable option. Although adding measures of prior medication adherence improved the C statistics, the overall discriminative ability of the models remained low, likely owing to the limitations of claims data in capturing measures of adherence and its predictors (see Study Limitations section). The modest C statistics limit the utility of claims data for identifying patients who are likely to become nonadherent at the time of initiating a new treatment.

Results from aim 2 suggest that adopting the approach of stratifying CER studies by predicted adherence could potentially help to assess heterogeneity of treatment effects; however, this should be done with caution, as we saw patterns of results consistent with our hypothesis in only one use case.

Study Limitations Although measures of prior medication adherence were relatively strong predictors of future adherence, when compared with typical claims-based predictors of adherence, their inclusion resulted in models with only modest discriminative ability. As emphasized by the PAB, administrative claims data lack information on potentially important predictors of medication adherence, such as social support, adverse effects, and use of adherence interventions, such as pill boxes or other reminders, which may explain the ability to achieve only modest discrimination. Measures of prior adherence do not contribute to adherence prediction of all patients, as not all patients have had prior use of a chronic medication for assessing prior adherence. Furthermore, while claims data are accurate for identifying who has received a prescription dispensation that generated an insurance claim, they do not provide information about whether patients actually consume the medications that they receive. As such, claims- based measures of adherence, such as PDC, reflect only a proxy of patients’ true medication adherence behavior. Although prior studies have found reasonable concordance between claims-based measures of adherence and pill count (ie, when an interviewer counts the number of pills that a patient has available and compares this with the expected number had the patient taken the medication according to the prescribing instructions),37 the discrepancy between claims-based adherence and true adherence could also partly explain the modest discriminative ability that we observed.

There are at least 3 reasons that stratifying by adherence prediction did not lead to meaningful differences in treatment effects across strata for any of the osteoporosis examples in aim 2. First, it is possible that the modest discriminative ability of the claims-based adherence prediction scores precluded our ability to observe differences. For example, while the prior mean PDCs were 0.53, 0.85, and 0.91 across tertiles of adherence prediction score in the statins use case, differences in observed follow-up statin PDC were smaller at 0.40, 0.57, and 0.63, respectively. It is possible that larger differences in actual adherence are necessary to yield evidence of heterogeneity of treatment effects. Second, stratification by adherence prediction score not only sorted patients based on their predicted adherence but also resulted in groups of patients that differed in other ways across strata. In the statins use case, comorbidity burden,

as assessed by the combined comorbidity score, increased monotonically across tertiles of adherence prediction score. Outcome rates were similar in the first and second tertiles but were considerably larger in the third tertile. In the osteoporosis example, patients in the first tertile had the highest comorbidity burden and patients in the third tertile had the lowest burden. Those in the second tertile tended to have the lowest rates of facture, while those in the first and third tertiles had similar rates. These differences in patient characteristics and outcome event rates across strata complicate assessments of heterogeneity of treatment effects due to adherence prediction score. Finally, it is possible that there are no or only very small true differences in effects between compared treatments, especially over the relatively short 1-year duration of follow-up in these studies, which could have precluded us from observing differences in treatment effects across strata.

Furthermore, the CER studies in aim 2 are subject to potential biases common in claims- based observational studies, including unmeasured confounding and misclassification of exposures and outcomes. While the active comparator new-user design can help address both measured and unmeasured confounding, we cannot rule out differences between groups in certain potentially important confounders, such as the severity of the condition for which the drugs are prescribed.

Across all 3 aims, we focused on a limited set of drug cohorts and clinical outcomes, which may limit the generalizability of our findings.

Future Research To our knowledge, this was the first systematic evaluation of the use of prior medication adherence to predict future adherence and the first examination of whether predicted adherence can explain heterogeneity of treatment effects. Future research should assess whether our findings apply to other drugs, other clinical contexts, and other populations and settings. Future work should also consider whether complementary data sources may enrich claims-based medication adherence prediction. Recent studies suggest that electronic health record (EHR) data may add value to adherence prediction.38 Patient-reported information or data collected from mobile health applications should also be considered.

Although we found little evidence that incorporating post-initiation data meaningfully improves prediction of treatment discontinuation, more work is needed to predict when patients are at risk of discontinuing their medications, as this may be the optimal timing for administering adherence improvement interventions. This may be a particularly valuable setting for enriching claims data with complementary data sources, such as EHR data or patient-reported data. Recent work has found that medication adherence and life events among family members can be associated with an individual’s medication adherence.39,40 Future work should evaluate the extent to which this information can improve claims-based adherence prediction. It is possible that better adherence prediction could improve confounding adjustment in claims-based observational studies or could help inform treatment decisions at the time of prescribing if, for example, a patient is likely be more adherent to one treatment vs another.

CONCLUSIONS

Measures of prior medication adherence appear to be stronger predictors of future adherence compared with other usual claims-based predictors of adherence. Incorporating measures of prior adherence can improve claims-based adherence prediction models. Drug- specific adherence prediction models are likely to perform better than universal adherence prediction scores; however, when it is infeasible to develop and validate a drug- or cohort- specific score in a given population or setting, transporting a universal adherence prediction score may be viable. Overall, the ability of claims data to discriminate who will and will not adhere to a new medication remains modest, likely due to a lack of several important correlates of medication adherence in claims data as well as to misclassification of claims-based measures of adherence and their predictors.

Although future activities involving adherence prediction should consider the measure of prior adherence, we found little evidence that claims-based adherence scores can explain heterogeneity of treatment effects in CER studies. This is likely due, in part, to the fact that patients with different predicted adherence also differ in other important clinical characteristics, making it difficult to disentangle potential effect measure modification due to differences in adherence vs differences in other patient characteristics. We also found little evidence that incorporating post-initiation data meaningfully improves prediction of treatment discontinuation; however, this area deserves attention in future work, especially where claims data can be linked to complementary data sources.

REFERENCES

1. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353:487-497.

2. Benner JS, Glynn RJ, Mogun H, Neumann PJ, Weinstein MC, Avorn J. Long-term persistence in use of statin therapy in elderly patients. JAMA. 2002;288:455-461.

3. Antman EM, Hand M, Armstrong PW, et al. 2007 Focused update of the ACC/AHA 2004 Guidelines for the Management of Patients With ST-Elevation Myocardial Infarction: a report of the American College of Cardiology/American Heart Association Task Force on practice guidelines; developed in collaboration with the Canadian Cardiovascular Society endorsed by the American Academy of Family Physicians; 2007 writing group to review new evidence and update the ACC/AHA 2004 Guidelines for the Management of Patients With ST-Elevation Myocardial Infarction, writing on behalf of the 2004 writing committee. Circulation. 2008;117:296-329.

4. Desai NR, Choudhry NK. Impediments to adherence to post myocardial infarction medications. Curr Cardiol Rep. 2013;15:322.

5. Cutrona SL, Choudhry NK, Fischer MA, et al. Targeting cardiovascular medication adherence interventions. J Am Pharm Assoc. 2012;52:381-397.

6. Misono AS, Cutrona SL, Choudhry NK, et al. Healthcare information technology interventions to improve cardiovascular and diabetes medication adherence. Am J Manag Care. 2010;16:SP82-92.

7. Viswanathan M, Golin CE, Jones CD, et al. Interventions to improve adherence to self- administered medications for chronic diseases in the United States: a systematic review. Ann Intern Med. 2012;157:785-795.

8. Choudhry NK, Avorn J, Glynn RJ, et al; Post-Myocardial Infarction Free Rx Event and Economic Evaluation (MI FREEE) Trial. Full coverage for preventive medications after myocardial infarction. N Engl J Med. 2011;365:2088-2097.

9. Gagne JJ, Choudhry NK, Kesselheim AS, et al. Comparative effectiveness of generic and brand-name statins on patient outcomes: a cohort study. Ann Intern Med. 2014;161:400-407.

10. Glynn RJ, Buring JE, Hennekens CH. Concerns about run-in periods in randomized trials. JAMA. 1998;279:1526.

11. Ammassari A, Trotta MP, Murri R, et al. Correlates and predictors of adherence to highly active antiretroviral therapy: overview of published literature. J Acquir Immune Defic Syndr. 2002;31(Suppl 3):S123-127.

12. Borah B, Sacco P, Zarotsky V. Predictors of adherence among Alzheimer’s disease patients receiving oral therapy. Curr Med Res Opin. 2010;26:1957-1965.

13. Chapman AB, Schwartz GL, Boerwinkle E, Turner ST. Predictors of antihypertensive response to a standard dose of hydrochlorothiazide for essential hypertension. Kidney Int. 2002;61:1047-1055.

14. Chapman RH, Benner JS, Petrilla AA, et al. Predictors of adherence with antihypertensive and lipid-lowering therapy. Arch Intern Med. 2005;165:1147-1152.

15. Huetsch JC, Uman JE, Udris EM, Au DH. Predictors of adherence to inhaled medications among veterans with COPD. J Gen Intern Med. 2012;27:1506-1512.

16. Riegel B, Lee CS, Ratcliffe SJ, et al. Predictors of objectively measured medication nonadherence in adults with heart failure. Circ Heart Fail. 2012;5:430-436.

17. Solomon DH, Brookhart MA, Tsao P, et al. Predictors of very low adherence with medications for osteoporosis: towards development of a clinical prediction rule. Osteoporos Int. 2011;22:1737-1743.

18. Curtis JR, Xi J, Westfall AO, et al. Improving the prediction of medication compliance: the example of bisphosphonates for osteoporosis. Med Care. 2009;47:334-341.

19. Andrade SE, Kahler KH, Frech F, Chan KA. Methods for evaluation of medication adherence and persistence using automated databases. Pharmacoepidemiol Drug Saf. 2006;15:565-574.

20. Choudhry NK, Shrank WH, Levin RL, et al. Measuring concurrent adherence to multiple related medications. Am J Manag Care. 2009;15:457-464.

21. Choudhry NK, Fischer MA, Avorn J, et al. The implications of therapeutic complexity on adherence to cardiovascular medications. Arch Intern Med. 2011;171:814-822.

22. Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159:702-706.

23. Steyerberg EW. Clinical Prediction Models: A Practical Approach to Development, Validation, and Updating (Statistics for Biology and Health). New York, NY: Springer Science Business Media; 2009.

24. Ridgeway G. The state of boosting. Comput Sci Statistics. 1999;31:172-181.

25. Friedman J. Greedy function approximation: a gradient boosting machine. Ann Statist. 2001;29:1189-1232.

26. Larsen K, Petersen JH, Budtz‐Jørgensen E, Endahl L. Interpreting parameters in the logistic regression model with random effects. Biometrics. 2000;56:909-914.

27. Choudhry NK, Levin R, Winkelmayer WC. Statins in elderly patients with acute coronary syndrome: an analysis of dose and class effects in typical practice. Heart. 2007;93:945- 951.

28. Wahl PM, Rodgers K, Schneeweiss S, et al. Validation of claims-based diagnostic and procedure codes for cardiovascular and gastrointestinal serious adverse events in a commercially-insured population. Pharmacoepidemiol Drug Saf. 2010;19(6):596-603.

29. Ray WA, Griffin MR, Fought RL, Adams ML. Identification of fractures from computerized Medicare files. J Clin Epidemiol. 1992;45:703-714.

30. Gagne JJ, Glynn RJ, Avorn J, Levin R, Schneeweiss S. A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol. 2011;64:749-759.

31. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41-55.

32. Desai RJ, Rothman KJ, Bateman BT, Hernandez-Diaz S, Huybrechts KF. A propensity- score-based fine stratification approach for confounding adjustment when exposure is infrequent. Epidemiology. 2017;28:249-257.

33. Shah ND, Steyerberg EW, Kent DM. Big data and predictive analytics: recalibrating expectations. JAMA. 2018;320:27-28.

34. Van Klaveren D, Gonen M, Steyerberg EW, Vergouwe Y. A new concordance measure for risk prediction models in external validation settings. Stat Med. 2016;35:4136-4152.

35. Gagne JJ, Glynn RJ, Avorn J, Levin R, Schneeweiss S. A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol. 2011;64:749-759.

36. Simard M, Sirois C, Candas B. Validation of the combined comorbidity index of Charlson and Elixhauser to predict 30-day mortality across ICD-9 and ICD-10. Med Care. 2018;56:441-447.

37. Grymonpre R, Cheang M, Fraser M, Metge C, Sitar DS. Validity of a prescription claims database to estimate medication adherence in older persons. Med Care. 2006;44:471- 477.

38. Franklin JM, Copalakrishnan C, Krumme AA, et al. The relative benefits of claims and electronic health record data for predicting medication adherence trajectory. Am Heart J. 2018;197:153-162.

39. Bushnell GA, Broolhart MA, Gaynes BN, Compton SN, Dusetzina SB, Sturmer T. Examining parental medication adherence as a predictor of child medication adherence in pediatric anxiety disorders. Med Care. 2018;56(6):510-519.

40. Lauffenburger JC, Gagne JJ, Song Z, Brill G, Choudhry NK. Potentially disruptive life events: what are the immediate impacts on chronic disease management? A case- crossover analysis. BMJ Open. 2016;6:e010958. https://bmjopen.bmj.com/content/6/8/e010958

75 APPENDICES

Appendix 1. Distribution of baseline characteristics among patients in the training and testing cohorts for the Aim 1, Part 1, primary analysis of statin initiators

Characteristics Training Testing Cohort Cohort Number of patients n = 44745 n = 44745 Age, mean (SD) 54.5 (10.2) 54.6 (10.2) Female, n (%) 24289 (54.3) 24424 (54.6)

Regions, n (%) Midwest 10422 (23.3) 10471 (23.4) Northeast 3431 (7.7) 3566 (8.0) South 25054 (56.0) 25013 (55.9) West 5838 (13.0) 5695 (12.7)

Use of preventive services Fecal occult blood tests, n (%) 4166 (9.3) 4178 (9.3) Colonoscopy, n (%) 3984 (8.9) 3904 (8.7) Mammography, n (% of females) 8585 (35.3) 8540 (35.0) Number of hospitalizations in the prior year, mean 0.17 (0.6) 0.17 (0.6) (SD) Total days in hospital in prior year, mean (SD) 0.88 (4.6) 0.9 (4.5)

Comorbidities, n (%) Peripheral vascular disease 1529 (3.4) 1502 (3.4) Liver disease 1938 (4.3) 1957 (4.4) Renal disease 2399 (5.4) 2434 (5.4) Recent myocardial infarction (MI) 588 (1.3) 593 (1.3) Prior MI 754 (1.7) 770 (1.7) Recent stroke 289 (0.7) 313 (0.7) Prior stroke 391 (0.9) 418 (0.9)

Characteristics Training Testing Cohort Cohort Ischemic heart disease 5191 (11.6) 5207 (11.6) Transient ischemic attack 158 (0.4) 132 (0.3) Hypertension 25051 (56.0) 25156 (56.2) Diabetes 11653 (26.0) 11579 (25.9) Depression 6517 (14.6) 6544 (14.6) Cancer 5978 (13.4) 5882 (13.2) Combined comorbidity score, mean (SD) 0.16 (1.5) 0.17 (1.5)

Baseline medications, n (%) ACEIs 15548 (34.8) 15480 (34.6) ARBs 8580 (19.2) 8576 (19.2) Beta-blockers 11182 (25.0) 11445 (25.6) Calcium channel blockers 9253 (20.7) 9266 (20.7) Thiazides 14850 (33.2) 14861 (33.2) Oral anticoagulants 1484 (3.3) 1537 (3.4) Antiplatelets 1708 (3.8) 1713 (3.8) Antidiabetics 11286 (25.2) 11229 (25.1) NSAIDs 10200 (22.8) 10105 (22.6) SSRIs 9652 (21.6) 9698 (21.7) Other lipid lowering agents 2169 (4.9) 2223 (5.0)

Index statin, n (%) Atorvastatin 7466 (16.7) 7601 (17.0) Fluvastatin 47 (0.1) 57 (0.1) Lovastatin 2208 (4.9) 2174 (4.9) Pitavastatin 501 (1.1) 545 (1.2) Pravastatin 8257 (18.5) 8127 (18.2) Rosuvastatin 6440 (14.4) 6271 (14.0) 77

Characteristics Training Testing Cohort Cohort Simvastatin 19826 (44.3) 19970 (44.6) High-intensity dose 4090 (9.1) 3931 (8.9)

Medication burden, mean (SD) Number of drug dispensations in prior year 26.4 (21.9) 26.5 (21.9) Number of unique drugs dispensed in prior year 7.6 (5.2) 7.6 (5.2) Number of drugs dispensed with days’ supply 2.4 (2.3) 2.4 (2.3) overlapping index

Financial burden: Health insurance plan type, n (%) POS 30753 (68.7) 30747 (68.7) EPO 7118 (15.9) 7028 (15.7) HMO 3706 (8.3) 3733 (8.3) IND 1753 (3.9) 1856 (4.2) PPO 1357 (3.1) 1336 (3.0) Other 58 (0.1) 45 (0.1) Total deductibles in US dollars, mean (SD) 557.8 (547.6) 559.6 (560.6) Copay of index statin in US dollars, mean (SD) 17 (19.5) 16.9 (19.3)

Appendix 2. Patient flow chart for Aim 1, Part 1 statin cohort

Intercept –2.9828 Demographics Age 0.0172 1.017 1.015 1.02 Female sex –0.1338 0.765 0.733 0.799 Drug use ACEI –0.0179 0.982 0.936 1.031 Antidiabetics –0.1815 0.834 0.79 0.881 Antiparkinson agents –0.2029 0.816 0.69 0.966 Antiplatelets –0.0441 0.957 0.859 1.066 ARBs –0.1227 0.885 0.834 0.938 Calcium channel –0.1638 0.849 0.804 0.897 blockers Digoxin –0.2985 0.742 0.583 0.945 Loop diuretics 0.0418 1.043 0.943 1.153 NSAIDs –0.0794 0.924 0.876 0.973 Oral anticoagulants 0.3474 1.415 1.259 1.591 SSRIs –0.00291 0.997 0.944 1.053 TCAs 0.1549 1.168 1.032 1.321 Thiazide diuretics –0.0246 0.976 0.927 1.027 Nonstatin lipid- 0.3958 1.486 1.351 1.634 lowering drugs Number of any drug 0.00988 1.01 1.008 1.012 Prescription Number of unique –0.0388 0.962 0.955 0.969 burden drug type Number of 0.0826 1.086 1.072 1.1 concurrently prescribed drugs as the index statin Refill synchronization –0.192 0.825 0.742 0.918

Appendix 3. Coefficients and corresponding odds ratios for the variables in the model predicting high statin adherence (PDC > 80%) without medical data components in Aim 1, Part I Number of drugs –0.00188 0.998 0.995 1.001 among those listed above Baseline financial Plan type EPO –0.3248 0.723 0.636 0.821 burden Plan type HMO -0.1051 0.9 0.787 1.03 Plan type IND 0.0504 1.052 0.901 1.228 Plan type OTH 0.1693 1.184 0.68 2.064 Plan type POS –0.1006 0.904 0.804 1.017 Index statin Copay amount –0.00399 0.996 0.995 0.997 information Days-supply 0.00641 1.006 1.005 1.007 High-dose statin (No vs 0.1984 1.219 1.129 1.317 Yes) Prior medication Mean PDC 1.8051 6.081 5.5 6.722 adherence

Appendix 4. Variables specific to the bisphosphonate cohort BMD tests Index year Other osteoporosis drugs Non-metastatic cancer Metastatic cancer Long-term use of steroids Osteoporosis Inhaled glucocorticoid Osteoporotic fracture Lipid-lowering agents Dose regimen: monthly

Appendix Table 5. Variables specific to the ACEI/ARB cohort Coronary artery bypass grafting Percutaneous coronary intervention COPD Other atherosclerotic diseases Lipid-lowering agents Primary non-adherence to anti-hyperlipidemia

Appendix 6. Frequency of selection of the predictors in the lasso model during 10-fold cross- validation across the study cohorts in Aim 1, Part II* Statin Bisphosphonate ACEI/ARB Common predictors

Age 100 100 100 Female 100 100 100 Region-Northeast 100 100 100 Region-South 100 100 100 Region-West 20 20 0 Physician visits 100 100 100 Hospitalizations 100 90 100

Total days in hospital in prior year 100 80 100 International Normalized Ratio (INR) tests 50 20 100 Fecal occult blood test 100 100 100 Colonoscopy 100 100 100 Mammography 100 100 100 Vaccinations 100 100 100 Peripheral vascular disease (PVD) 100 60 100 Liver disease 100 100 50 Renal disease 90 100 100 Prior myocardial infarction (MI) 100 50 100 Prior stroke 100 100 100 Ischemic heart disease 100 100 100 Transient ischemic attack 100 100 100 Hypertension 90 100 40 Diabetes 100 100 100 Hyperlipidemia 40 10 100

Appendix 6. Frequency of selection of the predictors in the lasso model during 10-fold cross- validation across the study cohorts in Aim 1, Part II* Congestive heart failure (CHF) 80 40 100 Parkinson 50 60 100 Alzheimer’s disease or dementia 40 100 100 Depression 100 100 100 Schizophrenia 50 20 100 Rheumatoid arthritis 100 100 100 Cancer 100 100 60 Combined comorbidity score 100 100 100 Angiotensin-converting enzyme inhibitors 100 100 − (ACEIs) Dementia treatment 100 50 100 Antidiabetics 100 100 100 Antiparkinson agents 100 100 100 Antiplatelets 100 100 100 Angiotensin II receptor blockers (ARBs) 100 100 − Atypical antipsychotics 100 80 100 Beta-blockers 100 100 100 Biologic disease-modifying antirheumatic 30 100 30 drugs (DMARDs) Calcium channel blockers 100 100 100 Digoxin 100 100 50 Heparin/low-molecular-weight heparin 60 100 50 (LMWH) Lithium 90 60 100 Loop diuretics 40 100 100 Nitrates 20 100 100 Nonbiologic DMARDs 60 100 100

Appendix 6. Frequency of selection of the predictors in the lasso model during 10-fold cross- validation across the study cohorts in Aim 1, Part II* Nonsteroidal anti-inflammatory drugs 100 100 100 (NSAIDs) Oral anticoagulants 100 80 100 Other anticoagulants 80 70 60 Other antihypertensives 50 100 100 Other antidepressants 100 100 90 Potassium-sparing diuretics 100 100 100 Renin inhibitor 60 60 100 Selective COX-2 inhibitor 100 100 50 Selective serotonin-norepinephrine 100 100 100 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 20 100 100 (SSRIs) Tricyclic antidepressants (TCAs) 100 100 100 Thiazides 100 100 60 Typical antipsychotics 80 100 20 Number of drug dispensations in prior 100 100 100 year Number of unique drugs dispensed in 100 100 100 prior year Number of concurrent medications 100 100 100 Copayment 100 100 100 Primary nonadherence to 50 100 − antihypertensives Mail synchronization metric 100 100 100 Number of chronic medications 100 100 100 dispensed Days’ supply of the index drug 100 100 100

Appendix 6. Frequency of selection of the predictors in the lasso model during 10-fold cross- validation across the study cohorts in Aim 1, Part II* Insurance plan type: health maintenance 100 90 30 organization(HMO) Insurance plan type: individual health 100 100 100 plan (IND) Insurance plan type: other (OTH) 100 100 100 Insurance plan type: point-of-service 100 60 20 (POS) Insurance plan type: preferred provider 100 70 100 organization (PPO) Copay of the index drug 100 100 100 Average proportion of days covered (PDC) 100 100 100 to prior medications

Appendix 7. Relative influence of variables estimated in GBM across the study cohorts in Aim 1, Part II Statin Bisphosphonate ACEI/ARB Variable Relative Variable Relative Variable Relative influence influence influence (%) (%) (%) Adherence to 48.482 Adherence to 24.994 Adherence to 54.675 prior medications prior medications prior medications Number of 6.417 Days’ supply 8.046 Number of 6.792 concurrently used concurrently medications used medications Age 5.741 Refill 6.339 Copay of the 5.198 synchronization index drug metric Regions 4.783 Number of BMD 5.917 Number of 4.235 tests chronic medications Days’ supply 3.438 Number of unique 5.804 Age 3.782 medications filled Copay of the 2.855 Number of 5.465 Days’ supply 3.257 index drug concurrently used medications Number of any 2.506 Other 4.821 Combined 2.152 medication fills osteoporosis comorbidity drugs score Female 2.497 Index year 4.763 Regions 1.732 Total copayment 2.410 Regions 3.638 Number of 1.684 unique medications filled Refill 2.285 Age 3.486 Refill 1.434 synchronization synchronization metric metric Number of unique 2.281 Number of 2.411 Total copayment 1.317 medications filled physician office visits

Appendix 7. Relative influence of variables estimated in GBM across the study cohorts in Aim 1, Part II Number of 1.783 Number of chronic 1.970 Number of any 1.301 hospitalizations in medications medication fills the prior year Number of 1.707 SSRIs 1.684 Antidiabetics 1.069 physician office visits Myocardial 1.542 Total copayment 1.565 Diabetes 0.886 infarction Number of 1.430 Copay of the index 1.422 Hospitalizations 0.781 chronic drug medications Insurance type 1.347 SNRIs 1.323 Number of 0.760 physician office visits Total days in 0.821 Insurance type 1.236 Insurance type 0.705 hospital in prior year Mammography 0.738 Mammography 1.225 Vaccinations 0.587 Oral 0.734 Female 1.095 Mammography 0.549 anticoagulants Calcium channel 0.612 Number of any 0.998 Coronary artery 0.428 blockers medication fills bypass graft (CABG) Vaccinations 0.608 Depression 0.719 Nitrate 0.392 Stroke 0.529 Hypertension 0.699 Lipid-lowering 0.348 agents Diabetes 0.509 Primary 0.617 Female 0.335 nonadherence to anti- hyperlipidemic drugs Number of INR 0.402 Cancer 0.600 Renal 0.332 tests

Appendix 7. Relative influence of variables estimated in GBM across the study cohorts in Aim 1, Part II ARBs 0.374 Combined 0.550 Ischemic heart 0.329 comorbidity score disease Combined 0.353 Antidiabetics 0.545 Total days in 0.318 comorbidity score hospital in prior year Antidiabetics 0.326 Number of INR 0.509 SSRI 0.279 tests Ischemic heart 0.298 Fecal occult blood 0.475 Renin inhibitor 0.273 disease test NSAIDs 0.258 Rheumatoid 0.435 Other 0.246 arthritis atherosclerotic diseases Antiplatelets 0.229 Beta-blockers 0.399 Stroke 0.240 Tricyclic 0.184 Ischemic heart 0.365 Hyperlipidemia 0.239 antidepressants disease Atypical 0.133 Long-term use of 0.338 COPD 0.238 antipsychotics steroids Cancer 0.130 Colonoscopy 0.324 Heart failure 0.219 Depression 0.118 Other 0.286 PVD 0.211 antidepressants Liver disease 0.105 Antiplatelets 0.284 Antiplatelets 0.181 Selective 0.102 Vaccinations 0.270 NSAIDs 0.179 serotonin- norepinephrine reuptake inhibitors Dementia 0.101 Osteoporosis 0.270 Number of 0.179 treatment antihypertensives Fecal occult blood 0.093 Inhaled 0.262 Antiparkinson 0.177 test glucocorticoid agents Thiazides 0.093 Diabetes 0.250 Depression 0.163

Appendix 7. Relative influence of variables estimated in GBM across the study cohorts in Aim 1, Part II Colonoscopy 0.091 Calcium channel 0.245 Number of INR 0.159 blockers tests Potassium- 0.061 Liver disease 0.240 Percutaneous 0.129 sparing diuretics coronary intervention Peripheral 0.057 Osteoporotic 0.239 Number of renal 0.123 vascular disease fracture function tests Transient 0.048 Alzheimer’s 0.228 Beta-blockers 0.118 ischemic attack disease or dementia Antiparkinson 0.043 Number of 0.206 Alzheimer’s 0.096 agents hospitalizations in disease or the prior year dementia Hypertension 0.032 Total days in 0.180 Oral 0.096 hospital in prior anticoagulants year Alzheimer’s 0.027 Oral 0.176 Fecal occult 0.090 disease or anticoagulants blood test dementia Renal disease 0.025 Dose regimen 0.165 Parkinson’s 0.089 disease Loop diuretics 0.024 ARBs 0.164 Transient 0.085 ischemic attack Rheumatoid 0.024 Thiazides 0.162 Calcium channel 0.077 arthritis blockers Other 0.022 Lipid-lowering 0.161 Atypical 0.076 anticoagulants agents antipsychotics Other 0.022 DMARDs-biologic 0.147 Colonoscopy 0.071 antidepressants COX-2 inhibitor 0.021 Heparin or LMWH 0.140 Dementia 0.068 treatment

Appendix 7. Relative influence of variables estimated in GBM across the study cohorts in Aim 1, Part II Hyperlipidemia 0.021 Primary 0.111 Rheumatoid 0.058 nonadherence to arthritis antihypertensives Selective 0.017 COX-2 inhibitor 0.105 DMARDs- 0.058 serotonin nonbiologic reuptake inhibitors Heart failure 0.013 TCAs 0.104 MI 0.054 Parkinson’s 0.012 Transient ischemic 0.096 Potassium- 0.051 disease attack sparing agents DMARDs-biologic 0.011 Other 0.085 Other 0.047 antihypertensives antihypertensives Typical 0.010 Nitrates 0.067 TCAs 0.047 antipsychotics Schizophrenia 0.010 DMARDs- 0.062 Angioedema 0.034 nonbiologic DMARDs- 0.009 Antiparkinson 0.060 Loop diuretics 0.027 nonbiologic agents Heparin or LMWH 0.009 Renal disease 0.049 Lithium 0.026 Digoxin 0.008 Dementia 0.045 Other 0.023 treatment antidepressants Lithium 0.008 Stroke 0.041 SNRIs 0.020 Other 0.006 Digoxin 0.038 Hyperlipidemia 0.016 antihypertensives Beta-blockers 0.005 Heart failure 0.035 Cancer 0.013 ACE inhibitors 0.004 Loop diuretics 0.034 COX-2 inhibitor 0.008 Renin inhibitor 0.002 Typical 0.033 DMARDs-biologic 0.008 antipsychotics Primary 0.002 NSAIDs 0.021 Atrial fibrillation 0.005 nonadherence to antihypertensives

Appendix 7. Relative influence of variables estimated in GBM across the study cohorts in Aim 1, Part II Nitrates 0.002 Potassium-sparing 0.021 Liver disease 0.005 diuretics

Bisphosphonate 0.020 Hypertension 0.005 initiated

Drug-induced 0.018 Digoxin 0.004 osteoporosis

Peripheral 0.017 Asthma 0.003 vascular disease

Other 0.017 Typical 0.003 anticoagulants antipsychotics

Renin inhibitor 0.015 Thiazides 0.003

Other fractures 0.012 Heparin or 0.002 LMWH

MI 0.012 Other 0.001 anticoagulants

Hyperlipidemia 0.008 Schizophrenia 0.000

Parkinson’s 0.008 Parkinson’s 0.008 disease disease

Lithium 0.005 Lithium 0.005

Atypical 0.004 Atypical 0.004 antipsychotics antipsychotics

ACEIs 0.004 ACEIs 0.004

Schizophrenia 0.003 Schizophrenia 0.003

Falls 0.002 Falls 0.002

Appendix 8. Estimated odds ratios from the fixed effects and random intercept unified models from Aim 1, Part III* Fixed-effects Model Random Intercept Model

OR 95% CI OR 95% CI

Common Variables Age 1.010 1.009 1.011 1.008 1.007 1.009 Female 0.774 0.758 0.789 0.836 0.819 0.853

Regions West Ref Ref Ref Ref Midwest 1.017 0.991 1.044 1.071 1.044 1.100 Northeast 0.953 0.920 0.986 0.975 0.942 1.009 South 0.810 0.791 0.829 0.843 0.823 0.863 Physician visits 0.999 0.998 1.000 1.006 1.005 1.007 Hospitalizations 0.993 0.971 1.016 0.988 0.966 1.010 Total days in hospital in 1.005 1.002 1.007 1.002 1.000 1.004 prior year Fecal occult blood test 1.084 1.055 1.113 1.102 1.073 1.132 Colonoscopy 1.094 1.065 1.125 1.105 1.075 1.135 Mammography 1.221 1.195 1.248 1.225 1.199 1.252 Vaccinations 1.068 1.054 1.082 1.051 1.037 1.065 PVD 0.935 0.899 0.972 0.917 0.881 0.954 Liver disease 0.972 0.935 1.012 0.959 0.921 0.998 Renal disease 0.939 0.907 0.973 0.905 0.874 0.938 Prior MI 1.542 1.436 1.656 1.560 1.452 1.676 Prior stroke 1.488 1.351 1.638 1.495 1.357 1.646 Ischemic heart disease 0.973 0.947 1.001 0.960 0.933 0.987 Transient ischemic attack 0.960 0.892 1.033 0.924 0.859 0.994

Appendix 8. Estimated odds ratios from the fixed effects and random intercept unified models from Aim 1, Part III* Diabetes 0.976 0.946 1.006 0.916 0.888 0.945 Depression 0.935 0.912 0.959 0.903 0.880 0.926 Rheumatoid arthritis 0.854 0.803 0.909 0.859 0.807 0.914 Cancer 1.164 1.141 1.187 1.054 1.033 1.076 Combined comorbidity 0.969 0.962 0.975 0.972 0.966 0.979 score Antidiabetics 0.904 0.874 0.934 0.882 0.853 0.912 Antiparkinson agents 0.883 0.834 0.936 0.875 0.825 0.928 Antiplatelets 0.881 0.849 0.915 0.868 0.836 0.902 Atypical antipsychotics 1.066 1.008 1.126 1.033 0.977 1.092 Beta-blockers 1.045 1.025 1.065 0.988 0.969 1.008 Calcium channel blockers 0.982 0.962 1.003 0.939 0.920 0.959 Digoxin 0.947 0.884 1.015 0.934 0.871 1.001 Lithium 0.950 0.815 1.108 0.946 0.811 1.104 Nonbiologic DMARDs 0.945 0.891 1.003 0.956 0.900 1.015 NSAIDs 0.987 0.967 1.008 0.970 0.950 0.991 Oral anticoagulants 1.154 1.109 1.201 1.079 1.036 1.123 Other anticoagulants 1.079 0.808 1.440 1.077 0.808 1.435 Other antihypertensives 0.977 0.94 1.016 0.959 0.922 0.997 Other antidepressants 0.969 0.938 1.001 0.968 0.937 1.000 Potassium-sparing 1.037 1.003 1.072 1.013 0.979 1.047 diuretics Renin inhibitor 1.355 1.209 1.519 1.336 1.192 1.498 COX-2 inhibitor 0.969 0.923 1.018 0.980 0.934 1.030 SNRIs 0.973 0.936 1.011 0.974 0.938 1.013 TCAs 0.995 0.951 1.040 0.998 0.954 1.044

Appendix 8. Estimated odds ratios from the fixed effects and random intercept unified models from Aim 1, Part III* Thiazides 0.996 0.977 1.015 0.950 0.932 0.969 Number of drug 1.006 1.006 1.007 1.005 1.004 1.005 dispensations in prior year Number of unique drugs 0.962 0.959 0.965 0.959 0.956 0.962 dispensed in prior year Number of concurrent 1.076 1.071 1.082 1.096 1.090 1.101 medications Copayment 1.000 1.000 1.000 1.000 1.000 1.000 Mail synchronization 1.014 0.978 1.051 1.034 0.996 1.074 metric Number of chronic 0.999 0.998 1.000 1.003 1.002 1.004 medications dispensed Days’ supply of the index 1.006 1.006 1.006 1.006 1.006 1.006 drug

Insurance plan type PPO Ref Ref Ref Ref EPO 0.818 0.783 0.855 0.874 0.836 0.913 HMO 0.988 0.949 1.029 0.912 0.875 0.950 IND 0.947 0.901 0.994 1.108 1.054 1.165 OTH 1.176 1.126 1.227 0.964 0.923 1.007 POS 0.906 0.873 0.941 0.961 0.926 0.998 Copay of the index drug 0.995 0.995 0.995 0.997 0.997 0.997 Average PDC to prior 4.668 4.485 4.859 4.580 4.399 4.767 medications

*Predictors with a selection frequency of ≥ 50% from Part II (Appendix 6) were included in the logistic regression models.

Appendix 9. Patient flow chart for Aim 2 statin cohort

Patients with at least 1 statin claim with no statin claim 365 days prior (n = 5,582,204)

Continuous enrollment in the database 365 days prior to the index date (n = 2,096,873)

Statin initiators at least 18 years old on the index date (n = 2,095,160)

Patients with only 1 statin claim on index date (n = 2,090,624)

High-intensity statin Low-intensity statin initiators initiators (n = 566,398) (n = 1,523,949)

Appendix 10. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 before propensity score matching in Aim 2 Low-intensity Statin High-intensity Statin Initiators Initiators (n = 379,214) (n = 157,075) Age, mean (SD)* 55 13.0 56 12.7 Sex: female* 214,779 56.6 79,595 50.7

Regions (%) Midwest 77,778 20.5 30,021 19.1 Northeast 28,895 7.6 12,211 7.8 South 208,569 55.0 91,749 58.4 West 62,261 16.4 22,485 14.3 Unknown 1,711 0.5 609 0.4

Use of preventive services Colonoscopy*, n (%) 29,420 7.8 12,269 7.8 Fecal occult blood test, n (%) 34,706 9.2 13,690 8.7 Mammography*, n (% of females) 49,665 23.1 18,164 22.8 Any prior lab test, n (%) 141,160 37.2 59,714 38.0 Lipid test*, n (%) 280,922 74.1 115,159 73.3

Benefit plan types, n (%) EPO* 57,567 15.2 23,588 15.0 HMO* 78,286 20.6 30,858 19.6 IND* 4,303 1.1 1,848 1.2 OTH* 23,233 6.1 10,613 6.8 POS* 194,984 51.4 80,813 51.4 Use of wheelchair, walker, crutches, 8,689 2.3 3,651 2.3 or cane

Index statin, n (%) Atorvastatin 64,813 17.1 91,966 58.5

Appendix 10. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 before propensity score matching in Aim 2 Fluvastatin 2,279 0.6 0 0.0 Lovastatin 34,465 9.1 147 0.1 Pitavastatin 2,245 0.6 0 0.0 Pravastatin 63,499 16.7 0 0.0 Rosuvastatin 15,157 4.0 55,304 35.2 Simvastatin 196,756 51.9 9,658 6.1

Comorbidities, n (%) Atrial fibrillation 6,284 1.7 2,870 1.8 Acute coronary syndrome, with or 25,276 6.7 15,763 10.0 without revascularization* Acute coronary syndrome, with 12,134 3.2 8,783 5.6 revascularization* Alzheimer’s disease 10,072 2.7 3,878 2.5 Angina 21,013 5.5 12,888 8.2 Coronary atherosclerosis 42,046 11.1 27,441 17.5 Bone mineral density test 6,750 1.8 2,719 1.7 Coronary artery bypass graft 2,322 0.6 1,574 1.0 (CABG), new CABG, old 5,702 1.5 4,132 2.6 Heart failure 22,521 5.9 11,421 7.3 Heart failure hospitalization 3,586 0.9 1,743 1.1 Chronic obstructive pulmonary 39,777 10.5 17,233 11.0 disease (COPD) Any cancer* 42,411 11.2 17,765 11.3 Cardiovascular system symptom 45,942 12.1 20,641 13.1 Chest pain 90,559 23.9 42,794 27.2 Conduction disorders 7,998 2.1 3,945 2.5

Appendix 10. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 before propensity score matching in Aim 2 Depression 55,756 14.7 22,453 14.3 Diabetes* 138,028 36.4 57,287 36.5 Electrocardiogram 165,687 43.7 74,573 47.5 History of falls 7,951 2.1 3,229 2.1 HIV 798 0.2 422 0.3 Hip fracture 513 0.1 172 0.1 Hyperlipidemia 291,546 76.9 125,513 79.9 Hyperparathyroidism 1,879 0.5 796 0.5 Hypertension 244,652 64.5 105,728 67.3 Hyperthyroidism 5,518 1.5 2,299 1.5 Ischemic heart disease* 44,241 11.7 28,591 18.2 Prior liver disease* 21,981 5.8 8,921 5.7 Osteoporosis 17,570 4.6 6,423 4.1 Use of oxygen 6,326 1.7 2,839 1.8 Peripheral vascular disease (PVD) or 21,197 5.6 10,523 6.7 PVD surgery* Palpitations 21,929 5.8 9,554 6.1 Parkinson’s disease 1,518 0.4 570 0.4 Post-myocardial infarction/acute 12,088 3.2 8,336 5.3 coronary syndromes (MI/ACS)* Preventive care 177,812 46.9 68,473 43.6 Prior MI 4,068 1.1 3,014 1.9 Prior stroke* 2,685 0.7 1,470 0.9 Rheumatoid arthritis 5,831 1.5 2,314 1.5 Recent MI* 492 0.1 564 0.4 Recent stroke* 1,053 0.3 622 0.4 Renal dysfunction* 33,640 8.9 15,504 9.9 99

Appendix 10. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 before propensity score matching in Aim 2 Schizophrenia 1,397 0.4 484 0.3 Transient ischemic attack 10,899 2.9 4,865 3.1 Urinary tract infection 45,162 11.9 17,682 11.3

Baseline medication use, n (%) Angiotensin-converting enzyme 135,837 35.8 54,111 34.4 inhibitors (ACEIs)* Agents for dementia 3,768 1.0 1,418 0.9 Antidiabetics* 120,140 31.7 48,674 31.0 Antiparkinson agents* 7,237 1.9 2,776 1.8 Antiplatelets* 23,649 6.2 16,186 10.3 Angiotensin II receptor blockers 75,683 20.0 36,816 23.4 (ARBs)* Atypical antipsychotics 9,048 2.4 3,733 2.4 Beta-blockers 96,609 25.5 46,521 29.6 Biologic disease-modifying 1,048 0.3 445 0.3 antirheumatic drugs (DMARDs) Calcium channel blockers* 87,602 23.1 40,779 26.0 Digoxin* 7,030 1.9 3,141 2.0 Heparin and low-molecular-weight 3,877 1.0 1,682 1.1 heparin Lithium 1,109 0.3 488 0.3 Loop diuretics* 29,345 7.7 13,673 8.7 Monoamine oxidase inhibitors 22 0.0 9 0.0 (MOAs) Nitrates 15,108 4.0 9,571 6.1 Nonbiologic DMARDs 5,071 1.3 1,939 1.2

Nonselective nonsteroidal anti- 90,561 23.9 36,412 23.2 inflammatory drugs (NSAIDs)*

100

Appendix 10. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 before propensity score matching in Aim 2 Opioids* 119,837 31.6 50,886 32.4 Oral anticoagulants 14,584 3.8 6,720 4.3 Other anticoagulants 285 0.1 105 0.1 Other antihypertensives 18,248 4.8 8,033 5.1 Other lipid-lowering agents* 28,116 7.4 14,075 9.0 Other newer and atypical 25,203 6.6 10,270 6.5 antidepressants Potassium-sparing diuretics 25,195 6.6 9,883 6.3 Renin inhibitor 1,044 0.3 663 0.4 Selective COX-2 inhibitors* 9,068 2.4 4,087 2.6 Selective serotonin-norepinephrine 15,000 4.0 6,502 4.1 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 80,595 21.3 32,126 20.5 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs) 11,914 3.1 4,787 3.0 Thiazide diuretics 120,049 31.7 49,638 31.6 Typical antipsychotics 1,304 0.3 475 0.3

Health care utilization, mean (SD) Number of distinct cardiovascular 2 3.3 3 3.7 diagnoses Number of International 0 2.0 0 2.1 Normalized Ratio (INR) tests Number of vaccinations* 0 0.5 0 0.5 Number of dispensations of any 23 18.5 23 19.0 drugs* Number of dispensations of any 4 2.2 4 2.3 drugs with days’ supply that overlap the index date* Number of office visits 15 17.2 15 17.7

101

Appendix 10. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 before propensity score matching in Aim 2 Number of physician visits* 7 5.8 7 5.9 Number of cardiovascular 0 0.6 0 0.6 hospitalizations Number of office visits with a 5 8.3 5 8.9 cardiovascular diagnosis Combined comorbidity score 1 1.9 1 2.0 Number of different drugs 9 5.6 9 5.7 dispensed* Total days in hospital in prior year 2 58.4 2 7.7 Number of hospitalizations in the 0 0.7 0 0.8 prior year* Index Rx copayment* 17 21.2 31 28.4 Index Rx days’ supply* 35 16.4 36 17.9 Number of dispensations of the 11 9 11 9.2 drugs of interest* Medication synchronization 0 0.2 0 0.2 metrics* Total baseline out-of-pocket 378 443.4 410 451.4 prescription drug costs

102

Appendix 11. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 before propensity score matching in Aim 2 Low-intensity High-intensity Statin Initiators Statin Initiators (n = 393,345) (n = 142,945) Age, mean (SD)* 60 12.1 60 11.7 Sex: female* 206,269 52.4 64,941 45.4 Regions (%) Midwest 91,927 23.4 31,921 22.3 Northeast 32,092 8.2 12,454 8.7 South 187,298 47.6 72,301 50.6 West 79,575 20.2 25,548 17.9 Unknown 2,453 0.6 721 0.5 Use of preventive services Colonoscopy*, n (%) 33,016 8.4 11,886 8.3 Fecal occult blood test, n (%) 40,411 10.3 13,758 9.6 Mammography*, n (% of females) 63,080 30.6 19,690 30.3 Any prior lab test, n (%) 113,964 29.0 41,927 29.3 Lipid test*, n (%) 281,132 71.5 98,559 68.9 Benefit plan types, n (%) EPO* 30,505 7.8 10,544 7.4 HMO* 100,753 25.6 33,445 23.4 IND* 7,754 2.0 3,190 2.2 OTH* 36,397 9.3 15,034 10.5 POS* 191,846 48.8 70,489 49.3 Use of wheelchair, walker, crutches, 6,633 1.7 2,626 1.8 or cane Index statin, n (%) Atorvastatin 64,497 16.4 88,457 61.9

103

Appendix 11. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 before propensity score matching in Aim 2 Fluvastatin 1,874 0.5 0 0.0 Lovastatin 38,184 9.7 121 0.1 Pitavastatin 1,591 0.4 0 0.0 Pravastatin 67,707 17.2 0 0.0 Rosuvastatin 13,843 3.5 44,674 31.3 Simvastatin 205,649 52.3 9,693 6.8 Comorbidities, n (%) Atrial fibrillation 5,335 1.4 2,331 1.6 Acute coronary syndrome, with or 24,219 6.2 15,673 11 without revascularization* Acute coronary syndrome, with 12,000 3.1 9,305 6.5 revascularization* Alzheimer’s disease 10,256 2.6 3,529 2.5 Angina 19,807 5.0 12,514 8.8 Coronary atherosclerosis 47,091 12.0 29,690 20.8 Bone mineral density test 7,389 1.9 2,472 1.7 Coronary artery bypass graft 2,414 0.6 1,650 1.2 (CABG), new CABG, old 5,540 1.4 3,656 2.6

Heart failure 20,832 5.3 9,623 6.7 Heart failure hospitalization 2,791 0.7 1,179 0.8 Chronic obstructive pulmonary 36,432 9.3 14,045 9.8 disease (COPD) Any cancer* 49,649 12.6 18,524 13.0 Cardiovascular system symptom 40,191 10.2 16,294 11.4 Chest pain 77,817 19.8 35,321 24.7 Conduction disorders 8,268 2.1 3,720 2.6

104

Appendix 11. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 before propensity score matching in Aim 2 Depression 47,162 12.0 16,114 11.3 Diabetes* 114,009 29.0 41,765 29.2 Electrocardiogram 159,942 40.7 65,288 45.7 History of falls 7,436 1.9 2,598 1.8 HIV 672 0.2 401 0.3 Hip fracture 307 0.1 91 0.1 Hyperlipidemia 295,191 75.0 110,145 77.1 Hyperparathyroidism 1,727 0.4 668 0.5 Hypertension 254,893 64.8 94,256 65.9 Hyperthyroidism 5,115 1.3 1,805 1.3 Ischemic heart disease* 49,747 12.6 30,977 21.7 Prior liver disease* 14,723 3.7 5,170 3.6 Osteoporosis 21,589 5.5 6,729 4.7 Use of oxygen 5,509 1.4 2,224 1.6 Peripheral vascular disease (PVD) or 19,111 4.9 8,342 5.8 PVD surgery* Palpitations 18,895 4.8 7,374 5.2 Parkinson’s disease 1,731 0.4 593 0.4 Post-myocardial infarction/acute 12,229 3.1 8,438 5.9 coronary syndromes (MI/ACS)* Preventive care 202,494 51.5 67,991 47.6 Prior MI 3,588 0.9 2,662 1.9 Prior stroke* 4,173 1.1 2,106 1.5 Rheumatoid arthritis 5,487 1.4 1,961 1.4 Recent MI* 1,377 0.4 1,432 1.0 Recent stroke* 2,088 0.5 1,192 0.8 Renal dysfunction* 29,951 7.6 11,804 8.3 105

Appendix 11. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 before propensity score matching in Aim 2 Schizophrenia 1,204 0.3 355 0.2 Transient ischemic attack 10,951 2.8 4,440 3.1 Urinary tract infection 36,729 9.3 12,374 8.7 Baseline medication use, n (%) Angiotensin-converting enzyme 156,600 39.8 55,048 38.5 inhibitors (ACEIs)* Agents for dementia 4,637 1.2 1,466 1.0 Antidiabetics* 100,646 25.6 35,983 25.2 Antiparkinson agents* 6,367 1.6 2,150 1.5 Antiplatelets* 25,617 6.5 16,963 11.9 Angiotensin II receptor blockers 67,772 17.2 27,701 19.4 (ARBs)* Atypical antipsychotics 7,544 1.9 2,516 1.8 Beta-blockers 114,018 29.0 49,850 34.9 Biologic disease-modifying 1,011 0.3 396 0.3 antirheumatic drugs (DMARDs) Calcium channel blockers* 86,959 22.1 33,617 23.5 Digoxin* 7,456 1.9 3,018 2.1 Heparin and low-molecular-weight 3,267 0.8 1,303 0.9 heparin Lithium 951 0.2 292 0.2 Loop diuretics* 28,133 7.2 11,924 8.3 Monoamine oxidase inhibitors 18 0.0 3 0.0 (MOAs) Nitrates 17,657 4.5 10,988 7.7 Nonbiologic DMARDs 4,946 1.3 1,753 1.2

Nonselective nonsteroidal anti- 75,586 19.2 26,366 18.4 inflammatory drugs (NSAIDs)*

106

Appendix 11. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 before propensity score matching in Aim 2 Opioids* 113,533 28.9 42,641 29.8 Oral anticoagulants 17,351 4.4 7,445 5.2 Other anticoagulants 177 0.0 55 0.0 Other antihypertensives 18,679 4.7 7,036 4.9 Other lipid-lowering agents* 37,274 9.5 16,591 11.6 Other newer and atypical 22,784 5.8 8,129 5.7 antidepressants Potassium-sparing diuretics 25,424 6.5 8,592 6.0 Renin inhibitor 843 0.2 398 0.3 Selective COX-2 inhibitors* 9,297 2.4 3,619 2.5 Selective serotonin-norepinephrine 14,613 3.7 5,759 4.0 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 68,969 17.5 23,676 16.6 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs) 11,473 2.9 3,939 2.8 Thiazide diuretics 128,052 32.6 44,291 31.0 Typical antipsychotics 1,153 0.3 375 0.3 Health care utilization, mean (SD) Number of distinct cardiovascular 2 3.1 3 3.5 diagnoses Number of International 0 2.2 0 2.4 Normalized Ratio (INR) tests Number of vaccinations* 0 0.6 0 0.6 Number of dispensations of any 25 21.5 24 22.1 drugs* Number of dispensations of any 5 2.4 5 2.6 drugs with days’ supply that overlap the index date* Number of office visits 12 15.3 13 15.5

107

Appendix 11. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 before propensity score matching in Aim 2 Number of physician visits* 6 5.4 6 5.5 Number of cardiovascular 0 0.5 0 0.5 hospitalizations Number of office visits with a 4 7.5 5 7.9 cardiovascular diagnosis Combined comorbidity score 0 1.7 0 1.8 Number of different drugs 8 5.1 8 5.2 dispensed* Total days in hospital in prior year 1 33.0 1 5.6 Number of hospitalizations in the 0 0.6 0 0.6 prior year* Index Rx copayment* 16 19.8 31 27.3 Index Rx days’ supply* 39 21.3 40 22.6 Number of dispensations of the 13 11.6 13 11.8 drugs of interest* Medication synchronization 0 0.2 0 0.2 metrics* Total baseline out-of-pocket 405 493.1 423 508.9 prescription drug costs

108

Appendix 12. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 before propensity score matching in Aim 2 Low-intensity Statin High-intensity Statin Initiators Initiators (n = 401,068) (n = 135,221) Age, mean (SD)* 69 10.7 67 11.0 Sex: female* 206,274 51.4 55,300 40.9 Regions (%) Midwest 95,717 23.9 32,947 24.4 Northeast 35,711 8.9 14,264 10.5 South 163,469 40.8 59,273 43.8 West 102,494 25.6 27,852 20.6 Unknown 3,677 0.9 885 0.7 Use of preventive services Colonoscopy*, n (%) 40,231 10.0 13,508 10.0 Fecal occult blood test, n (%) 39,410 9.8 11,876 8.8 Mammography*, n (% of females) 79,917 38.7 20,807 37.6 Any prior lab test, n (%) 56,067 14.0 20,936 15.5 Lipid test*, n (%) 266,360 66.4 82,083 60.7 Benefit plan types, n (%) EPO* 8,544 2.1 3,908 2.9 HMO* 133,144 33.2 36,241 26.8 IND* 20,701 5.2 7,335 5.4 OTH* 112,338 28.0 37,752 27.9 POS* 93,548 23.3 39,101 28.9 Use of wheelchair, walker, crutches, 10,064 2.5 3,696 2.7 or cane Index statin, n (%) Atorvastatin 51,189 12.8 89,751 66.4

109

Appendix 12. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 before propensity score matching in Aim 2 Fluvastatin 1,306 0.3 0 0.0 Lovastatin 43,593 10.9 57 0.0 Pitavastatin 1,104 0.3 0 0.0 Pravastatin 80,744 20.1 0 0.0 Rosuvastatin 10,269 2.6 29,232 21.6 Simvastatin 212,863 53.1 16,181 12.0 Comorbidities, n (%) Atrial fibrillation 8,438 2.1 3,937 2.9 Acute coronary syndrome, with or 47,166 11.8 32,980 24.4 without revascularization* Acute coronary syndrome, with 28,931 7.2 24,520 18.1 revascularization* Alzheimer’s disease 19,915 5.0 5,884 4.4 Angina 30,859 7.7 18,594 13.8 Coronary atherosclerosis 96,252 24.0 55,620 41.1 Bone mineral density test 6,453 1.6 2,062 1.5 Coronary artery bypass graft 7,216 1.8 4,520 3.3 (CABG), new CABG, old 11,988 3.0 6,938 5.1

Heart failure 37,450 9.3 16,687 12.3 Heart failure hospitalization 3,389 0.8 1,354 1.0 Chronic obstructive pulmonary 53,458 13.3 19,438 14.4 disease (COPD) Any cancer* 78,450 19.6 25,878 19.1 Cardiovascular system symptom 45,542 11.4 17,218 12.7 Chest pain 93,496 23.3 46,107 34.1 Conduction disorders 14,766 3.7 6,516 4.8

110

Appendix 12. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 before propensity score matching in Aim 2 Depression 48,971 12.2 15,802 11.7 Diabetes* 111,510 27.8 38,663 28.6 Electrocardiogram 184,637 46.0 75,065 55.5 History of falls 10,681 2.7 3,406 2.5 HIV 949 0.2 496 0.4 Hip fracture 329 0.1 110 0.1 Hyperlipidemia 300,025 74.8 101,868 75.3 Hyperparathyroidism 2,282 0.6 751 0.6 Hypertension 279,374 69.7 93,998 69.5 Hyperthyroidism 5,259 1.3 1,610 1.2 Ischemic heart disease* 101,095 25.2 57,664 42.6 Prior liver disease* 12,487 3.1 4,252 3.1 Osteoporosis 33,177 8.3 8,279 6.1 Use of oxygen 9,432 2.4 3,608 2.7 Peripheral vascular disease (PVD) or 26,752 6.7 9,747 7.2 PVD surgery* Palpitations 18,058 4.5 6,161 4.6 Parkinson’s disease 3,010 0.8 877 0.6 Post-myocardial infarction/acute 34,774 8.7 26,895 19.9 coronary syndromes (MI/ACS)* Preventive care 227,837 56.8 68,236 50.5 Prior MI 20,878 5.2 19,762 14.6 Prior stroke* 11,986 3.0 5,619 4.2 Rheumatoid arthritis 8,662 2.2 2,854 2.1 Recent MI* 18,796 4.7 18,841 13.9 Recent stroke* 8,952 2.2 4,479 3.3 Renal dysfunction* 43,090 10.7 15,551 11.5 111

Appendix 12. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 before propensity score matching in Aim 2 Schizophrenia 1,810 0.5 425 0.3 Transient ischemic attack 16,260 4.1 5,865 4.3 Urinary tract infection 43,570 10.9 13,259 9.8 Baseline medication use, n (%) Angiotensin-converting enzyme 177,919 44.4 62,119 45.9 inhibitors (ACEIs)* Agents for dementia 10,789 2.7 2,940 2.2 Antidiabetics* 88,810 22.1 30,312 22.4 Antiparkinson agents* 8,152 2.0 2,549 1.9 Antiplatelets* 46,632 11.6 31,123 23 Angiotensin II receptor blockers 55,731 13.9 19,774 14.6 (ARBs)* Atypical antipsychotics 10,832 2.7 3,329 2.5 Beta-blockers 165,886 41.4 70,272 52.0 Biologic disease-modifying 1,258 0.3 504 0.4 antirheumatic drugs (DMARDs) Calcium channel blockers* 92,304 23.0 30,276 22.4 Digoxin* 9,535 2.4 3,361 2.5 Heparin and low-molecular-weight 4,378 1.1 1,743 1.3 heparin Lithium 1,115 0.3 362 0.3 Loop diuretics* 49,160 12.3 18,792 13.9 Monoamine oxidase inhibitors 21 0.0 4 0.0 (MOAs) Nitrates 34,979 8.7 20,846 15.4 Nonbiologic DMARDs 7,934 2.0 2,563 1.9

Nonselective nonsteroidal anti- 69,680 17.4 22,744 16.8 inflammatory drugs (NSAIDs)*

112

Appendix 12. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 before propensity score matching in Aim 2 Opioids* 152,859 38.1 53,400 39.5 Oral anticoagulants 30,929 7.7 11,571 8.6 Other anticoagulants 191 0.0 86 0.1 Other antihypertensives 30,905 7.7 10,007 7.4 Other lipid-lowering agents* 70,751 17.6 27,822 20.6 Other newer and atypical 27,428 6.8 9,157 6.8 antidepressants Potassium-sparing diuretics 31,390 7.8 9,760 7.2 Renin inhibitor 878 0.2 378 0.3 Selective COX-2 inhibitors* 12,049 3.0 4,294 3.2 Selective serotonin-norepinephrine 21,777 5.4 7,633 5.6 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 64,444 16.1 20,365 15.1 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs) 16,280 4.1 5,067 3.7 Thiazide diuretics 129,553 32.3 38,434 28.4 Typical antipsychotics 1,666 0.4 477 0.4 Health care utilization, mean (SD) Number of distinct cardiovascular 3 3.7 4 4.2 diagnoses Number of International 1 3.3 1 3.3 Normalized Ratio (INR) tests Number of vaccinations* 1 0.8 0 0.8 Number of dispensations of any 33 30.9 32 31.6 drugs* Number of dispensations of any 6 3.4 6 3.5 drugs with days’ supply that overlap the index date* Number of office visits 15 17.4 15 17.8

113

Appendix 12. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 before propensity score matching in Aim 2 Number of physician visits* 7 6.2 7 6.4 Number of cardiovascular 0 0.5 0 0.6 hospitalizations Number of office visits with a 5 9.0 6 9.0 cardiovascular diagnosis Combined comorbidity score 1 2.0 1 2.1 Number of different drugs 9 5.4 9 5.6 dispensed* Total days in hospital in prior year 2 5.7 2 6.6 Number of hospitalizations in the 0 0.6 0 0.7 prior year* Index Rx copayment* 13 16.9 27 26.8 Index Rx days’ supply* 52 29.2 49 28.4 Number of dispensations of the 17 15.7 15 15.5 drugs of interest* Medication synchronization 0 0.2 0 0.2 metrics* Total baseline out-of-pocket 487 668.4 484 693.6 prescription drug costs

114

Appendix 13. Characteristics of low- and high-intensity statin initiators overall across all tertiles before propensity score matching in Aim 2 Low-intensity Statin High-intensity Statin Initiators Initiators (n = 1,173,627) (n = 435,241) Age, mean (SD)* 62 13.2 61 12.8 Sex: female* 627,322 53.5 199,836 45.9 Regions (%) Midwest 265,422 22.6 94,889 21.8 Northeast 96,698 8.2 38,929 8.9 South 559,336 47.7 223,323 51.3 West 244,330 20.8 75,885 17.4 Unknown 7,841 0.7 2,215 0.5 Use of preventive services Colonoscopy*, n (%) 102,667 8.7 37,663 8.7 Fecal occult blood test, n (%) 114,527 9.8 39,324 9.0 Mammography*, n (% of females) 192,662 30.7 58,661 29.4 Any prior lab test, n (%) 311,191 26.5 122,577 28.2 Lipid test*, n (%) 828,414 70.6 295,801 68.0 Benefit plan types, n (%) EPO* 96,616 8.2 38,040 8.7 HMO* 312,183 26.6 100,544 23.1 IND* 32,758 2.8 12,373 2.8 OTH* 171,968 14.7 63,399 14.6 POS* 480,378 40.9 190,403 43.7 Use of wheelchair, walker, crutches, 25,386 2.2 9,973 2.3 or cane Index statin, n (%) Atorvastatin 180,499 15.4 270,174 62.1

115

Appendix 13. Characteristics of low- and high-intensity statin initiators overall across all tertiles before propensity score matching in Aim 2 Fluvastatin 5,459 0.5 0 0.0 Lovastatin 116,242 9.9 325 0.1 Pitavastatin 4,940 0.4 0 0.0 Pravastatin 211,950 18.1 0 0.0 Rosuvastatin 39,269 3.3 129,210 29.7 Simvastatin 615,268 52.4 35,532 8.2 Comorbidities, n (%) Atrial fibrillation 20,057 1.7 9,138 2.1 Acute coronary syndrome, with or 96,661 8.2 64,416 14.8 without revascularization* Acute coronary syndrome, with 53,065 4.5 42,608 9.8 revascularization* Alzheimer’s disease 40,243 3.4 13,291 3.1 Angina 71,679 6.1 43,996 10.1 Coronary atherosclerosis 185,389 15.8 112,751 25.9 Bone mineral density test 20,592 1.8 7,253 1.7 Coronary artery bypass graft 11,952 1.0 7,744 1.8 (CABG), new CABG, old 23,230 2.0 14,726 3.4

Heart failure 80,803 6.9 37,731 8.7 Heart failure hospitalization 9,766 0.8 4,276 1.0 Chronic obstructive pulmonary 129,667 11.0 50,716 11.7 disease (COPD) Any cancer* 170,510 14.5 62,167 14.3 Cardiovascular system symptom 131,675 11.2 54,153 12.4 Chest pain 261,872 22.3 124,222 28.5 Conduction disorders 31,032 2.6 14,181 3.3

116

Appendix 13. Characteristics of low- and high-intensity statin initiators overall across all tertiles before propensity score matching in Aim 2 Depression 151,889 12.9 54,369 12.5 Diabetes* 363,547 31 137,715 31.6 Electrocardiogram 510,266 43.5 214,926 49.4 History of falls 26,068 2.2 9,233 2.1 HIV 2419 0.2 1319 0.3 Hip fracture 1149 0.1 373 0.1 Hyperlipidemia 886,762 75.6 337,526 77.5 Hyperparathyroidism 5,888 0.5 2215 0.5 Hypertension 778,919 66.4 293,982 67.5 Hyperthyroidism 15,892 1.4 5,714 1.3 Ischemic heart disease* 195,083 16.6 117,232 26.9 Prior liver disease* 49,191 4.2 18,343 4.2 Osteoporosis 72,336 6.2 21,431 4.9 Use of oxygen 21,267 1.8 8,671 2.0 Peripheral vascular disease (PVD) or 67,060 5.7 28,612 6.6 PVD surgery* Palpitations 58,882 5.0 23,089 5.3 Parkinson’s disease 6,259 0.5 2040 0.5 Post-myocardial infarction/acute 59,091 5.0 43,669 10.0 coronary syndromes (MI/ACS)* Preventive care 608,143 51.8 204,700 47.0 Prior MI 28,534 2.4 25,438 5.8 Prior stroke* 18,844 1.6 9,195 2.1 Rheumatoid arthritis 19,980 1.7 7,129 1.6 Recent MI* 20665 1.8 20837 4.8 Recent stroke* 12,093 1.0 6293 1.4 Renal dysfunction* 106,681 9.1 42,859 9.8 117

Appendix 13. Characteristics of low- and high-intensity statin initiators overall across all tertiles before propensity score matching in Aim 2 Schizophrenia 4,411 0.4 1264 0.3 Transient ischemic attack 38,110 3.2 15,170 3.5 Urinary tract infection 125,461 10.7 43,315 10.0 Baseline medication use, n (%) Angiotensin-converting enzyme 470,356 40.1 171,278 39.4 inhibitors (ACEIs)* Agents for dementia 19,194 1.6 5,824 1.3 Antidiabetics* 309,596 26.4 114,969 26.4 Antiparkinson agents* 21,756 1.9 7,475 1.7 Antiplatelets* 95,898 8.2 64,272 14.8 Angiotensin II receptor blockers 199,186 17.0 84,291 19.4 (ARBs)* Atypical antipsychotics 27,424 2.3 9,578 2.2 Beta-blockers 376,513 32.1 166,643 38.3 Biologic disease-modifying 3,317 0.3 1345 0.3 antirheumatic drugs (DMARDs) Calcium channel blockers* 266,865 22.7 104,672 24 Digoxin* 24,021 2.0 9,520 2.2 Heparin and low-molecular-weight 11,522 1.0 4,728 1.1 heparin Lithium 3,175 0.3 1142 0.3 Loop diuretics* 106,638 9.1 44,389 10.2 Monoamine oxidase inhibitors 61 0.0 16 0.0 (MOAs) Nitrates 67,744 5.8 41,405 9.5 Nonbiologic DMARDs 17,951 1.5 6,255 1.4 Nonselective nonsteroidal anti- 235,827 20.1 85,522 19.6 inflammatory drugs (NSAIDs)*

118

Appendix 13. Characteristics of low- and high-intensity statin initiators overall across all tertiles before propensity score matching in Aim 2 Opioids* 386,229 32.9 146,927 33.8 Oral anticoagulants 62,864 5.4 25,736 5.9 Other anticoagulants 653 0.1 246 0.1 Other antihypertensives 67,832 5.8 25,076 5.8 Other lipid-lowering agents* 136,141 11.6 58,488 13.4 Other newer and atypical 75,415 6.4 27,556 6.3 antidepressants Potassium-sparing diuretics 82,009 7.0 28,235 6.5 Renin inhibitor 2,765 0.2 1439 0.3 Selective COX-2 inhibitors* 30,414 2.6 12,000 2.8 Selective serotonin-norepinephrine 51,390 4.4 19,894 4.6 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 214,008 18.2 76,167 17.5 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs) 39,667 3.4 13,793 3.2 Thiazide diuretics 377,654 32.2 132,363 30.4 Typical antipsychotics 4,123 0.4 1327 0.3 Health care utilization, mean (SD) Number of distinct cardiovascular 3 3.4 3 3.9 diagnoses Number of International 1 2.6 1 2.6 Normalized Ratio (INR) tests Number of vaccinations* 0 0.7 0 0.6 Number of dispensations of any 27 24.7 26 24.8 drugs* Number of dispensations of any 5 2.9 5 3.0 drugs with days’ supply that overlap the index date* Number of office visits 14 16.7 14 17.1

119

Appendix 13. Characteristics of low- and high-intensity statin initiators overall across all tertiles before propensity score matching in Aim 2 Number of physician visits* 7 5.8 7 6 Number of cardiovascular 0 0.5 0 0.6 hospitalizations Number of office visits with a 5 8.3 5 8.7 cardiovascular diagnosis Combined comorbidity score 1 1.9 1 2.0 Number of different drugs 9 5.4 9 5.6 dispensed* Total days in hospital in prior year 1 38.5 2 6.7 Number of hospitalizations in the 0 0.6 0 0.7 prior year* Index Rx copayment* 15 19.4 30 27.6 Index Rx days’ supply* 42 24.2 41 23.8 Number of dispensations of the 14 12.7 13 12.4 drugs of interest* Medication synchronization 0 0.2 0 0.2 metrics* Total baseline out-of-pocket 424 547.6 437 556.0 prescription drug costs

120

Appendix 14. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 after propensity score matching in Aim 2 Low-intensity Statin High-intensity Statin Initiators Initiators (n = 147 442) (n = 147 442) Age, mean (SD)* 55 12.8 55 12.6 Sex: female* 71,898 48.8 75,705 51.3 Regions (%) Midwest 27,257 18.5 28,450 19.3 Northeast 11,901 8.1 11,512 7.8 South 88,137 59.8 85,601 58.1 West 19,663 13.3 21,308 14.5 Unknown 484 0.3 571 0.4 Use of preventive services Colonoscopy*, n (%) 11,479 7.8 11,549 7.8 Fecal occult blood test, n (%) 13,004 8.8 12,969 8.8 Mammography*, n (% of females) 16,760 23.3 17,362 22.9 Any prior lab test, n (%) 58,252 39.5 56,346 38.2 Lipid test*, n (%) 107,818 73.1 108,354 73.5 Benefit plan types, n (%) EPO* 22,735 15.4 22,237 15.1 HMO* 26,681 18.1 28,861 19.6 IND* 1,651 1.1 1,749 1.2 OTH* 8,678 5.9 9,688 6.6 POS* 78,798 53.4 76,257 51.7 Use of wheelchair, walker, crutches, or 3,388 2.3 3,419 2.3 cane Index statin, n (%) Atorvastatin 36,850 25.0 87,239 59.2

121

Appendix 14. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 after propensity score matching in Aim 2 Fluvastatin 1,989 1.3 0 0.0 Lovastatin 8,394 5.7 140 0.1 Pitavastatin 2,036 1.4 0 0.0 Pravastatin 17,461 11.8 0 0.0 Rosuvastatin 11,660 7.9 50,679 34.4 Simvastatin 69,052 46.8 9,384 6.4 Comorbidities, n (%) Atrial fibrillation 2,640 1.8 2,646 1.8 Acute coronary syndrome, with or 15,369 10.4 14,033 9.5 without revascularization* Acute coronary syndrome, with 8,607 5.8 7,639 5.2 revascularization* Alzheimer’s disease 3,480 2.4 3,654 2.5 Angina 12,599 8.5 11,470 7.8 Coronary atherosclerosis 26,333 17.9 24,198 16.4 Bone mineral density test 2,751 1.9 2,610 1.8 Coronary artery bypass graft (CABG), 1,565 1.1 1,378 0.9 new CABG, old 3,833 2.6 3,634 2.5 Heart failure 10,667 7.2 10,376 7.0 Heart failure hospitalization 1,704 1.2 1,600 1.1 Chronic obstructive pulmonary 16,009 10.9 16,025 10.9 disease (COPD) Any cancer* 16,259 11.0 16,641 11.3 Cardiovascular system symptom 19,811 13.4 19,205 13.0 Chest pain 41,376 28.1 39,420 26.7 Conduction disorders 3,697 2.5 3,610 2.4

122

Appendix 14. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 after propensity score matching in Aim 2 Depression 21,071 14.3 21,178 14.4 Diabetes* 52,989 35.9 53,659 36.4 Electrocardiogram 71,043 48.2 69,176 46.9 History of falls 2,900 2.0 3,013 2.0 HIV 441 0.3 390 0.3 Hip fracture 148 0.1 165 0.1 Hyperlipidemia 118,557 80.4 117,305 79.6 Hyperparathyroidism 723 0.5 752 0.5 Hypertension 98,819 67.0 98,600 66.9 Hyperthyroidism 2,228 1.5 2,167 1.5 Ischemic heart disease* 27,469 18.6 25,258 17.1 Prior liver disease* 8,565 5.8 8,419 5.7 Osteoporosis 5,604 3.8 6,116 4.1 Use of oxygen 2,599 1.8 2,612 1.8 Peripheral vascular disease (PVD) or 9,750 6.6 9,595 6.5 PVD surgery* Palpitations 9,216 6.3 8,951 6.1 Parkinson’s disease 515 0.3 537 0.4 Post-myocardial infarction/acute 7,864 5.3 7,349 5.0 coronary syndromes (MI/ACS)* Preventive care 61,945 42.0 64,883 44 Prior MI 2,805 1.9 2,662 1.8 Prior stroke* 1,479 1.0 1,309 0.9 Rheumatoid arthritis 2,117 1.4 2,182 1.5 Recent MI* 455 0.3 490 0.3 Recent stroke* 662 0.4 547 0.4 Renal dysfunction* 14,286 9.7 14,226 9.6 123

Appendix 14. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 after propensity score matching in Aim 2 Schizophrenia 404 0.3 464 0.3 Transient ischemic attack 4,539 3.1 4,497 3.1 Urinary tract infection 16,199 11.0 16,651 11.3 Baseline medication use, n (%) Angiotensin-converting enzyme 49,151 33.3 50,925 34.5 inhibitors (ACEIs)* Agents for dementia 1,183 0.8 1,330 0.9 Antidiabetics* 45,046 30.6 45,784 31.1 Antiparkinson agents* 2,435 1.7 2,611 1.8 Antiplatelets* 15,272 10.4 14,199 9.6 Angiotensin II receptor blockers 36,119 24.5 33,967 23.0 (ARBs)* Atypical antipsychotics 3,547 2.4 3,504 2.4 Beta-blockers 43,822 29.7 42,777 29 Biologic disease-modifying 457 0.3 415 0.3 antirheumatic drugs (DMARDs) Calcium channel blockers* 37,665 25.5 37,525 25.5 Digoxin* 2,903 2.0 2,924 2.0 Heparin and low-molecular-weight 1,579 1.1 1,547 1.0 heparin Lithium 489 0.3 464 0.3 Loop diuretics* 12,579 8.5 12,622 8.6 Monoamine oxidase inhibitors 11 0.0 9 0.0 (MOAs) Nitrates 8,919 6.0 8,499 5.8 Nonbiologic DMARDs 1,790 1.2 1,832 1.2

Nonselective nonsteroidal anti- 34,123 23.1 34,317 23.3 inflammatory drugs (NSAIDs)*

124

Appendix 14. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 after propensity score matching in Aim 2 Opioids* 47,769 32.4 47,631 32.3 Oral anticoagulants 6,219 4.2 6,244 4.2 Other anticoagulants 97 0.1 100 0.1 Other antihypertensives 7,409 5.0 7,451 5.1 Other lipid-lowering agents* 13,926 9.4 13,012 8.8 Other newer and atypical 9,708 6.6 9,644 6.5 antidepressants Potassium-sparing diuretics 9,013 6.1 9,282 6.3 Renin inhibitor 619 0.4 573 0.4 Selective COX-2 inhibitors* 3,975 2.7 3,814 2.6 Selective serotonin-norepinephrine 6,299 4.3 6,115 4.1 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 29,986 20.3 30,328 20.6 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs) 4,413 3.0 4,505 3.1 Thiazide diuretics 46,296 31.4 46,574 31.6 Typical antipsychotics 424 0.3 446 0.3 Health care utilization, mean (SD) Number of distinct cardiovascular 3 3.8 3 3.7 diagnoses Number of International Normalized 0 2.1 0 2.1 Ratio (INR) tests Number of vaccinations* 0 0.5 0 0.5 Number of dispensations of any 23 18.9 23 19.0 drugs* Number of dispensations of any drugs 4 2.3 4 2.3 with days’ supply that overlap the index date* Number of office visits 15 17.1 15 17.7

125

Appendix 14. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 1 after propensity score matching in Aim 2 Number of physician visits* 7 5.8 7 5.9 Number of cardiovascular 0 0.6 0 0.6 hospitalizations Number of office visits with a 5 8.6 5 8.9 cardiovascular diagnosis Combined comorbidity score 1 2.0 1 2.0 Number of different drugs 9 5.7 9 5.7 dispensed* Total days in hospital in prior year 2 6.5 2 7.7 Number of hospitalizations in the 0 0.8 0 0.8 prior year* Index Rx copayment* 29 28.1 30 27.8 Index Rx days’ supply* 35 16.5 35 17.6 Number of dispensations of the drugs 11 9.0 11 9.2 of interest* Medication synchronization metrics* 0 0.2 0 0.2 Total baseline out-of-pocket 403 421.7 407 451.8 prescription drug costs

126

Appendix 15. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 after propensity score matching in Aim 2 Low-intensity Statin High-intensity Statin Initiators Initiators (n = 135,118) (n = 135,118) Age, mean (SD)* 60 11.9 60 11.7 Sex: female* 59,523 44.1 62,161 46 Regions (%) Midwest 30,231 22.4 30,409 22.5 Northeast 12,331 9.1 11,779 8.7 South 69,508 51.4 68,145 50.4 West 22,445 16.6 24,111 17.8 Unknown 603 0.4 674 0.5 Use of preventive services Colonoscopy*, n (%) 11,117 8.2 11,277 8.3 Fecal occult blood test, n (%) 13,050 9.7 13,113 9.7 Mammography*, n (% of females) 18,190 30.6 18,881 30.4 Any prior lab test, n (%) 41,531 30.7 40,116 29.7 Lipid test*, n (%) 92,027 68.1 93,540 69.2 Benefit plan types, n (%) EPO* 10,208 7.6 10,105 7.5 HMO* 29,417 21.8 31,292 23.2 IND* 3,205 2.4 3,022 2.2 OTH* 12,659 9.4 13,777 10.2 POS* 69,791 51.7 67,331 49.8 Use of wheelchair, walker, crutches, 2,546 1.9 2,432 1.8 or cane Index statin, n (%) Atorvastatin 32,490 24.0 84,086 62.2

127

Appendix 15. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 after propensity score matching in Aim 2 Fluvastatin 1,546 1.1 0 0.0 Lovastatin 8,387 6.2 116 0.1 Pitavastatin 1,306 1.0 0 0.0 Pravastatin 17,114 12.7 0 0.0 Rosuvastatin 9,958 7.4 41,398 30.6 Simvastatin 64,317 47.6 9,518 7.0 Comorbidities, n (%) Atrial fibrillation 2,276 1.7 2,138 1.6 Acute coronary syndrome, with or 15,351 11.4 13,713 10.1 without revascularization* Acute coronary syndrome, with 8,988 6.7 7,927 5.9 revascularization* Alzheimer’s disease 3,275 2.4 3,308 2.4 Angina 12,337 9.1 10,982 8.1 Coronary atherosclerosis 28,454 21.1 26,133 19.3 Bone mineral density test 2,498 1.8 2,394 1.8 Coronary artery bypass graft 1,649 1.2 1,452 1.1 (CABG), new CABG, old 3,404 2.5 3,230 2.4 Heart failure 9,364 6.9 8,752 6.5 Heart failure hospitalization 1,190 0.9 1,077 0.8 Chronic obstructive pulmonary 13,268 9.8 13,067 9.7 disease (COPD) Any cancer* 17,298 12.8 17,371 12.9 Cardiovascular system symptom 15,823 11.7 15,224 11.3 Chest pain 34,510 25.5 32,416 24.0 Conduction disorders 3,583 2.7 3,397 2.5

128

Appendix 15. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 after propensity score matching in Aim 2 Depression 15,426 11.4 15,391 11.4 Diabetes* 39,130 29.0 39,198 29.0 Electrocardiogram 62,845 46.5 60,805 45.0 History of falls 2,452 1.8 2,443 1.8 HIV 421 0.3 367 0.3 Hip fracture 95 0.1 87 0.1 Hyperlipidemia 103,967 76.9 103,867 76.9 Hyperparathyroidism 612 0.5 615 0.5 Hypertension 88,485 65.5 88,695 65.6 Hyperthyroidism 1,734 1.3 1,720 1.3 Ischemic heart disease* 29,703 22 27,313 20.2 Prior liver disease* 5,043 3.7 4,887 3.6 Osteoporosis 6,121 4.5 6,400 4.7 Use of oxygen 2,135 1.6 2,037 1.5 Peripheral vascular disease (PVD) or 7,894 5.8 7,644 5.7 PVD surgery* Palpitations 7,216 5.3 6,969 5.2 Parkinson’s disease 525 0.4 560 0.4 Post-myocardial infarction/acute 8,009 5.9 7,305 5.4 coronary syndromes (MI/ACS)* Preventive care 62,375 46.2 64,662 47.9 Prior MI 2,449 1.8 2,286 1.7 Prior stroke* 2,252 1.7 1,927 1.4 Rheumatoid arthritis 1,889 1.4 1,859 1.4 Recent MI* 1,231 0.9 1,178 0.9 Recent stroke* 1,288 1.0 1,072 0.8 Renal dysfunction* 10,991 8.1 10,888 8.1 129

Appendix 15. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 after propensity score matching in Aim 2 Schizophrenia 310 0.2 340 0.3 Transient ischemic attack 4,306 3.2 4,107 3.0 Urinary tract infection 11,484 8.5 11,704 8.7 Baseline medication use, n (%) Angiotensin-converting enzyme 50,932 37.7 51,970 38.5 inhibitors (ACEIs)* Agents for dementia 1,275 0.9 1,373 1.0 Antidiabetics* 33,830 25.0 33,962 25.1 Antiparkinson agents* 2,023 1.5 2,035 1.5 Antiplatelets* 16,126 11.9 14,773 10.9 Angiotensin II receptor blockers 26,968 20.0 25,909 19.2 (ARBs)* Atypical antipsychotics 2,438 1.8 2,394 1.8 Beta-blockers 47,473 35.1 45,849 33.9 Biologic disease-modifying 412 0.3 380 0.3 antirheumatic drugs (DMARDs) Calcium channel blockers* 31,414 23.2 31,355 23.2 Digoxin* 2,807 2.1 2,779 2.1 Heparin and low-molecular-weight 1,271 0.9 1,210 0.9 heparin Lithium 275 0.2 281 0.2 Loop diuretics* 11,285 8.4 10,953 8.1 Monoamine oxidase inhibitors 3 0.0 3 0.0 (MOAs) Nitrates 10,431 7.7 9,659 7.1 Nonbiologic DMARDs 1,708 1.3 1,668 1.2

Nonselective nonsteroidal anti- 25,062 18.5 25,018 18.5 inflammatory drugs (NSAIDs)*

130

Appendix 15. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 after propensity score matching in Aim 2 Opioids* 40,875 30.3 40,132 29.7 Oral anticoagulants 7,013 5.2 6,822 5.0 Other anticoagulants 47 0.0 51 0.0 Other antihypertensives 6,511 4.8 6,565 4.9 Other lipid-lowering agents* 16,320 12.1 15,490 11.5 Other newer and atypical 7,883 5.8 7,693 5.7 antidepressants Potassium-sparing diuretics 7,993 5.9 8,120 6.0 Renin inhibitor 388 0.3 364 0.3 Selective COX-2 inhibitors* 3,580 2.6 3,440 2.5 Selective serotonin-norepinephrine 5,662 4.2 5,444 4.0 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 22,612 16.7 22,616 16.7 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs) 3,753 2.8 3,746 2.8 Thiazide diuretics 41,307 30.6 42,081 31.1 Typical antipsychotics 344 0.3 351 0.3 Health care utilization, mean (SD) Number of distinct cardiovascular 3 3.6 3 3.5 diagnoses Number of International 0 2.3 0 2.3 Normalized Ratio (INR) tests Number of vaccinations* 0 0.5 0 0.6 Number of dispensations of any 24 22.0 24 22.2 drugs* Number of dispensations of any 5 2.6 5 2.5 drugs with days’ supply that overlap the index date* Number of office visits 13 15.3 13 15.5

131

Appendix 15. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 2 after propensity score matching in Aim 2 Number of physician visits* 6 5.5 6 5.5 Number of cardiovascular 0 0.6 0 0.5 hospitalizations Number of office visits with a 5 7.9 4 7.9 cardiovascular diagnosis Combined comorbidity score 0 1.9 0 1.8 Number of different drugs 8 5.3 8 5.2 dispensed* Total days in hospital in prior year 1 7.4 1 5.6 Number of hospitalizations in the 0 0.6 0 0.6 prior year* Index Rx copayment* 28 27.0 29 25.5 Index Rx days’ supply* 39 21.5 40 22.0 Number of dispensations of the 13 11.4 13 11.8 drugs of interest* Medication synchronization 0 0.2 0 0.2 metrics* Total baseline out-of-pocket 422 480.1 419 507.1 prescription drug costs

132

Appendix 16. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 after propensity score matching in Aim 2 Low-intensity Statin High-intensity Statin Initiators Initiators (n = 122,870) (n = 122,870) Age, mean (SD)* 68 11.2 67 11.0 Sex: female* 50,395 41.0 51,944 42.3 Regions (%) Midwest 30,461 24.8 30,022 24.4 Northeast 13,078 10.6 12,861 10.5 South 53,686 43.7 53,568 43.6 West 24,894 20.3 25,608 20.8 Unknown 751 0.6 811 0.7 Use of preventive services Colonoscopy*, n (%) 12,338 10.0 12,366 10.1 Fecal occult blood test, n (%) 10,761 8.8 10,976 8.9 Mammography*, n (% of females) 18,891 37.5 19,626 37.8 Any prior lab test, n (%) 20,134 16.4 19,256 15.7 Lipid test*, n (%) 73,741 60.0 75,631 61.5 Benefit plan types, n (%) EPO* 3,633 3.0 3,393 2.8 HMO* 32,156 26.2 33,311 27.1 IND* 7,142 5.8 6,794 5.5 OTH* 33,164 27.0 34,564 28.1 POS* 36,984 30.1 34,920 28.4 Use of wheelchair, walker, crutches, or 3,554 2.9 3,386 2.8 cane Index statin, n (%) Atorvastatin 23,031 18.7 81,372 66.2

133

Appendix 16. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 after propensity score matching in Aim 2 Fluvastatin 904 0.7 0 0.0 Lovastatin 8,871 7.2 51 0.0 Pitavastatin 811 0.7 0 0.0 Pravastatin 19,939 16.2 0 0.0 Rosuvastatin 6,502 5.3 25,655 20.9 Simvastatin 62,812 51.1 15,792 12.9 Comorbidities, n (%) Atrial fibrillation 3,730 3.0 3,469 2.8 Acute coronary syndrome, with or 29,117 23.7 26,699 21.7 without revascularization* Acute coronary syndrome, with 21,006 17.1 19,126 15.6 revascularization* Alzheimer’s disease 5,567 4.5 5,499 4.5 Angina 16,959 13.8 15,683 12.8 Coronary atherosclerosis 49,646 40.4 47,004 38.3 Bone mineral density test 2,035 1.7 1,945 1.6 Coronary artery bypass graft (CABG), 4,294 3.5 3,866 3.1 new CABG, old 6,324 5.1 5,952 4.8

Heart failure 15,319 12.5 14,710 12.0 Heart failure hospitalization 1,299 1.1 1,241 1.0 Chronic obstructive pulmonary 17,968 14.6 17,625 14.3 disease (COPD) Any cancer* 23,468 19.1 23,674 19.3 Cardiovascular system symptom 16,046 13.1 15,559 12.7 Chest pain 41,700 33.9 39,472 32.1 Conduction disorders 6,028 4.9 5,782 4.7

134

Appendix 16. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 after propensity score matching in Aim 2 Depression 14,801 12.0 14,706 12.0 Diabetes* 35,297 28.7 35,285 28.7 Electrocardiogram 68,228 55.5 66,530 54.1 History of falls 3,247 2.6 3,182 2.6 HIV 515 0.4 440 0.4 Hip fracture 99 0.1 105 0.1 Hyperlipidemia 91,978 74.9 92,582 75.3 Hyperparathyroidism 680 0.6 701 0.6 Hypertension 85,422 69.5 85,634 69.7 Hyperthyroidism 1,571 1.3 1,502 1.2 Ischemic heart disease* 51,519 41.9 48,834 39.7 Prior liver disease* 4,066 3.3 3,891 3.2 Osteoporosis 7,408 6.0 7,808 6.4 Use of oxygen 3,357 2.7 3,285 2.7 Peripheral vascular disease (PVD) or 9,046 7.4 8,807 7.2 PVD surgery* Palpitations 5,816 4.7 5,676 4.6 Parkinson’s disease 783 0.6 827 0.7 Post-myocardial infarction/acute 23,218 18.9 21,206 17.3 coronary syndromes (MI/ACS)* Preventive care 61,562 50.1 63,084 51.3 Prior MI 16,387 13.3 14,924 12.1 Prior stroke* 5,565 4.5 5,144 4.2 Rheumatoid arthritis 2,620 2.1 2,632 2.1 Recent MI* 15,489 12.6 14,075 11.5 Recent stroke* 4,460 3.6 4,074 3.3 Renal dysfunction* 14,275 11.6 14,062 11.4 135

Appendix 16. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 after propensity score matching in Aim 2 Schizophrenia 399 0.3 410 0.3 Transient ischemic attack 5,544 4.5 5,433 4.4 Urinary tract infection 12,150 9.9 12,278 10.0 Baseline medication use, n (%) Angiotensin-converting enzyme 55,763 45.4 55,671 45.3 inhibitors (ACEIs)* Agents for dementia 2,723 2.2 2,758 2.2 Antidiabetics* 27,924 22.7 27,744 22.6 Antiparkinson agents* 2,347 1.9 2,375 1.9 Antiplatelets* 27,165 22.1 25,223 20.5 Angiotensin II receptor blockers 18,128 14.8 18,076 14.7 (ARBs)* Atypical antipsychotics 3,283 2.7 3,141 2.6 Beta-blockers 62,937 51.2 61,558 50.1 Biologic disease-modifying 490 0.4 460 0.4 antirheumatic drugs (DMARDs) Calcium channel blockers* 27,616 22.5 27,704 22.5 Digoxin* 3,165 2.6 3,054 2.5 Heparin and low-molecular-weight 1,623 1.3 1,557 1.3 heparin Lithium 333 0.3 338 0.3 Loop diuretics* 17,248 14.0 16,898 13.8 Monoamine oxidase inhibitors 4 0.0 4 0.0 (MOAs) Nitrates 18,463 15.0 17,258 14.0 Nonbiologic DMARDs 2,352 1.9 2,381 1.9

Nonselective nonsteroidal anti- 20,961 17.1 20,944 17 inflammatory drugs (NSAIDs)*

136

Appendix 16. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 after propensity score matching in Aim 2 Opioids* 49,215 40.1 48,797 39.7 Oral anticoagulants 10,573 8.6 10,492 8.5 Other anticoagulants 77 0.1 80 0.1 Other antihypertensives 9,199 7.5 9,213 7.5 Other lipid-lowering agents* 25,482 20.7 25,126 20.4 Other newer and atypical 8,514 6.9 8,463 6.9 antidepressants Potassium-sparing diuretics 8,833 7.2 8,882 7.2 Renin inhibitor 365 0.3 337 0.3 Selective COX-2 inhibitors* 3,959 3.2 3,967 3.2 Selective serotonin-norepinephrine 7,223 5.9 7,081 5.8 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 18,821 15.3 18,941 15.4 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs) 4,676 3.8 4,693 3.8 Thiazide diuretics 34,950 28.4 35,639 29.0 Typical antipsychotics 425 0.3 437 0.4 Health care utilization, mean (SD) Number of distinct cardiovascular 4 4.2 4 4.1 diagnoses Number of International Normalized 1 3.2 1 3.3 Ratio (INR) tests Number of vaccinations* 0 0.8 0 0.8 Number of dispensations of any 32 31.6 32 32.0 drugs* Number of dispensations of any drugs 6 3.6 6 3.5 with days’ supply that overlap the index date* Number of office visits 16 17.6 16 18.0

137

Appendix 16. Characteristics of low- and high-intensity statin initiators in adherence prediction score tertile 3 after propensity score matching in Aim 2 Number of physician visits* 7 6.3 7 6.4 Number of cardiovascular 0 0.7 0 0.6 hospitalizations Number of office visits with a 6 9.0 6 9.2 cardiovascular diagnosis Combined comorbidity score 1 2.1 1 2.1 Number of different drugs 9 5.7 9 5.7 dispensed* Total days in hospital in prior year 2 6.8 2 6.6 Number of hospitalizations in the 0 0.7 0 0.7 prior year* Index Rx copayment* 22 25.3 23 23.4 Index Rx days’ supply* 48 28.0 49 28.3 Number of dispensations of the drugs 16 15.1 16 15.7 of interest* Medication synchronization metrics* 0 0.2 0 0.2 Total baseline out-of-pocket 489 619.0 484 701.7 prescription drug costs

138

Appendix 17. Characteristics of low- and high-intensity statin initiators overall across all tertiles after propensity score matching in Aim 2 Low-intensity Statin High-intensity Statin Initiators Initiators (n = 407,121) (n = 407,121) Age, mean (SD)* 60 13.1 61 12.8 Sex: female* 182,907 44.9 190,299 46.7 Regions (%) Midwest 88,189 21.7 89,167 21.9 Northeast 37,709 9.3 36,341 8.9 South 212,148 52.1 208,332 51.2 West 67,327 16.5 71,242 17.5 Unknown 1,748 0.4 2,039 0.5 Use of preventive services Colonoscopy*, n (%) 35,099 8.6 35,292 8.7 Fecal occult blood test, n (%) 36,782 9.0 37,198 9.1 Mammography*, n (% of females) 53,964 29.5 56,068 29.5 Any prior lab test, n (%) 121,073 29.7 116,207 28.5 Lipid test*, n (%) 274,407 67.4 278,465 68.4 Benefit plan types, n (%) EPO* 37,070 9.1 35,975 8.8 HMO* 87,693 21.5 93,564 23 IND* 12,251 3.0 11,583 2.8 OTH* 54,053 13.3 58,307 14.3 POS* 187,571 46.1 179,439 44.1 Use of wheelchair, walker, crutches, or 9,519 2.3 9,244 2.3 cane Index statin, n (%) Atorvastatin 92,003 22.6 253,608 62.3

139

Appendix 17. Characteristics of low- and high-intensity statin initiators overall across all tertiles after propensity score matching in Aim 2 Fluvastatin 4,487 1.1 0 0.0 Lovastatin 25,570 6.3 310 0.1 Pitavastatin 4,191 1.0 0 0.0 Pravastatin 54,979 13.5 0 0.0 Rosuvastatin 28,058 6.9 118,565 29.1 Simvastatin 197,833 48.6 34,638 8.5 Comorbidities, n (%) Atrial fibrillation 8,632 2.1 8,284 2.0 Acute coronary syndrome, with or 60,915 15.0 55,172 13.6 without revascularization* Acute coronary syndrome, with 39,543 9.7 35,299 8.7 revascularization* Alzheimer’s disease 12,231 3.0 12,414 3.0 Angina 42,503 10.4 38,395 9.4 Coronary atherosclerosis 105,696 26.0 98,065 24.1 Bone mineral density test 7,470 1.8 6,945 1.7 Coronary artery bypass graft (CABG), 7,789 1.9 6,741 1.7 new CABG, old 13,858 3.4 12,764 3.1 Heart failure 35,454 8.7 33,921 8.3 Heart failure hospitalization 4,225 1.0 3,928 1.0 Chronic obstructive pulmonary disease 47,534 11.7 46,898 11.5 (COPD) Any cancer* 57,297 14.1 57,922 14.2 Cardiovascular system symptom 51,687 12.7 50,143 12.3 Chest pain 118,915 29.2 112,015 27.5 Conduction disorders 13,347 3.3 12,769 3.1

140

Appendix 17. Characteristics of low- and high-intensity statin initiators overall across all tertiles after propensity score matching in Aim 2 Depression 51,844 12.7 51,447 12.6 Diabetes* 128,273 31.5 128,537 31.6 Electrocardiogram 203,420 50.0 197,403 48.5 History of falls 8,672 2.1 8,609 2.1 HIV 1397 0.3 1210 0.3 Hip fracture 352 0.1 353 0.1 Hyperlipidemia 315,937 77.6 315,091 77.4 Hyperparathyroidism 2127 0.5 2058 0.5 Hypertension 273,275 67.1 273,916 67.3 Hyperthyroidism 5,360 1.3 5,383 1.3 Ischemic heart disease* 109,960 27.0 102,148 25.1 Prior liver disease* 17,899 4.4 17,292 4.2 Osteoporosis 19,208 4.7 20,278 5.0 Use of oxygen 8,152 2.0 7,990 2.0 Peripheral vascular disease (PVD) or 26,683 6.6 26,121 6.4 PVD surgery* Palpitations 22,251 5.5 21,631 5.3 Parkinson’s disease 1845 0.5 1925 0.5 Post-myocardial infarction/acute 40,007 9.8 36,514 9.0 coronary syndromes (MI/ACS)* Preventive care 186,833 45.9 193,393 47.5 Prior MI 22,522 5.5 20,501 5.0 Prior stroke* 9,524 2.3 8,462 2.1 Rheumatoid arthritis 6,780 1.7 6,692 1.6 Recent MI* 18068 4.4 16438 4.0 Recent stroke* 6633 1.6 5755 1.4 Renal dysfunction* 39,715 9.8 39,427 9.7 141

Appendix 17. Characteristics of low- and high-intensity statin initiators overall across all tertiles after propensity score matching in Aim 2 Schizophrenia 1197 0.3 1214 0.3 Transient ischemic attack 14,474 3.6 14,116 3.5 Urinary tract infection 40,154 9.9 40,772 10.0 Baseline medication use, n (%) Angiotensin-converting enzyme 156,645 38.5 159,266 39.1 inhibitors (ACEIs)* Agents for dementia 5,137 1.3 5,448 1.3 Antidiabetics* 107,549 26.4 107,820 26.5 Antiparkinson agents* 6,794 1.7 7,028 1.7 Antiplatelets* 59,442 14.6 54,771 13.5 Angiotensin II receptor blockers 81,101 19.9 78,182 19.2 (ARBs)* Atypical antipsychotics 9,218 2.3 9,063 2.2 Beta-blockers 154,986 38.1 151,013 37.1 Biologic disease-modifying 1386 0.3 1262 0.3 antirheumatic drugs (DMARDs) Calcium channel blockers* 96,547 23.7 96,760 23.8 Digoxin* 8,750 2.1 8,739 2.1 Heparin and low-molecular-weight 4,501 1.1 4,338 1.1 heparin Lithium 1129 0.3 1080 0.3 Loop diuretics* 41,308 10.1 40,652 10 Monoamine oxidase inhibitors (MOAs) 21 0.0 16 0.0 Nitrates 38,386 9.4 35,779 8.8 Nonbiologic DMARDs 5,912 1.5 5,888 1.4

Nonselective nonsteroidal anti- 80,377 19.7 80,575 19.8 inflammatory drugs (NSAIDs)* Opioids* 138,763 34.1 137,044 33.7 142

Appendix 17. Characteristics of low- and high-intensity statin initiators overall across all tertiles after propensity score matching in Aim 2 Oral anticoagulants 23,792 5.8 23,595 5.8 Other anticoagulants 220 0.1 228 0.1 Other antihypertensives 23,263 5.7 23,336 5.7 Other lipid-lowering agents* 56,276 13.8 54,096 13.3 Other newer and atypical 26,304 6.5 25,886 6.4 antidepressants Potassium-sparing diuretics 25,960 6.4 26,428 6.5 Renin inhibitor 1383 0.3 1303 0.3 Selective COX-2 inhibitors* 11,657 2.9 11,213 2.8 Serotonin and norepinephrine 19,181 4.7 18,723 4.6 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 72,259 17.7 72,090 17.7 (SSRIs)* Tricyclic antidepressants (TCAs) 13,035 3.2 13,000 3.2 Thiazide diuretics 122,640 30.1 124,696 30.6 Typical antipsychotics 1215 0.3 1244 0.3 Health care utilization, mean (SD) Number of distinct cardiovascular 3 4.0 3 3.8 diagnoses Number of International Normalized 1 2.5 1 2.6 Ratio (INR) tests Number of vaccinations* 0 0.6 0 0.6 Number of dispensations of any 26 24.6 26 24.9 drugs* Number of dispensations of any drugs 5 3.0 5 3.0 with days’ supply that overlap the index date* Number of office visits 15 16.7 14 17.1 Number of physician visits* 7 5.9 7 6.0

143

Appendix 17. Characteristics of low- and high-intensity statin initiators overall across all tertiles after propensity score matching in Aim 2 Number of cardiovascular 0 0.6 0 0.6 hospitalizations Number of office visits with a 5 8.5 5 8.7 cardiovascular diagnosis Combined comorbidity score 1 2.0 1 1.9 Number of different drugs dispensed* 9 5.6 9 5.6 Total days in hospital in prior year 2 7.0 2 6.7 Number of hospitalizations in the prior 0 0.7 0 0.7 year* Index Rx copayment* 26 27.1 28 25.7 Index Rx days’ supply* 40 22.6 41 23.4 Number of dispensations of the drugs 13 12.1 13 12.5 of interest* Medication synchronization metrics* 0 0.2 0 0.2 Total baseline out-of-pocket 435 516.8 433 555.9 prescription drug costs

144

Appendix 18. Descriptive statistics for adherence in baseline and follow-up after propensity score matching in the statin cohort Exposure Variable Mean Median Lower Upper Min Max Group Quartile Quartile Tertile 1 Low- intensity Adherence score 1.78 1.89 1.50 2.16 -2.22 2.36 statins Mean PDC (baseline) 0.53 0.54 0.33 0.73 0.00 1.00 Mean PDC (follow-up) 0.40 0.38 0.16 0.61 0.00 1.00 High- Adherence score 1.79 1.89 1.51 2.16 -1.13 2.36 intensity Mean PDC (baseline) 0.53 0.54 0.33 0.73 0.00 1.00 statins Mean PDC (follow-up) 0.40 0.38 0.33 0.61 0.00 1.00 Tertile 2 Low- intensity Adherence score 2.66 2.66 0.33 2.80 2.36 2.95 statins Mean PDC (baseline) 0.85 0.91 0.33 1.00 0.01 1.00 Mean PDC (follow-up) 0.57 0.60 0.33 0.83 0.00 1.00 High- Adherence score 2.66 2.66 0.33 2.80 2.36 2.95 intensity Mean PDC (baseline) 0.85 0.91 0.33 1.00 0.00 1.00 statins Mean PDC (follow-up) 0.57 0.60 0.33 0.83 0.00 1.00 Tertile 3 Low- intensity Adherence score 3.42 3.32 0.33 3.63 2.95 7.26 statins Mean PDC (baseline) 0.91 0.96 0.33 1.00 0.01 1.00 Mean PDC (follow-up) 0.63 0.67 0.33 0.88 0.00 1.00 Adherence score 3.41 3.31 0.33 3.61 2.95 7.53

145

Appendix 18. Descriptive statistics for adherence in baseline and follow-up after propensity score matching in the statin cohort High- Mean PDC (baseline) 0.91 0.96 0.33 1.00 0.01 1.00 intensity Mean PDC (follow-up) 0.63 0.67 0.33 0.88 0.00 1.00 statins Overall Low- intensity Adherence score 2.57 2.60 0.33 3.06 -2.22 7.14 data statins Mean PDC (baseline) 0.75 0.83 0.33 0.99 0.00 1.00 Mean PDC (follow-up) 0.52 0.53 0.33 0.79 0.00 1.00 High- Adherence score 2.57 2.61 0.33 3.07 -1.13 7.53 intensity Mean PDC (baseline) 0.75 0.84 0.33 0.99 0.00 1.00 statins Mean PDC (follow-up) 0.53 0.54 0.33 0.78 0.00 1.00

146

Appendix 19. Myocardial infarction incidence rates (IR) and IR differences between low- and high-intensity statin initiators after propensity score matching in as-treated analysis

Low-intensity Statins High-intensity Statins

No. of IR/1000 No. of IR/1000 IR difference events person-years events person-years

Tertile 1 556 6.50 558 6.76 0.26 Tertile 2 578 6.68 564 6.81 0.13 Tertile 3 1,150 13.44 1,090 13.38 -0.06 Overall data 2,267 8.77 2,203 8.89 0.12

147

Appendix 20. Cerebrovascular incidence rates (IR) and IR differences between low- and high-intensity statin initiators after propensity score matching in as-treated analysis

Low-intensity Statins High-intensity Statins

No. of events IR/1000 No. of IR/1000 IR difference person-years events person-years

Tertile 1 861 10.08 839 10.18 0.11 Tertile 2 906 10.49 898 10.86 0.37 Tertile 3 1,495 17.51 1,532 18.85 1.34 Overall data 3,294 12.76 3,285 13.28 0.5

148

Appendix 21. Patient flow chart for Aim 2 osteoporosis drug cohort

149

Appendix 22. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the bisphosphonate vs calcitonin drug cohort Exposure group Variable Mean Median Lower Upper Min Max Quartile Quartile Tertile 1 Bisphosphonates Adherence –145.62 –145.53 –145.81 –145.34 –148.47 –145.18 score Mean PDC 0.59 0.62 0.42 0.79 0.00 1.00 (baseline)

Mean PDC 0.78 0.84 0.64 0.97 0.01 1.00 (follow-up)

Calcitonin Adherence –145.65 –145.56 –145.86 –145.35 –147.77 –145.18 score Mean PDC 0.62 0.64 0.45 0.81 0.01 1.00 (baseline)

Mean PDC 0.76 0.81 0.63 0.95 0.06 1.00 (follow-up)

Tertile 2 Bisphosphonates Adherence –144.91 –144.91 –145.04 –144.78 –145.18 –144.65 score Mean PDC 0.80 0.85 0.70 0.95 0.01 1.00 (baseline)

Mean PDC 0.84 0.91 0.76 0.99 0.01 1.00 (follow-up)

150

Appendix 22. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the bisphosphonate vs calcitonin drug cohort Calcitonin Adherence –144.92 –144.93 –145.05 –144.79 –145.18 –144.65 score Mean PDC 0.80 0.85 0.70 0.95 0.09 1.00 (baseline)

Mean PDC 0.82 0.86 0.72 0.97 0.03 1.00 (follow-up)

Tertile 3 Bisphosphonates Adherence –144.24 –144.32 –144.50 –144.06 –144.65 –140.47 score Mean PDC 0.89 0.93 0.84 0.98 0.02 1.00 (baseline)

Mean PDC 0.88 0.94 0.82 0.99 0.01 1.00 (follow-up)

Calcitonin Adherence –144.22 –144.31 –144.50 –144.03 –144.65 –141.78 score Mean PDC 0.89 0.93 0.84 0.98 0.13 1.00 (baseline)

Mean PDC 0.85 0.90 0.77 0.98 0.09 1.00 (follow-up)

Overall Bisphosphonates Adherence –144.92 –144.91 –145.34 –144.50 –148.47 –140.47 data score

151

Appendix 22. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the bisphosphonate vs calcitonin drug cohort Mean PDC 0.76 0.83 0.63 0.95 0.00 1.00 (baseline)

Mean PDC 0.82 0.90 0.73 0.99 0.00 1.00 (follow-up)

Calcitonin Adherence –144.96 –144.96 –145.39 –144.53 –147.77 –141.78 score Mean PDC 0.76 0.83 0.64 0.95 0.01 1.00 (baseline)

Mean PDC 0.78 0.84 0.67 0.96 0.01 1.00 (follow-up)

152

Appendix 23. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the raloxifene vs calcitonin drug cohort Exposure Variable Mean Median Lower Upper Min Max Group Quartile Quartile Tertile 1 Raloxifene Adherence –145.63 –145.55 –145.83 –145.36 –147.36 –145.20 score Mean PDC 0.57 0.58 0.38 0.78 0.00 1.00 (baseline)

Mean PDC 0.75 0.79 0.62 0.94 0.08 1.00 (follow-up)

Calcitonin Adherence –145.66 –145.58 –145.87 –145.37 –147.77 –145.20 score Mean PDC 0.61 0.64 0.44 0.81 0.01 1.00 (baseline)

Mean PDC 0.76 0.81 0.63 0.95 0.06 1.00 (follow-up)

Tertile 2 Raloxifene Adherence –144.93 –144.92 –145.06 –144.80 –145.20 –144.66 score Mean PDC 0.78 0.84 0.67 0.95 0.09 1.00 (baseline)

153

Appendix 23. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the raloxifene vs calcitonin drug cohort Mean PDC 0.83 0.87 0.74 0.97 0.09 1.00 (follow-up)

Calcitonin Adherence –144.93 –144.93 –145.06 –144.79 –145.20 –144.66 score Mean PDC 0.80 0.84 0.70 0.95 0.09 1.00 (baseline)

Mean PDC 0.82 0.86 0.72 0.97 0.03 1.00 (follow-up)

Tertile 3 Raloxifene Adherence –144.24 –144.32 –144.51 –144.06 –144.66 –142.13 score Mean PDC 0.89 0.93 0.84 0.98 0.15 1.00 (baseline)

Mean PDC 0.87 0.92 0.80 0.98 0.09 1.00 (follow-up)

Calcitonin Adherence –144.23 –144.31 –144.51 –144.03 –144.66 –141.78 score Mean PDC 0.89 0.93 0.84 0.98 0.13 1.00 (baseline)

Mean PDC 0.84 0.90 0.76 0.98 0.09 1.00 (follow-up)

154

Appendix 23. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the raloxifene vs calcitonin drug cohort Overall Raloxifene Adherence –144.92 –144.91 –145.34 –144.49 –147.36 –142.13 data score Mean PDC 0.75 0.83 0.61 0.95 0.00 1.00 (baseline)

Mean PDC 0.81 0.88 0.71 0.97 0.03 1.00 (follow-up)

Calcitonin Adherence –144.96 –144.96 –145.39 –144.53 –147.77 –141.78 score Mean PDC 0.76 0.83 0.64 0.95 0.01 1.00 (baseline)

Mean PDC 0.78 0.84 0.67 0.96 0.01 1.00 (follow-up)

155

Appendix 24. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the bisphosphonates vs raloxifene drug cohort Exposure Group Variable Mean Median Lower Upper Min Max Quartile Quartile Tertile 1 Bisphosphonates Adherence –145.62 –145.53 –145.81 –145.34 –148.47 –145.18 score Mean PDC 0.60 0.62 0.42 0.79 0.00 1.00 (baseline)

Mean PDC 0.78 0.84 0.64 0.97 0.01 1.00 (follow-up)

Raloxifene Adherence –145.62 –145.54 –145.81 –145.34 –147.36 –145.18 score Mean PDC 0.58 0.59 0.39 0.78 0.00 1.00 (baseline)

Mean PDC 0.76 0.80 0.63 0.94 0.08 1.00 (follow-up)

Tertile 2 Bisphosphonates Adherence –144.91 –144.91 –145.04 –144.78 –145.18 –144.65 score Mean PDC 0.80 0.85 0.70 0.95 0.01 1.00 (baseline)

Mean PDC 0.84 0.91 0.76 0.99 0.01 1.00 (follow-up)

156

Appendix 24. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the bisphosphonates vs raloxifene drug cohort Raloxifene Adherence –144.91 –144.91 –145.04 –144.79 –145.18 –144.65 score Mean PDC 0.79 0.84 0.68 0.95 0.09 1.00 (baseline)

Mean PDC 0.83 0.87 0.74 0.97 0.09 1.00 (follow-up)

Tertile 3 Bisphosphonates Adherence –144.24 –144.32 –144.50 –144.06 –144.65 –140.47 score Mean PDC 0.89 0.93 0.84 0.98 0.02 1.00 (baseline)

Mean PDC 0.88 0.94 0.82 0.99 0.01 1.00 (follow-up)

Raloxifene Adherence –144.24 –144.31 –144.50 –144.05 –144.65 –142.13 score Mean PDC 0.89 0.93 0.84 0.98 0.15 1.00 (baseline)

Mean PDC 0.87 0.92 0.80 0.98 0.09 1.00 (follow-up)

Overall Bisphosphonates Adherence –144.92 –144.91 –145.34 –144.50 –148.47 –140.47 data score

157

Appendix 24. Descriptive statistics for adherence in baseline and follow-up before propensity score fine stratification in the bisphosphonates vs raloxifene drug cohort Mean PDC 0.76 0.83 0.63 0.95 0.00 1.00 (baseline)

Mean PDC 0.82 0.90 0.73 0.99 0.00 1.00 (follow-up)

Raloxifene Adherence –144.92 –144.91 –145.34 –144.49 –147.36 –142.13 score Mean PDC 0.75 0.83 0.61 0.95 0.00 1.00 (baseline)

Mean PDC 0.81 0.88 0.71 0.97 0.03 1.00 (follow-up)

158

Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 69 (12.9) 66 (11.5) 3.2 26.5 Sex: female* 2396 (87.1) 65252 (91.2) –4.1 –13.1 Combined comorbidity score* 2 (2.7) 1 (2) 0.9 37.2 Commercial plan type, n (%) 933 (33.9) 30399 (42.5) –8.6 –17.7 Benefit plan types, n (%) EPO* 111 (4) 4718 (6.6) –2.6 -11.4 HMO* 1201 (43.7) 25537 (35.7) 8 16.4 IND* 73 (2.7) 945 (1.3) 1.3 9.6 OTH* 570 (20.7) 14902 (20.8) –0.1 –0.2 POS* 582 (21.2) 20425 (28.5) –7.4 –17.1 Comorbidities, n (%) Acute coronary syndrome, 232 (8.4) 4015 (5.6) 2.8 11.1 with or without revascularization Acute coronary syndrome, 69 (2.5) 1127 (1.6) 0.9 6.6 with revascularization Atrial fibrillation 67 (2.4) 880 (1.2) 1.2 9

159

Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Alzheimer’s disease* 206 (7.5) 2521 (3.5) 4 17.5 Angina 197 (7.2) 3444 (4.8) 2.3 9.9 Coronary atherosclerosis* 583 (21.2) 10462 (14.6) 6.6 17.2 Coronary artery bypass graft 16 (0.6) 211 (0.3) 0.3 4.3 (CABG), new CABG, old 75 (2.7) 1136 (1.6) 1.1 7.8 Any cancer 576 (20.9) 12264 (17.1) 3.8 9.7 Any malignancy, including 527 (19.2) 11411 (15.9) 3.2 8.5 lymphoma and leukemia, except malignant neoplasm of skin* Cardiovascular system 465 (16.9) 10015 (14) 2.9 8.1 symptom Chest pain 850 (30.9) 16528 (23.1) 7.8 17.7 Heart failure* 414 (15) 5430 (7.6) 7.5 23.7 Heart failure hospitalization 40 (1.5) 350 (0.5) 1 9.9 Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2

160

Calcitonin Bisphosphonates Absolute Standardized Difference Differences Conduction disorders 104 (3.8) 1758 (2.5) 1.3 7.6 Chronic obstructive pulmonary 700 (25.4) 13101 (18.3) 7.1 17.3 disease (COPD) Depression* 690 (25.1) 16607 (23.2) 1.9 4.4 Diabetes* 695 (25.3) 17518 (24.5) 0.8 1.8 Drug-induced osteoporosis* 60 (2.2) 1515 (2.1) 0.1 0.4 Falls* 173 (6.3) 2155 (3) 3.3 15.6 History of falls 153 (5.6) 1825 (2.5) 3 15.3 HIV infection 1 (0) 80 (0.1) –0.1 –2.8 Hyperlipidemia 1605 (58.3) 44927 (62.8) –4.4 –9.1 Hyperparathyroidism 62 (2.3) 1170 (1.6) 0.6 4.5 Hypertension* 1896 (68.9) 45806 (64) 4.9 10.4 Hyperthyroidism 54 (2) 1629 (2.3) -0.3 -2.2 Inflammatory bowel disease 64 (2.3) 1334 (1.9) 0.5 3.2 (IBD)* Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences

161

Ischemic heart disease* 608 (22.1) 11044 (15.4) 6.7 17.2 Liver disease* 222 (8.1) 4683 (6.5) 1.5 5.9 Metastatic cancer* 47 (1.7) 568 (0.8) 0.9 8.2 Osteoporotic fracture: non- — — — — vertebral* Osteoporotic fracture: — — — — vertebral* Osteoporosis* 1011 (36.8) 32706 (45.7) –8.9 –18.2 Other fractures* — — — — Palpitations 221 (8) 4710 (6.6) 1.5 5.6 Parkinson’s disease 49 (1.8) 549 (0.8) 1 9.1 Post-myocardial 98 (3.6) 1794 (2.5) 1.1 6.2 infarction/acute coronary syndromes (MI/ACS)*

Preventive care 1547 (56.2) 48155 (67.3) –11 –22.9 Prior MI 34 (1.2) 500 (0.7) 0.5 5.5 Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Prior stroke 14 (0.5) 273 (0.4) 0.1 1.9

162

Peripheral vascular disease 306 (11.1) 5175 (7.2) 3.9 13.5 (PVD) or PVD surgery Rheumatoid arthritis* 181 (6.6) 5830 (8.1) –1.6 –6 Recent MI 2 (0.1) 19 (0) 0 2.1 Recent stroke 1 (0) 27 (0) 0 -0.1 Renal dysfunction 578 (21) 8898 (12.4) 8.6 23.1 Schizophrenia 12 (0.4) 203 (0.3) 0.2 2.5 Transient ischemic attack* 136 (4.9) 1857 (2.6) 2.3 12.4 Upper GI diseases* 1004 (36.5) 18645 (26) 10.4 22.7 Urinary tract infection 661 (24) 13353 (18.7) 5.4 13.1 Baseline medication use, n (%) Systemic steroid* 411 (14.9) 10028 (14) 0.9 2.6 Bile acid sequestrants 59 (2.1) 1169 (1.6) 0.5 3.8 IV osteoclast inhibitors* 29 (1.1) 426 (0.6) 0.5 5.1 Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Inhaled glucocorticoid* 383 (13.9) 7692 (10.7) 3.2 9.7 Lipid-lowering agents 1155 (42) 34717 (48.5) –6.5 –13.1 Niacin and fibrates 108 (3.9) 3095 (4.3) –0.4 –2 163

Angiotensin-converting 906 (32.9) 22399 (31.3) 1.6 3.5 enzyme inhibitors (ACEIs)* Agents for dementia 127 (4.6) 1875 (2.6) 2 10.7 Antidiabetics* 505 (18.4) 12984 (18.1) 0.2 0.6 Antiparkinson agents* 141 (5.1) 2609 (3.6) 1.5 7.2 Antiplatelets* 4 (0.1) 133 (0.2) 0 –1 Angiotensin II receptor 554 (20.1) 13328 (18.6) 1.5 3.8 blockers (ARBs)* Aromatase inhibitors 28 (1) 1402 (2) –0.9 –7.8 Atypical antipsychotics 132 (4.8) 2232 (3.1) 1.7 8.6 Beta-blockers* 1138 (41.4) 22684 (31.7) 9.7 20.2

164

Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Biologic disease-modifying 32 (1.2) 1187 (1.7) –0.5 –4.2 antirheumatic drugs (DMARDs)* Calcium channel blockers* 832 (30.2) 18354 (25.6) 4.6 10.3 Digoxin* 119 (4.3) 1433 (2) 2.3 13.3 Heparin and low-molecular- 52 (1.9) 715 (1) 0.9 7.5 weight heparin* Lithium 10 (0.4) 210 (0.3) 0.1 1.2 Loop diuretics* 611 (22.2) 9021 (12.6) 9.6 25.5 Monoamine oxidase inhibitors 0 (0) 6 (0) 0 –1.3 (MOAs) Nitrates* 313 (11.4) 4192 (5.9) 5.5 19.8 Nonbiologic DMARDs 171 (6.2) 5826 (8.1) –1.9 –7.5

Nonselective nonsteroidal 749 (27.2) 20686 (28.9) –1.7 –3.7 anti-inflammatory drugs (NSAIDs)* Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2

165

Calcitonin Bisphosphonates Absolute Standardized Difference Differences Opioids* 1785 (64.9) 34724 (48.5) 16.4 33.5 Oral anticoagulants* 252 (9.2) 3735 (5.2) 3.9 15.3 Other anticoagulants* 6 (0.2) 38 (0.1) 0.2 4.5 Other antihypertensives 219 (8) 3268 (4.6) 3.4 14 Other newer and atypical 416 (15.1) 8658 (12.1) 3 8.8 antidepressants* Potassium-sparing diuretics 277 (10.1) 5509 (7.7) 2.4 8.3 Renin inhibitor 11 (0.4) 179 (0.3) 0.1 2.6 Selective COX-2 inhibitors* 139 (5.1) 3366 (4.7) 0.4 1.6 Selective estrogen receptor 8 (0.3) 404 (0.6) –0.3 –4.2 blockers Selective serotonin- 259 (9.4) 6383 (8.9) 0.5 1.7 norepinephrine reuptake inhibitors (SNRIs)* Selective serotonin reuptake 961 (34.9) 23401 (32.7) 2.2 4.7 inhibitors (SSRIs)* Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences

166

Tricyclic antidepressants 232 (8.4) 4175 (5.8) 2.6 10.1 (TCAs)* Thiazide diuretics* 755 (27.4) 21326 (29.8) –2.3 –5.2 Typical antipsychotics* 11 (0.4) 258 (0.4) 0 0.6 No lipid-lowering drug despite 648 (23.6) 15328 (21.4) 2.1 5.1 diagnosis of hyperlipidemia* Long-term use of opioids* 725 (26.4) 16705 (23.3) 3 7 Long-term current use of 44 (1.6) 1640 (2.3) –0.7 –5 steroids* Lack of 2nd RX among all the 144 (5.2) 5221 (7.3) –2.1 –8.5 interested drugs groups* Opioids baseline days’ supply ≥ 758 (27.6) 12540 (17.5) 10 24.2 90* Index days’ supply* 32 (10.3) 36 (18.8) –3.9 –25.9 Use of preventive services Colonoscopy*, n (%) 297 (10.8) 7566 (10.6) 0.2 0.7 Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Fecal occult blood test*, n (%) 274 (10) 8600 (12) –2.1 –6.6

167

Mammography*, n (% of 638 (26.6) 23210 (35.6) –9.0 –20.7 females) Any prior lab test, n (%) 0 (0.4) 0 (0.4) –0.1 –12.5 Bone mineral density test*, n 91 (3.3) 2707 (3.8) –0.5 –2.6 (%) Electrocardiography, n (%) 1351 (49.1) 31850 (44.5) 4.6 9.3 INR test, n (%) 1 (4.1) 0 (2.7) 0.4 12.6 Use of wheelchair, walker, 147 (5.3) 2132 (3) 2.4 11.9 crutches, or cane, n (%) Use of oxygen, n (%) 148 (5.4) 2306 (3.2) 2.2 10.7 Lipid test, n (%) 1509 (54.9) 46413 (64.8) –10 –20.5 Health care utilization, mean (SD) Number of distinct 4 (4.1) 3 (3.2) 1 27.4 cardiovascular diagnoses Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Total days in hospital in prior 2 (6.1) 1 (4.5) 1.1 20.3 year* Medication synchronization 0 (0.2) 0 (0.2) 0 10.7 metrics

168

Copayment 805 (883.3) 595 (730.6) 210.1 25.9 Number of office visits 24 (27.1) 18 (19.8) 6.3 26.7 Number of unique drugs of 6 (3.6) 4 (3.3) 1.7 48.5 interest (by NDC) in prior year or on index date Number of office visits with a 7 (11.7) 5 (9.1) 2.2 20.5 cardiovascular diagnosis Number of drugs of interest 18 (14.4) 15 (12.9) 3.1 22.7 dispensations in 365 days in prior year or on index date* Number of hospitalizations in 0 (0.9) 0 (0.7) 0.2 24.3 the prior year Medication synchronization 0 (0.2) 0 (0.2) 0 9.9 metrics (non-mail)* Appendix 25. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Number of cardiovascular 0 (0.7) 0 (0.5) 0.1 20.9 hospitalizations Index copayment 31 (25.2) 20 (26.9) 11.3 43.4 Vaccinations* 0 (0.8) 0 (0.8) 0 –1 Physician visits* 11 (9) 10 (7.4) 1.8 21.9

169

Number of drugs of interest 3 (1.7) 2 (1.7) 1.1 67 dispensations with days’ supply that overlap index date Number unique (by NDC) 5 (3) 4 (2.7) 0.9 31.7 drugs in 365 days before or on index date with days’ supply that overlap index date Number of dispensations in 6 (3.1) 5 (2.9) 0.9 31 baseline period with days’ supply that overlap index date*

170

171

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 71 (11.1) 68 (10.9) 3.1 28.1 Sex: female* 2246 (87.5) 64847 (90.4) –2.9 –9.3 Combined comorbidity score* 1 (2.5) 1 (1.8) 0.8 34.9 Commercial plan type, n (%) 861 (33.5) 27903 (38.9) –5.4 –11.2 Benefit plan types, n (%) EPO* 78 (3) 3448 (4.8) –1.8 –9.1 HMO* 1259 (49) 29193 (40.7) 8.3 16.8 IND* 119 (4.6) 2008 (2.8) 1.8 9.7 OTH* 467 (18.2) 14921 (20.8) –2.6 –6.6 POS* 472 (18.4) 17055 (23.8) –5.4 –13.2 Comorbidities, n (%) Acute coronary syndrome, 138 (5.4) 2655 (3.7) 1.7 8.1 with or without revascularization Acute coronary syndrome, 34 (1.3) 680 (0.9) 0.4 3.6 with revascularization

172

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Atrial fibrillation 55 (2.1) 728 (1) 1.1 9.1 Alzheimer’s disease* 243 (9.5) 2839 (4) 5.5 22.1 Angina 112 (4.4) 2276 (3.2) 1.2 6.3 Coronary atherosclerosis* 431 (16.8) 7820 (10.9) 5.9 17.1 Coronary artery bypass graft 4 (0.2) 89 (0.1) 0 0.8 (CABG), new CABG, old 51 (2) 882 (1.2) 0.8 6 Any cancer 512 (19.9) 13756 (19.2) 0.8 1.9 Any malignancy, including 474 (18.5) 12757 (17.8) 0.7 1.8 lymphoma and leukemia, except malignant neoplasm of skin* Cardiovascular system 311 (12.1) 7643 (10.7) 1.5 4.6 symptom Chest pain 594 (23.1) 11475 (16) 7.1 18.1 Heart failure* 304 (11.8) 3965 (5.5) 6.3 22.6 Heart failure hospitalization 19 (0.7) 242 (0.3) 0.4 5.5 Conduction disorders 82 (3.2) 1471 (2) 1.1 7.2 Chronic obstructive pulmonary 491 (19.1) 9653 (13.5) 5.7 15.4 disease (COPD)

173

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Depression* 380 (14.8) 9231 (12.9) 1.9 5.6 Diabetes* 511 (19.9) 14454 (20.1) –0.2 –0.6 Drug-induced osteoporosis* 50 (1.9) 1402 (2) 0 0 Falls* 115 (4.5) 1646 (2.3) 2.2 12.1 History of falls 94 (3.7) 1424 (2) 1.7 10.1 HIV infection 1 (0) 77 (0.1) –0.1 –2.5 Hyperlipidemia 1418 (55.2) 45050 (62.8) –7.6 –15.4 Hyperparathyroidism 33 (1.3) 1108 (1.5) –0.3 –2.2 Hypertension* 1718 (66.9) 44431 (61.9) 5 10.4 Hyperthyroidism 51 (2) 1455 (2) 0 –0.3 Inflammatory bowel disease 35 (1.4) 756 (1.1) 0.3 2.8 (IBD)* Ischemic heart disease* 459 (17.9) 8331 (11.6) 6.3 17.7 Liver disease* 150 (5.8) 3003 (4.2) 1.7 7.6 Metastatic cancer* 50 (1.9) 764 (1.1) 0.9 7.3 Osteoporotic fracture: non- — — — — vertebral* Osteoporotic fracture: — — — — vertebral*

174

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Osteoporosis* 1044 (40.7) 33444 (46.6) –6 –12 Other fractures* — — — — Palpitations 126 (4.9) 3526 (4.9) 0 0 Parkinson’s disease 57 (2.2) 569 (0.8) 1.4 11.7 Post-myocardial 70 (2.7) 1350 (1.9) 0.8 5.6 infarction/acute coronary syndromes (MI/ACS)* Preventive care 1485 (57.8) 49684 (69.2) –11.4 –23.9 Prior MI 27 (1.1) 291 (0.4) 0.6 7.6 Prior stroke 16 (0.6) 230 (0.3) 0.3 4.4 Peripheral vascular disease 257 (10) 4321 (6) 4 14.7 (PVD) or PVD surgery Rheumatoid arthritis* 109 (4.2) 3514 (4.9) –0.7 –3.1 Recent MI 2 (0.1) 19 (0) 0.1 2.3 Recent stroke 2 (0.1) 16 (0) 0.1 2.5 Renal dysfunction 449 (17.5) 7599 (10.6) 6.9 19.9 Schizophrenia 15 (0.6) 205 (0.3) 0.3 4.5 Transient ischemic attack* 80 (3.1) 1459 (2) 1.1 6.8 Upper GI diseases* 720 (28) 13405 (18.7) 9.4 22.3

175

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Urinary tract infection 518 (20.2) 10310 (14.4) 5.8 15.4 Baseline medication use, n (%) Systemic steroid* 289 (11.3) 6713 (9.4) 1.9 6.3 Bile acid sequestrants 40 (1.6) 878 (1.2) 0.3 2.9 IV osteoclast inhibitors* 31 (1.2) 549 (0.8) 0.4 4.5 Inhaled glucocorticoid* 224 (8.7) 5257 (7.3) 1.4 5.1 Lipid-lowering agents 1214 (47.3) 40578 (56.5) –9.3 –18.6 Niacin and fibrates 112 (4.4) 3675 (5.1) -0.8 -3.6 Angiotensin-converting 898 (35) 23652 (33) 2 4.2 enzyme inhibitors (ACEIs)* Agents for dementia 131 (5.1) 1998 (2.8) 2.3 11.9 Antidiabetics* 395 (15.4) 10987 (15.3) 0.1 0.2 Antiparkinson agents* 103 (4) 1826 (2.5) 1.5 8.2 Antiplatelets* 0 (0) 91 (0.1) –0.1 –5 Angiotensin II receptor 507 (19.7) 12505 (17.4) 2.3 6 blockers (ARBs)* Aromatase inhibitors 32 (1.2) 1803 (2.5) –1.3 –9.3 Atypical antipsychotics 112 (4.4) 1579 (2.2) 2.2 12.2 Beta-blockers* 1105 (43) 22436 (31.3) 11.8 24.5 176

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Biologic disease-modifying 14 (0.5) 453 (0.6) –0.1 –1.1 antirheumatic drugs (DMARDs)* Calcium channel blockers* 810 (31.5) 18685 (26) 5.5 12.2 Digoxin* 135 (5.3) 1611 (2.2) 3 15.9 Heparin and low-molecular- 46 (1.8) 624 (0.9) 0.9 8.1 weight heparin* Lithium 4 (0.2) 153 (0.2) –0.1 –1.3 Loop diuretics* 502 (19.5) 7066 (9.8) 9.7 27.7 Monoamine oxidase inhibitors 0 (0) 1 (0) 0 –0.5 (MOAs) Nitrates* 207 (8.1) 3024 (4.2) 3.8 16.1 Nonbiologic DMARDs 136 (5.3) 4218 (5.9) –0.6 –2.5

Nonselective nonsteroidal 512 (19.9) 15307 (21.3) –1.4 –3.4 anti-inflammatory drugs (NSAIDs)* Opioids* 1302 (50.7) 25403 (35.4) 15.3 31.3 Oral anticoagulants* 258 (10) 4263 (5.9) 4.1 15.2 Other anticoagulants* 1 (0) 38 (0.1) 0 –0.7 Other antihypertensives 176 (6.9) 3012 (4.2) 2.7 11.6 177

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Other newer and atypical 277 (10.8) 5439 (7.6) 3.2 11.1 antidepressants* Potassium-sparing diuretics 244 (9.5) 5917 (8.2) 1.3 4.4 Renin inhibitor 8 (0.3) 125 (0.2) 0.1 2.8 Selective COX-2 inhibitors* 103 (4) 2766 (3.9) 0.2 0.8 Selective estrogen receptor 14 (0.5) 494 (0.7) –0.1 –1.8 blockers Selective serotonin- 113 (4.4) 2828 (3.9) 0.5 2.3 norepinephrine reuptake inhibitors (SNRIs)* Selective serotonin reuptake 618 (24.1) 14928 (20.8) 3.3 7.8 inhibitors (SSRIs)* Tricyclic antidepressants 168 (6.5) 3114 (4.3) 2.2 9.7 (TCAs)* Thiazide diuretics* 805 (31.3) 22700 (31.6) –0.3 –0.6 Typical antipsychotics* 17 (0.7) 245 (0.3) 0.3 4.5 No lipid-lowering drug despite 466 (18.1) 10955 (15.3) 2.9 7.7 diagnosis of hyperlipidemia* Long-term use of opioids* 564 (22) 14800 (20.6) 1.3 3.3 Long-term current use of 27 (1.1) 1185 (1.7) –0.6 –5.2 steroids*

178

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Lack of 2nd RX among all the 17 (0.7) 1023 (1.4) –0.8 –7.5 interested drugs groups* Opioids baseline days’ supply ≥ 413 (16.1) 6294 (8.8) 7.3 22.3 90* Index days’ supply* 35 (17.1) 44 (25.2) –8.7 –40.5 Use of preventive services Colonoscopy*, n (%) 261 (10.2) 7630 (10.6) –0.5 –1.5 Fecal occult blood test*, n (%) 290 (11.3) 10135 (14.1) –2.8 –8.5 Mammography*, n (% of 709 (31.6) 27107 (41.8) –10.2 –21.8 females) Any prior lab test, n (%) 0 (0.3) 0 (0.4) 0 –12.6 Bone mineral density test*, n 215 (8.4) 6819 (9.5) –1.1 –4 (%) Electrocardiography, n (%) 1099 (42.8) 26956 (37.6) 5.2 10.7 International Normalized Ratio 1 (4.5) 1 (3) 0.4 10.8 (INR) test, n (%) Use of wheelchair, walker, 92 (3.6) 1407 (2) 1.6 9.9 crutches, or cane, n (%) Use of oxygen, n (%) 113 (4.4) 1571 (2.2) 2.2 12.4 Lipid test, n (%) 1292 (50.3) 45529 (63.4) –13.1 –26.8

179

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Health care utilization, mean (SD) Number of distinct 3 (3.5) 2 (2.7) 0.9 28.6 cardiovascular diagnoses Total days in hospital in prior 2 (6.4) 1 (3.9) 1 19.5 year* Medication synchronization 0 (0.2) 0 (0.2) 0 15.6 metrics Copayment 793 (846.9) 624 (701.1) 168.6 21.7 Number of office visits 20 (23.8) 15 (16.5) 5 24.5 Number of unique drugs of 6 (3.6) 4 (3) 1.8 55 interest (by NDC) in prior year or on index date Number of office visits with a 6 (10.2) 4 (7.6) 2 22 cardiovascular diagnosis Number of drugs of interest 23 (16.7) 18 (14.4) 4.4 28 dispensations in 365 days in prior year or on index date* Number of hospitalizations in 0 (1.1) 0 (0.5) 0.2 17.6 the prior year Medication synchronization 0 (0.2) 0 (0.2) 0 16.8 metrics (non-mail)*

180

Appendix 26. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Number of cardiovascular 0 (0.6) 0 (0.4) 0.1 18.5 hospitalizations Index copayment 35 (31.4) 25 (32.8) 10.1 31.3 Vaccinations* 0 (0.8) 0 (0.8) 0 -0.4 Physician visits* 9 (7.6) 8 (6.3) 0.9 12.3 Number of drugs of interest 4 (1.8) 2 (1.8) 1.3 70.9 dispensations with days’ supply that overlap index date Number unique (by NDC) 6 (3.1) 5 (2.8) 1.2 39.6 drugs in 365 days before or on index date with days’ supply that overlap index date Number of dispensations in 7 (3.4) 5 (2.9) 1.2 39 baseline period with days’ supply that overlap index date* Number of unique (by NDC) 15 (8.7) 12 (7.3) 3.1 39.2 drugs in 365 days before or on index date* Number of any drug 47 (33.5) 36 (26.6) 11.1 36.7 dispensations in 365 days before or on index date*

181

182

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 71 (10.4) 69 (10.5) 2.8 27 Sex: female* 2078 (86.2) 63634 (88.5) –2.3 –6.9 Combined comorbidity score* 1 (2.4) 1 (1.8) 0.7 32.7 Commercial plan type, n (%) 1196 (49.6) 35408 (49.2) 0.4 0.7 Benefit plan types, n (%) EPO* 74 (3.1) 3008 (4.2) –1.1 -6 HMO* 1018 (42.2) 28665 (39.9) 2.4 4.8 IND* 356 (14.8) 6375 (8.9) 5.9 18.4 OTH* 287 (11.9) 10562 (14.7) –2.8 –8.2 POS* 442 (18.3) 17276 (24) –5.7 –14 Comorbidities, n (%) Acute coronary syndrome, with or 90 (3.7) 2081 (2.9) 0.8 4.7 without revascularization Acute coronary syndrome, with 23 (1) 497 (0.7) 0.3 2.9 revascularization Atrial fibrillation 43 (1.8) 676 (0.9) 0.8 7.3 Alzheimer’s disease* 219 (9.1) 2990 (4.2) 4.9 19.9

183

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Angina 72 (3) 1781 (2.5) 0.5 3.1 Coronary atherosclerosis* 298 (12.4) 6152 (8.6) 3.8 12.5 Coronary artery bypass graft (CABG), new 4 (0.2) 60 (0.1) 0.1 2.3 CABG, old 36 (1.5) 751 (1) 0.4 4 Any cancer 544 (22.6) 15804 (22) 0.6 1.4 Any malignancy, including lymphoma and 492 (20.4) 14690 (20.4) 0 0 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 264 (10.9) 6482 (9) 1.9 6.5 Chest pain 420 (17.4) 9195 (12.8) 4.6 13 Heart failure* 248 (10.3) 3355 (4.7) 5.6 21.5 Heart failure hospitalization 12 (0.5) 238 (0.3) 0.2 2.6 Conduction disorders 67 (2.8) 1242 (1.7) 1.1 7.1 Chronic obstructive pulmonary disease 355 (14.7) 7958 (11.1) 3.7 10.9 (COPD) Depression* 224 (9.3) 5826 (8.1) 1.2 4.2 Diabetes* 439 (18.2) 11938 (16.6) 1.6 4.2 Drug-induced osteoporosis* 47 (1.9) 1425 (2) 0 -0.2 Falls* 91 (3.8) 1336 (1.9) 1.9 11.6

184

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 History of falls 80 (3.3) 1170 (1.6) 1.7 10.9 HIV infection 5 (0.2) 106 (0.1) 0.1 1.4 Hyperlipidemia 1145 (47.5) 43163 (60) –12.5 –25.3 Hyperparathyroidism 46 (1.9) 1016 (1.4) 0.5 3.9 Hypertension* 1376 (57.1) 39550 (55) 2.1 4.2 Hyperthyroidism 47 (1.9) 1354 (1.9) 0.1 0.5 Inflammatory bowel disease (IBD)* 31 (1.3) 586 (0.8) 0.5 4.6 Ischemic heart disease* 318 (13.2) 6661 (9.3) 3.9 12.5 Liver disease* 121 (5) 2265 (3.1) 1.9 9.5 Metastatic cancer* 65 (2.7) 1140 (1.6) 1.1 7.7 Osteoporotic fracture - non-vertebral* — — — — Osteoporotic fracture - vertebral* — — — — Osteoporosis* 944 (39.2) 32294 (44.9) –5.8 –11.7 Other fractures* — — — — Palpitations 104 (4.3) 2832 (3.9) 0.4 1.9 Parkinson’s disease 50 (2.1) 575 (0.8) 1.3 10.7 Post-MI/ACS 49 (2) 1066 (1.5) 0.6 4.2 Preventive care 1286 (53.3) 48949 (68.1) –14.7 –30.5

185

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Prior MI 12 (0.5) 203 (0.3) 0.2 3.5 Prior stroke 18 (0.7) 174 (0.2) 0.5 7.2 Peripheral vascular disease (PVD) or PVD 232 (9.6) 3627 (5) 4.6 17.6 surgery Rheumatoid arthritis* 88 (3.6) 2706 (3.8) –0.1 –0.6 Recent MI 1 (0) 13 (0) 0 1.4 Recent stroke 3 (0.1) 19 (0) 0.1 3.6 Renal dysfunction 331 (13.7) 5832 (8.1) 5.6 18.1 Schizophrenia 12 (0.5) 200 (0.3) 0.2 3.5 Transient ischemic attack* 52 (2.2) 1153 (1.6) 0.6 4.1 Upper GI diseases* 597 (24.8) 10670 (14.8) 9.9 25.1 Urinary tract infection 414 (17.2) 8553 (11.9) 5.3 15 Baseline medication use, n (%) Systemic steroid* 246 (10.2) 5993 (8.3) 1.9 6.5 Bile acid sequestrants 43 (1.8) 825 (1.1) 0.6 5.3 IV osteoclast inhibitors* 28 (1.2) 672 (0.9) 0.2 2.2 Inhaled glucocorticoid* 197 (8.2) 4500 (6.3) 1.9 7.4 Lipid-lowering agents 1233 (51.1) 44543 (61.9) –10.8 –21.9

186

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Niacin and fibrates 137 (5.7) 4696 (6.5) –0.8 –3.5 Angiotensin-converting enzyme inhibitors 818 (33.9) 23257 (32.3) 1.6 3.4 (ACEIs)* Agents for dementia 125 (5.2) 2074 (2.9) 2.3 11.7 Antidiabetics* 360 (14.9) 9828 (13.7) 1.3 3.6 Antiparkinson agents* 99 (4.1) 1553 (2.2) 1.9 11.2 Antiplatelets* 1 (0) 58 (0.1) 0 –1.6 Angiotensin II receptor blockers (ARBs)* 431 (17.9) 11938 (16.6) 1.3 3.4 Aromatase inhibitors 52 (2.2) 2200 (3.1) –0.9 –5.7 Atypical antipsychotics 100 (4.1) 1501 (2.1) 2.1 11.9 Beta-blockers* 950 (39.4) 22012 (30.6) 8.8 18.5 Biologic disease-modifying antirheumatic 4 (0.2) 279 (0.4) –0.2 –4.2 drugs (DMARDs)* Calcium channel blockers* 738 (30.6) 18253 (25.4) 5.2 11.7 Digoxin* 134 (5.6) 2046 (2.8) 2.7 13.6 Heparin and low-molecular-weight 47 (1.9) 696 (1) 1 8.2 heparin* Lithium 6 (0.2) 174 (0.2) 0 0.1 Loop diuretics* 505 (20.9) 7169 (10) 11 30.7

187

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Monoamine oxidase inhibitors (MOAs) 1 (0) 3 (0) 0 2.5 Nitrates* 176 (7.3) 2543 (3.5) 3.8 16.7 Nonbiologic DMARDs 148 (6.1) 4158 (5.8) 0.4 1.5

Nonselective nonsteroidal anti- 412 (17.1) 12462 (17.3) –0.2 –0.6 inflammatory drugs (NSAIDs)* Opioids* 1049 (43.5) 20150 (28) 15.5 32.7 Oral anticoagulants* 317 (13.1) 5112 (7.1) 6 20.1 Other anticoagulants* 3 (0.1) 46 (0.1) 0.1 2 Other antihypertensives 202 (8.4) 3836 (5.3) 3 12.1 Other newer and atypical 211 (8.8) 4234 (5.9) 2.9 11 antidepressants* Potassium-sparing diuretics 321 (13.3) 7275 (10.1) 3.2 10 Renin inhibitor 5 (0.2) 109 (0.2) 0.1 1.3 Selective COX-2 inhibitors* 101 (4.2) 3106 (4.3) –0.1 –0.6 Selective estrogen receptor blockers 18 (0.7) 738 (1) -0.3 -3 Selective serotonin-norepinephrine 71 (2.9) 1886 (2.6) 0.3 2 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 443 (18.4) 11225 (15.6) 2.8 7.4 (SSRIs)*

188

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Tricyclic antidepressants (TCAs)* 142 (5.9) 3246 (4.5) 1.4 6.2 Thiazide diuretics* 790 (32.8) 23596 (32.8) 0 -0.1 Typical antipsychotics* 11 (0.5) 305 (0.4) 0 0.5 No lipid-lowering drug despite diagnosis 294 (12.2) 7576 (10.5) 1.7 5.2 of hyperlipidemia* Long-term use of opioids* 510 (21.2) 14737 (20.5) 0.7 1.6 Long-term current use of steroids* 16 (0.7) 1119 (1.6) –0.9 –8.5 Lack of 2nd RX among all the interested 9 (0.4) 279 (0.4) 0 -0.2 drugs groups* Opioids baseline days’ supply >= 90* 278 (11.5) 3898 (5.4) 6.1 22.1 Index days’ supply* 46 (26.4) 57 (28.9) –11.6 –41.9 Use of preventive services Colonoscopy*, n (%) 239 (9.9) 8319 (11.6) -1.7 -5.3 Fecal occult blood test*, n (%) 301 (12.5) 12116 (16.8) –4.4 –12.4 Mammography*, n (% of females) 782 (37.6) 32345 (50.8) –13.2 –26 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 -8.2 Bone mineral density test*, n (%) 544 (22.6) 20704 (28.8) –6.2 –14.3 Electrocardiography, n (%) 833 (34.5) 23788 (33.1) 1.5 3.1 INR test, n (%) 1 (5.3) 1 (3.1) 0.6 14.8 189

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Use of wheelchair, walker, crutches, or 64 (2.7) 1128 (1.6) 1.1 7.6 cane, n (%) Use of oxygen, n (%) 85 (3.5) 1411 (2) 1.6 9.6 Lipid test, n (%) 1021 (42.3) 41240 (57.3) –15 –30.3 Health care utilization, mean (SD) Number of distinct cardiovascular 3 (3.2) 2 (2.5) 0.6 21.7 diagnoses Total days in hospital in prior year* 2 (12.3) 1 (5.1) 1.3 13.7 Medication synchronization metrics 1 (0.2) 1 (0.3) 0 9.6 Copayment 938 (985.8) 738 (806.9) 199.9 22.2 Number of office visits 18 (24.4) 14 (16.2) 4.6 22 Number of unique drugs of interest (by 6 (3.4) 4 (2.9) 1.7 53.7 NDC) in prior year or on index date Number of office visits with a 6 (12.7) 4 (7.7) 2 19.1 cardiovascular diagnosis Number of drugs of interest dispensations 25 (19.7) 19 (16.6) 5.2 28.4 in 365 days in prior year or on index date* Number of hospitalizations in the prior 0 (0.7) 0 (0.5) 0.1 17.7 year

190

Appendix 27. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 21.7 mail)* Number of cardiovascular hospitalizations 0 (0.5) 0 (0.4) 0.1 13.5 Index copayment 42 (41) 35 (38.7) 6.8 17.1 Vaccinations* 0 (0.7) 0 (0.8) –0.1 –11.6 Physician visits* 8 (7.3) 7 (6) 0.6 8.9 Number of drugs of interest dispensations 4 (2.2) 3 (2.1) 1.4 63 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 7 (3.9) 6 (3.2) 1.6 44.4 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 8 (4.4) 6 (3.7) 1.7 42 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 15 (9) 12 (7.1) 3.3 41.1 days before or on index date* Number of any drug dispensations in 365 53 (42.8) 38 (31) 14.5 38.9 days before or on index date*

191

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 68 (12.2) 65 (11.4) 3 25.5 Sex: female* 14746 (89) 396290 (91.6) –2.6 –8.7 Combined comorbidity score* 1 (2.2) 1 (1.6) 0.5 26.6 Commercial plan type, n (%) 8948 (54) 250028 (57.8) –3.8 –7.6 Benefit plan types, n (%) EPO* 778 (4.7) 28577 (6.6) –1.9 –8.3 HMO* 6075 (36.7) 140558 (32.5) 4.2 8.8 IND* 1625 (9.8) 23354 (5.4) 4.4 16.7 OTH* 2285 (13.8) 63927 (14.8) –1 –2.8 POS* 4137 (25) 137715 (31.8) –6.9 –15.2 Comorbidities, n (%) Acute coronary syndrome, with or 634 (3.8) 11331 (2.6) 1.2 6.8 without revascularization Acute coronary syndrome, with 175 (1.1) 2952 (0.7) 0.4 4 revascularization Atrial fibrillation 222 (1.3) 2916 (0.7) 0.7 6.7 Alzheimer’s disease* 1006 (6.1) 11814 (2.7) 3.3 16.4

192

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Angina 523 (3.2) 9756 (2.3) 0.9 5.6 Coronary atherosclerosis* 1765 (10.7) 30722 (7.1) 3.6 12.5 Coronary artery bypass graft (CABG), 33 (0.2) 439 (0.1) 0.1 2.5 new CABG, old 207 (1.3) 3377 (0.8) 0.5 4.7 Any cancer 2741 (16.6) 70745 (16.4) 0.2 0.5 Any malignancy, including lymphoma 2517 (15.2) 65555 (15.2) 0 0.1 and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 1590 (9.6) 36729 (8.5) 1.1 3.9 Chest pain 2824 (17.1) 55997 (12.9) 4.1 11.5 Heart failure* 1268 (7.7) 15892 (3.7) 4 17.3 Heart failure hospitalization 100 (0.6) 1035 (0.2) 0.4 5.6 Conduction disorders 347 (2.1) 5904 (1.4) 0.7 5.6 Chronic obstructive pulmonary disease 2332 (14.1) 45198 (10.4) 3.6 11.1 (COPD) Depression* 1883 (11.4) 44868 (10.4) 1 3.2 Diabetes* 2171 (13.1) 54699 (12.6) 0.5 1.4 Drug-induced osteoporosis* 249 (1.5) 7349 (1.7) –0.2 –1.6

193

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Falls* 512 (3.1) 7012 (1.6) 1.5 9.7 History of falls 449 (2.7) 6141 (1.4) 1.3 9.1 HIV infection 14 (0.1) 431 (0.1) 0 -0.5 Hyperlipidemia 6680 (40.3) 192966 (44.6) –4.3 –8.7 Hyperparathyroidism 205 (1.2) 4611 (1.1) 0.2 1.6 Hypertension* 6634 (40.1) 166342 (38.5) 1.6 3.3 Hyperthyroidism 252 (1.5) 7027 (1.6) –0.1 –0.8 Inflammatory bowel disease (IBD)* 235 (1.4) 4573 (1.1) 0.4 3.3 Ischemic heart disease* 1869 (11.3) 32935 (7.6) 3.7 12.6 Liver disease* 749 (4.5) 15192 (3.5) 1 5.1 Metastatic cancer* 271 (1.6) 4026 (0.9) 0.7 6.3 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 5356 (32.3) 168354 (38.9) –6.6 –13.8 Other fractures* — — — — Palpitations 697 (4.2) 17320 (4) 0.2 1 Parkinson’s disease 197 (1.2) 2171 (0.5) 0.7 7.5

194

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 276 (1.7) 5081 (1.2) 0.5 4.2 coronary syndromes (MI/ACS)* Preventive care 7409 (44.7) 251630 (58.2) –13.4 –27.1 Prior MI 105 (0.6) 1226 (0.3) 0.4 5.2 Prior stroke 79 (0.5) 929 (0.2) 0.3 4.5 Peripheral vascular disease (PVD) or PVD 1092 (6.6) 17108 (4) 2.6 11.8 surgery Rheumatoid arthritis* 572 (3.5) 16684 (3.9) –0.4 –2.1 Recent MI 8 (0) 84 (0) 0 1.6 Recent stroke 10 (0.1) 102 (0) 0 1.8 Renal dysfunction 1713 (10.3) 27037 (6.3) 4.1 14.9 Schizophrenia 45 (0.3) 763 (0.2) 0.1 2 Transient ischemic attack* 358 (2.2) 6033 (1.4) 0.8 5.8 Upper GI diseases* 3822 (23.1) 64280 (14.9) 8.2 21.1 Urinary tract infection 2440 (14.7) 49957 (11.5) 3.2 9.4 Baseline medication use, n (%) Systemic steroid* 1314 (7.9) 32017 (7.4) 0.5 2 Bile acid sequestrants 205 (1.2) 3470 (0.8) 0.4 4.3 IV osteoclast inhibitors* 128 (0.8) 2269 (0.5) 0.2 3.1 195

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 1250 (7.5) 26073 (6) 1.5 6 Lipid-lowering agents 5367 (32.4) 150603 (34.8) –2.4 –5.1 Niacin and fibrates 521 (3.1) 13925 (3.2) -0.1 -0.4 Angiotensin-converting enzyme 3449 (20.8) 86061 (19.9) 0.9 2.3 inhibitors (ACEIs)* Agents for dementia 610 (3.7) 8331 (1.9) 1.8 10.7 Antidiabetics* 1610 (9.7) 40513 (9.4) 0.4 1.2 Antiparkinson agents* 437 (2.6) 7287 (1.7) 1 6.6 Antiplatelets* 9 (0.1) 347 (0.1) 0 -1 Angiotensin II receptor blockers (ARBs)* 1940 (11.7) 46798 (10.8) 0.9 2.8 Aromatase inhibitors 191 (1.2) 8948 (2.1) –0.9 –7.3 Atypical antipsychotics 498 (3) 6527 (1.5) 1.5 10.1 Beta-blockers* 4148 (25) 82942 (19.2) 5.9 14.2 Biologic disease-modifying antirheumatic 77 (0.5) 2677 (0.6) –0.2 –2.1 drugs (DMARDs)* Calcium channel blockers* 3116 (18.8) 68321 (15.8) 3 8 Digoxin* 534 (3.2) 6495 (1.5) 1.7 11.4 Heparin and low-molecular-weight 182 (1.1) 2608 (0.6) 0.5 5.4 heparin*

196

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Lithium 30 (0.2) 756 (0.2) 0 0.2 Loop diuretics* 2077 (12.5) 28351 (6.6) 6 20.5 Monoamine oxidase inhibitors (MOAs) 1 (0) 26 (0) 0 0 Nitrates* 911 (5.5) 11813 (2.7) 2.8 14 Nonbiologic DMARDs 643 (3.9) 18749 (4.3) –0.5 –2.3

Nonselective nonsteroidal anti- 2754 (16.6) 74266 (17.2) –0.5 –1.4 inflammatory drugs (NSAIDs)* Opioids* 6420 (38.8) 118648 (27.4) 11.3 24.3 Oral anticoagulants* 1074 (6.5) 15945 (3.7) 2.8 12.8 Other anticoagulants* 12 (0.1) 155 (0) 0 1.6 Other antihypertensives 764 (4.6) 12407 (2.9) 1.7 9.2 Other newer and atypical 1459 (8.8) 27109 (6.3) 2.5 9.6 antidepressants* Potassium-sparing diuretics 1088 (6.6) 23536 (5.4) 1.1 4.8 Renin inhibitor 26 (0.2) 478 (0.1) 0 1.3 Selective COX-2 inhibitors* 656 (4) 14909 (3.4) 0.5 2.7 Selective estrogen receptor blockers 97 (0.6) 3487 (0.8) –0.2 –2.7 Selective serotonin-norepinephrine 704 (4.3) 17345 (4) 0.2 1.2 reuptake inhibitors (SNRIs)* 197

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 2766 (16.7) 63022 (14.6) 2.1 5.9 (SSRIs)* Tricyclic antidepressants (TCAs)* 866 (5.2) 15914 (3.7) 1.6 7.5 Thiazide diuretics* 3070 (18.5) 84703 (19.6) –1 –2.7 Typical antipsychotics* 63 (0.4) 1004 (0.2) 0.1 2.7 No lipid-lowering drug despite diagnosis 2876 (17.4) 75333 (17.4) 0 -0.1 of hyperlipidemia* Long-term use of opioids* 2749 (16.6) 66319 (15.3) 1.3 3.5 Long-term current use of steroids* 123 (0.7) 5393 (1.2) –0.5 –5.1 Lack of 2nd RX among all the interested 170 (1) 6523 (1.5) –0.5 –4.3 drugs groups* Opioids baseline days’ supply ≥ 90* 1901 (11.5) 28259 (6.5) 4.9 17.3 Index days’ supply* 38 (20.9) 45 (25.8) –6.3 –26.9 Use of preventive services Colonoscopy*, n (%) 1388 (8.4) 39674 (9.2) –0.8 –2.8 Fecal occult blood test*, n (%) 1646 (9.9) 57850 (13.4) –3.4 –10.7 Mammography*, n (% of females) 3936 (26.7) 146873 (37.1) –13.4 –22.6 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 –11.7 Bone mineral density test*, n (%) 1755 (10.6) 65556 (15.2) –4.6 –13.6

198

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 5126 (31) 127535 (29.5) 1.5 3.2 INR test, n (%) 1 (3.5) 0 (2.2) 0.3 8.8 Use of wheelchair, walker, crutches, or 437 (2.6) 6250 (1.4) 1.2 8.4 cane, n (%) Use of oxygen, n (%) 497 (3) 7321 (1.7) 1.3 8.7 Lipid test, n (%) 6427 (38.8) 213768 (49.4) –10.6 –21.5 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (3.2) 1 (2.4) 0.5 17.8 diagnoses Total days in hospital in prior year* 1 (7.3) 1 (3.8) 0.8 13.3 Medication synchronization metrics 0 (0.3) 0 (0.3) 0 11.5 Copayment 535 (752.1) 428 (630.2) 107.4 15.5 Number of office visits 15 (21.9) 12 (15.5) 3 15.9 Number of unique drugs of interest (by 4 (3.4) 2 (2.9) 1.4 45.1 NDC) in prior year or on index date Number of office visits with a 4 (9.3) 3 (6.6) 1.2 14.4 cardiovascular diagnosis Number of drugs of interest 12 (15.5) 9 (13.6) 2.4 16.5 dispensations in 365 days in prior year or on index date*

199

Appendix 28. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (0.7) 0 (0.5) 0.1 16 year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 15.8 (non-mail)* Number of cardiovascular 0 (0.5) 0 (0.3) 0.1 13.7 hospitalizations Index copayment 35 (30.4) 29 (31.4) 6 19.5 Vaccinations* 0 (0.7) 0 (0.7) 0 –4.6 Physician visits* 7 (7.5) 6 (6.2) 0.5 7.3 Number of drugs of interest 3 (1.9) 2 (1.6) 0.8 46.9 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 5 (3.4) 4 (2.9) 1 30.6 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 5 (3.7) 4 (3.1) 1 29.6 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 11 (9.4) 9 (7.7) 2.2 25.2 days before or on index date* Number of any drug dispensations in 365 30 (33.4) 23 (26.1) 6.4 21.2 days before or on index date*

200

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 69 (12.9) 69 (11.4) 0 0 Sex: female* 2394 (87.1) 61919 (87.3) –0.2 –0.6 Combined comorbidity score* 2 (2.7) 2 (2.7) 0 -0.9 Commercial plan type, n (%) 933 (33.9) 23589 (33.3) 0.7 1.4 Benefit plan types, n (%) EPO* 111 (4) 2670 (3.8) 0.3 1.4 HMO* 1199 (43.6) 31139 (43.9) –0.3 –0.6 IND* 73 (2.7) 1933 (2.7) -0.1 -0.4 OTH* 570 (20.7) 14806 (20.9) –0.1 –0.3 POS* 582 (21.2) 14604 (20.6) 0.6 1.4 Comorbidities, n (%) Acute coronary syndrome, with or without 232 (8.4) 6136 (8.7) –0.2 –0.8 revascularization Acute coronary syndrome, with 69 (2.5) 1759 (2.5) 0 0.2 revascularization Atrial fibrillation 67 (2.4) 1802 (2.5) –0.1 –0.7

201

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Alzheimer’s disease* 205 (7.5) 5160 (7.3) 0.2 0.7 Angina 197 (7.2) 5215 (7.4) –0.2 –0.7 Coronary atherosclerosis* 583 (21.2) 15506 (21.9) –0.7 –1.6 Coronary artery bypass graft (CABG), new 16 (0.6) 427 (0.6) 0 -0.3 CABG, old 75 (2.7) 2027 (2.9) –0.1 –0.8 Any cancer 576 (21) 15095 (21.3) -0.3 -0.8 Any malignancy, including lymphoma and 527 (19.2) 13883 (19.6) –0.4 –1 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 465 (16.9) 12329 (17.4) –0.5 –1.2 Chest pain 849 (30.9) 22400 (31.6) –0.7 –1.5 Heart failure* 414 (15.1) 10700 (15.1) 0 –0.1 Heart failure hospitalization 40 (1.5) 1111 (1.6) –0.1 –0.9 Conduction disorders 104 (3.8) 2901 (4.1) –0.3 –1.6 Chronic obstructive pulmonary disease 699 (25.4) 18372 (25.9) –0.5 –1.1 (COPD) Depression* 690 (25.1) 18056 (25.5) –0.4 –0.8 Diabetes* 693 (25.2) 18549 (26.1) –0.9 –2.2 Drug-induced osteoporosis* 60 (2.2) 1574 (2.2) 0 –0.2

202

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Falls* 172 (6.3) 4325 (6.1) 0.2 0.7 History of falls 152 (5.5) 3773 (5.3) 0.2 0.9 HIV infection 1 (0) 26 (0) 0 0 Hyperlipidemia 1604 (58.3) 41974 (59.2) –0.8 –1.7 Hyperparathyroidism 62 (2.3) 1669 (2.4) –0.1 –0.6 Hypertension* 1894 (68.9) 49685 (70) –1.1 –2.5 Hyperthyroidism 54 (2) 1401 (2) 0 –0.1 Inflammatory bowel disease (IBD)* 64 (2.3) 1603 (2.3) 0.1 0.5 Ischemic heart disease* 608 (22.1) 16175 (22.8) –0.7 –1.6 Liver disease* 222 (8.1) 5953 (8.4) –0.3 –1.2 Metastatic cancer* 47 (1.7) 1275 (1.8) –0.1 –0.7 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1010 (36.7) 26271 (37) –0.3 –0.6 Other fractures* — — — — Palpitations 221 (8) 5669 (8) 0 0.2 Parkinson’s disease 49 (1.8) 1206 (1.7) 0.1 0.6

203

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute coronary 98 (3.6) 2716 (3.8) –0.3 –1.4 syndromes (MI/ACS)* Preventive care 1546 (56.2) 39783 (56.1) 0.2 0.3 Prior MI 34 (1.2) 885 (1.2) 0 -0.1 Prior stroke 14 (0.5) 325 (0.5) 0.1 0.7 Peripheral vascular disease (PVD) or PVD 305 (11.1) 7904 (11.1) 0 -0.2 surgery Rheumatoid arthritis* 181 (6.6) 4902 (6.9) –0.3 –1.3 Recent MI 2 (0.1) 59 (0.1) 0 -0.4 Recent stroke 1 (0) 29 (0) 0 -0.2 Renal dysfunction 577 (21) 14975 (21.1) –0.1 –0.3 Schizophrenia 12 (0.4) 335 (0.5) 0 -0.5 Transient ischemic attack* 136 (4.9) 3500 (4.9) 0 0.1 Upper GI diseases* 1003 (36.5) 26528 (37.4) –0.9 –1.9 Urinary tract infection 660 (24) 17103 (24.1) –0.1 –0.2 Baseline medication use, n (%) Systemic steroid* 411 (15) 10974 (15.5) –0.5 –1.4 Bile acid sequestrants 59 (2.1) 1635 (2.3) –0.2 –1.1 IV osteoclast inhibitors* 29 (1.1) 795 (1.1) –0.1 –0.6 204

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 383 (13.9) 10167 (14.3) –0.4 –1.2 Lipid-lowering agents 1154 (42) 30645 (43.2) –1.2 –2.5 Niacin and fibrates 107 (3.9) 3031 (4.3) –0.4 –1.9 Angiotensin-converting enzyme inhibitors 904 (32.9) 24226 (34.2) –1.3 –2.7 (ACEIs)* Agents for dementia 127 (4.6) 3287 (4.6) 0 –0.1 Antidiabetics* 503 (18.3) 13740 (19.4) –1.1 –2.7 Antiparkinson agents* 141 (5.1) 3761 (5.3) –0.2 –0.8 Antiplatelets* 4 (0.1) 119 (0.2) 0 –0.6 Angiotensin II receptor blockers (ARBs)* 553 (20.1) 14700 (20.7) –0.6 –1.5 Aromatase inhibitors 28 (1) 717 (1) 0 0.1 Atypical antipsychotics 131 (4.8) 3509 (4.9) –0.2 –0.8 Beta-blockers* 1137 (41.4) 30197 (42.6) –1.2 –2.5 Biologic disease-modifying antirheumatic 32 (1.2) 822 (1.2) 0 0.1 drugs (DMARDs)* Calcium channel blockers* 831 (30.2) 22198 (31.3) –1.1 –2.3 Digoxin* 119 (4.3) 3215 (4.5) -0.2 -1 Heparin and low-molecular-weight 52 (1.9) 1314 (1.9) 0 0.3 heparin*

205

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Lithium 10 (0.4) 244 (0.3) 0 0.3 Loop diuretics* 611 (22.2) 16261 (22.9) –0.7 –1.7 Monoamine oxidase inhibitors (MOAs) Nitrates* 313 (11.4) 8163 (11.5) –0.1 –0.4 Nonbiologic DMARDs 171 (6.2) 4630 (6.5) –0.3 –1.3 Nonselective nonsteroidal anti- 749 (27.2) 19735 (27.8) –0.6 –1.3 inflammatory drugs (NSAIDs)* Opioids* 1784 (64.9) 46892 (66.1) –1.2 –2.5 Oral anticoagulants* 252 (9.2) 6729 (9.5) –0.3 –1.1 Other anticoagulants* 6 (0.2) 110 (0.2) 0.1 1.4 Other antihypertensives 219 (8) 5670 (8) 0 –0.1 Other newer and atypical antidepressants* 416 (15.1) 10930 (15.4) –0.3 –0.8 Potassium-sparing diuretics 277 (10.1) 7496 (10.6) –0.5 –1.6 Renin inhibitor 11 (0.4) 338 (0.5) –0.1 –1.2 Selective COX-2 inhibitors* 139 (5.1) 3709 (5.2) –0.2 –0.8 Selective estrogen receptor blockers 8 (0.3) 188 (0.3) 0 0.5 Selective serotonin-norepinephrine 258 (9.4) 6796 (9.6) –0.2 –0.7 reuptake inhibitors (SNRIs)*

206

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 960 (34.9) 25350 (35.7) –0.8 –1.7 (SSRIs)* Tricyclic antidepressants (TCAs)* 232 (8.4) 6382 (9) –0.6 –2 Thiazide diuretics* 754 (27.4) 20374 (28.7) –1.3 –2.9 Typical antipsychotics* 11 (0.4) 288 (0.4) 0 –0.1 No lipid-lowering drug despite diagnosis of 648 (23.6) 16536 (23.3) 0.3 0.6 hyperlipidemia* Long-term use of opioids* 723 (26.3) 19213 (27.1) –0.8 –1.8 Long-term current use of steroids* 44 (1.6) 1169 (1.6) 0 -0.4 Lack of 2nd RX among all the interested 144 (5.2) 3399 (4.8) 0.4 2 drugs groups* Opioids baseline days’ supply ≥ 90* 757 (27.5) 20395 (28.8) –1.2 –2.7 Index days’ supply* 32 (10.3) 32 (13.1) 0 -0.3 Use of preventive services Colonoscopy*, n (%) 297 (10.8) 7775 (11) –0.2 –0.5 Fecal occult blood test*, n (%) 274 (10) 7056 (9.9) 0 0.1 Mammography*, n (% of females) 638 (26.6) 16717 (27.9) –0.4 –0.8 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 0.4 Bone mineral density test*, n (%) 91 (3.3) 2290 (3.2) 0.1 0.5

207

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 1350 (49.1) 35453 (50) –0.9 –1.7 INR test, n (%) 1 (4.1) 1 (4.1) 0 -0.9 Use of wheelchair, walker, crutches, or 147 (5.3) 3943 (5.6) –0.2 –0.9 cane, n (%) Use of oxygen, n (%) 148 (5.4) 3892 (5.5) –0.1 –0.5 Lipid test, n (%) 1508 (54.9) 39289 (55.4) –0.5 –1.1 Health care utilization, mean (SD) Number of distinct cardiovascular 4 (4.1) 4 (4.2) –0.1 –1.8 diagnoses Total days in hospital in prior year* 2 (6.1) 2 (6.4) –0.1 –1 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 -4.8 Copayment 803 (879.3) 816 (1280) –13 –1.2 Number of office visits 24 (26.3) 25 (28.8) –0.6 –2.3 Number of unique drugs of interest (by 6 (3.6) 6 (4.9) –0.2 –4.9 NDC) in prior year or on index date Number of office visits with a 7 (11.7) 7 (12.9) –0.2 –1.7 cardiovascular diagnosis Number of drugs of interest dispensations 18 (14.4) 18 (15.8) –0.6 –3.7 in 365 days in prior year or on index date*

208

Appendix 29. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (0.9) 0 (1) 0 –1 year Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 –4.9 mail)* Number of cardiovascular hospitalizations 0 (0.7) 0 (0.7) 0 –0.9 Index copayment 31 (25.2) 32 (57.8) –0.1 –0.3 Vaccinations* 0 (0.8) 0 (0.8) 0 –0.1 Physician visits* 11 (9) 12 (9.5) –0.3 –3.3 Number of drugs of interest dispensations 3 (1.7) 3 (2.3) –0.1 –4.4 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 days 5 (3) 6 (3.2) –0.2 –5 before or on index date with days’ supply that overlap index date Number of dispensations in baseline 6 (3.1) 6 (3.3) –0.2 –5.3 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 19 (10.7) 19 (11.4) –0.5 –4.5 days before or on index date* Number of any drug dispensations in 365 47 (32.9) 48 (35.6) –1.1 –3.3 days before or on index date*

209

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 71 (11.1) 71 (10.5) 0 –0.3 Sex: female* 2246 (87.5) 62263 (87.2) 0.3 1 Combined comorbidity score* 1 (2.5) 1 (2.5) 0 –1 Commercial plan type, n (%) 861 (33.5) 24003 (33.6) –0.1 –0.1 Benefit plan types, n (%) EPO* 78 (3) 2128 (3) 0.1 0.3 HMO* 1258 (49) 34910 (48.9) 0.1 0.3 IND* 119 (4.6) 3466 (4.9) –0.2 –1 OTH* 467 (18.2) 12971 (18.2) 0 0.1 POS* 472 (18.4) 13087 (18.3) 0.1 0.2 Comorbidities, n (%) Acute coronary syndrome, with or 138 (5.4) 4137 (5.8) –0.4 –1.8 without revascularization Acute coronary syndrome, with 34 (1.3) 1146 (1.6) –0.3 –2.3 revascularization Atrial fibrillation 55 (2.1) 1642 (2.3) –0.2 –1.1 Alzheimer’s disease* 243 (9.5) 6263 (8.8) 0.7 2.4

210

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Angina 112 (4.4) 3298 (4.6) –0.3 –1.2 Coronary atherosclerosis* 431 (16.8) 12581 (17.6) –0.8 –2.2 Coronary artery bypass graft (CABG), new 4 (0.2) 139 (0.2) 0 -0.9 CABG, old 51 (2) 1590 (2.2) –0.2 –1.7 Any cancer 511 (19.9) 14689 (20.6) –0.7 –1.6 Any malignancy, including lymphoma and 473 (18.4) 13610 (19.1) –0.6 –1.6 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 310 (12.1) 8923 (12.5) –0.4 –1.3 Chest pain 593 (23.1) 16870 (23.6) –0.5 –1.2 Heart failure* 303 (11.8) 8539 (12) –0.2 –0.5 Heart failure hospitalization 19 (0.7) 529 (0.7) 0 0 Conduction disorders 82 (3.2) 2411 (3.4) –0.2 –1 Chronic obstructive pulmonary disease 490 (19.1) 13924 (19.5) –0.4 –1 (COPD) Depression* 379 (14.8) 10647 (14.9) –0.1 –0.4 Diabetes* 511 (19.9) 15226 (21.3) –1.4 –3.5 Drug-induced osteoporosis* 50 (1.9) 1464 (2) –0.1 –0.7 Falls* 115 (4.5) 3326 (4.7) –0.2 –0.8

211

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 History of falls 94 (3.7) 2703 (3.8) –0.1 –0.6 HIV infection 1 (0) 26 (0) 0 0.1 Hyperlipidemia 1418 (55.2) 39865 (55.8) –0.6 –1.2 Hyperparathyroidism 33 (1.3) 938 (1.3) 0 -0.3 Hypertension* 1717 (66.9) 48621 (68.1) –1.2 –2.5 Hyperthyroidism 51 (2) 1439 (2) 0 –0.2 Inflammatory bowel disease (IBD)* 35 (1.4) 1012 (1.4) –0.1 –0.5 Ischemic heart disease* 459 (17.9) 13350 (18.7) –0.8 –2.1 Liver disease* 150 (5.8) 4373 (6.1) –0.3 –1.2 Metastatic cancer* 50 (1.9) 1511 (2.1) –0.2 –1.2 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1044 (40.7) 29418 (41.2) –0.5 –1.1 Other fractures* — — — — Palpitations 126 (4.9) 3487 (4.9) 0 0.1 Parkinson’s disease 57 (2.2) 1435 (2) 0.2 1.5 Post-myocardial infarction/acute coronary 70 (2.7) 2017 (2.8) –0.1 –0.6 syndromes (MI/ACS)*

212

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Preventive care 1484 (57.8) 41246 (57.7) 0.1 0.1 Prior MI 27 (1.1) 760 (1.1) 0 –0.1 Prior stroke 16 (0.6) 469 (0.7) 0 –0.4 Peripheral vascular disease (PVD) or PVD 256 (10) 7147 (10) 0 –-0.1 surgery Rheumatoid arthritis* 109 (4.2) 3227 (4.5) –0.3 –1.3 Recent MI 2 (0.1) 38 (0.1) 0 1 Recent stroke 2 (0.1) 48 (0.1) 0 0.4 Renal dysfunction 448 (17.5) 12945 (18.1) –0.7 –1.8 Schizophrenia 15 (0.6) 416 (0.6) 0 0 Transient ischemic attack* 80 (3.1) 2202 (3.1) 0 0.2 Upper GI diseases* 719 (28) 20272 (28.4) –0.4 –0.8 Urinary tract infection 518 (20.2) 14653 (20.5) –0.3 –0.8 Baseline medication use, n (%) Systemic steroid* 288 (11.2) 8312 (11.6) –0.4 –1.3 Bile acid sequestrants 40 (1.6) 1202 (1.7) –0.1 –1 IV osteoclast inhibitors* 31 (1.2) 1021 (1.4) –0.2 –1.9 Inhaled glucocorticoid* 223 (8.7) 6446 (9) –0.3 –1.2

213

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Lipid-lowering agents 1214 (47.3) 34600 (48.4) –1.2 –2.3 Niacin and fibrates 112 (4.4) 3204 (4.5) –0.1 –0.6 Angiotensin-converting enzyme inhibitors 898 (35) 25850 (36.2) –1.2 –2.5 (ACEIs)* Agents for dementia 131 (5.1) 3336 (4.7) 0.4 2 Antidiabetics* 395 (15.4) 12074 (16.9) –1.5 –4.1 Antiparkinson agents* 103 (4) 2893 (4.1) 0 -0.2 Antiplatelets* Angiotensin II receptor blockers (ARBs)* 507 (19.8) 15014 (21) –1.3 –3.2 Aromatase inhibitors 32 (1.2) 856 (1.2) 0 0.4 Atypical antipsychotics 112 (4.4) 3088 (4.3) 0 0.2 Beta-blockers* 1104 (43) 31583 (44.2) –1.2 –2.4 Biologic disease-modifying antirheumatic 14 (0.5) 424 (0.6) 0 -0.6 drugs (DMARDs)* Calcium channel blockers* 810 (31.6) 23123 (32.4) –0.8 –1.8 Digoxin* 135 (5.3) 3700 (5.2) 0.1 0.4 Heparin and low-molecular-weight 46 (1.8) 1441 (2) –0.2 –1.7 heparin* Lithium 4 (0.2) 130 (0.2) 0 -0.6

214

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Loop diuretics* 501 (19.5) 14701 (20.6) –1.1 –2.7 Monoamine oxidase inhibitors (MOAs) Nitrates* 207 (8.1) 5745 (8) 0 0.1 Nonbiologic DMARDs 136 (5.3) 3990 (5.6) –0.3 –1.3

Nonselective nonsteroidal anti- 512 (19.9) 14753 (20.7) –0.7 –1.8 inflammatory drugs (NSAIDs)* Opioids* 1301 (50.7) 36958 (51.7) –1.1 –2.1 Oral anticoagulants* 258 (10.1) 7539 (10.6) –0.5 –1.7 Other anticoagulants* 1 (0) 30 (0) 0 -0.1 Other antihypertensives 176 (6.9) 5172 (7.2) –0.4 –1.5 Other newer and atypical 277 (10.8) 7694 (10.8) 0 0.1 antidepressants* Potassium-sparing diuretics 244 (9.5) 7266 (10.2) –0.7 –2.2 Renin inhibitor 8 (0.3) 231 (0.3) 0 -0.2 Selective COX-2 inhibitors* 103 (4) 2991 (4.2) –0.2 –0.9 Selective estrogen receptor blockers 14 (0.5) 368 (0.5) 0 0.4 Selective serotonin-norepinephrine 113 (4.4) 3242 (4.5) –0.1 –0.7 reuptake inhibitors (SNRIs)*

215

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 617 (24) 17268 (24.2) –0.1 –0.3 (SSRIs)* Tricyclic antidepressants (TCAs)* 168 (6.5) 4908 (6.9) –0.3 –1.3 Thiazide diuretics* 805 (31.4) 23444 (32.8) –1.5 –3.1 Typical antipsychotics* 17 (0.7) 509 (0.7) –0.1 –0.6 No lipid-lowering drug despite diagnosis 466 (18.2) 12853 (18) 0.2 0.4 of hyperlipidemia* Long-term use of opioids* 564 (22) 15787 (22.1) –0.1 –0.3 Long-term current use of steroids* 26 (1) 833 (1.2) –0.2 –1.5 Lack of 2nd RX among all the interested 17 (0.7) 427 (0.6) 0.1 0.8 drugs groups* Opioids baseline days’ supply ≥ 90* 412 (16) 11831 (16.6) –0.5 –1.4 Index days’ supply* 35 (17.1) 35 (18.3) 0 –0.2 Use of preventive services Colonoscopy*, n (%) 261 (10.2) 7402 (10.4) –0.2 –0.6 Fecal occult blood test*, n (%) 290 (11.3) 8007 (11.2) 0.1 0.3 Mammography*, n (% of females) 709 (31.6) 19685 (31.6) 0.0 0.1 Any prior lab test, n (%) 0 (0.3) 0 (0.3) 0 -0.2 Bone mineral density test*, n (%) 215 (8.4) 5987 (8.4) 0 0

216

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 1098 (42.8) 31029 (43.4) –0.7 –1.4 INR test, n (%) 1 (4.5) 1 (4.2) 0 –0.8 Use of wheelchair, walker, crutches, or 92 (3.6) 2786 (3.9) –0.3 –1.7 cane, n (%) Use of oxygen, n (%) 112 (4.4) 3336 (4.7) –0.3 –1.5 Lipid test, n (%) 1292 (50.3) 36075 (50.5) –0.2 –0.4 Health care utilization, mean (SD) Number of distinct cardiovascular 3 (3.5) 3 (3.7) –0.1 –2.7 diagnoses Total days in hospital in prior year* 2 (6.4) 2 (7.4) 0 -0.5 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –4.1 Copayment 793 (846.9) 819 (895.8) –26 –3 Number of office visits 20 (23.8) 20 (24.5) –0.4 –1.5 Number of unique drugs of interest (by 6 (3.6) 6 (4.5) –0.2 –5.5 NDC) in prior year or on index date Number of office visits with a 6 (10.2) 6 (11.6) –0.2 –2.1 cardiovascular diagnosis Number of drugs of interest dispensations 23 (16.7) 23 (17.4) –0.7 –4 in 365 days in prior year or on index date*

217

Appendix 30. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (1.1) 0 (1) 0 0.4 year Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 –4.3 mail)* Number of cardiovascular hospitalizations 0 (0.6) 0 (0.7) 0 –0.9 Index copayment 35 (31.4) 35 (63.3) –0.4 –0.7 Vaccinations* 0 (0.8) 0 (0.8) 0 –1.3 Physician visits* 9 (7.6) 9 (7.6) –0.3 –3.3 Number of drugs of interest dispensations 4 (1.8) 4 (2.5) –0.1 –5.9 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 6 (3.1) 6 (3.4) –0.2 –5.8 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 7 (3.4) 7 (3.6) –0.2 –6.2 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 15 (8.7) 16 (9.4) –0.5 –5.2 days before or on index date* Number of any drug dispensations in 365 47 (33.3) 48 (37.5) –1.1 –3 days before or on index date*

218

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 71 (10.4) 72 (9.9) –0.2 –1.6 Sex: female* 2078 (86.2) 60865 (85.1) 1.1 3.2 Combined comorbidity score* 1 (2.4) 1 (2.4) 0 –2 Commercial plan type, n (%) 1196 (49.6) 35379 (49.4) 0.2 0.3 Benefit plan types, n (%) EPO* 74 (3.1) 2172 (3) 0 0.2 HMO* 1018 (42.2) 29793 (41.6) 0.6 1.2 IND* 356 (14.8) 11113 (15.5) –0.8 –2.1 OTH* 287 (11.9) 8819 (12.3) –0.4 –1.3 POS* 442 (18.3) 12466 (17.4) 0.9 2.4 Comorbidities, n (%) Acute coronary syndrome, with or 90 (3.7) 2879 (4) –0.3 –1.5 without revascularization Acute coronary syndrome, with 23 (1) 775 (1.1) –0.1 –1.3 revascularization Atrial fibrillation 43 (1.8) 1429 (2) –0.2 –1.6 Alzheimer’s disease* 219 (9.1) 6519 (9.1) 0 –0.1

219

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Angina 72 (3) 2241 (3.1) –0.1 –0.8 Coronary atherosclerosis* 298 (12.4) 9283 (13) –0.6 –1.8 Coronary artery bypass graft (CABG), 4 (0.2) 139 (0.2) 0 –0.7 new CABG, old 36 (1.5) 1296 (1.8) –0.3 –2.5 Any cancer 544 (22.6) 16024 (22.4) 0.2 0.4 Any malignancy, including lymphoma 492 (20.4) 14521 (20.3) 0.1 0.3 and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 264 (10.9) 8095 (11.3) –0.4 –1.2 Chest pain 420 (17.4) 12621 (17.6) –0.2 –0.6 Heart failure* 248 (10.3) 7566 (10.6) –0.3 –0.9 Heart failure hospitalization 12 (0.5) 497 (0.7) –0.2 –2.6 Conduction disorders 67 (2.8) 2051 (2.9) –0.1 –0.5 Chronic obstructive pulmonary disease 355 (14.7) 11023 (15.4) –0.7 –1.9 (COPD) Depression* 224 (9.3) 6610 (9.2) 0.1 0.2 Diabetes* 439 (18.2) 14410 (20.1) –1.9 –4.9 Drug-induced osteoporosis* 47 (1.9) 1350 (1.9) 0.1 0.5

220

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Falls* 91 (3.8) 2628 (3.7) 0.1 0.5 History of falls 80 (3.3) 2299 (3.2) 0.1 0.6 HIV infection 5 (0.2) 168 (0.2) 0 –0.6 Hyperlipidemia 1145 (47.5) 34231 (47.8) –0.4 –0.7 Hyperparathyroidism 46 (1.9) 1519 (2.1) –0.2 –1.5 Hypertension* 1376 (57.1) 41527 (58) –1 –2 Hyperthyroidism 47 (1.9) 1480 (2.1) –0.1 –0.9 Inflammatory bowel disease (IBD)* 31 (1.3) 894 (1.2) 0 0.3 Ischemic heart disease* 318 (13.2) 9853 (13.8) –0.6 –1.7 Liver disease* 121 (5) 3878 (5.4) –0.4 –1.8 Metastatic cancer* 65 (2.7) 1979 (2.8) –0.1 –0.4 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 944 (39.2) 28144 (39.3) –0.2 –0.4 Other fractures* — — — — Palpitations 104 (4.3) 3166 (4.4) –0.1 –0.5 Parkinson’s disease 50 (2.1) 1695 (2.4) –0.3 –2

221

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 49 (2) 1459 (2) 0 0 coronary syndromes (MI/ACS)* Preventive care 1286 (53.3) 37013 (51.7) 1.6 3.2 Prior MI 12 (0.5) 452 (0.6) –0.1 –1.8 Prior stroke 18 (0.7) 532 (0.7) 0 0 Peripheral vascular disease (PVD) or 232 (9.6) 6958 (9.7) –0.1 –0.3 PVD surgery Rheumatoid arthritis* 88 (3.6) 2820 (3.9) –0.3 –1.5 Recent MI 1 (0) 42 (0.1) 0 –0.7 Recent stroke 3 (0.1) 102 (0.1) 0 –0.5 Renal dysfunction 331 (13.7) 10148 (14.2) –0.5 –1.3 Schizophrenia 12 (0.5) 320 (0.4) 0.1 0.7 Transient ischemic attack* 52 (2.2) 1663 (2.3) –0.2 –1.1 Upper GI diseases* 597 (24.8) 18218 (25.5) –0.7 –1.6 Urinary tract infection 414 (17.2) 11854 (16.6) 0.6 1.6 Baseline medication use, n (%) Systemic steroid* 246 (10.2) 7873 (11) –0.8 –2.6 Bile acid sequestrants 43 (1.8) 1345 (1.9) –0.1 –0.7 IV osteoclast inhibitors* 28 (1.2) 981 (1.4) –0.2 –1.9 222

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 197 (8.2) 6200 (8.7) –0.5 –1.8 Lipid-lowering agents 1233 (51.1) 37012 (51.7) –0.6 –1.2 Niacin and fibrates 137 (5.7) 4410 (6.2) –0.5 –2 Angiotensin-converting enzyme 818 (33.9) 25014 (35) –1 –2.2 inhibitors (ACEIs)* Agents for dementia 125 (5.2) 3789 (5.3) –0.1 –0.5 Antidiabetics* 360 (14.9) 12256 (17.1) –2.2 –6 Antiparkinson agents* 99 (4.1) 3120 (4.4) –0.3 –1.3 Antiplatelets* 1 (0) 51 (0.1) 0 –1.2 Angiotensin II receptor blockers 431 (17.9) 13619 (19) –1.2 –3 (ARBs)* Aromatase inhibitors 52 (2.2) 1496 (2.1) 0.1 0.5 Atypical antipsychotics 100 (4.1) 3020 (4.2) –0.1 –0.4 Beta-blockers* 950 (39.4) 29437 (41.1) –1.7 –3.5 Biologic disease-modifying 4 (0.2) 169 (0.2) –0.1 –1.6 antirheumatic drugs (DMARDs)* Calcium channel blockers* 738 (30.6) 23323 (32.6) –2 –4.3 Digoxin* 134 (5.6) 4232 (5.9) –0.4 –1.5

223

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Heparin and low-molecular-weight 47 (1.9) 1495 (2.1) –0.1 –1 heparin* Lithium 6 (0.2) 183 (0.3) 0 -0.1 Loop diuretics* 505 (20.9) 16013 (22.4) –1.4 –3.5 Monoamine oxidase inhibitors (MOAs) 1 (0) 55 (0.1) 0 –1.4 Nitrates* 176 (7.3) 5539 (7.7) –0.4 –1.7 Nonbiologic DMARDs 148 (6.1) 4604 (6.4) –0.3 –1.2

Nonselective nonsteroidal anti- 412 (17.1) 12548 (17.5) –0.5 –1.2 inflammatory drugs (NSAIDs)* Opioids* 1049 (43.5) 32213 (45) –1.5 –3 Oral anticoagulants* 317 (13.1) 10195 (14.2) –1.1 –3.2 Other anticoagulants* 3 (0.1) 94 (0.1) 0 –0.2 Other antihypertensives 202 (8.4) 6461 (9) –0.7 –2.3 Other newer and atypical 211 (8.8) 6299 (8.8) –0.1 –0.2 antidepressants* Potassium-sparing diuretics 321 (13.3) 10302 (14.4) –1.1 –3.1 Renin inhibitor 5 (0.2) 154 (0.2) 0 –0.2 Selective COX-2 inhibitors* 101 (4.2) 3051 (4.3) –0.1 –0.4 Selective estrogen receptor blockers 18 (0.7) 496 (0.7) 0.1 0.6 224

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Selective serotonin-norepinephrine 71 (2.9) 2088 (2.9) 0 0.2 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 443 (18.4) 13378 (18.7) –0.3 –0.8 (SSRIs)* Tricyclic antidepressants (TCAs)* 142 (5.9) 4358 (6.1) –0.2 –0.8 Thiazide diuretics* 790 (32.8) 24579 (34.4) –1.6 –3.4 Typical antipsychotics* 11 (0.5) 318 (0.4) 0 0.2 No lipid-lowering drug despite 294 (12.2) 8410 (11.8) 0.4 1.4 diagnosis of hyperlipidemia* Long-term use of opioids* 510 (21.2) 14919 (20.9) 0.3 0.7 Long-term current use of steroids* 16 (0.7) 589 (0.8) –0.2 –1.9 Lack of 2nd RX among all the 9 (0.4) 252 (0.4) 0 0.3 interested drugs groups* Opioids baseline days’ supply ≥ 90* 278 (11.5) 8875 (12.4) –0.9 –2.7 Index days’ supply* 46 (26.4) 46 (26.6) –0.8 –3 Use of preventive services Colonoscopy*, n (%) 239 (9.9) 7208 (10.1) –0.2 –0.5 Fecal occult blood test*, n (%) 301 (12.5) 8432 (11.8) 0.7 2.1 Mammography*, n (% of females) 782 (37.6) 22015 (36.2) 1.4 3.6

225

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 2.6 Bone mineral density test*, n (%) 544 (22.6) 14975 (20.9) 1.6 4 Electrocardiography, n (%) 833 (34.5) 25317 (35.4) –0.8 –1.8 International Normalized Ratio (INR) 1 (5.3) 1 (5) 0 0.6 test, n (%) Use of wheelchair, walker, crutches, or 64 (2.7) 2078 (2.9) –0.2 –1.5 cane, n (%) Use of oxygen, n (%) 85 (3.5) 2797 (3.9) –0.4 –2 Lipid test, n (%) 1021 (42.3) 29875 (41.8) 0.6 1.2 Health care utilization, mean (SD) Number of distinct cardiovascular 3 (3.2) 3 (3.3) –0.1 –3.4 diagnoses Total days in hospital in prior year* 2 (12.3) 2 (8.6) 0.2 1.6 Medication synchronization metrics 1 (0.2) 1 (0.2) 0 –2.7 Copayment 938 (985.8) 991 (1120.7) –53 –5 Number of office visits 18 (24.4) 19 (24.2) –0.2 –1 Number of unique drugs of interest 6 (3.4) 6 (4.4) –0.3 –6.5 (by NDC) in prior year or on index date Number of office visits with a 6 (12.7) 6 (13.8) –0.3 –2.1 cardiovascular diagnosis 226

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Number of drugs of interest 25 (19.7) 25 (20.6) –0.8 –4.1 dispensations in 365 days in prior year or on index date* Number of hospitalizations in the prior 0 (0.7) 0 (0.7) 0 –2.1 year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –7.6 (non-mail)* Number of cardiovascular 0 (0.5) 0 (0.5) 0 –3 hospitalizations Index copayment 42 (41) 42 (63.5) –0.2 –0.4 Vaccinations* 0 (0.7) 0 (0.7) 0 –0.6 Physician visits* 8 (7.3) 8 (7.7) –0.4 –5.4 Number of drugs of interest 4 (2.2) 5 (3.2) –0.2 –8.2 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 7 (3.9) 8 (4.3) –0.3 –7.4 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 8 (4.4) 9 (4.8) –0.4 –8.8 period with days’ supply that overlap index date*

227

Appendix 31. Characteristics of calcitonin vs bisphosphonate initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Number of unique (by NDC) drugs in 15 (9) 16 (9.9) –0.5 –5.3 365 days before or on index date* Number of any drug dispensations in 53 (42.8) 54 (46.6) –1.1 –2.3 365 days before or on index date*

228

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Calcitonin Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 71 (11.6) 71 (10.7) 0 –0.2 Sex: female* 6718 (86.9) 186778 (86.8) 0.1 0.3 Combined comorbidity score* 2 (2.6) 2 (2.5) 0 –0.5 Commercial plan type, n (%) 2989 (38.7) 82949 (38.6) 0.1 0.3 Benefit plan types, n (%) EPO* 263 (3.4) 7036 (3.3) 0.1 0.7 HMO* 3476 (45) 96629 (44.9) 0.1 0.1 IND* 548 (7.1) 15667 (7.3) –0.2 –0.7 OTH* 1324 (17.1) 37103 (17.2) -0.1 -0.3 POS* 1496 (19.4) 41075 (19.1) 0.3 0.7 Comorbidities, n (%) Acute coronary syndrome, with or 460 (6) 13255 (6.2) –0.2 –0.9 without revascularization Acute coronary syndrome, with 126 (1.6) 3700 (1.7) –0.1 –0.7 revascularization Atrial fibrillation 165 (2.1) 4834 (2.2) –0.1 –0.8 Alzheimer’s disease* 668 (8.6) 17514 (8.1) 0.5 1.8

229

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Angina 381 (4.9) 10936 (5.1) –0.2 –0.7 Coronary atherosclerosis* 1311 (17) 37445 (17.4) –0.4 –1.2 Coronary artery bypass graft 24 (0.3) 759 (0.4) 0 –0.7 (CABG), new CABG, old 161 (2.1) 4903 (2.3) –0.2 –1.3 Any cancer 1631 (21.1) 46146 (21.4) –0.3 –0.8 Any malignancy, including 1492 (19.3) 42272 (19.6) –0.3 –0.9 lymphoma and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 1040 (13.5) 29667 (13.8) –0.3 –1 Chest pain 1863 (24.1) 52976 (24.6) –0.5 –1.2 Heart failure* 965 (12.5) 26696 (12.4) 0.1 0.2 Heart failure hospitalization 71 (0.9) 2220 (1) –0.1 –1.1 Conduction disorders 253 (3.3) 7264 (3.4) –0.1 –0.6 Chronic obstructive pulmonary 1544 (20) 43995 (20.4) –0.5 –1.2 disease (COPD) Depression* 1294 (16.7) 36164 (16.8) –0.1 –0.2 Diabetes* 1644 (21.3) 48047 (22.3) –1.1 –2.6 Drug-induced osteoporosis* 157 (2) 4547 (2.1) –0.1 –0.6

230

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Falls* 379 (4.9) 10330 (4.8) 0.1 0.5 History of falls 327 (4.2) 8835 (4.1) 0.1 0.6 HIV infection 7 (0.1) 223 (0.1) 0 –0.4 Hyperlipidemia 4167 (53.9) 117137 (54.4) –0.5 –1.1 Hyperparathyroidism 141 (1.8) 4128 (1.9) –0.1 –0.7 Hypertension* 4988 (64.5) 141117 (65.6) –1 –2.2 Hyperthyroidism 152 (2) 4284 (2) 0 –0.2 Inflammatory bowel disease (IBD)* 130 (1.7) 3606 (1.7) 0 0 Ischemic heart disease* 1384 (17.9) 39505 (18.4) –0.5 –1.2 Liver disease* 493 (6.4) 14298 (6.6) –0.3 –1.1 Metastatic cancer* 162 (2.1) 4671 (2.2) –0.1 –0.5 Osteoporotic fracture: non- — — — — vertebral* Osteoporotic fracture: vertebral* — — — — Osteoporosis* 2999 (38.8) 84166 (39.1) –0.3 –0.6 Other fractures* — — — — Palpitations 451 (5.8) 12759 (5.9) –0.1 –0.4 Parkinson’s disease 156 (2) 4082 (1.9) 0.1 0.9

231

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 217 (2.8) 6302 (2.9) –0.1 –0.7 coronary syndromes (MI/ACS)* Preventive care 4318 (55.9) 119574 (55.6) 0.3 0.6 Prior MI 73 (0.9) 2108 (1) 0 –0.4 Prior stroke 48 (0.6) 1322 (0.6) 0 0.1 Peripheral vascular disease (PVD) or 795 (10.3) 21821 (10.1) 0.1 0.5 PVD surgery Rheumatoid arthritis* 378 (4.9) 11040 (5.1) –0.2 –1.1 Recent MI 5 (0.1) 133 (0.1) 0 0.1 Recent stroke 6 (0.1) 168 (0.1) 0 0 Renal dysfunction 1358 (17.6) 38469 (17.9) –0.3 –0.8 Schizophrenia 39 (0.5) 984 (0.5) 0 0.7 Transient ischemic attack* 268 (3.5) 7614 (3.5) –0.1 –0.4 Upper GI diseases* 2320 (30) 65971 (30.7) –0.6 –1.4 Urinary tract infection 1592 (20.6) 44361 (20.6) 0 0 Baseline medication use, n (%) Systemic steroid* 946 (12.2) 27425 (12.7) –0.5 –1.5 Bile acid sequestrants 142 (1.8) 4212 (2) –0.1 –0.9 IV osteoclast inhibitors* 88 (1.1) 2736 (1.3) –0.1 –1.2 232

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 804 (10.4) 23113 (10.7) –0.3 –1.1 Lipid-lowering agents 3601 (46.6) 101586 (47.2) –0.6 –1.2 Niacin and fibrates 357 (4.6) 10517 (4.9) –0.3 –1.3 Angiotensin-converting enzyme 2621 (33.9) 75176 (34.9) –1 –2.2 inhibitors (ACEIs)* Agents for dementia 383 (5) 10234 (4.8) 0.2 0.9 Antidiabetics* 1259 (16.3) 37625 (17.5) –1.2 –3.2 Antiparkinson agents* 343 (4.4) 9581 (4.5) 0 –0.1 Antiplatelets* 5 (0.1) 178 (0.1) 0 –0.7 Angiotensin II receptor blockers 1491 (19.3) 43170 (20.1) –0.8 –1.9 (ARBs)* Aromatase inhibitors 112 (1.4) 3095 (1.4) 0 0.1 Atypical antipsychotics 344 (4.5) 9329 (4.3) 0.1 0.6 Beta-blockers* 3193 (41.3) 91445 (42.5) –1.2 –2.4 Biologic disease-modifying 50 (0.6) 1406 (0.7) 0 –0.1 antirheumatic drugs (DMARDs)* Calcium channel blockers* 2379 (30.8) 68507 (31.8) –1.1 –2.3 Digoxin* 388 (5) 10863 (5) 0 –0.1

233

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Heparin and low-molecular-weight 145 (1.9) 4194 (1.9) –0.1 –0.5 heparin* Lithium 20 (0.3) 548 (0.3) 0 0.1 Loop diuretics* 1618 (20.9) 46328 (21.5) –0.6 –1.5 Monoamine oxidase inhibitors 1 (0) 35 (0) 0 –0.3 (MOAs) Nitrates* 695 (9) 19669 (9.1) –0.1 –0.5 Nonbiologic DMARDs 455 (5.9) 13155 (6.1) –0.2 –1

Nonselective nonsteroidal anti- 1673 (21.6) 47803 (22.2) –0.6 –1.4 inflammatory drugs (NSAIDs)* Opioids* 4134 (53.5) 117414 (54.6) –1.1 –2.2 Oral anticoagulants* 827 (10.7) 23718 (11) –0.3 –1 Other anticoagulants* 10 (0.1) 266 (0.1) 0 0.2 Other antihypertensives 597 (7.7) 17130 (8) –0.2 –0.9 Other newer and atypical 904 (11.7) 25387 (11.8) –0.1 –0.3 antidepressants* Potassium-sparing diuretics 842 (10.9) 24908 (11.6) –0.7 –2.2 Renin inhibitor 24 (0.3) 731 (0.3) 0 –0.5 Selective COX-2 inhibitors* 343 (4.4) 9933 (4.6) –0.2 –0.9

234

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Selective estrogen receptor 40 (0.5) 1054 (0.5) 0 0.4 blockers Selective serotonin-norepinephrine 442 (5.7) 12626 (5.9) –0.1 –0.6 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 2022 (26.2) 56940 (26.5) –0.3 –0.7 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs)* 540 (7) 15924 (7.4) –0.4 –1.6 Thiazide diuretics* 2350 (30.4) 68326 (31.8) –1.3 –2.9 Typical antipsychotics* 39 (0.5) 1101 (0.5) 0 –0.1 No lipid-lowering drug despite 1408 (18.2) 39115 (18.2) 0 0.1 diagnosis of hyperlipidemia* Long-term use of opioids* 1798 (23.3) 50553 (23.5) –0.2 –0.5 Long-term current use of steroids* 87 (1.1) 2509 (1.2) 0 –0.4 Lack of 2nd RX among all the 170 (2.2) 4368 (2) 0.2 1.2 interested drugs groups* Opioids baseline days’ supply ≥ 90* 1447 (18.7) 41813 (19.4) –0.7 –1.8 Index days’ supply* 37 (19.6) 37 (20.7) –0.2 –1 Use of preventive services Colonoscopy*, n (%) 797 (10.3) 22580 (10.5) –0.2 –0.6 Fecal occult blood test*, n (%) 864 (11.2) 23811 (11.1) 0.1 0.4 235

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Mammography*, n (% of females) 2129 (31.7) 58903 (31.5) 0.2 0.4 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 0.2 Bone mineral density test*, n (%) 850 (11) 23463 (10.9) 0.1 0.3 Electrocardiography, n (%) 3282 (42.5) 92958 (43.2) –0.7 –1.5 INR test, n (%) 1 (4.7) 1 (4.4) 0 0.1 Use of wheelchair, walker, crutches, 303 (3.9) 8802 (4.1) –0.2 –0.9 or cane, n (%) Use of oxygen, n (%) 346 (4.5) 9921 (4.6) –0.1 –0.6 Lipid test, n (%) 3821 (49.4) 106561 (49.5) –0.1 –0.2 Health care utilization, mean (SD) Number of distinct cardiovascular 3 (3.7) 3 (3.8) –0.1 –1.7 diagnoses Total days in hospital in prior year* 2 (8.6) 2 (7.8) 0 –0.5 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –3.4 Copayment 842 (906.8) 864 (1077.4) –22.2 –2.2 Number of office visits 21 (25.3) 21 (25.9) –0.3 –1.1 Number of unique drugs of interest 6 (3.5) 6 (4.6) –0.2 –3.9 (by NDC) in prior year or on index date

236

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Number of office visits with a 6 (11.6) 6 (12.3) –0.2 –1.4 cardiovascular diagnosis Number of drugs of interest 22 (17.2) 22 (18.1) –0.4 –2.2 dispensations in 365 days in prior year or on index date* Number of hospitalizations in the 0 (0.9) 0 (0.9) 0 –0.6 prior year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –4.1 (non-mail)* Number of cardiovascular 0 (0.6) 0 (0.6) 0 –1.2 hospitalizations Index copayment 36 (33.1) 36 (61.7) –0.3 –0.6 Vaccinations* 0 (0.8) 0 (0.8) 0 –0.4 Physician visits* 9 (8.2) 10 (8.5) –0.3 –3.5 Number of drugs of interest 4 (2) 4 (2.8) –0.1 –4.2 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 6 (3.4) 6 (3.6) –0.1 –3.9 365 days before or on index date with days’ supply that overlap index date

237

Appendix 32. Characteristics of calcitonin vs bisphosphonate initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Number of dispensations in 7 (3.8) 7 (4) –0.2 –4.3 baseline period with days’ supply that overlap index date* Number of unique (by NDC) drugs 16 (9.7) 17 (10.3) -0.3 -3.5 in 365 days before or on index date* Number of any drug dispensations 49 (36.5) 49 (39.6) -0.5 -1.3 in 365 days before or on index date*

238

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 69 (12.9) 62 (10.7) 6.5 54.9 Sex: female* 2320 (86.9) 4355 (99.4) –12.5 –51.1 Combined comorbidity score* 2 (2.7) 1 (1.8) 1.2 52.8 Commercial plan type, n (%) 905 (33.9) 2582 (58.9) –25 –51.8 Benefit plan types, n (%) EPO* 111 (4.2) 389 (8.9) –4.7 –19.2 HMO* 1165 (43.6) 1112 (25.4) 18.3 39.2 IND* 70 (2.6) 40 (0.9) 1.7 13 OTH* 550 (20.6) 755 (17.2) 3.4 8.6 POS* 562 (21.1) 1776 (40.5) –19.5 –43.2 Comorbidities, n (%) Acute coronary syndrome, with or 225 (8.4) 210 (4.8) 3.6 14.7 without revascularization Acute coronary syndrome, with 69 (2.6) 69 (1.6) 1 7.1 revascularization Atrial fibrillation 66 (2.5) 43 (1) 1.5 11.5 Alzheimer’s disease* 203 (7.6) 77 (1.8) 5.8 28

239

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Angina 190 (7.1) 187 (4.3) 2.9 12.3 Coronary atherosclerosis* 567 (21.2) 441 (10.1) 11.2 31.1 Coronary artery bypass graft (CABG), 16 (0.6) 5 (0.1) 0.5 8.1 new CABG, old 70 (2.6) 32 (0.7) 1.9 14.8 Any cancer 553 (20.7) 994 (22.7) –2 –4.8 Any malignancy, including lymphoma 507 (19) 923 (21.1) –2.1 –5.2 and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 455 (17) 623 (14.2) 2.8 7.8 Chest pain 826 (30.9) 927 (21.2) 9.8 22.4 Heart failure* 404 (15.1) 207 (4.7) 10.4 35.4 Heart failure hospitalization 39 (1.5) 17 (0.4) 1.1 11.2 Conduction disorders 100 (3.7) 85 (1.9) 1.8 10.9 Chronic obstructive pulmonary 684 (25.6) 542 (12.4) 13.3 34.3 disease (COPD) Depression* 677 (25.4) 1038 (23.7) 1.7 3.9 Diabetes* 671 (25.1) 929 (21.2) 3.9 9.3 Drug-induced osteoporosis* 57 (2.1) 85 (1.9) 0.2 1.4

240

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Falls* 167 (6.3) 92 (2.1) 4.2 20.9 History of falls 149 (5.6) 84 (1.9) 3.7 19.4 HIV infection 1 (0) 6 (0.1) –0.1 –3.4 Hyperlipidemia 1551 (58.1) 2638 (60.2) –2.1 –4.3 Hyperparathyroidism 59 (2.2) 48 (1.1) 1.1 8.8 Hypertension* 1838 (68.9) 2686 (61.3) 7.6 15.9 Hyperthyroidism 53 (2) 97 (2.2) –0.2 –1.6 Inflammatory bowel disease (IBD)* 63 (2.4) 66 (1.5) 0.9 6.2 Ischemic heart disease* 592 (22.2) 468 (10.7) 11.5 31.4 Liver disease* 215 (8.1) 305 (7) 1.1 4.2 Metastatic cancer* 45 (1.7) 36 (0.8) 0.9 7.8 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 974 (36.5) 1450 (33.1) 3.4 7.1 Other fractures* — — — — Palpitations 215 (8.1) 335 (7.6) 0.4 1.5 Parkinson’s disease 46 (1.7) 28 (0.6) 1.1 10

241

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 95 (3.6) 51 (1.2) 2.4 15.8 coronary syndromes (MI/ACS)* Preventive care 1496 (56.1) 3051 (69.6) –13.6 –28.4 Prior MI 33 (1.2) 19 (0.4) 0.8 8.8 Prior stroke 14 (0.5) 9 (0.2) 0.3 5.3 Peripheral vascular disease (PVD) or 300 (11.2) 201 (4.6) 6.7 24.8 PVD surgery Rheumatoid arthritis* 177 (6.6) 232 (5.3) 1.3 5.6 Recent MI 2 (0.1) 1 (0) 0.1 2.4 Recent stroke 1 (0) 4 (0.1) –0.1 –2.1 Renal dysfunction 568 (21.3) 381 (8.7) 12.6 35.8 Schizophrenia 12 (0.4) 15 (0.3) 0.1 1.7 Transient ischemic attack* 133 (5) 67 (1.5) 3.5 19.6 Upper GI diseases* 975 (36.5) 1338 (30.5) 6 12.7 Urinary tract infection 648 (24.3) 783 (17.9) 6.4 15.8 Baseline medication use, n (%) Systemic steroid* 397 (14.9) 292 (6.7) 8.2 26.7 Bile acid sequestrants 56 (2.1) 64 (1.5) 0.6 4.8 IV osteoclast inhibitors* 28 (1) 19 (0.4) 0.6 7.2 242

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 375 (14.1) 395 (9) 5 15.8 Lipid-lowering agents 1111 (41.6) 1644 (37.5) 4.1 8.4 Niacin and fibrates 104 (3.9) 144 (3.3) 0.6 3.3 Angiotensin-converting enzyme 877 (32.9) 1153 (26.3) 6.5 14.4 inhibitors (ACEIs)* Agents for dementia 127 (4.8) 54 (1.2) 3.5 20.8 Antidiabetics* 487 (18.2) 691 (15.8) 2.5 6.6 Antiparkinson agents* 137 (5.1) 162 (3.7) 1.4 7 Antiplatelets* 4 (0.1) 2 (0) 0.1 3.3 Angiotensin II receptor blockers 537 (20.1) 924 (21.1) –1 –2.4 (ARBs)* Aromatase inhibitors 27 (1) 83 (1.9) –0.9 –7.4 Atypical antipsychotics 129 (4.8) 166 (3.8) 1 5.1 Beta-blockers* 1106 (41.4) 1274 (29.1) 12.4 26.1 Biologic disease-modifying 32 (1.2) 30 (0.7) 0.5 5.3 antirheumatic drugs (DMARDs)* Calcium channel blockers* 805 (30.2) 932 (21.3) 8.9 20.4 Digoxin* 116 (4.3) 63 (1.4) 2.9 17.4

243

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Heparin and low-molecular-weight 48 (1.8) 33 (0.8) 1 9.3 heparin* Lithium 10 (0.4) 17 (0.4) 0 –0.2 Loop diuretics* 599 (22.4) 413 (9.4) 13 36.1 Monoamine oxidase inhibitors (MOAs) Nitrates* 306 (11.5) 168 (3.8) 7.6 29 Nonbiologic DMARDs 167 (6.3) 210 (4.8) 1.5 6.4

Nonselective nonsteroidal anti- 731 (27.4) 1241 (28.3) –0.9 –2.1 inflammatory drugs (NSAIDs)* Opioids* 1740 (65.2) 2041 (46.6) 18.6 38.1 Oral anticoagulants* 242 (9.1) 126 (2.9) 6.2 26.4 Other anticoagulants* 6 (0.2) 2 (0) 0.2 4.9 Other antihypertensives 215 (8.1) 159 (3.6) 4.4 19 Other newer and atypical 408 (15.3) 556 (12.7) 2.6 7.5 antidepressants* Potassium-sparing diuretics 262 (9.8) 387 (8.8) 1 3.4 Renin inhibitor 10 (0.4) 22 (0.5) –0.1 –1.9 Selective COX-2 inhibitors* 136 (5.1) 207 (4.7) 0.4 1.7

244

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Selective estrogen receptor blockers 7 (0.3) 121 (2.8) –2.5 –20.6 Selective serotonin-norepinephrine 255 (9.6) 455 (10.4) –0.8 –2.8 reuptake inhibitors (SNRIs)* Selective serotonin reuptake 935 (35) 1716 (39.2) –4.1 –8.6 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs)* 228 (8.5) 270 (6.2) 2.4 9.1 Thiazide diuretics* 726 (27.2) 1426 (32.5) –5.3 –11.7 Typical antipsychotics* 11 (0.4) 17 (0.4) 0 0.4 No lipid-lowering drug despite 632 (23.7) 1208 (27.6) –3.9 –8.9 diagnosis of hyperlipidemia* Long-term use of opioids* 699 (26.2) 973 (22.2) 4 9.3 Long-term current use of steroids* 44 (1.6) 24 (0.5) 1.1 10.6 Lack of 2nd RX among all the 144 (5.4) 386 (8.8) –3.4 –13.3 interested drugs groups* Opioids baseline days’ supply ≥ 90* 746 (28) 584 (13.3) 14.6 36.7 Index days’ supply* 32 (10.2) 34 (15.6) –2.8 –21.2 Use of preventive services Colonoscopy*, n (%) 288 (10.8) 500 (11.4) –0.6 –2 Fecal occult blood test*, n (%) 264 (9.9) 661 (15.1) –5.2 –15.8

245

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Mammography*, n (% of females) 616 (26.6) 1645 (37.8) –11.2 –31.9 Any prior lab test, n (%) 0 (0.4) 0 (0.5) –0.1 –34.3 Bone mineral density test*, n (%) 84 (3.1) 192 (4.4) –1.2 –6.5 Electrocardiography, n (%) 1308 (49) 1864 (42.5) 6.5 13 INR test, n (%) 1 (4) 0 (1.9) 0.6 20.5 Use of wheelchair, walker, crutches, 144 (5.4) 74 (1.7) 3.7 20.2 or cane, n (%) Use of oxygen, n (%) 144 (5.4) 63 (1.4) 4 21.9 Lipid test, n (%) 1459 (54.7) 2791 (63.7) –9 –18.5 Health care utilization, mean (SD) Number of distinct cardiovascular 4 (4.1) 2 (2.7) 1.6 44.7 diagnoses Total days in hospital in prior year* 2 (6.1) 1 (3.6) 1.5 29.9 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 16.3 Copayment 809 (890.7) 615 (641.1) 193.5 24.9 Number of office visits 25 (27.2) 19 (18) 6 26 Number of unique drugs of interest 6 (3.7) 5 (2.9) 1.2 36.3 (by NDC) in prior year or on index date

246

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Number of office visits with a 7 (11.8) 4 (7) 3 30.8 cardiovascular diagnosis Number of drugs of interest 18 (14.3) 14 (12) 3.3 24.7 dispensations in 365 days in prior year or on index date* Number of hospitalizations in the 0 (0.9) 0 (0.5) 0.3 35.9 prior year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 18 (non-mail)* Number of cardiovascular 0 (0.7) 0 (0.4) 0.2 30.4 hospitalizations Index copayment 31 (25.1) 39 (38.6) –7.9 –24.2 Vaccinations* 0 (0.8) 0 (0.7) 0 4.5 Physician visits* 11 (9.1) 10 (7.1) 1.7 20.7 Number of drugs of interest 3 (1.7) 3 (1.6) 0.3 17.7 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 5 (3) 4 (2.6) 1.3 48 days before or on index date with days’ supply that overlap index date

247

Appendix 33. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Number of dispensations in baseline 6 (3.1) 4 (2.7) 1.4 47 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 19 (10.8) 14 (8.6) 4.7 48.2 365 days before or on index date* Number of any drug dispensations in 47 (33) 33 (25.2) 13.4 45.7 365 days before or on index date*

248

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 71 (11.1) 66 (10.3) 5.9 54.9 Sex: female* 2309 (87.6) 4383 (99.3) –11.6 –48.4 Combined comorbidity score* 1 (2.5) 0 (1.6) 1.1 52 Commercial plan type, n (%) 882 (33.5) 2333 (52.8) –19.4 –39.9 Benefit plan types, n (%) EPO* 77 (2.9) 284 (6.4) –3.5 –16.7 HMO* 1288 (48.9) 1305 (29.6) 19.3 40.4 IND* 120 (4.6) 131 (3) 1.6 8.3 OTH* 486 (18.4) 883 (20) –1.6 –3.9 POS* 489 (18.6) 1458 (33) –14.5 –33.5 Comorbidities, n (%) Acute coronary syndrome, with or without 144 (5.5) 133 (3) 2.5 12.2 revascularization Acute coronary syndrome, with 34 (1.3) 35 (0.8) 0.5 4.9 revascularization Atrial fibrillation 56 (2.1) 25 (0.6) 1.6 13.6 Alzheimer’s disease* 245 (9.3) 95 (2.2) 7.1 31.1

249

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Angina 118 (4.5) 120 (2.7) 1.8 9.5 Coronary atherosclerosis* 444 (16.9) 337 (7.6) 9.2 28.4 Coronary artery bypass graft (CABG), new 4 (0.2) 2 (0) 0.1 3.4 CABG, old 56 (2.1) 16 (0.4) 1.8 16 Any cancer 533 (20.2) 998 (22.6) –2.4 –5.8 Any malignancy, including lymphoma and 492 (18.7) 926 (21) –2.3 –5.8 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 320 (12.1) 471 (10.7) 1.5 4.6 Chest pain 614 (23.3) 671 (15.2) 8.1 20.7 Heart failure* 312 (11.8) 164 (3.7) 8.1 30.7 Heart failure hospitalization 20 (0.8) 6 (0.1) 0.6 9.3 Conduction disorders 86 (3.3) 61 (1.4) 1.9 12.5 Chronic obstructive pulmonary disease 504 (19.1) 438 (9.9) 9.2 26.4 (COPD) Depression* 391 (14.8) 615 (13.9) 0.9 2.6 Diabetes* 532 (20.2) 846 (19.2) 1 2.6 Drug-induced osteoporosis* 52 (2) 68 (1.5) 0.4 3.3 Falls* 120 (4.6) 63 (1.4) 3.1 18.4 History of falls 97 (3.7) 60 (1.4) 2.3 14.9 250

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 HIV infection 1 (0) 2 (0) 0 -0.4 Hyperlipidemia 1463 (55.5) 2627 (59.5) –4 –8.1 Hyperparathyroidism 36 (1.4) 40 (0.9) 0.5 4.3 Hypertension* 1766 (67) 2836 (64.2) 2.8 5.9 Hyperthyroidism 52 (2) 100 (2.3) –0.3 –2 Inflammatory bowel disease (IBD)* 35 (1.3) 47 (1.1) 0.3 2.4 Ischemic heart disease* 470 (17.8) 361 (8.2) 9.7 29 Liver disease* 157 (6) 178 (4) 1.9 8.9 Metastatic cancer* 51 (1.9) 37 (0.8) 1.1 9.4 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1072 (40.7) 1579 (35.8) 4.9 10.1 Other fractures* - - - - Palpitations 132 (5) 260 (5.9) –0.9 –3.9 Parkinson’s disease 60 (2.3) 25 (0.6) 1.7 14.5 Post-myocardial infarction/acute coronary 73 (2.8) 34 (0.8) 2 15.2 syndromes (MI/ACS)* Preventive care 1526 (57.9) 3140 (71.1) –13.2 –27.9

251

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Prior MI 28 (1.1) 8 (0.2) 0.9 11.2 Prior stroke 16 (0.6) 1 (0) 0.6 10.4 Peripheral vascular disease (PVD) or PVD 259 (9.8) 167 (3.8) 6 24.2 surgery Rheumatoid arthritis* 113 (4.3) 130 (2.9) 1.3 7.2 Recent MI 2 (0.1) 0 (0) 0.1 3.9 Recent stroke 2 (0.1) 0 (0) 0.1 3.9 Renal dysfunction 458 (17.4) 305 (6.9) 10.5 32.5 Schizophrenia 15 (0.6) 11 (0.2) 0.3 5 Transient ischemic attack* 82 (3.1) 62 (1.4) 1.7 11.5 Upper GI diseases* 744 (28.2) 1041 (23.6) 4.7 10.6 Urinary tract infection 529 (20.1) 683 (15.5) 4.6 12.1 Baseline medication use, n (%) Systemic steroid* 302 (11.5) 190 (4.3) 7.2 26.8 Bile acid sequestrants 43 (1.6) 48 (1.1) 0.5 4.7 IV osteoclast inhibitors* 32 (1.2) 31 (0.7) 0.5 5.3 Inhaled glucocorticoid* 231 (8.8) 262 (5.9) 2.8 10.9 Lipid-lowering agents 1252 (47.5) 2058 (46.6) 0.9 1.8

252

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Niacin and fibrates 114 (4.3) 199 (4.5) –0.2 –0.9 Angiotensin-converting enzyme inhibitors 920 (34.9) 1298 (29.4) 5.5 11.8 (ACEIs)* Agents for dementia 131 (5) 61 (1.4) 3.6 20.6 Antidiabetics* 411 (15.6) 638 (14.5) 1.1 3.2 Antiparkinson agents* 106 (4) 102 (2.3) 1.7 9.8 Antiplatelets* 0 (0) 1 (0) 0 -2.1 Angiotensin II receptor blockers (ARBs)* 523 (19.8) 972 (22) –2.2 –5.3 Aromatase inhibitors 33 (1.3) 81 (1.8) –0.6 –4.7 Atypical antipsychotics 114 (4.3) 110 (2.5) 1.8 10.1 Beta-blockers* 1133 (43) 1386 (31.4) 11.6 24.2 Biologic disease-modifying antirheumatic 14 (0.5) 16 (0.4) 0.2 2.5 drugs (DMARDs)* Calcium channel blockers* 830 (31.5) 1173 (26.6) 4.9 10.9 Digoxin* 137 (5.2) 85 (1.9) 3.3 17.7 Heparin and low-molecular-weight heparin* 50 (1.9) 21 (0.5) 1.4 13.2 Lithium 4 (0.2) 12 (0.3) –0.1 –2.6 Loop diuretics* 511 (19.4) 327 (7.4) 12 35.7 Monoamine oxidase inhibitors (MOAs) 0 (0) 1 (0) 0 –2.1 253

Nonselective nonsteroidal anti-inflammatory 527 (20) 989 (22.4) –2.4 –5.9 drugs (NSAIDs)* Opioids* 1340 (50.9) 1548 (35.1) 15.8 32.3 Oral anticoagulants* 267 (10.1) 135 (3.1) 7.1 28.8 Other anticoagulants* 1 (0) 2 (0) 0 -0.4 Other antihypertensives 179 (6.8) 141 (3.2) 3.6 16.6 Other newer and atypical antidepressants* 283 (10.7) 327 (7.4) 3.3 11.6 Potassium-sparing diuretics 257 (9.8) 440 (10) –0.2 –0.7 Renin inhibitor 9 (0.3) 11 (0.2) 0.1 1.7 Selective COX-2 inhibitors* 106 (4) 181 (4.1) -0.1 -0.4 Selective estrogen receptor blockers 15 (0.6) 95 (2.2) –1.6 –13.7 Selective serotonin-norepinephrine reuptake 115 (4.4) 207 (4.7) –0.3 –1.6 inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 642 (24.4) 1171 (26.5) –2.2 –5 (SSRIs)* Tricyclic antidepressants (TCAs)* 172 (6.5) 220 (5) 1.5 6.6 Thiazide diuretics* 828 (31.4) 1608 (36.4) –5 –10.6 254

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Typical antipsychotics* 17 (0.6) 5 (0.1) 0.5 8.7 No lipid-lowering drug despite diagnosis of 479 (18.2) 858 (19.4) –1.3 –3.2 hyperlipidemia* Long-term use of opioids* 587 (22.3) 887 (20.1) 2.2 5.4 Long-term current use of steroids* 27 (1) 14 (0.3) 0.7 8.7 Lack of 2nd RX among all the interested drugs 17 (0.6) 56 (1.3) –0.6 –6.4 groups* Opioids baseline days’ supply ≥ 90* 422 (16) 279 (6.3) 9.7 31.2 Index days’ supply* 35 (16.9) 43 (24.3) –7.5 –35.9 Use of preventive services Colonoscopy*, n (%) 269 (10.2) 527 (11.9) –1.7 –5.5 Fecal occult blood test*, n (%) 298 (11.3) 824 (18.7) –7.4 –20.7 Mammography*, n (% of females) 723 (31.3) 1990 (45.4) –14.1 –37.3 Any prior lab test, n (%) 0 (0.4) 0 (0.4) –0.1 –33.2 Bone mineral density test*, n (%) 219 (8.3) 493 (11.2) –2.9 –9.6 Electrocardiography, n (%) 1134 (43) 1740 (39.4) 3.6 7.4 INR test, n (%) 1 (4.6) 0 (1.8) 0.7 21.5 Use of wheelchair, walker, crutches, or cane, 95 (3.6) 77 (1.7) 1.9 11.6 n (%)

255

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Use of oxygen, n (%) 116 (4.4) 39 (0.9) 3.5 22.1 Lipid test, n (%) 1336 (50.7) 2835 (64.2) –13.5 –27.6 Health care utilization, mean (SD) Number of distinct cardiovascular diagnoses 3 (3.5) 2 (2.2) 1.2 42.3 Total days in hospital in prior year* 2 (6.3) 1 (2.4) 1.2 26 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 14.3 Copayment 791 (841.8) 650 (638.5) 140.8 18.8 Number of office visits 20 (23.7) 15 (15) 4.6 23 Number of unique drugs of interest (by NDC) 6 (3.6) 5 (2.7) 1.1 35.3 in prior year or on index date Number of office visits with a cardiovascular 6 (10.2) 4 (5.7) 2.3 28 diagnosis Number of drugs of interest dispensations in 23 (16.7) 18 (13.5) 4 26.4 365 days in prior year or on index date* Number of hospitalizations in the prior year 0 (1.1) 0 (0.4) 0.2 21.2 Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 16.9 mail)* Number of cardiovascular hospitalizations 0 (0.6) 0 (0.3) 0.1 24.6 Index copayment 35 (31.2) 44 (48.6) –8.8 –21.6

256

Appendix 34. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Vaccinations* 0 (0.8) 0 (0.7) 0.1 7.7 Physician visits* 9 (7.6) 8 (6) 0.8 11.7 Number of drugs of interest dispensations 4 (1.8) 3 (1.6) 0.4 21.8 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 days 6 (3.1) 5 (2.5) 1.5 52.4 before or on index date with days’ supply that overlap index date Number of dispensations in baseline period 6 (3.4) 5 (2.7) 1.5 50.5 with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 days 15 (8.6) 12 (6.9) 3.5 44.9 before or on index date* Number of any drug dispensations in 365 days 47 (33.3) 34 (25.1) 12.9 43.6 before or on index date*

257

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 71 (10.4) 66 (9.7) 5.5 54.6 Sex: female* 2091 (86.2) 4594 (99.4) –13.2 –52.5 Combined comorbidity score* 1 (2.5) 0 (1.5) 1 50.1 Commercial plan type, n (%) 1203 (49.6) 2864 (61.9) –12.3 –25.1 Benefit plan types, n (%) EPO* 75 (3.1) 287 (6.2) –3.1 –14.8 HMO* 1025 (42.3) 1395 (30.2) 12.1 25.3 IND* 358 (14.8) 404 (8.7) 6 18.8 OTH* 288 (11.9) 617 (13.3) –1.5 –4.4 POS* 445 (18.3) 1464 (31.7) –13.3 –31.1 Comorbidities, n (%) Acute coronary syndrome, with or without 91 (3.8) 104 (2.2) 1.5 8.8 revascularization Acute coronary syndrome, with 23 (0.9) 35 (0.8) 0.2 2.1 revascularization Atrial fibrillation 43 (1.8) 18 (0.4) 1.4 13.4 Alzheimer’s disease* 220 (9.1) 110 (2.4) 6.7 29.1

258

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Angina 73 (3) 92 (2) 1 6.5 Coronary atherosclerosis* 301 (12.4) 283 (6.1) 6.3 21.8 Coronary artery bypass graft (CABG), new 4 (0.2) 2 (0) 0.1 3.8 CABG, old 36 (1.5) 25 (0.5) 0.9 9.4 Any cancer 546 (22.5) 995 (21.5) 1 2.4 Any malignancy, including lymphoma and 494 (20.4) 915 (19.8) 0.6 1.4 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 265 (10.9) 430 (9.3) 1.6 5.4 Chest pain 424 (17.5) 631 (13.6) 3.8 10.6 Heart failure* 250 (10.3) 148 (3.2) 7.1 28.6 Heart failure hospitalization 12 (0.5) 2 (0) 0.5 8.7 Conduction disorders 67 (2.8) 53 (1.1) 1.6 11.7 Chronic obstructive pulmonary disease 358 (14.8) 412 (8.9) 5.8 18.2 (COPD) Depression* 226 (9.3) 441 (9.5) –0.2 –0.8 Diabetes* 442 (18.2) 729 (15.8) 2.5 6.5 Drug-induced osteoporosis* 48 (2) 64 (1.4) 0.6 4.6 Falls* 92 (3.8) 54 (1.2) 2.6 16.9

259

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 History of falls 81 (3.3) 58 (1.3) 2.1 13.9 HIV infection 5 (0.2) 1 (0) 0.2 5.5 Hyperlipidemia 1154 (47.6) 2625 (56.8) –9.2 –18.5 Hyperparathyroidism 46 (1.9) 39 (0.8) 1.1 9.1 Hypertension* 1386 (57.1) 2716 (58.7) –1.6 –3.3 Hyperthyroidism 47 (1.9) 95 (2.1) –0.1 –0.8 Inflammatory bowel disease (IBD)* 32 (1.3) 35 (0.8) 0.6 5.5 Ischemic heart disease* 323 (13.3) 309 (6.7) 6.6 22.2 Liver disease* 121 (5) 130 (2.8) 2.2 11.3 Metastatic cancer* 66 (2.7) 55 (1.2) 1.5 11.1 Osteoporotic fracture: non-vertebral* Osteoporotic fracture: vertebral* Osteoporosis* 953 (39.3) 1502 (32.5) 6.8 14.2 Other fractures* Palpitations 104 (4.3) 217 (4.7) –0.4 –2 Parkinson’s disease 50 (2.1) 32 (0.7) 1.4 11.8 Post-myocardial infarction/acute coronary 49 (2) 34 (0.7) 1.3 11 syndromes (MI/ACS)*

260

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Preventive care 1296 (53.4) 3199 (69.2) –15.8 –32.8 Prior MI 12 (0.5) 7 (0.2) 0.3 6.1 Prior stroke 18 (0.7) 9 (0.2) 0.5 8 Peripheral vascular disease (PVD) or PVD 236 (9.7) 141 (3) 6.7 27.6 surgery Rheumatoid arthritis* 88 (3.6) 129 (2.8) 0.8 4.8 Recent MI Recent stroke 3 (0.1) 2 (0) 0.1 2.8 Renal dysfunction 332 (13.7) 264 (5.7) 8 27.2 Schizophrenia 12 (0.5) 6 (0.1) 0.4 6.5 Transient ischemic attack* 53 (2.2) 45 (1) 1.2 9.7 Upper GI diseases* 602 (24.8) 889 (19.2) 5.6 13.5 Urinary tract infection 416 (17.1) 584 (12.6) 4.5 12.7 Baseline medication use, n (%) Systemic steroid* 247 (10.2) 186 (4) 6.2 24.2 Bile acid sequestrants 43 (1.8) 48 (1) 0.7 6.2 IV osteoclast inhibitors* 28 (1.2) 27 (0.6) 0.6 6.1 Inhaled glucocorticoid* 198 (8.2) 269 (5.8) 2.3 9.2

261

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Lipid-lowering agents 1239 (51.1) 2393 (51.8) –0.7 –1.4 Niacin and fibrates 139 (5.7) 304 (6.6) –0.8 –3.5 Angiotensin-converting enzyme inhibitors 825 (34) 1482 (32.1) 2 4.2 (ACEIs)* Agents for dementia 125 (5.2) 82 (1.8) 3.4 18.6 Antidiabetics* 362 (14.9) 585 (12.7) 2.3 6.6 Antiparkinson agents* 100 (4.1) 92 (2) 2.1 12.4 Antiplatelets* 1 (0) 2 (0) 0 –0.1 Angiotensin II receptor blockers (ARBs)* 432 (17.8) 928 (20.1) –2.3 –5.8 Aromatase inhibitors 52 (2.1) 74 (1.6) 0.5 4 Atypical antipsychotics 101 (4.2) 91 (2) 2.2 12.8 Beta-blockers* 954 (39.3) 1508 (32.6) 6.7 14 Biologic disease-modifying antirheumatic 4 (0.2) 14 (0.3) –0.1 –2.9 drugs (DMARDs)* Calcium channel blockers* 745 (30.7) 1195 (25.8) 4.9 10.8 Digoxin* 135 (5.6) 89 (1.9) 3.6 19.3 Heparin and low-molecular-weight heparin* 47 (1.9) 23 (0.5) 1.4 13.2 Lithium 6 (0.2) 10 (0.2) 0 0.6 Loop diuretics* 508 (20.9) 366 (7.9) 13 37.7 262

Nonselective nonsteroidal anti-inflammatory 415 (17.1) 872 (18.9) –1.8 –4.6 drugs (NSAIDs)* Opioids* 1056 (43.5) 1273 (27.5) 16 33.9 Oral anticoagulants* 318 (13.1) 166 (3.6) 9.5 34.9 Other anticoagulants* 3 (0.1) 3 (0.1) 0.1 1.9 Other antihypertensives 203 (8.4) 195 (4.2) 4.2 17.2 Other newer and atypical antidepressants* 213 (8.8) 269 (5.8) 3 11.4 Potassium-sparing diuretics 323 (13.3) 571 (12.3) 1 2.9 Renin inhibitor 5 (0.2) 11 (0.2) 0 -0.7 Selective COX-2 inhibitors* 101 (4.2) 237 (5.1) –1 –4.6 Selective estrogen receptor blockers 18 (0.7) 70 (1.5) –0.8 –7.3 Selective serotonin-norepinephrine reuptake 73 (3) 153 (3.3) –0.3 –1.7 inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 445 (18.3) 948 (20.5) –2.2 –5.5 (SSRIs)* Tricyclic antidepressants (TCAs)* 142 (5.9) 229 (5) 0.9 4 263

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Thiazide diuretics* 796 (32.8) 1828 (39.5) –6.7 –14 Typical antipsychotics* 11 (0.5) 19 (0.4) 0 0.6 No lipid-lowering drug despite diagnosis of 297 (12.2) 662 (14.3) –2.1 –6.1 hyperlipidemia* Long-term use of opioids* 513 (21.1) 843 (18.2) 2.9 7.3 Long-term current use of steroids* 16 (0.7) 19 (0.4) 0.2 3.4 Lack of 2nd RX among all the interested 9 (0.4) 14 (0.3) 0.1 1.2 drugs groups* Opioids baseline days’ supply ≥ 90* 281 (11.6) 214 (4.6) 7 25.7 Index days’ supply* 46 (26.3) 58 (29.9) –12.6 –44.6 Use of preventive services Colonoscopy*, n (%) 240 (9.9) 595 (12.9) –3 –9.4 Fecal occult blood test*, n (%) 303 (12.5) 943 (20.4) –7.9 –21.4 Mammography*, n (% of females) 790 (37.8) 2386 (51.9) –14.1 –39.3 Any prior lab test, n (%) 0 (0.4) 0 (0.4) –0.1 –25.6 Bone mineral density test*, n (%) 547 (22.5) 1453 (31.4) –8.9 –20.1 Electrocardiography, n (%) 841 (34.7) 1534 (33.2) 1.5 3.2 INR test, n (%) 1 (5.3) 0 (1.7) 1 25.6

264

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Use of wheelchair, walker, crutches, or cane, 64 (2.6) 43 (0.9) 1.7 12.9 n (%) Use of oxygen, n (%) 86 (3.5) 57 (1.2) 2.3 15.2 Lipid test, n (%) 1027 (42.3) 2611 (56.5) –14.1 –28.6 Health care utilization, mean (SD) Number of distinct cardiovascular diagnoses 3 (3.2) 2 (1.9) 0.9 34.1 Total days in hospital in prior year* 2 (12.3) 0 (2.6) 1.6 17.7 Medication synchronization metrics 1 (0.2) 1 (0.3) 0 4 Copayment 936 (983.7) 726 (699.6) 210 24.6 Number of office visits 19 (24.4) 13 (12.7) 5.2 26.9 Number of unique drugs of interest (by NDC) 6 (3.4) 5 (2.5) 0.9 30.9 in prior year or on index date Number of office visits with a cardiovascular 6 (12.7) 3 (4.2) 2.6 27.6 diagnosis Number of drugs of interest dispensations in 25 (19.7) 19 (15) 5.3 30.4 365 days in prior year or on index date* Number of hospitalizations in the prior year 0 (0.7) 0 (0.4) 0.1 24.3 Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 19.5 mail)* Number of cardiovascular hospitalizations 0 (0.5) 0 (0.3) 0.1 19.2 265

Appendix 35. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Index copayment 42 (41) 48 (48.9) –6.6 –14.6 Vaccinations* 0 (0.7) 0 (0.7) 0 –1.5 Physician visits* 8 (7.3) 7 (5.4) 0.9 13.4 Number of drugs of interest dispensations 7 (3.8) 6 (3) 1.8 53.3 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 days 8 (4.4) 6 (3.4) 2 50.6 before or on index date with days’ supply that overlap index date Number of dispensations in baseline period 4 (2.2) 4 (2) 0.4 17.1 with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 15 (9) 11 (6.7) 3.8 48.1 days before or on index date* Number of any drug dispensations in 365 53 (42.8) 35 (27.8) 17.9 49.6 days before or on index date*

266

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 68 (12.2) 62 (9.6) 5.3 48.8 Sex: female* 14746 (89) 48939 (99.5) –10.4 –46 Combined comorbidity score* 1 (2.2) 0 (1.1) 0.8 46.6 Commercial plan type, n (%) 8948 (54) 35977 (73.1) –19.1 –40.5 Benefit plan types, n (%) EPO* 778 (4.7) 3978 (8.1) –3.4 –13.9 HMO* 6075 (36.7) 11860 (24.1) 12.6 27.6 IND* 1625 (9.8) 3284 (6.7) 3.1 11.4 OTH* 2285 (13.8) 6014 (12.2) 1.6 4.7 POS* 4137 (25) 18584 (37.8) –12.8 –27.8 Comorbidities, n (%) Acute coronary syndrome, with or without 634 (3.8) 761 (1.5) 2.3 14.1 revascularization Acute coronary syndrome, with 175 (1.1) 220 (0.4) 0.6 7.1 revascularization Atrial fibrillation 222 (1.3) 146 (0.3) 1 11.6 Alzheimer’s disease* 1006 (6.1) 589 (1.2) 4.9 26.3

267

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Angina 523 (3.2) 683 (1.4) 1.8 11.9 Coronary atherosclerosis* 1765 (10.7) 1828 (3.7) 6.9 27.1 Coronary artery bypass graft (CABG), new 33 (0.2) 15 (0) 0.2 5 CABG, old 207 (1.3) 140 (0.3) 1 11.1 Any cancer 2741 (16.6) 6876 (14) 2.6 7.2 Any malignancy, including lymphoma and 2517 (15.2) 6305 (12.8) 2.4 6.9 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 1590 (9.6) 3048 (6.2) 3.4 12.7 Chest pain 2824 (17.1) 4540 (9.2) 7.8 23.3 Heart failure* 1268 (7.7) 805 (1.6) 6 28.9 Heart failure hospitalization 100 (0.6) 42 (0.1) 0.5 8.9 Conduction disorders 347 (2.1) 347 (0.7) 1.4 11.9 Chronic obstructive pulmonary disease 2332 (14.1) 2870 (5.8) 8.2 27.8 (COPD) Depression* 1883 (11.4) 3796 (7.7) 3.7 12.5 Diabetes* 2171 (13.1) 4109 (8.4) 4.8 15.4 Drug-induced osteoporosis* 249 (1.5) 473 (1) 0.5 4.9 Falls* 512 (3.1) 364 (0.7) 2.4 17.2

268

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 History of falls 449 (2.7) 361 (0.7) 2 15.2 HIV infection 14 (0.1) 28 (0.1) 0 1 Hyperlipidemia 6680 (40.3) 17011 (34.6) 5.8 11.9 Hyperparathyroidism 205 (1.2) 252 (0.5) 0.7 7.8 Hypertension* 6634 (40.1) 13323 (27.1) 13 27.7 Hyperthyroidism 252 (1.5) 613 (1.2) 0.3 2.4 Inflammatory bowel disease (IBD)* 235 (1.4) 305 (0.6) 0.8 8 Ischemic heart disease* 1869 (11.3) 1976 (4) 7.3 27.6 Liver disease* 749 (4.5) 1180 (2.4) 2.1 11.6 Metastatic cancer* 271 (1.6) 266 (0.5) 1.1 10.6 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 5356 (32.3) 10672 (21.7) 10.7 24.2 Other fractures* — — — — Palpitations 697 (4.2) 1651 (3.4) 0.9 4.5 Parkinson’s disease 197 (1.2) 146 (0.3) 0.9 10.4 Post-myocardial infarction/acute coronary 276 (1.7) 199 (0.4) 1.3 12.5 syndromes (MI/ACS)*

269

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Preventive care 7409 (44.7) 22977 (46.7) –2 –3.9 Prior MI 105 (0.6) 59 (0.1) 0.5 8.4 Prior stroke 79 (0.5) 33 (0.1) 0.4 7.9 Peripheral vascular disease (PVD) or PVD 1092 (6.6) 897 (1.8) 4.8 23.9 surgery Rheumatoid arthritis* 572 (3.5) 856 (1.7) 1.7 10.8 Recent MI 8 (0) 3 (0) 0 2.6 Recent stroke 10 (0.1) 8 (0) 0 2.3 Renal dysfunction 1713 (10.3) 1359 (2.8) 7.6 31 Schizophrenia 45 (0.3) 45 (0.1) 0.2 4.2 Transient ischemic attack* 358 (2.2) 319 (0.6) 1.5 12.9 Upper GI diseases* 3822 (23.1) 6815 (13.9) 9.2 23.9 Urinary tract infection 2440 (14.7) 4252 (8.6) 6.1 19 Baseline medication use, n (%) Systemic steroid* 1314 (7.9) 1065 (2.2) 5.8 26.6 Bile acid sequestrants 205 (1.2) 298 (0.6) 0.6 6.6 IV osteoclast inhibitors* 128 (0.8) 146 (0.3) 0.5 6.5 Inhaled glucocorticoid* 1250 (7.5) 1835 (3.7) 3.8 16.6

270

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Lipid-lowering agents 5367 (32.4) 13426 (27.3) 5.1 11.2 Niacin and fibrates 521 (3.1) 1304 (2.7) 0.5 3 Angiotensin-converting enzyme inhibitors 3449 (20.8) 6869 (14) 6.9 18.2 (ACEIs)* Agents for dementia 610 (3.7) 434 (0.9) 2.8 18.8 Antidiabetics* 1610 (9.7) 3276 (6.7) 3.1 11.2 Antiparkinson agents* 437 (2.6) 534 (1.1) 1.6 11.5 Antiplatelets* 9 (0.1) 9 (0) 0 1.9 Angiotensin II receptor blockers (ARBs)* 1940 (11.7) 4779 (9.7) 2 6.5 Aromatase inhibitors 191 (1.2) 393 (0.8) 0.4 3.6 Atypical antipsychotics 498 (3) 542 (1.1) 1.9 13.5 Beta-blockers* 4148 (25) 7282 (14.8) 10.2 25.9 Biologic disease-modifying antirheumatic 77 (0.5) 110 (0.2) 0.2 4.1 drugs (DMARDs)* Calcium channel blockers* 3116 (18.8) 5656 (11.5) 7.3 20.5 Digoxin* 534 (3.2) 462 (0.9) 2.3 16.1 Heparin and low-molecular-weight 182 (1.1) 120 (0.2) 0.9 10.5 heparin* Lithium 30 (0.2) 68 (0.1) 0 1.1

271

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Loop diuretics* 2077 (12.5) 1825 (3.7) 8.8 32.8 Monoamine oxidase inhibitors (MOAs) 1 (0) 5 (0) 0 -0.5 Nitrates* 911 (5.5) 700 (1.4) 4.1 22.4 Nonbiologic DMARDs 643 (3.9) 842 (1.7) 2.2 13.2

Nonselective nonsteroidal anti- 2754 (16.6) 6302 (12.8) 3.8 10.8 inflammatory drugs (NSAIDs)* Opioids* 6420 (38.8) 9427 (19.2) 19.6 44.3 Oral anticoagulants* 1074 (6.5) 662 (1.3) 5.1 26.7 Other anticoagulants* 12 (0.1) 10 (0) 0.1 2.4 Other antihypertensives 764 (4.6) 809 (1.6) 3 17.1 Other newer and atypical antidepressants* 1459 (8.8) 2340 (4.8) 4.1 16.2 Potassium-sparing diuretics 1088 (6.6) 2514 (5.1) 1.5 6.2 Renin inhibitor 26 (0.2) 51 (0.1) 0.1 1.5 Selective COX-2 inhibitors* 656 (4) 1684 (3.4) 0.5 2.9 Selective estrogen receptor blockers 97 (0.6) 550 (1.1) -0.5 -5.8 Selective serotonin-norepinephrine 704 (4.3) 1791 (3.6) 0.6 3.1 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 2766 (16.7) 6496 (13.2) 3.5 9.8 (SSRIs)*

272

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Tricyclic antidepressants (TCAs)* 866 (5.2) 1598 (3.2) 2 9.8 Thiazide diuretics* 3070 (18.5) 8679 (17.6) 0.9 2.3 Typical antipsychotics* 63 (0.4) 82 (0.2) 0.2 4.1 No lipid-lowering drug despite diagnosis of 2876 (17.4) 7831 (15.9) 1.4 3.9 hyperlipidemia* Long-term use of opioids* 2749 (16.6) 5442 (11.1) 5.5 16.1 Long-term current use of steroids* 123 (0.7) 99 (0.2) 0.5 7.9 Lack of 2nd RX among all the interested 170 (1) 456 (0.9) 0.1 1 drugs groups* Opioids baseline days’ supply ≥ 90* 1901 (11.5) 1544 (3.1) 8.3 32.5 Index days’ supply* 38 (20.9) 47 (26.8) –8.3 –34.5 Use of preventive services Colonoscopy*, n (%) 1388 (8.4) 3787 (7.7) 0.7 2.5 Fecal occult blood test*, n (%) 1646 (9.9) 6442 (13.1) –3.2 –9.9 Mammography*, n (% of females) 3936 (26.7) 14761 (30.2) –3.5 –14.1 Any prior lab test, n (%) 0 (0.4) 0 (0.4) –0.1 –13.3 Bone mineral density test*, n (%) 1755 (10.6) 5643 (11.5) –0.9 –2.8 Electrocardiography, n (%) 5126 (31) 10746 (21.8) 9.1 20.8 INR test, n (%) 1 (3.5) 0 (1) 0.5 19.1 273

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Use of wheelchair, walker, crutches, or 437 (2.6) 362 (0.7) 1.9 14.8 cane, n (%) Use of oxygen, n (%) 497 (3) 302 (0.6) 2.4 18 Lipid test, n (%) 6427 (38.8) 18591 (37.8) 1 2.1 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (3.2) 1 (1.7) 1.1 43.3 diagnoses Total days in hospital in prior year* 1 (7.3) 0 (2) 1.1 21.1 Medication synchronization metrics 0 (0.3) 0 (0.3) 0 2.6 Copayment 535 (752.1) 305 (472.9) 230.7 36.7 Number of office visits 15 (21.9) 9 (12.4) 6.1 34.3 Number of unique drugs of interest (by 4 (3.4) 3 (2.2) 1.2 42.8 NDC) in prior year or on index date Number of office visits with a 4 (9.3) 2 (3.9) 2.3 31.7 cardiovascular diagnosis Number of drugs of interest dispensations 12 (15.5) 7 (10.2) 5.3 40.5 in 365 days in prior year or on index date* Number of hospitalizations in the prior 0 (0.7) 0 (0.3) 0.2 27.6 year

274

Appendix 36. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 13.2 mail)* Number of cardiovascular hospitalizations 0 (0.5) 0 (0.2) 0.1 24.9 Index copayment 35 (30.4) 41 (38.9) –5.9 –17 Vaccinations* 0 (0.7) 0 (0.5) 0.1 14 Physician visits* 7 (7.5) 4 (5.5) 2.1 32.1 Number of drugs of interest dispensations 3 (1.9) 2 (1.6) 0.5 26.7 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 days 5 (3.4) 3 (2.4) 1.6 54.5 before or on index date with days’ supply that overlap index date Number of dispensations in baseline period 5 (3.7) 3 (2.6) 1.7 53.6 with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 11 (9.4) 7 (6.3) 4.7 58.7 days before or on index date* Number of any drug dispensations in 365 30 (33.4) 14 (20.2) 15.1 54.6 days before or on index date*

275

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 69 (12.9) 69 (11.7) 0.4 3.3 Sex: female* 2317 (87.4) 3933 (90.9) –3.6 –11.5 Combined comorbidity score* 2 (2.7) 2 (3) -0.2 -8.7 Commercial plan type, n (%) 901 (34) 1287 (29.8) 4.2 9 Benefit plan types, n (%) EPO* 111 (4.2) 162 (3.8) 0.4 2.2 HMO* 1158 (43.7) 1930 (44.6) –1 –1.9 IND* 68 (2.6) 72 (1.7) 0.9 6.3 OTH* 544 (20.5) 1002 (23.2) –2.7 –6.4 POS* 561 (21.2) 828 (19.1) 2 5 Comorbidities, n (%) Acute coronary syndrome, with or 221 (8.3) 433 (10) –1.7 –5.8 without revascularization Acute coronary syndrome, with 69 (2.6) 97 (2.2) 0.4 2.3 revascularization Atrial fibrillation 66 (2.5) 86 (2) 0.5 3.4 Alzheimer’s* 197 (7.4) 361 (8.4) –0.9 –3.4

276

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Angina 188 (7.1) 373 (8.6) –1.5 –5.7 Coronary atherosclerosis* 555 (20.9) 1101 (25.4) –4.5 –10.7 Coronary artery bypass graft (CABG), new 16 (0.6) 17 (0.4) 0.2 3.1 CABG, old 68 (2.6) 132 (3) –0.5 –2.9 Any cancer 546 (20.6) 926 (21.4) –0.8 –2 Any malignancy, including lymphoma and 500 (18.9) 881 (20.4) –1.5 –3.8 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 445 (16.8) 882 (20.4) –3.6 –9.3 Chest pain 817 (30.8) 1665 (38.5) –7.7 –16.2 Heart failure* 393 (14.8) 726 (16.8) –2 –5.4 Heart failure hospitalization 39 (1.5) 109 (2.5) –1.1 –7.6 Conduction disorders 99 (3.7) 265 (6.1) –2.4 –11 Chronic obstructive pulmonary disease 674 (25.4) 1366 (31.6) –6.2 –13.7 (COPD) Depression* 670 (25.3) 1245 (28.8) –3.5 –7.9 Diabetes* 665 (25.1) 1120 (25.9) –0.8 –1.9 Drug-induced osteoporosis* 55 (2.1) 94 (2.2) –0.1 –0.7 Falls* 163 (6.1) 283 (6.5) –0.4 –1.6

277

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 History of falls 146 (5.5) 346 (8) –2.5 –10 HIV infection 1 (0) 2 (0) 0 0 Hyperlipidemia 1541 (58.1) 2631 (60.8) –2.7 –5.6 Hyperparathyroidism 59 (2.2) 140 (3.2) –1 –6.2 Hypertension* 1823 (68.7) 3081 (71.2) –2.5 –5.4 Hyperthyroidism 53 (2) 181 (4.2) –2.2 –12.6 Inflammatory bowel disease (IBD)* 61 (2.3) 130 (3) –0.7 –4.4 Ischemic heart disease* 580 (21.9) 1126 (26) –4.2 –9.8 Liver disease* 212 (8) 294 (6.8) 1.2 4.5 Metastatic cancer* 44 (1.7) 46 (1.1) 0.6 5.2 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 967 (36.5) 1815 (42) –5.5 –11.3 Other fractures* — — — — Palpitations 209 (7.9) 379 (8.8) –0.9 -3.2 Parkinson’s disease 46 (1.7) 131 (3) -1.3 –8.5 Post-MI/ACS 89 (3.4) 149 (3.4) –0.1 –0.4 Preventive care 1493 (56.3) 2434 (56.3) 0 0

278

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Prior MI 32 (1.2) 38 (0.9) 0.3 3.2 Prior stroke 14 (0.5) 24 (0.6) 0 –0.4 Peripheral vascular disease (PVD) or PVD 294 (11.1) 585 (13.5) –2.4 –7.4 surgery Rheumatoid arthritis* 176 (6.6) 374 (8.7) –2 –7.6 Recent MI 2 (0.1) 1 (0) 0 1.9 Recent stroke 1 (0) 2 (0) 0 –0.3 Renal dysfunction 556 (21) 1014 (23.4) –2.5 –6 Schizophrenia 12 (0.5) 11 (0.3) 0.2 3.3 Transient ischemic attack* 129 (4.9) 313 (7.2) –2.4 –10 Upper GI diseases* 969 (36.5) 1732 (40) –3.5 –7.2 Urinary tract infection 644 (24.3) 1164 (26.9) –2.6 –6.1 Baseline medication use, n (%) Systemic steroid* 387 (14.6) 711 (16.4) –1.8 –5.1 Bile acid sequestrants 56 (2.1) 138 (3.2) –1.1 –6.7 IV osteoclast inhibitors* 27 (1) 48 (1.1) –0.1 –0.8 Inhaled glucocorticoid* 372 (14) 660 (15.3) –1.2 –3.5 Lipid-lowering agents 1101 (41.5) 1939 (44.8) –3.3 -6.7

279

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Niacin and fibrates 104 (3.9) 192 (4.4) -0.5 -2.5 Angiotensin-converting enzyme 867 (32.7) 1409 (32.6) 0.1 0.3 inhibitors (ACEIs)* Agents for dementia 125 (4.7) 152 (3.5) 1.2 6 Antidiabetics* 483 (18.2) 770 (17.8) 0.4 1.1 Antiparkinson agents* 137 (5.2) 394 (9.1) -3.9 -15.3 Antiplatelets* 4 (0.2) 3 (0.1) 0.1 2.2 Angiotensin II receptor blockers (ARBs)* 536 (20.2) 838 (19.4) 0.8 2.1 Aromatase inhibitors 26 (1) 45 (1.1) -0.1 -0.7 Atypical antipsychotics 129 (4.9) 265 (6.1) -1.3 -5.5 Beta-blockers* 1093 (41.2) 1831 (42.3) -1.1 -2.3 Biologic disease-modifying antirheumatic 31 (1.2) 54 (1.3) -0.1 -0.8 drugs (DMARDs)* Calcium channel blockers* 797 (30.1) 1234 (28.5) 1.5 3.4 Digoxin* 113 (4.3) 130 (3) 1.2 6.7 Heparin and low-molecular-weight 48 (1.8) 92 (2.1) -0.3 -2.3 heparin* Lithium 10 (0.4) 8 (0.2) 0.2 3.6 Loop diuretics* 588 (22.2) 949 (22) 0.2 0.5

280

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Monoamine oxidase inhibitors (MOAs) 0 (0.0) 0 (0.0) 0 0 Nitrates* 301 (11.3) 460 (10.6) 0.7 2.3 Nonbiologic DMARDs 165 (6.2) 350 (8.1) -1.9 -7.3

Nonselective nonsteroidal anti- 730 (27.5) 1170 (27) 0.5 1.1 inflammatory drugs (NSAIDs)* Opioids* 1725 (65) 2757 (63.7) 1.3 2.7 Oral anticoagulants* 240 (9) 527 (12.2) -3.1 -10.2 Other anticoagulants* 6 (0.2) 3 (0.1) 0.2 4.3 Other antihypertensives 211 (8) 285 (6.6) 1.4 5.3 Other newer and atypical 403 (15.2) 743 (17.2) -2 -5.4 antidepressants* Potassium-sparing diuretics 258 (9.7) 385 (8.9) 0.8 2.8 Renin inhibitor 10 (0.4) 11 (0.3) 0.1 2.2 Selective COX-2 inhibitors* 136 (5.1) 269 (6.2) -1.1 -4.7 Selective estrogen receptor blockers 6 (0.2) 15 (0.3) -0.1 -2.3 Selective serotonin-norepinephrine 255 (9.6) 399 (9.2) 0.4 1.3 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 928 (35) 1554 (35.9) -0.9 -2 (SSRIs)*

281

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Tricyclic antidepressants (TCAs)* 225 (8.5) 447 (10.3) -1.9 -6.4 Thiazide diuretics* 721 (27.2) 1127 (26) 1.1 2.6 Typical antipsychotics* 11 (0.4) 10 (0.2) 0.2 3.4 No lipid-lowering drug despite diagnosis 630 (23.8) 1024 (23.7) 0.1 0.2 of hyperlipidemia* Long-term use of opioids* 689 (26) 1299 (30) -4 -9 Long-term current use of steroids* 41 (1.5) 126 (2.9) -1.4 -9.2 Lack of 2nd RX among all the interested 144 (5.4) 212 (4.9) 0.5 2.4 drugs groups* Opioids baseline days’ supply ≥ 90* 737 (27.8) 1363 (31.5) -3.7 -8.2 Index days’ supply* 32 (10.2) 32 (10.3) -0.2 -2 Use of preventive services Colonoscopy*, n (%) 285 (10.7) 617 (14.3) -3.5 -10.7 Fecal occult blood test*, n (%) 263 (9.9) 427 (9.9) 0.1 0.2 Mammography*, n (% of females) 616 (26.6) 1059 (26.9) -0.3 -3 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 3.9 Bone mineral density test*, n (%) 84 (3.2) 129 (3) 0.2 1 Electrocardiography, n (%) 1294 (48.8) 2306 (53.3) -4.5 -9.1 INR test, n (%) 1 (4) 1 (3.2) 0.1 1.6 282

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Use of wheelchair, walker, crutches, or 142 (5.4) 191 (4.4) 0.9 4.3 cane, n (%) Use of oxygen, n (%) 140 (5.3) 181 (4.2) 1.1 5.2 Lipid test, n (%) 1453 (54.8) 2494 (57.7) -2.9 -5.8 Health care utilization, mean (SD) Number of distinct cardiovascular 4 (4.1) 4 (5) -0.7 -14.2 diagnoses Total days in hospital in prior year* 2 (6.1) 3 (8.3) -1 -13.7 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 2.1 Copayment 805 (885.8) 876 (1025.1) -71.1 -7.4 Number of office visits 24 (26.9) 27 (32.3) -3 -10.1 Number of unique drugs of interest (by 6 (3.6) 6 (3.7) -0.1 -2.4 NDC) in prior year or on index date Number of office visits with a 7 (11.7) 9 (16.5) -1.6 -11.1 cardiovascular diagnosis Number of drugs of interest 18 (14.3) 17 (13.3) 0.3 2.3 dispensations in 365 days in prior year or on index date* Number of hospitalizations in the prior 0 (0.9) 0 (1) -0.1 -6.5 year

283

Appendix 37. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 -0.9 (non-mail)* Number of cardiovascular 0 (0.7) 0 (0.7) 0 -2.2 hospitalizations Index copayment 31 (25.1) 32 (25) -0.1 -0.6 Vaccinations* 0 (0.8) 0 (0.8) 0 -4.8 Physician visits* 11 (9) 12 (8.6) -0.9 -9.8 Number of drugs of interest 3 (1.7) 3 (1.7) -0.1 -3.7 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 5 (3) 6 (3.5) -0.3 -9.2 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 6 (3.1) 6 (3.7) -0.3 -9.8 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 19 (10.7) 20 (11.6) -1 -9.1 days before or on index date* Number of any drug dispensations in 365 47 (32.9) 48 (33.4) -1.6 -4.9 days before or on index date*

284

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 71 (11.1) 71 (10.1) 0.2 2.3 Sex: female* 2283 (90.8) 4057 (92.6) –1.8 –6.6 Combined comorbidity score* 1 (2.3) 1 (2.3) 0 –0.4 Commercial plan type, n (%) 857 (34.1) 1472 (33.6) 0.5 1 Benefit plan types, n (%) EPO* 76 (3) 131 (3) 0 0.2 HMO* 1207 (48) 2087 (47.6) 0.4 0.7 IND* 113 (4.5) 208 (4.7) –0.2 –1.2 OTH* 471 (18.7) 819 (18.7) 0 0.1 POS* 476 (18.9) 814 (18.6) 0.3 0.9 Comorbidities, n (%) Acute coronary syndrome, with or without 128 (5.1) 204 (4.7) 0.4 2 revascularization Acute coronary syndrome, with 30 (1.2) 56 (1.3) –0.1 –0.7 revascularization Atrial fibrillation 43 (1.7) 69 (1.6) 0.1 1.1 Alzheimer’s disease* 206 (8.2) 301 (6.9) 1.3 5

285

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Angina 110 (4.4) 176 (4) 0.3 1.7 Coronary atherosclerosis* 392 (15.6) 675 (15.4) 0.2 0.5 Coronary artery bypass graft (CABG), new 3 (0.1) 6 (0.1) 0 –0.2 CABG, old 41 (1.6) 66 (1.5) 0.1 0.9 Any cancer 502 (20) 832 (19) 1 2.5 Any malignancy, including lymphoma and 463 (18.4) 770 (17.6) 0.8 2.2 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 287 (11.4) 463 (10.6) 0.8 2.7 Chest pain 556 (22.1) 990 (22.6) –0.5 –1.2 Heart failure* 265 (10.5) 458 (10.5) 0.1 0.3 Heart failure hospitalization 15 (0.6) 27 (0.6) 0 –0.2 Conduction disorders 69 (2.7) 157 (3.6) –0.8 –4.8 Chronic obstructive pulmonary disease 447 (17.8) 836 (19.1) –1.3 –3.4 (COPD) Depression* 360 (14.3) 663 (15.1) –0.8 –2.3 Diabetes* 496 (19.7) 888 (20.3) –0.5 –1.4 Drug-induced osteoporosis* 48 (1.9) 86 (2) –0.1 –0.4 Falls* 103 (4.1) 175 (4) 0.1 0.5

286

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 History of falls 85 (3.4) 132 (3) 0.4 2.1 HIV infection 1 (0) 0 (0) 0 2.6 Hyperlipidemia 1398 (55.6) 2430 (55.5) 0.1 0.3 Hyperparathyroidism 33 (1.3) 79 (1.8) –0.5 –4 Hypertension* 1667 (66.3) 2939 (67.1) –0.8 –1.7 Hyperthyroidism 50 (2) 96 (2.2) –0.2 –1.4 Inflammatory bowel disease (IBD)* 31 (1.2) 59 (1.3) –0.1 –1 Ischemic heart disease* 418 (16.6) 720 (16.4) 0.2 0.5 Liver disease* 140 (5.6) 274 (6.3) –0.7 –2.9 Metastatic cancer* 42 (1.7) 70 (1.6) 0.1 0.5 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1039 (41.3) 1871 (42.7) –1.4 –2.8 Other fractures* — — — — Palpitations 126 (5) 186 (4.3) 0.8 3.6 Parkinson’s disease 50 (2) 65 (1.5) 0.5 3.8 Post-myocardial infarction/acute coronary 59 (2.3) 93 (2.1) 0.2 1.5 syndromes (MI/ACS)*

287

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Preventive care 1484 (59) 2613 (59.6) –0.6 –1.3 Prior MI 17 (0.7) 13 (0.3) 0.4 5.6 Prior stroke 9 (0.4) 30 (0.7) –0.3 –4.4 Peripheral vascular disease (PVD) or PVD 225 (8.9) 349 (8) 1 3.5 surgery Rheumatoid arthritis* 105 (4.2) 209 (4.8) –0.6 –2.9 Recent MI 0 (0.0) 0 (0.0) 0 0 Recent stroke 0 (0.0) 0 (0.0) 0 0 Renal dysfunction 398 (15.8) 664 (15.1) 0.7 1.9 Schizophrenia 13 (0.5) 12 (0.3) 0.3 4.1 Transient ischemic attack* 76 (3) 115 (2.6) 0.4 2.4 Upper GI diseases* 695 (27.6) 1303 (29.7) –2.1 –4.6 Urinary tract infection 485 (19.3) 896 (20.5) –1.2 –3 Baseline medication use, n (%) Systemic steroid* 261 (10.4) 565 (12.9) –2.5 –7.9 Bile acid sequestrants 38 (1.5) 71 (1.6) –0.1 –0.9 IV osteoclast inhibitors* 28 (1.1) 33 (0.8) 0.4 3.7 Inhaled glucocorticoid* 201 (8) 328 (7.5) 0.5 1.9

288

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Lipid-lowering agents 1189 (47.3) 2028 (46.3) 1 2 Niacin and fibrates 110 (4.4) 196 (4.5) –0.1 –0.5 Angiotensin-converting enzyme inhibitors 870 (34.6) 1435 (32.7) 1.8 3.9 (ACEIs)* Agents for dementia 116 (4.6) 154 (3.5) 1.1 5.5 Antidiabetics* 384 (15.3) 649 (14.8) 0.5 1.3 Antiparkinson agents* 97 (3.9) 162 (3.7) 0.2 0.8 Antiplatelets* 0 (0.0) 0 (0.0) 0 0 Angiotensin II receptor blockers (ARBs)* 500 (19.9) 808 (18.4) 1.4 3.7 Aromatase inhibitors 32 (1.3) 49 (1.1) 0.2 1.4 Atypical antipsychotics 102 (4.1) 215 (4.9) –0.8 –4.1 Beta-blockers* 1074 (42.7) 1947 (44.5) –1.7 –3.5 Biologic disease-modifying antirheumatic 13 (0.5) 28 (0.6) –0.1 –1.6 drugs (DMARDs)* Calcium channel blockers* 787 (31.3) 1386 (31.6) –0.3 –0.7 Digoxin* 123 (4.9) 215 (4.9) 0 -0.1 Heparin and low-molecular-weight 40 (1.6) 95 (2.2) –0.6 –4.2 heparin* Lithium 4 (0.2) 5 (0.1) 0 1

289

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Loop diuretics* 454 (18.1) 765 (17.5) 0.6 1.6 Monoamine oxidase inhibitors (MOAs) 0 (0.0) 0 (0.0) 0 0 Nitrates* 186 (7.4) 288 (6.6) 0.8 3.3 Nonbiologic DMARDs 128 (5.1) 256 (5.8) –0.7 –3.3

Nonselective nonsteroidal anti- 514 (20.4) 905 (20.7) –0.2 –0.5 inflammatory drugs (NSAIDs)* Opioids* 1247 (49.6) 2155 (49.2) 0.4 0.8 Oral anticoagulants* 234 (9.3) 413 (9.4) –0.1 –0.5 Other anticoagulants* 1 (0) 1 (0) 0 1.1 Other antihypertensives 167 (6.6) 382 (8.7) –2.1 –7.8 Other newer and atypical 256 (10.2) 427 (9.7) 0.4 1.5 antidepressants* Potassium-sparing diuretics 247 (9.8) 429 (9.8) 0 0.1 Renin inhibitor 9 (0.4) 14 (0.3) 0 0.5 Selective COX-2 inhibitors* 101 (4) 202 (4.6) –0.6 –2.9 Selective estrogen receptor blockers 15 (0.6) 22 (0.5) 0.1 1.2 Selective serotonin-norepinephrine 108 (4.3) 238 (5.4) –1.1 –5.3 reuptake inhibitors (SNRIs)*

290

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 606 (24.1) 1134 (25.9) –1.8 –4.1 (SSRIs)* Tricyclic antidepressants (TCAs)* 164 (6.5) 355 (8.1) –1.6 –6.1 Thiazide diuretics* 803 (31.9) 1387 (31.7) 0.3 0.6 Typical antipsychotics* 14 (0.6) 34 (0.8) –0.2 –2.8 No lipid-lowering drug despite diagnosis 461 (18.3) 844 (19.3) –0.9 –2.4 of hyperlipidemia* Long-term use of opioids* 542 (21.6) 907 (20.7) 0.8 2.1 Long-term current use of steroids* 22 (0.9) 24 (0.6) 0.3 3.8 Lack of 2nd RX among all the interested 16 (0.6) 26 (0.6) 0 0.5 drugs groups* Opioids baseline days’ supply ≥ 90* 388 (15.4) 796 (18.2) –2.7 –7.4 Index days’ supply* 35 (17.3) 36 (17.7) –0.7 –3.8 Use of preventive services Colonoscopy*, n (%) 254 (10.1) 468 (10.7) –0.6 –1.9 Fecal occult blood test*, n (%) 294 (11.7) 573 (13.1) –1.4 –4.2 Mammography*, n (% of females) 719 (31.4) 1367 (33.7) –2.6 –5.7 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 –0.8 Bone mineral density test*, n (%) 213 (8.5) 427 (9.7) –1.3 –4.4

291

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 1055 (41.9) 1853 (42.3) –0.4 –0.7 International Normalized Ratio (INR) test, 1 (3.9) 1 (3.2) 0 1 n (%) Use of wheelchair, walker, crutches, or 76 (3) 141 (3.2) –0.2 –1.1 cane, n (%) Use of oxygen, n (%) 94 (3.7) 170 (3.9) –0.1 –0.8 Lipid test, n (%) 1287 (51.2) 2164 (49.4) 1.8 3.6 Health care utilization, mean (SD) Number of distinct cardiovascular 3 (3.3) 3 (3.4) –0.1 –1.5 diagnoses Total days in hospital in prior year* 1 (5.2) 2 (5.1) –0.2 –3.7 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –3.8 Copayment 779 (817.5) 803 (810.8) –23.6 –2.9 Number of office visits 19 (22.3) 20 (20.8) –0.8 –3.6 Number of unique drugs of interest (by 6 (3.3) 6 (3.4) 0 0.1 NDC) in prior year or on index date Number of office visits with a 6 (9.2) 6 (8.4) –0.3 –3.1 cardiovascular diagnosis Number of drugs of interest dispensations 22 (16.2) 22 (15.4) 0.5 3.4 in 365 days in prior year or on index date*

292

Appendix 38. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (1.1) 0 (0.8) 0 –3.3 year Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 –5 mail)* Number of cardiovascular hospitalizations 0 (0.5) 0 (0.5) 0 –2.4 Index copayment 35 (31.4) 35 (28.6) 0.4 1.2 Vaccinations* 0 (0.8) 0 (0.8) 0 1.6 Physician visits* 9 (7.5) 9 (7.4) –0.4 –5.5 Number of drugs of interest dispensations 4 (1.8) 4 (1.8) 0 –0.1 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 6 (3) 6 (3.5) –0.2 –6.8 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 6 (3.2) 7 (3.8) –0.2 –6.4 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 15 (8.1) 15 (9.5) –0.5 –5.8 days before or on index date* Number of any drug dispensations in 365 46 (31.6) 47 (34.3) –1 –3 days before or on index date*

293

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 71 (10.4) 70 (10.1) 0.7 6.4 Sex: female* 2057 (90.1) 4076 (89.1) 1 3.2 Combined comorbidity score* 1 (2.2) 1 (2.4) –0.2 –7.5 Commercial plan type, n (%) 1146 (50.2) 2425 (53) –2.8 –5.7 Benefit plan types, n (%) EPO* 74 (3.2) 142 (3.1) 0.1 0.7 HMO* 948 (41.5) 1769 (38.7) 2.8 5.8 IND* 328 (14.4) 780 (17.1) –2.7 –7.4 OTH* 275 (12) 536 (11.7) 0.3 1 POS* 432 (18.9) 959 (21) –2.1 –5.2 Comorbidities, n (%) Acute coronary syndrome, with or without 80 (3.5) 184 (4) –0.5 –2.7 revascularization Acute coronary syndrome, with 23 (1) 47 (1) 0 –0.2 revascularization Atrial fibrillation 28 (1.2) 51 (1.1) 0.1 0.9 Alzheimer’s disease* 179 (7.8) 372 (8.1) –0.3 –1.1

294

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Angina 65 (2.8) 164 (3.6) –0.7 –4.2 Coronary atherosclerosis* 256 (11.2) 518 (11.3) –0.1 –0.3 Coronary artery bypass graft (CABG), new 4 (0.2) 2 (0) 0.1 3.8 CABG, old 24 (1.1) 27 (0.6) 0.5 5 Any cancer 486 (21.3) 1000 (21.8) –0.6 –1.4 Any malignancy, including lymphoma and 442 (19.4) 925 (20.2) –0.9 –2.2 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 239 (10.5) 509 (11.1) –0.7 –2.1 Chest pain 376 (16.5) 775 (16.9) –0.5 –1.3 Heart failure* 192 (8.4) 448 (9.8) –1.4 –4.8 Heart failure hospitalization 7 (0.3) 8 (0.2) 0.1 2.7 Conduction disorders 50 (2.2) 92 (2) 0.2 1.2 Chronic obstructive pulmonary disease 312 (13.7) 833 (18.2) –4.6 –12.5 (COPD) Depression* 198 (8.7) 358 (7.8) 0.8 3.1 Diabetes* 394 (17.3) 772 (16.9) 0.4 1 Drug-induced osteoporosis* 45 (2) 86 (1.9) 0.1 0.6 Falls* 75 (3.3) 125 (2.7) 0.5 3.2

295

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 History of falls 69 (3) 125 (2.7) 0.3 1.8 HIV infection 3 (0.1) 2 (0.1) 0.1 2.6 Hyperlipidemia 1097 (48) 2310 (50.5) –2.5 –4.9 Hyperparathyroidism 42 (1.8) 112 (2.4) –0.6 –4.2 Hypertension* 1291 (56.5) 2575 (56.3) 0.2 0.5 Hyperthyroidism 44 (1.9) 97 (2.1) –0.2 –1.4 Inflammatory bowel disease (IBD)* 28 (1.2) 57 (1.2) 0 -0.2 Ischemic heart disease* 273 (12) 634 (13.9) –1.9 –5.7 Liver disease* 104 (4.6) 211 (4.6) –0.1 –0.2 Metastatic cancer* 55 (2.4) 197 (4.3) –1.9 –10.6 Osteoporotic fracture: non-vertebral* 0 (0.0) 0 (0.0) 0 0 Osteoporotic fracture: vertebral* 0 (0.0) 0 (0.0) 0 0 Osteoporosis* 914 (40) 1797 (39.3) 0.7 1.5 Other fractures* 0 (0.0) 0 (0.0) 0 0 Palpitations 102 (4.5) 204 (4.5) 0 0 Parkinson’s 42 (1.8) 104 (2.3) –0.4 –3.1 Post-myocardial infarction/acute coronary 38 (1.7) 52 (1.1) 0.5 4.5 syndromes (MI/ACS)*

296

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Preventive care 1245 (54.5) 2503 (54.7) –0.2 –0.4 Prior MI 9 (0.4) 10 (0.2) 0.2 3.3 Prior stroke 13 (0.6) 25 (0.6) 0 0.2 Peripheral vascular disease (PVD) or PVD 181 (7.9) 367 (8) –0.1 –0.4 surgery Rheumatoid arthritis* 85 (3.7) 185 (4) –0.3 –1.7 Recent MI // 0 (0.0) 0 0 Recent stroke 2 (0.1) 11 (0.3) –0.2 –4 Renal dysfunction 264 (11.6) 536 (11.7) –0.2 –0.5 Schizophrenia 11 (0.5) 15 (0.3) 0.1 2.3 Transient ischemic attack* 45 (2) 111 (2.4) –0.4 –3 Upper GI diseases* 556 (24.3) 1137 (24.9) –0.5 –1.2 Urinary tract infection 377 (16.5) 679 (14.8) 1.7 4.6 Baseline medication use, n (%) Systemic steroid* 211 (9.2) 451 (9.9) –0.6 –2.1 Bile acid sequestrants 39 (1.7) 69 (1.5) 0.2 1.5 IV osteoclast inhibitors* 26 (1.1) 141 (3.1) –1.9 –13.5 Inhaled glucocorticoid* 180 (7.9) 315 (6.9) 1 3.8

297

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Lipid-lowering agents 1160 (50.8) 2364 (51.7) –0.9 –1.8 Niacin and fibrates 130 (5.7) 251 (5.5) 0.2 0.9 Angiotensin-converting enzyme inhibitors 774 (33.9) 1550 (33.9) 0 0 (ACEIs)* Agents for dementia 110 (4.8) 240 (5.2) –0.4 –1.9 Antidiabetics* 326 (14.3) 640 (14) 0.3 0.8 Antiparkinson agents* 89 (3.9) 190 (4.2) –0.3 –1.3 Antiplatelets* 1 (0) 3 (0.1) 0 –0.7 Angiotensin II receptor blockers (ARBs)* 403 (17.6) 905 (19.8) –2.1 –5.5 Aromatase inhibitors 50 (2.2) 96 (2.1) 0.1 0.7 Atypical antipsychotics 84 (3.7) 124 (2.7) 1 5.5 Beta-blockers* 880 (38.5) 1740 (38) 0.5 1 Biologic disease-modifying antirheumatic 4 (0.2) 6 (0.1) 0 1.1 drugs (DMARDs)* Calcium channel blockers* 695 (30.4) 1333 (29.1) 1.3 2.8 Digoxin* 113 (4.9) 187 (4.1) 0.9 4.2 Heparin and low-molecular-weight 37 (1.6) 37 (0.8) 0.8 7.5 heparin* Lithium 5 (0.2) 9 (0.2) 0 0.5

298

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Loop diuretics* 432 (18.9) 921 (20.1) –1.2 –3.1 Monoamine oxidase inhibitors (MOAs) 1 (0) 2 (0) 0 0.5 Nitrates* 157 (6.9) 332 (7.3) –0.4 –1.5 Nonbiologic DMARDs 137 (6) 291 (6.4) –0.4 –1.5

Nonselective nonsteroidal anti- 406 (17.8) 790 (17.3) 0.5 1.3 inflammatory drugs (NSAIDs)* Opioids* 957 (41.9) 1771 (38.7) 3.2 6.5 Oral anticoagulants* 257 (11.3) 671 (14.7) –3.4 –10.2 Other anticoagulants* 2 (0.1) 1 (0) 0.1 2.6 Other antihypertensives 170 (7.4) 382 (8.4) –0.9 –3.4 Other newer and atypical 192 (8.4) 364 (8) 0.5 1.7 antidepressants* Potassium-sparing diuretics 298 (13) 665 (14.5) –1.5 –4.3 Renin inhibitor 5 (0.2) 9 (0.2) 0 0.4 Selective COX-2 inhibitors* 97 (4.2) 189 (4.1) 0.1 0.5 Selective estrogen receptor blockers 18 (0.8) 32 (0.7) 0.1 1 Selective serotonin-norepinephrine 70 (3.1) 192 (4.2) –1.1 –6.1 reuptake inhibitors (SNRIs)*

299

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 417 (18.3) 746 (16.3) 1.9 5.1 (SSRIs)* Tricyclic antidepressants (TCAs)* 133 (5.8) 232 (5.1) 0.7 3.3 Thiazide diuretics* 763 (33.4) 1522 (33.3) 0.1 0.3 Typical antipsychotics* 10 (0.4) 11 (0.2) 0.2 3.4 No lipid-lowering drug despite diagnosis 285 (12.5) 648 (14.2) –1.7 –5 of hyperlipidemia* Long-term use of opioids* 466 (20.4) 1077 (23.5) –3.1 –7.6 Long-term current use of steroids* 12 (0.5) 25 (0.6) 0 -0.4 Lack of 2nd RX among all the interested 8 (0.4) 13 (0.3) 0.1 1.3 drugs groups* Opioids baseline days’ supply ≥ 90* 257 (11.3) 460 (10.1) 1.2 3.9 Index days’ supply* 46 (26.7) 48 (27.6) –2.2 –7.9 Use of preventive services Colonoscopy*, n (%) 224 (9.8) 421 (9.2) 0.6 2.1 Fecal occult blood test*, n (%) 290 (12.7) 560 (12.2) 0.5 1.4 Mammography*, n (% of females) 788 (38.3) 1619 (39.7) –1.4 –1.9 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 -9 Bone mineral density test*, n (%) 535 (23.4) 1043 (22.8) 0.6 1.5

300

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 764 (33.5) 1589 (34.7) –1.3 –2.7 INR test, n (%) 1 (3.7) 1 (3.6) –0.1 –3.9 Use of wheelchair, walker, crutches, or 54 (2.4) 87 (1.9) 0.5 3.2 cane, n (%) Use of oxygen, n (%) 70 (3.1) 110 (2.4) 0.7 4 Lipid test, n (%) 986 (43.2) 2040 (44.6) –1.4 –2.9 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (2.8) 2 (2.6) –0.1 –3.6 diagnoses Total days in hospital in prior year* 1 (6.5) 1 (5.1) 0 0.6 Medication synchronization metrics 1 (0.2) 1 (0.2) 0 0.1 Copayment 918 (972.3) 971 (938.5) –52.7 –5.5 Number of office visits 17 (20.4) 19 (24.9) –2.1 –9.3 Number of unique drugs of interest (by 5 (3.1) 5 (3.2) 0 –1 NDC) in prior year or on index date Number of office visits with a 4 (7.7) 4 (6) 0.1 2.2 cardiovascular diagnosis Number of drugs of interest dispensations 24 (18.8) 25 (19.1) –0.7 –3.6 in 365 days in prior year or on index date*

301

Appendix 39. Characteristics of calcitonin vs raloxifene initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (0.6) 0 (0.5) 0 1.7 year Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 –0.2 mail)* Number of cardiovascular hospitalizations 0 (0.5) 0 (0.4) 0 2.3 Index copayment 42 (40.9) 45 (46.4) –3.2 –7.4 Vaccinations* 0 (0.7) 0 (0.6) 0 1.5 Physician visits* 7 (7) 8 (6.7) –0.4 –6.1 Number of drugs of interest dispensations 7 (3.6) 7 (4.1) –0.1 –2.8 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 8 (4.1) 8 (4.6) –0.1 –2 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 4 (2.1) 4 (2.2) 0 –0.8 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 14 (8.4) 15 (9.3) –0.2 –1.9 days before or on index date* Number of any drug dispensations in 365 51 (40.9) 51 (40.3) –0.3 –0.7 days before or on index date*

302

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Calcitonin Raloxifene Absolute Standardized Difference Differences Age, mean (SD)* 71 (11.6) 71 (10.6) 0 0.1 Sex: female* 6720 (87.3) 11900 (89.2) –1.9 –5.9 Combined comorbidity score* 2 (2.5) 2 (2.7) –0.1 –5.1 Commercial plan type, n (%) 2984 (38.8) 5032 (37.7) 1.1 2.2 Benefit plan types, n (%) EPO* 262 (3.4) 423 (3.2) 0.2 1.3 HMO* 3451 (44.9) 6095 (45.7) –0.9 –1.7 IND* 545 (7.1) 1020 (7.7) -0.6 -2.2 OTH* 1320 (17.2) 2218 (16.6) 0.5 1.4 POS* 1496 (19.4) 2498 (18.7) 0.7 1.8 Comorbidities, n (%) Acute coronary syndrome, with or without 457 (5.9) 802 (6) –0.1 –0.3 revascularization Acute coronary syndrome, with 126 (1.6) 211 (1.6) 0.1 0.4 revascularization Atrial fibrillation 160 (2.1) 303 (2.3) –0.2 –1.3 Alzheimer’s disease* 651 (8.5) 999 (7.5) 1 3.6

303

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Angina 380 (4.9) 668 (5) –0.1 –0.3 Coronary atherosclerosis* 1295 (16.8) 2546 (19.1) –2.3 –5.9 Coronary artery bypass graft (CABG), new 24 (0.3) 28 (0.2) 0.1 1.9 CABG, old 159 (2.1) 210 (1.6) 0.5 3.7 Any cancer 1619 (21) 2990 (22.4) –1.4 –3.4 Any malignancy, including lymphoma and 1481 (19.2) 2804 (21) –1.8 –4.4 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 1030 (13.4) 1744 (13.1) 0.3 0.9 Chest pain 1839 (23.9) 3563 (26.7) –2.8 –6.5 Heart failure* 942 (12.2) 1733 (13) –0.8 –2.3 Heart failure hospitalization 70 (0.9) 112 (0.8) 0.1 0.7 Conduction disorders 250 (3.2) 479 (3.6) –0.3 –1.9 Chronic obstructive pulmonary disease 1526 (19.8) 3098 (23.2) –3.4 –8.3 (COPD) Depression* 1279 (16.6) 2457 (18.4) –1.8 –4.7 Diabetes* 1633 (21.2) 2841 (21.3) –0.1 –0.2 Drug-induced osteoporosis* 154 (2) 256 (1.9) 0.1 0.6 Falls* 369 (4.8) 643 (4.8) 0 –0.1

304

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 History of falls 320 (4.2) 556 (4.2) 0 0 HIV infection 7 (0.1) 7 (0.1) 0 1.5 Hyperlipidemia 4145 (53.9) 7283 (54.6) –0.7 –1.5 Hyperparathyroidism 138 (1.8) 289 (2.2) –0.4 –2.7 Hypertension* 4959 (64.5) 8756 (65.7) –1.2 –2.5 Hyperthyroidism 150 (1.9) 293 (2.2) –0.3 –1.8 Inflammatory bowel disease (IBD)* 126 (1.6) 256 (1.9) –0.3 –2.1 Ischemic heart disease* 1367 (17.8) 2694 (20.2) –2.4 –6.2 Liver disease* 488 (6.3) 794 (6) 0.4 1.6 Metastatic cancer* 161 (2.1) 355 (2.7) –0.6 –3.7 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 2986 (38.8) 5624 (42.2) –3.4 –6.9 Other fractures* — — — — Palpitations 446 (5.8) 688 (5.2) 0.6 2.8 Parkinson’s disease 153 (2) 222 (1.7) 0.3 2.4 Post-myocardial infarction/acute coronary 205 (2.7) 398 (3) –0.3 –1.9 syndromes (MI/ACS)*

305

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Preventive care 4307 (56) 7397 (55.5) 0.5 1 Prior MI 70 (0.9) 77 (0.6) 0.3 3.8 Prior stroke 47 (0.6) 100 (0.7) –0.1 –1.7 Peripheral vascular disease (PVD) or PVD 775 (10.1) 1365 (10.2) –0.2 –0.5 surgery Rheumatoid arthritis* 377 (4.9) 893 (6.7) –1.8 –7.7 Recent MI 5 (0.1) 3 (0) 0 2.2 Recent stroke 6 (0.1) 9 (0.1) 0 0.5 Renal dysfunction 1337 (17.4) 2304 (17.3) 0.1 0.3 Schizophrenia 39 (0.5) 37 (0.3) 0.2 3.6 Transient ischemic attack* 267 (3.5) 422 (3.2) 0.3 1.7 Upper GI diseases* 2303 (29.9) 4160 (31.2) –1.3 –2.7 Urinary tract infection 1575 (20.5) 3153 (23.6) –3.2 –7.7 Baseline medication use, n (%) Systemic steroid* 932 (12.1) 1909 (14.3) –2.2 –6.5 Bile acid sequestrants 140 (1.8) 262 (2) –0.1 –1.1 IV osteoclast inhibitors* 87 (1.1) 185 (1.4) –0.3 –2.3 Inhaled glucocorticoid* 793 (10.3) 1432 (10.7) –0.4 –1.4

306

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Lipid-lowering agents 3580 (46.5) 6308 (47.3) –0.8 –1.6 Niacin and fibrates 356 (4.6) 590 (4.4) 0.2 1 Angiotensin-converting enzyme inhibitors 2606 (33.9) 4577 (34.3) –0.5 –1 (ACEIs)* Agents for dementia 376 (4.9) 595 (4.5) 0.4 2 Antidiabetics* 1251 (16.3) 2152 (16.1) 0.1 0.3 Antiparkinson agents* 339 (4.4) 675 (5.1) –0.7 –3.1 Antiplatelets* 5 (0.1) 8 (0.1) 0 0.2 Angiotensin II receptor blockers (ARBs)* 1489 (19.4) 2511 (18.8) 0.5 1.3 Aromatase inhibitors 111 (1.4) 193 (1.4) 0 0 Atypical antipsychotics 339 (4.4) 536 (4) 0.4 1.9 Beta-blockers* 3170 (41.2) 5684 (42.6) –1.4 –2.9 Biologic disease-modifying antirheumatic 49 (0.6) 92 (0.7) –0.1 –0.7 drugs (DMARDs)* Calcium channel blockers* 2364 (30.7) 4003 (30) 0.7 1.5 Digoxin* 384 (5) 681 (5.1) –0.1 –0.5 Heparin and low-molecular-weight heparin* 141 (1.8) 208 (1.6) 0.3 2.1 Lithium 20 (0.3) 23 (0.2) 0.1 1.8 Loop diuretics* 1596 (20.7) 2907 (21.8) –1.1 –2.6 307

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Monoamine oxidase inhibitors (MOAs) 1 (0) 2 (0) 0 0 Nitrates* 687 (8.9) 1126 (8.4) 0.5 1.7 Nonbiologic DMARDs 453 (5.9) 1009 (7.6) –1.7 –6.7

Nonselective nonsteroidal anti- 1671 (21.7) 2921 (21.9) –0.2 –0.5 inflammatory drugs (NSAIDs)* Opioids* 4105 (53.4) 6797 (51) 2.4 4.8 Oral anticoagulants* 812 (10.6) 1801 (13.5) –3 –9.1 Other anticoagulants* 10 (0.1) 6 (0) 0.1 3 Other antihypertensives 586 (7.6) 1076 (8.1) –0.5 –1.7 Other newer and atypical antidepressants* 895 (11.6) 1507 (11.3) 0.3 1 Potassium-sparing diuretics 835 (10.9) 1525 (11.4) –0.6 –1.8 Renin inhibitor 24 (0.3) 35 (0.3) 0.1 1 Selective COX-2 inhibitors* 343 (4.5) 658 (4.9) –0.5 –2.3 Selective estrogen receptor blockers 40 (0.5) 69 (0.5) 0 0.1 Selective serotonin-norepinephrine 443 (5.8) 786 (5.9) –0.1 –0.6 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 2007 (26.1) 3628 (27.2) –1.1 –2.5 (SSRIs)* Tricyclic antidepressants (TCAs)* 538 (7) 994 (7.5) –0.5 –1.8 308

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Thiazide diuretics* 2337 (30.4) 4126 (30.9) –0.6 –1.2 Typical antipsychotics* 39 (0.5) 41 (0.3) 0.2 3.2 No lipid-lowering drug despite diagnosis of 1403 (18.2) 2450 (18.4) –0.1 –0.4 hyperlipidemia* Long-term use of opioids* 1785 (23.2) 3162 (23.7) –0.5 –1.2 Long-term current use of steroids* 81 (1.1) 182 (1.4) –0.3 –2.8 Lack of 2nd RX among all the interested 170 (2.2) 267 (2) 0.2 1.4 drugs groups* Opioids baseline days’ supply ≥ 90* 1435 (18.7) 2762 (20.7) –2.1 –5.2 Index days’ supply* 37 (19.7) 38 (20.2) –0.7 –3.6 Use of preventive services Colonoscopy*, n (%) 789 (10.3) 1515 (11.4) –1.1 –3.6 Fecal occult blood test*, n (%) 862 (11.2) 1472 (11) 0.2 0.5 Mammography*, n (% of females) 2129 (31.7) 3856 (32.4) –0.7 –2.8 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 –0.2 Bone mineral density test*, n (%) 849 (11) 1478 (11.1) 0 –0.1 Electrocardiography, n (%) 3257 (42.3) 5962 (44.7) –2.4 –4.8 INR test, n (%) 1 (4.5) 1 (3.6) 0 0.2

309

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Use of wheelchair, walker, crutches, or 299 (3.9) 472 (3.5) 0.4 1.9 cane, n (%) Use of oxygen, n (%) 338 (4.4) 585 (4.4) 0 0 Lipid test, n (%) 3811 (49.5) 6369 (47.8) 1.8 3.5 Health care utilization, mean (SD) Number of distinct cardiovascular diagnoses 3 (3.7) 3 (3.9) –0.2 –5.2 Total days in hospital in prior year* 2 (8.4) 2 (6.3) 0 –0.1 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –0.5 Copayment 839 (902.1) 895 (968.5) –56.2 –6 Number of office visits 21 (25) 23 (29) –1.8 –6.5 Number of unique drugs of interest (by 6 (3.5) 6 (3.5) –0.1 –3.5 NDC) in prior year or on index date Number of office visits with a cardiovascular 6 (11.4) 6 (11.2) –0.2 –1.8 diagnosis Number of drugs of interest dispensations 21 (17.1) 21 (16.2) –0.1 –0.4 in 365 days in prior year or on index date* Number of hospitalizations in the prior year 0 (0.9) 0 (0.8) 0 –1.3 Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 –1.4 mail)* Number of cardiovascular hospitalizations 0 (0.6) 0 (0.6) 0 –1.5 310

Appendix 40. Characteristics of calcitonin vs raloxifene initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Index copayment 36 (33.1) 36 (31.9) –0.6 –1.8 Vaccinations* 0 (0.8) 0 (0.7) 0 1.5 Physician visits* 9 (8.2) 10 (8.3) –0.7 –8.7 Number of drugs of interest dispensations 4 (2) 4 (2) –0.1 –3.8 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 days 6 (3.4) 6 (3.8) –0.1 –4.1 before or on index date with days’ supply that overlap index date Number of dispensations in baseline period 7 (3.7) 7 (4.2) –0.2 –4.5 with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 16 (9.6) 17 (10.7) –0.6 –5.8 days before or on index date* Number of any drug dispensations in 365 49 (36.2) 49 (37.4) –0.8 –2 days before or on index date*

311

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 63 (10.7) 66 (11.5) –3.2 –29 Sex: female* 4503 (99.4) 65359 (91.2) 8.2 39.3 Combined comorbidity score* 1 (1.8) 1 (2) –0.3 –16.1 Commercial plan type, n (%) 2659 (58.7) 30438 (42.5) 16.2 32.9 Benefit plan types, n (%) EPO* 396 (8.7) 4722 (6.6) 2.2 8.1 HMO* 1151 (25.4) 25590 (35.7) –10.3 –22.5 IND* 42 (0.9) 947 (1.3) –0.4 –3.7 OTH* 793 (17.5) 14924 (20.8) –3.3 –8.4 POS* 1833 (40.4) 20453 (28.5) 11.9 25.3 Comorbidities, n (%) Acute coronary syndrome, with or without 215 (4.7) 4019 (5.6) –0.9 –3.9 revascularization Acute coronary syndrome, with 73 (1.6) 1128 (1.6) 0 0.3 revascularization Atrial fibrillation 45 (1) 880 (1.2) –0.2 –2.2 Alzheimer’s disease* 83 (1.8) 2526 (3.5) –1.7 –10.5

312

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Angina 192 (4.2) 3448 (4.8) –0.6 –2.8 Coronary atherosclerosis* 455 (10) 10476 (14.6) –4.6 –13.9 Coronary artery bypass graft (CABG), new 6 (0.1) 211 (0.3) –0.2 –3.5 CABG, old 33 (0.7) 1136 (1.6) –0.9 –8 Any cancer 1039 (22.9) 12284 (17.1) 5.8 14.5 Any malignancy, including lymphoma and 967 (21.3) 11431 (15.9) 5.4 13.9 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 642 (14.2) 10031 (14) 0.2 0.5 Chest pain 950 (21) 16546 (23.1) –2.1 –5.1 Heart failure* 213 (4.7) 5438 (7.6) –2.9 –12 Heart failure hospitalization 18 (0.4) 350 (0.5) –0.1 –1.4 Conduction disorders 85 (1.9) 1761 (2.5) –0.6 –4 Chronic obstructive pulmonary disease 558 (12.3) 13126 (18.3) –6 –16.7 (COPD) Depression* 1062 (23.4) 16626 (23.2) 0.2 0.6 Diabetes* 967 (21.3) 17538 (24.5) –3.1 –7.4 Drug-induced osteoporosis* 89 (2) 1515 (2.1) –0.1 –1.1 Falls* 93 (2.1) 2156 (3) –1 –6.1

313

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 History of falls 85 (1.9) 1826 (2.5) –0.7 –4.6 HIV infection 6 (0.1) 80 (0.1) 0 0.6 Hyperlipidemia 2736 (60.4) 44998 (62.8) –2.4 –4.9 Hyperparathyroidism 48 (1.1) 1173 (1.6) –0.6 –5 Hypertension* 2798 (61.7) 45877 (64) –2.3 –4.7 Hyperthyroidism 99 (2.2) 1631 (2.3) –0.1 –0.6 Inflammatory bowel disease (IBD)* 67 (1.5) 1334 (1.9) –0.4 –3 Ischemic heart disease* 482 (10.6) 11058 (15.4) –4.8 –14.3 Liver disease* 311 (6.9) 4686 (6.5) 0.3 1.3 Metastatic cancer* 38 (0.8) 568 (0.8) 0 0.5 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1496 (33) 32763 (45.7) –12.7 –26.2 Other fractures* — — — — Palpitations 346 (7.6) 4717 (6.6) 1.1 4.1 Parkinson’s disease 28 (0.6) 550 (0.8) –0.1 –1.8 Post-myocardial infarction/acute coronary 51 (1.1) 1797 (2.5) –1.4 –10.4 syndromes (MI/ACS)*

314

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Preventive care 3163 (69.8) 48230 (67.3) 2.5 5.4 Prior MI 19 (0.4) 500 (0.7) –0.3 –3.7 Prior stroke 9 (0.2) 273 (0.4) –0.2 –3.4 Peripheral vascular disease (PVD) or PVD 207 (4.6) 5181 (7.2) –2.7 –11.3 surgery Rheumatoid arthritis* 237 (5.2) 5834 (8.1) –2.9 –11.7 Recent MI 1 (0) 19 (0) 0 -0.3 Recent stroke 4 (0.1) 27 (0) 0.1 2 Renal dysfunction 399 (8.8) 8909 (12.4) –3.6 –11.8 Schizophrenia 15 (0.3) 203 (0.3) 0 0.9 Transient ischemic attack* 69 (1.5) 1861 (2.6) –1.1 –7.6 Upper GI diseases* 1378 (30.4) 18666 (26) 4.4 9.7 Urinary tract infection 809 (17.9) 13372 (18.7) –0.8 –2.1 Baseline medication use, n (%) Systemic steroid* 303 (6.7) 10039 (14) –7.3 –24.2 Bile acid sequestrants 66 (1.5) 1170 (1.6) –0.2 –1.4 IV osteoclast inhibitors* 19 (0.4) 426 (0.6) –0.2 –2.5 Inhaled glucocorticoid* 407 (9) 7703 (10.7) –1.8 –5.9

315

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Lipid-lowering agents 1718 (37.9) 34781 (48.5) –10.6 –21.5 Niacin and fibrates 149 (3.3) 3099 (4.3) –1 –5.4 Angiotensin-converting enzyme inhibitors 1198 (26.4) 22432 (31.3) –4.9 –10.7 (ACEIs)* Agents for dementia 56 (1.2) 1878 (2.6) –1.4 –10.1 Antidiabetics* 722 (15.9) 12995 (18.1) –2.2 –5.8 Antiparkinson agents* 164 (3.6) 2614 (3.6) 0 -0.1 Antiplatelets* 2 (0) 133 (0.2) –0.1 –4.2 Angiotensin II receptor blockers (ARBs)* 951 (21) 13348 (18.6) 2.4 5.9 Aromatase inhibitors 84 (1.9) 1405 (2) –0.1 –0.8 Atypical antipsychotics 170 (3.8) 2235 (3.1) 0.6 3.5 Beta-blockers* 1323 (29.2) 22721 (31.7) –2.5 –5.4 Biologic disease-modifying antirheumatic 30 (0.7) 1187 (1.7) –1 –9.3 drugs (DMARDs)* Calcium channel blockers* 979 (21.6) 18386 (25.6) –4 –9.5 Digoxin* 67 (1.5) 1433 (2) –0.5 –4 Heparin and low-molecular-weight 33 (0.7) 717 (1) –0.3 –2.9 heparin* Lithium 18 (0.4) 210 (0.3) 0.1 1.8

316

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Loop diuretics* 424 (9.4) 9030 (12.6) –3.2 –10.4 Monoamine oxidase inhibitors (MOAs) 0 (0) 6 (0) 0 -1.3 Nitrates* 174 (3.8) 4198 (5.9) –2 –9.4 Nonbiologic DMARDs 214 (4.7) 5831 (8.1) –3.4 –13.9

Nonselective nonsteroidal anti- 1271 (28) 20717 (28.9) –0.9 –1.9 inflammatory drugs (NSAIDs)* Opioids* 2098 (46.3) 34784 (48.5) –2.2 –4.5 Oral anticoagulants* 133 (2.9) 3738 (5.2) –2.3 –11.5 Other anticoagulants* 2 (0) 38 (0.1) 0 –0.4 Other antihypertensives 163 (3.6) 3274 (4.6) –1 –4.9 Other newer and atypical antidepressants* 564 (12.4) 8667 (12.1) 0.4 1.1 Potassium-sparing diuretics 401 (8.8) 5518 (7.7) 1.2 4.2 Renin inhibitor 22 (0.5) 180 (0.3) 0.2 3.9 Selective COX-2 inhibitors* 209 (4.6) 3374 (4.7) –0.1 –0.4 Selective estrogen receptor blockers 126 (2.8) 404 (0.6) 2.2 17.4 Selective serotonin-norepinephrine 464 (10.2) 6393 (8.9) 1.3 4.5 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 1760 (38.8) 23428 (32.7) 6.2 12.9 (SSRIs)* 317

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Tricyclic antidepressants (TCAs)* 280 (6.2) 4181 (5.8) 0.3 1.5 Thiazide diuretics* 1481 (32.7) 21357 (29.8) 2.9 6.2 Typical antipsychotics* 17 (0.4) 259 (0.4) 0 0.2 No lipid-lowering drug despite diagnosis of 1240 (27.4) 15349 (21.4) 6 13.9 hyperlipidemia* Long-term use of opioids* 1005 (22.2) 16728 (23.3) –1.2 –2.8 Long-term current use of steroids* 25 (0.6) 1642 (2.3) –1.7 –14.7 Lack of 2nd RX among all the interested 389 (8.6) 5226 (7.3) 1.3 4.8 drugs groups* Opioids baseline days’ supply ≥ 90* 593 (13.1) 12553 (17.5) –4.4 –12.3 Index days’ supply* 35 (15.9) 36 (18.8) –1 –5.7 Use of preventive services Colonoscopy*, n (%) 520 (11.5) 7580 (10.6) 0.9 2.9 Fecal occult blood test*, n (%) 683 (15.1) 8612 (12) 3.1 8.9 Mammography*, n (% of females) 1704 (37.8) 23250 (35.6) 2.2 10.9 Any prior lab test, n (%) 0 (0.5) 0 (0.4) 0.1 21.5 Bone mineral density test*, n (%) 200 (4.4) 2716 (3.8) 0.6 3.2 Electrocardiography, n (%) 1933 (42.7) 31889 (44.5) –1.8 –3.7 INR test, n (%) 0 (2) 0 (2.7) –0.2 –9.2 318

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Use of wheelchair, walker, crutches, or 79 (1.7) 2134 (3) –1.2 –8.1 cane, n (%) Use of oxygen, n (%) 67 (1.5) 2309 (3.2) –1.7 –11.5 Lipid test, n (%) 2888 (63.7) 46486 (64.8) –1.1 –2.3 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (2.7) 3 (3.2) –0.5 –17.7 diagnoses Total days in hospital in prior year* 1 (3.6) 1 (4.5) –0.4 –9.6 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –4.7 Copayment 617 (640.4) 595 (730.4) 22.2 3.2 Number of office visits 19 (17.9) 18 (19.8) 0.4 1.9 Number of unique drugs of interest (by 5 (2.9) 4 (3.3) 0.5 16 NDC) in prior year or on index date Number of office visits with a 4 (7) 5 (9.1) –0.8 –9.7 cardiovascular diagnosis Number of drugs of interest dispensations 15 (12.1) 15 (12.9) –0.2 –1.7 in 365 days in prior year or on index date* Number of hospitalizations in the prior 0 (0.5) 0 (0.7) –0.1 –11.5 year

319

Appendix 41. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 1 before propensity score fine stratification and weighting in Aim 2 Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 –7.2 mail)* Number of cardiovascular hospitalizations 0 (0.4) 0 (0.5) 0 –10 Index copayment 39 (38.7) 20 (26.9) 19.3 58 Vaccinations* 0 (0.7) 0 (0.8) 0 –6 Physician visits* 10 (7.1) 10 (7.4) 0.1 1.8 Number of drugs of interest dispensations 3 (1.6) 2 (1.7) 0.8 51.5 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 days 4 (2.6) 4 (2.7) –0.4 –15.7 before or on index date with days’ supply that overlap index date Number of dispensations in baseline period 4 (2.7) 5 (2.9) –0.4 –15.3 with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 14 (8.6) 15 (9.1) –0.9 –9.9 days before or on index date* Number of any drug dispensations in 365 33 (25.2) 36 (27.2) –2.9 –11.1 days before or on index date*

320

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 66 (10.3) 68 (10.9) –2.8 –26.4 Sex: female* 4299 (99.3) 64963 (90.4) 9 41.3 Combined comorbidity score* 0 (1.5) 1 (1.8) –0.3 –19.5 Commercial plan type, n (%) 2291 (52.9) 27950 (38.9) 14 28.5 Benefit plan types, n (%) EPO* 281 (6.5) 3450 (4.8) 1.7 7.3 HMO* 1287 (29.7) 29252 (40.7) –11 –23.1 IND* 132 (3) 2017 (2.8) 0.2 1.4 OTH* 853 (19.7) 14957 (20.8) –1.1 –2.7 POS* 1421 (32.8) 17072 (23.7) 9.1 20.3 Comorbidities, n (%) Acute coronary syndrome, with or 130 (3) 2656 (3.7) –0.7 –3.8 without revascularization Acute coronary syndrome, with 32 (0.7) 679 (0.9) –0.2 –2.2 revascularization Atrial fibrillation 23 (0.5) 729 (1) –0.5 –5.5 Alzheimer’s disease* 91 (2.1) 2841 (4) –1.8 –10.8

321

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Angina 117 (2.7) 2276 (3.2) –0.5 –2.7 Coronary atherosclerosis* 329 (7.6) 7820 (10.9) –3.3 –11.3 Coronary artery bypass graft (CABG), 1 (0) 89 (0.1) –0.1 –3.7 new CABG, old 15 (0.3) 886 (1.2) –0.9 –10 Any cancer 965 (22.3) 13781 (19.2) 3.1 7.7 Any malignancy, including lymphoma 892 (20.6) 12778 (17.8) 2.8 7.2 and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 464 (10.7) 7656 (10.6) 0.1 0.2 Chest pain 659 (15.2) 11492 (16) –0.8 –2.1 Heart failure* 160 (3.7) 3968 (5.5) –1.8 –8.7 Heart failure hospitalization 5 (0.1) 243 (0.3) –0.2 –4.7 Conduction disorders 61 (1.4) 1471 (2) –0.6 –4.9 Chronic obstructive pulmonary disease 429 (9.9) 9653 (13.4) –3.5 –11 (COPD) Depression* 601 (13.9) 9235 (12.8) 1 3.1 Diabetes* 818 (18.9) 14476 (20.1) –1.2 –3.1 Drug-induced osteoporosis* 65 (1.5) 1406 (2) –0.5 –3.5

322

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Falls* 63 (1.5) 1651 (2.3) –0.8 –6.2 History of falls 59 (1.4) 1430 (2) –0.6 –4.9 HIV infection 2 (0) 77 (0.1) –0.1 –2.2 Hyperlipidemia 2571 (59.4) 45135 (62.8) –3.4 –7 Hyperparathyroidism 40 (0.9) 1105 (1.5) –0.6 –5.6 Hypertension* 2763 (63.8) 44506 (61.9) 1.9 4 Hyperthyroidism 98 (2.3) 1458 (2) 0.2 1.6 Inflammatory bowel disease (IBD)* 47 (1.1) 759 (1.1) 0 0.3 Ischemic heart disease* 353 (8.2) 8333 (11.6) –3.4 –11.5 Liver disease* 174 (4) 3007 (4.2) –0.2 –0.8 Metastatic cancer* 35 (0.8) 768 (1.1) –0.3 –2.7 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1563 (36.1) 33502 (46.6) –10.5 –21.4 Other fractures* — — — — Palpitations 256 (5.9) 3535 (4.9) 1 4.4 Parkinson’s disease 27 (0.6) 569 (0.8) –0.2 –2

323

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 35 (0.8) 1350 (1.9) –1.1 –9.3 coronary syndromes (MI/ACS)* Preventive care 3072 (71) 49791 (69.3) 1.7 3.7 Prior MI 8 (0.2) 291 (0.4) –0.2 –4.1 Prior stroke 1 (0) 231 (0.3) –0.3 –7.2 Peripheral vascular disease (PVD) or PVD 163 (3.8) 4326 (6) –2.3 –10.5 surgery Rheumatoid arthritis* 127 (2.9) 3520 (4.9) –2 –10.1 Recent MI 0 (0) 19 (0) 0 -2.3 Recent stroke 0 (0) 16 (0) 0 –2.1 Renal dysfunction 292 (6.7) 7614 (10.6) –3.8 –13.7 Schizophrenia 12 (0.3) 207 (0.3) 0 -0.2 Transient ischemic attack* 60 (1.4) 1458 (2) –0.6 –5 Upper GI diseases* 1013 (23.4) 13442 (18.7) 4.7 11.6 Urinary tract infection 667 (15.4) 10324 (14.4) 1 2.9 Baseline medication use, n (%) Systemic steroid* 181 (4.2) 6729 (9.4) –5.2 –20.7 Bile acid sequestrants 47 (1.1) 882 (1.2) –0.1 –1.3 IV osteoclast inhibitors* 32 (0.7) 549 (0.8) 0 –0.3 324

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 252 (5.8) 5262 (7.3) –1.5 –6 Lipid-lowering agents 2022 (46.7) 40658 (56.6) –9.8 –19.8 Niacin and fibrates 198 (4.6) 3681 (5.1) –0.5 –2.5 Angiotensin-converting enzyme 1269 (29.3) 23710 (33) –3.7 –7.9 inhibitors (ACEIs)* Agents for dementia 60 (1.4) 1999 (2.8) –1.4 –9.8 Antidiabetics* 617 (14.3) 11005 (15.3) –1.1 –3 Antiparkinson agents* 103 (2.4) 1825 (2.5) –0.2 –1 Antiplatelets* 1 (0) 91 (0.1) –0.1 –3.8 Angiotensin II receptor blockers (ARBs)* 959 (22.2) 12521 (17.4) 4.7 11.9 Aromatase inhibitors 82 (1.9) 1804 (2.5) –0.6 –4.2 Atypical antipsychotics 107 (2.5) 1579 (2.2) 0.3 1.8 Beta-blockers* 1362 (31.5) 22477 (31.3) 0.2 0.4 Biologic disease-modifying 16 (0.4) 453 (0.6) –0.3 –3.7 antirheumatic drugs (DMARDs)* Calcium channel blockers* 1148 (26.5) 18722 (26) 0.5 1.1 Digoxin* 82 (1.9) 1616 (2.2) –0.4 –2.5 Heparin and low-molecular-weight 21 (0.5) 625 (0.9) –0.4 -4.7 heparin*

325

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Lithium 11 (0.3) 153 (0.2) 0 0.9 Loop diuretics* 320 (7.4) 7080 (9.8) –2.5 –8.8 Monoamine oxidase inhibitors (MOAs) 1 (0) 1 (0) 0 2 Nitrates* 131 (3) 3026 (4.2) –1.2 –6.3 Nonbiologic DMARDs 143 (3.3) 4225 (5.9) –2.6 –12.3

Nonselective nonsteroidal anti- 968 (22.4) 15323 (21.3) 1 2.5 inflammatory drugs (NSAIDs)* Opioids* 1512 (34.9) 25419 (35.4) –0.4 –0.9 Oral anticoagulants* 129 (3) 4277 (5.9) –3 –14.4 Other anticoagulants* 2 (0) 38 (0.1) 0 –0.3 Other antihypertensives 138 (3.2) 3015 (4.2) –1 –5.3 Other newer and atypical 327 (7.6) 5446 (7.6) 0 -0.1 antidepressants* Potassium-sparing diuretics 431 (10) 5926 (8.2) 1.7 6 Renin inhibitor 11 (0.3) 124 (0.2) 0.1 1.8 Selective COX-2 inhibitors* 180 (4.2) 2767 (3.8) 0.3 1.6 Selective estrogen receptor blockers 90 (2.1) 495 (0.7) 1.4 11.9 Selective serotonin-norepinephrine 202 (4.7) 2820 (3.9) 0.7 3.7 reuptake inhibitors (SNRIs)* 326

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 1141 (26.4) 14936 (20.8) 5.6 13.2 (SSRIs)* Tricyclic antidepressants (TCAs)* 211 (4.9) 3124 (4.3) 0.5 2.5 Thiazide diuretics* 1571 (36.3) 22747 (31.6) 4.7 9.8 Typical antipsychotics* 6 (0.1) 246 (0.3) –0.2 –4.2 No lipid-lowering drug despite diagnosis 835 (19.3) 10976 (15.3) 4 10.7 of hyperlipidemia* Long-term use of opioids* 868 (20.1) 14833 (20.6) –0.6 –1.4 Long-term current use of steroids* 13 (0.3) 1185 (1.6) –1.3 –13.8 Lack of 2nd RX among all the interested 53 (1.2) 1018 (1.4) –0.2 –1.7 drugs groups* Opioids baseline days’ supply ≥ 90* 274 (6.3) 6296 (8.8) –2.4 –9.2 Index days’ supply* 43 (24.5) 44 (25.2) –1.1 –4.4 Use of preventive services Colonoscopy*, n (%) 514 (11.9) 7636 (10.6) 1.3 4 Fecal occult blood test*, n (%) 813 (18.8) 10160 (14.1) 4.6 12.6 Mammography*, n (% of females) 1957 (45.5) 27181 (41.8) 3.7 15.1 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0.1 20.6 Bone mineral density test*, n (%) 497 (11.5) 6846 (9.5) 2 6.4

327

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 1695 (39.2) 27006 (37.6) 1.6 3.3 INR test, n (%) 0 (1.7) 1 (3) –0.3 –13.3 Use of wheelchair, walker, crutches, or 73 (1.7) 1410 (2) –0.3 –2.1 cane, n (%) Use of oxygen, n (%) 36 (0.8) 1575 (2.2) –1.4 –11.2 Lipid test, n (%) 2779 (64.2) 45623 (63.5) 0.7 1.5 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (2.2) 2 (2.7) –0.3 –14.1 diagnoses Total days in hospital in prior year* 1 (2.4) 1 (3.9) –0.2 –7.3 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 1.4 Copayment 652 (645.1) 624 (700.9) 27.6 4.1 Number of office visits 15 (15) 15 (16.5) 0.4 2.6 Number of unique drugs of interest (by 5 (2.7) 4 (3) 0.7 24.8 NDC) in prior year or on index date Number of office visits with a 4 (5.7) 4 (7.6) –0.4 –5.7 cardiovascular diagnosis Number of drugs of interest 19 (13.5) 18 (14.4) 0.4 2.9 dispensations in 365 days in prior year or on index date*

328

Appendix 42. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 2 before propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (0.4) 0 (0.5) 0 –5.7 year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –0.2 (non-mail)* Number of cardiovascular 0 (0.3) 0 (0.4) 0 –6.2 hospitalizations Index copayment 44 (50.4) 25 (32.8) 18.9 44.5 Vaccinations* 0 (0.7) 0 (0.8) –0.1 –7.8 Physician visits* 8 (6) 8 (6.3) 0.1 1.1 Number of drugs of interest 3 (1.7) 2 (1.8) 0.9 52.9 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 5 (2.5) 5 (2.8) –0.3 –12.7 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 5 (2.7) 5 (2.9) –0.3 –11.5 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 12 (6.9) 12 (7.3) –0.4 –6 days before or on index date* Number of any drug dispensations in 34 (25.2) 36 (26.6) –1.9 –7.3 365 days before or on index date*

329

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 66 (9.7) 69 (10.5) –2.7 –26.6 Sex: female* 4530 (99.4) 63411 (88.5) 10.9 46.8 Combined comorbidity score* 0 (1.5) 1 (1.8) –0.3 –19.8 Commercial plan type, n (%) 2829 (62.1) 35322 (49.3) 12.8 25.9 Benefit plan types, n (%) EPO* 283 (6.2) 3002 (4.2) 2 9.1 HMO* 1374 (30.1) 28553 (39.8) –9.7 –20.5 IND* 401 (8.8) 6364 (8.9) –0.1 –0.3 OTH* 609 (13.4) 10504 (14.7) –1.3 –3.7 POS* 1444 (31.7) 17231 (24) 7.6 17.1 Comorbidities, n (%) Acute coronary syndrome, with or 102 (2.2) 2076 (2.9) –0.7 –4.2 without revascularization Acute coronary syndrome, with 34 (0.7) 497 (0.7) 0.1 0.6 revascularization Atrial fibrillation 18 (0.4) 675 (0.9) –0.5 –6.7

330

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Alzheimer’s disease* 108 (2.4) 2983 (4.2) –1.8 –10.1 Angina 90 (2) 1777 (2.5) –0.5 –3.4 Coronary atherosclerosis* 277 (6.1) 6138 (8.6) –2.5 –9.6 Coronary artery bypass graft (CABG), new 2 (0) 60 (0.1) 0 –1.6 CABG, old 25 (0.1) 265 (0.2) –0.1 –2.3 Any cancer 983 (21.6) 15759 (22) –0.4 –1 Any malignancy, including lymphoma and 1608 (6.8) 14496 (10.6) –3.8 –13.4 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 603 (2.6) 5547 (4) –1.5 –8.3 Chest pain 620 (13.6) 9160 (12.8) 0.8 2.4 Heart failure* 141 (0.6) 1562 (1.1) –0.5 –5.8 Heart failure hospitalization 2 (0) 237 (0.3) –0.3 –6.6 Conduction disorders 53 (1.2) 1239 (1.7) –0.6 –4.7 Chronic obstructive pulmonary disease 581 (2.5) 5688 (4.2) –1.7 –9.4 (COPD) Depression* 431 (9.5) 5803 (8.1) 1.4 4.8 Diabetes* 719 (15.8) 11896 (16.6) –0.8 –2.2 Drug-induced osteoporosis* 63 (1.4) 1421 (2) –0.6 –4.7

331

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Falls* 53 (1.2) 1330 (1.9) –0.7 –5.7 History of falls 65 (0.3) 818 (0.6) –0.3 –4.9 HIV infection 1 (0) 106 (0.1) –0.1 –4.3 Hyperlipidemia 3883 (16.5) 33867 (24.7) –8.2 –20.5 Hyperparathyroidism 39 (0.9) 1016 (1.4) –0.6 –5.3 Hypertension* 2677 (58.7) 39404 (55) 3.7 7.5 Hyperthyroidism 95 (2.1) 1349 (1.9) 0.2 1.4 Inflammatory bowel disease (IBD)* 34 (0.7) 583 (0.8) –0.1 –0.8 Ischemic heart disease* 303 (6.6) 6645 (9.3) –2.6 –9.7 Liver disease* 128 (2.8) 2258 (3.2) –0.3 –2 Metastatic cancer* 55 (1.2) 1136 (1.6) –0.4 –3.2 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1472 (32.3) 32179 (44.9) –12.6 –26.1 Other fractures* — — — — Palpitations 210 (4.6) 2816 (3.9) 0.7 3.3 Parkinson’s disease 30 (0.7) 574 (0.8) –0.1 –1.7

332

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 33 (0.7) 1063 (1.5) –0.8 –7.3 coronary syndromes (MI/ACS)* Preventive care 3155 (69.2) 48767 (68) 1.2 2.5 Prior MI 7 (0.2) 203 (0.3) –0.1 –2.8 Prior stroke 9 (0.2) 173 (0.2) 0 –0.9 Peripheral vascular disease (PVD) or PVD 139 (3) 3616 (5) –2 –10.1 surgery Rheumatoid arthritis* 127 (2.8) 2696 (3.8) –1 –5.5 Recent MI 0 (0) 13 (0) 0 –1.9 Recent stroke 2 (0) 19 (0) 0 0.9 Renal dysfunction 259 (5.7) 5806 (8.1) –2.4 –9.6 Schizophrenia 5 (0.1) 198 (0.3) –0.2 –3.8 Transient ischemic attack* 45 (1) 1150 (1.6) –0.6 –5.5 Upper GI diseases* 877 (19.2) 10612 (14.8) 4.4 11.8 Urinary tract infection 574 (12.6) 8520 (11.9) 0.7 2.1 Baseline medication use, n (%) Systemic steroid* 184 (4) 5966 (8.3) –4.3 –17.9 Bile acid sequestrants 47 (1) 820 (1.1) –0.1 –1.1 IV osteoclast inhibitors* 26 (0.6) 672 (0.9) –0.4 –4.2 333

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 267 (5.9) 4484 (6.3) –0.4 –1.7 Lipid-lowering agents 2355 (51.7) 44399 (62) –10.3 –20.9 Niacin and fibrates 300 (6.6) 4686 (6.5) 0 0.2 Angiotensin-converting enzyme 1466 (32.2) 23166 (32.3) –0.2 –0.4 inhibitors (ACEIs)* Agents for dementia 81 (1.8) 2070 (2.9) –1.1 –7.4 Antidiabetics* 575 (12.6) 9799 (13.7) –1.1 –3.1 Antiparkinson agents* 89 (2) 1549 (2.2) –0.2 –1.5 Antiplatelets* 2 (0) 58 (0.1) 0 -1.5 Angiotensin II receptor blockers (ARBs)* 914 (20) 11902 (16.6) 3.4 8.9 Aromatase inhibitors 72 (1.6) 2196 (3.1) –1.5 –9.9 Atypical antipsychotics 90 (2) 1498 (2.1) –0.1 –0.8 Beta-blockers* 1483 (32.5) 21934 (30.6) 1.9 4.1 Biologic disease-modifying antirheumatic 14 (0.3) 279 (0.4) –0.1 –1.4 drugs (DMARDs)* Calcium channel blockers* 1173 (25.7) 18184 (25.4) 0.4 0.8 Digoxin* 88 (1.9) 2041 (2.8) –0.9 –6 Heparin and low-molecular-weight 23 (0.5) 693 (1) –0.5 –5.4 heparin*

334

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Lithium 10 (0.2) 174 (0.2) 0 –0.5 Loop diuretics* 362 (7.9) 7146 (10) –2 –7.1 Monoamine oxidase inhibitors (MOAs) 1 (0) 3 (0) 0 1.6 Nitrates* 133 (2.9) 2535 (3.5) –0.6 –3.5 Nonbiologic DMARDs 176 (3.9) 4146 (5.8) –1.9 –9

Nonselective nonsteroidal anti- 863 (18.9) 12415 (17.3) 1.6 4.2 inflammatory drugs (NSAIDs)* Opioids* 1252 (27.5) 20074 (28) –0.5 –1.2 Oral anticoagulants* 165 (3.6) 5095 (7.1) –3.5 –15.5 Other anticoagulants* 3 (0.1) 46 (0.1) 0 0.1 Other antihypertensives 194 (4.3) 3827 (5.3) –1.1 –5.1 Other newer and atypical 261 (5.7) 4218 (5.9) –0.2 –0.7 antidepressants* Potassium-sparing diuretics 566 (12.4) 7257 (10.1) 2.3 7.2 Renin inhibitor 11 (0.2) 109 (0.2) 0.1 2 Selective COX-2 inhibitors* 236 (5.2) 3097 (4.3) 0.9 4 Selective estrogen receptor blockers 70 (1.5) 737 (1) 0.5 4.5 Selective serotonin-norepinephrine 149 (3.3) 1884 (2.6) 0.6 3.8 reuptake inhibitors (SNRIs)* 335

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 934 (20.5) 11190 (15.6) 4.9 12.7 (SSRIs)* Tricyclic antidepressants (TCAs)* 228 (5) 3230 (4.5) 0.5 2.3 Thiazide diuretics* 1810 (39.7) 23518 (32.8) 6.9 14.4 Typical antipsychotics* 18 (0.4) 303 (0.4) 0 –0.4 No lipid-lowering drug despite diagnosis 653 (14.3) 7534 (10.5) 3.8 11.6 of hyperlipidemia* Long-term use of opioids* 830 (18.2) 14681 (20.5) –2.3 –5.8 Long-term current use of steroids* 19 (0.4) 1117 (1.6) –1.1 –11.6 Lack of 2nd RX among all the interested 14 (0.3) 279 (0.4) –0.1 –1.4 drugs groups* Opioids baseline days’ supply ≥ 90* 210 (4.6) 3883 (5.4) –0.8 –3.7 Index days’ supply* 58 (29.9) 57 (28.9) 1 3.6 Use of preventive services Colonoscopy*, n (%) 1007 (4.3) 8423 (6.1) –1.9 –8.4 Fecal occult blood test*, n (%) 932 (20.4) 12079 (16.9) 3.6 9.2 Mammography*, n (% of females) 2360 (52.1) 32231 (50.8) 1.3 13.6 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0.1 17.3 Bone mineral density test*, n (%) 1441 (31.6) 20668 (28.8) 2.8 6

336

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 2300 (9.8) 20964 (15.3) –5.5 –16.8 INR test, n (%) 0 (1.7) 1 (3.1) –0.4 –14.5 Use of wheelchair, walker, crutches, or 42 (0.9) 1123 (1.6) –0.6 –5.8 cane, n (%) Use of oxygen, n (%) 56 (1.2) 1404 (2) –0.7 –5.8 Lipid test, n (%) 2570 (56.4) 41073 (57.3) –0.9 –1.9 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (1.9) 2 (2.5) –0.3 –12.3 diagnoses Total days in hospital in prior year* 0 (2.6) 1 (5.1) –0.3 –7 Medication synchronization metrics 1 (0.3) 1 (0.3) 0 5.2 Copayment 724 (695.5) 738 (807.5) –14.2 –1.9 Number of office visits 13 (12.8) 14 (16.2) –0.6 –4.3 Number of unique drugs of interest (by 5 (2.5) 4 (2.9) 0.8 27.8 NDC) in prior year or on index date Number of office visits with a 3 (4.1) 4 (7.7) –0.6 –9.6 cardiovascular diagnosis Number of drugs of interest 19 (15) 19 (16.6) –0.2 –1.1 dispensations in 365 days in prior year or on index date*

337

Appendix 43. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 before propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (0.4) 0 (0.5) 0 –7 year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 2.3 (non-mail)* Number of cardiovascular 0 (0.3) 0 (0.4) 0 -6 hospitalizations Index copayment 48 (47.4) 35 (38.7) 13.2 30.5 Vaccinations* 0 (0.7) 0 (0.8) –0.1 –10.4 Physician visits* 7 (5.4) 7 (6) –0.3 –4.7 Number of drugs of interest 4 (2) 3 (2.1) 1 48.3 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 6 (3) 6 (3.2) –0.2 –8.1 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 6 (3.4) 6 (3.7) –0.3 –7.8 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 11 (6.7) 12 (7.2) –0.5 –7 days before or on index date* Number of any drug dispensations in 365 35 (27.7) 38 (31) –3.4 –11.5 days before or on index date*

338

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 62 (9.6) 65 (11.4) –2.4 –22.3 Sex: female* 48939 (99.5) 396290 (91.6) 7.9 38.8 Combined comorbidity score* 0 (1.1) 1 (1.6) –0.3 –22.1 Commercial plan type, n (%) 35977 (73.1) 250028 (57.8) 15.3 32.7 Benefit plan types, n (%) EPO* 3978 (8.1) 28577 (6.6) 1.5 5.7 HMO* 11860 (24.1) 140558 (32.5) –8.4 –18.7 IND* 3284 (6.7) 23354 (5.4) 1.3 5.4 OTH* 6014 (12.2) 63927 (14.8) –2.6 –7.5 POS* 18584 (37.8) 137715 (31.8) 5.9 12.5 Comorbidities, n (%) Acute coronary syndrome, with or 761 (1.5) 11331 (2.6) –1.1 –7.5 without revascularization Acute coronary syndrome, with 220 (0.4) 2952 (0.7) –0.2 –3.1 revascularization Atrial fibrillation 146 (0.3) 2916 (0.7) –0.4 –5.4

339

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Alzheimer’s disease* 589 (1.2) 11814 (2.7) –1.5 –11.1 Angina 683 (1.4) 9756 (2.3) –0.9 –6.5 Coronary atherosclerosis* 1828 (3.7) 30722 (7.1) –3.4 –15 Coronary artery bypass graft 15 (0) 439 (0.1) –0.1 –2.8 (CABG), new CABG, old 140 (0.3) 3377 (0.8) –0.5 –6.8 Any cancer 6876 (14) 70745 (16.4) –2.4 –6.6 Any malignancy, including 6305 (12.8) 65555 (15.2) –2.3 –6.7 lymphoma and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 3048 (6.2) 36729 (8.5) –2.3 –8.8 Chest pain 4540 (9.2) 55997 (12.9) –3.7 –11.9 Heart failure* 805 (1.6) 15892 (3.7) –2 –12.7 Heart failure hospitalization 42 (0.1) 1035 (0.2) –0.2 –3.8 Conduction disorders 347 (0.7) 5904 (1.4) –0.7 –6.5 Chronic obstructive pulmonary 2870 (5.8) 45198 (10.4) –4.6 –16.9 disease (COPD) Depression* 3796 (7.7) 44868 (10.4) –2.7 –9.3 Diabetes* 4109 (8.4) 54699 (12.6) –4.3 –14

340

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Drug-induced osteoporosis* 473 (1) 7349 (1.7) –0.7 –6.4 Falls* 364 (0.7) 7012 (1.6) –0.9 –8.2 History of falls 361 (0.7) 6141 (1.4) –0.7 –6.6 HIV infection 28 (0.1) 431 (0.1) 0 –1.5 Hyperlipidemia 17011 (34.6) 192966 (44.6) –10 –20.6 Hyperparathyroidism 252 (0.5) 4611 (1.1) –0.6 –6.3 Hypertension* 13323 (27.1) 166342 (38.5) –11.4 –24.4 Hyperthyroidism 613 (1.2) 7027 (1.6) –0.4 –3.2 Inflammatory bowel disease (IBD)* 305 (0.6) 4573 (1.1) –0.4 –4.8 Ischemic heart disease* 1976 (4) 32935 (7.6) –3.6 –15.4 Liver disease* 1180 (2.4) 15192 (3.5) –1.1 –6.6 Metastatic cancer* 266 (0.5) 4026 (0.9) –0.4 –4.6 Osteoporotic fracture: non- — — — — vertebral* Osteoporotic fracture: vertebral* — — — — Osteoporosis* 10672 (21.7) 168354 (38.9) –17.2 –38.2 Other fractures* — — — — Palpitations 1651 (3.4) 17320 (4) –0.6 –3.4

341

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Parkinson’s disease 146 (0.3) 2171 (0.5) –0.2 –3.3 Post-myocardial infarction/acute 199 (0.4) 5081 (1.2) –0.8 –8.7 coronary syndromes (MI/ACS)* Preventive care 22977 (46.7) 251630 (58.2) –11.5 –23.1 Prior MI 59 (0.1) 1226 (0.3) –0.2 –3.6 Prior stroke 33 (0.1) 929 (0.2) –0.1 –3.9 Peripheral vascular disease (PVD) 897 (1.8) 17108 (4) –2.1 –12.8 or PVD surgery Rheumatoid arthritis* 856 (1.7) 16684 (3.9) –2.1 –12.9 Recent MI 3 (0) 84 (0) 0 –1.2 Recent stroke 8 (0) 102 (0) 0 –0.5 Renal dysfunction 1359 (2.8) 27037 (6.3) –3.5 –16.9 Schizophrenia 45 (0.1) 763 (0.2) –0.1 –2.3 Transient ischemic attack* 319 (0.6) 6033 (1.4) –0.7 –7.4 Upper GI diseases* 6815 (13.9) 64280 (14.9) –1 –2.9 Urinary tract infection 4252 (8.6) 49957 (11.5) –2.9 –9.7 Baseline medication use, n (%) Systemic steroid* 1065 (2.2) 32017 (7.4) –5.2 –24.7 Bile acid sequestrants 298 (0.6) 3470 (0.8) –0.2 –2.4 342

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 IV osteoclast inhibitors* 146 (0.3) 2269 (0.5) –0.2 –3.6 Inhaled glucocorticoid* 1835 (3.7) 26073 (6) –2.3 –10.7 Lipid-lowering agents 13426 (27.3) 150603 (34.8) –7.5 –16.3 Niacin and fibrates 1304 (2.7) 13925 (3.2) –0.6 –3.4 Angiotensin-converting enzyme 6869 (14) 86061 (19.9) –5.9 –15.9 inhibitors (ACEIs)* Agents for dementia 434 (0.9) 8331 (1.9) –1 –8.9 Antidiabetics* 3276 (6.7) 40513 (9.4) –2.7 –10 Antiparkinson agents* 534 (1.1) 7287 (1.7) –0.6 –5.1 Antiplatelets* 9 (0) 347 (0.1) –0.1 –2.8 Angiotensin II receptor blockers 4779 (9.7) 46798 (10.8) –1.1 –3.6 (ARBs)* Aromatase inhibitors 393 (0.8) 8948 (2.1) –1.3 –10.7 Atypical antipsychotics 542 (1.1) 6527 (1.5) –0.4 –3.6 Beta-blockers* 7282 (14.8) 82942 (19.2) –4.4 –11.7 Biologic disease-modifying 110 (0.2) 2677 (0.6) –0.4 –6.1 antirheumatic drugs (DMARDs)* Calcium channel blockers* 5656 (11.5) 68321 (15.8) –4.3 –12.5 Digoxin* 462 (0.9) 6495 (1.5) –0.6 –5.1

343

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Heparin and low-molecular-weight 120 (0.2) 2608 (0.6) –0.4 –5.5 heparin* Lithium 68 (0.1) 756 (0.2) 0 –0.9 Loop diuretics* 1825 (3.7) 28351 (6.6) –2.8 –12.9 Monoamine oxidase inhibitors 5 (0) 26 (0) 0 0.5 (MOAs) Nitrates* 700 (1.4) 11813 (2.7) –1.3 –9.2 Nonbiologic DMARDs 842 (1.7) 18749 (4.3) –2.6 –15.4

Nonselective nonsteroidal anti- 6302 (12.8) 74266 (17.2) –4.4 –12.2 inflammatory drugs (NSAIDs)* Opioids* 9427 (19.2) 118648 (27.4) –8.3 –19.7 Oral anticoagulants* 662 (1.3) 15945 (3.7) –2.3 –15 Other anticoagulants* 10 (0) 155 (0) 0 –0.9 Other antihypertensives 809 (1.6) 12407 (2.9) –1.2 –8.2 Other newer and atypical 2340 (4.8) 27109 (6.3) –1.5 –6.6 antidepressants* Potassium-sparing diuretics 2514 (5.1) 23536 (5.4) –0.3 –1.5 Renin inhibitor 51 (0.1) 478 (0.1) 0 –0.2 Selective COX-2 inhibitors* 1684 (3.4) 14909 (3.4) 0 –0.1

344

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Selective estrogen receptor 550 (1.1) 3487 (0.8) 0.3 3.2 blockers Selective serotonin- 1791 (3.6) 17345 (4) –0.4 –1.9 norepinephrine reuptake inhibitors (SNRIs)* Selective serotonin reuptake 6496 (13.2) 63022 (14.6) –1.4 –3.9 inhibitors (SSRIs)* Tricyclic antidepressants (TCAs)* 1598 (3.2) 15914 (3.7) –0.4 –2.4 Thiazide diuretics* 8679 (17.6) 84703 (19.6) –1.9 –5 Typical antipsychotics* 82 (0.2) 1004 (0.2) –0.1 –1.5 No lipid-lowering drug despite 7831 (15.9) 75333 (17.4) –1.5 –4 diagnosis of hyperlipidemia* Long-term use of opioids* 5442 (11.1) 66319 (15.3) –4.3 –12.6 Long-term current use of steroids* 99 (0.2) 5393 (1.2) –1 –12.4 Lack of 2nd RX among all the 456 (0.9) 6523 (1.5) –0.6 –5.3 interested drugs groups* Opioids baseline days’ supply ≥ 1544 (3.1) 28259 (6.5) –3.4 –15.9 90* Index days’ supply* 47 (26.8) 45 (25.8) 2 7.6 Use of preventive services

345

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Colonoscopy*, n (%) 3787 (7.7) 39674 (9.2) –1.5 –5.3 Fecal occult blood test*, n (%) 6442 (13.1) 57850 (13.4) –0.3 –0.8 Mammography*, n (% of females) 14761 (30.2) 146873 (37.1) –6.9 –8.5 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 1.6 Bone mineral density test*, n (%) 5643 (11.5) 65556 (15.2) –3.7 –10.9 Electrocardiography, n (%) 10746 (21.8) 127535 (29.5) –7.6 –17.6 INR test, n (%) 0 (1) 0 (2.2) –0.2 –13.2 Use of wheelchair, walker, 362 (0.7) 6250 (1.4) –0.7 –6.8 crutches, or cane, n (%) Use of oxygen, n (%) 302 (0.6) 7321 (1.7) –1.1 –10.1 Lipid test, n (%) 18591 (37.8) 213768 (49.4) –11.6 –23.6 Health care utilization, mean (SD) Number of distinct cardiovascular 1 (1.7) 1 (2.4) –0.6 –28.3 diagnoses Total days in hospital in prior year* 0 (2) 1 (3.8) –0.4 –11.8 Medication synchronization 0 (0.3) 0 (0.3) 0 8.1 metrics Copayment 305 (472.9) 428 (630.2) –123.3 –22.1 Number of office visits 9 (12.4) 12 (15.5) –3.1 –21.9

346

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 Number of unique drugs of 3 (2.2) 2 (2.9) 0.2 8 interest (by NDC) in prior year or on index date Number of office visits with a 2 (3.9) 3 (6.6) –1.1 –20.3 cardiovascular diagnosis Number of drugs of interest 7 (10.2) 9 (13.6) –2.9 –24.3 dispensations in 365 days in prior year or on index date* Number of hospitalizations in the 0 (0.3) 0 (0.5) –0.1 –14.4 prior year Medication synchronization 0 (0.2) 0 (0.2) 0 1.6 metrics (non-mail)* Number of cardiovascular 0 (0.2) 0 (0.3) 0 –12.7 hospitalizations Index copayment 41 (38.9) 29 (31.4) 12 33.9 Vaccinations* 0 (0.5) 0 (0.7) –0.1 –18.7 Physician visits* 4 (5.5) 6 (6.2) –1.6 –27.3 Number of drugs of interest 2 (1.6) 2 (1.6) 0.4 22.3 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 3 (2.4) 4 (2.9) –0.7 –24.5 365 days before or on index date

347

Appendix 44. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles before propensity score fine stratification and weighting in Aim 2 with days’ supply that overlap index date

Number of dispensations in 3 (2.6) 4 (3.1) -0.7 -24.6 baseline period with days’ supply that overlap index date* Number of unique (by NDC) drugs 7 (6.3) 9 (7.7) –2.5 –35.7 in 365 days before or on index date* Number of any drug dispensations 14 (20.2) 23 (26.1) –8.7 –37.4 in 365 days before or on index date*

348

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 63 (10.7) 63 (11.1) –0.7 –6.7 Sex: female* 4502 (99.4) 70616 (98.8) 0.6 6.3 Combined comorbidity score* 1 (1.8) 1 (1.9) –0.2 –8.7 Commercial plan type, n (%) 2659 (58.7) 39439 (55.2) 3.5 7.1 Benefit plan types, n (%) EPO* 396 (8.7) 6074 (8.5) 0.2 0.9 HMO* 1151 (25.4) 19738 (27.6) –2.2 –5 IND* 42 (0.9) 655 (0.9) 0 0.1 OTH* 792 (17.5) 13021 (18.2) –0.7 –1.9 POS* 1833 (40.5) 26678 (37.3) 3.1 6.5 Comorbidities, n (%) Acute coronary syndrome, with or 215 (4.7) 3957 (5.5) –0.8 –3.6 without revascularization Acute coronary syndrome, with 73 (1.6) 1102 (1.5) 0.1 0.6 revascularization Atrial fibrillation 45 (1) 965 (1.3) –0.4 –3.3 Alzheimer’s disease* 83 (1.8) 1686 (2.4) –0.5 –3.7

349

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Angina 192 (4.2) 3477 (4.9) –0.6 –3 Coronary atherosclerosis* 455 (10) 8473 (11.8) –1.8 –5.8 Coronary artery bypass graft (CABG), 6 (0.1) 115 (0.2) 0 –0.8 new CABG, old 33 (0.7) 614 (0.9) –0.1 –1.5 Any cancer 1038 (22.9) 15985 (22.4) 0.6 1.3 Any malignancy, including lymphoma 966 (21.3) 14918 (20.9) 0.5 1.1 and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 642 (14.2) 10636 (14.9) –0.7 –2 Chest pain 950 (21) 16253 (22.7) –1.8 –4.3 Heart failure* 213 (4.7) 4350 (6.1) –1.4 –6.1 Heart failure hospitalization 18 (0.4) 402 (0.6) –0.2 –2.4 Conduction disorders 85 (1.9) 1474 (2.1) –0.2 –1.3 Chronic obstructive pulmonary disease 558 (12.3) 9904 (13.8) –1.5 –4.6 (COPD) Depression* 1061 (23.4) 17297 (24.2) –0.8 –1.8 Diabetes* 967 (21.3) 18595 (26) –4.7 –11 Drug-induced osteoporosis* 89 (2) 1464 (2) –0.1 –0.6

350

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Falls* 93 (2.1) 1885 (2.6) –0.6 –3.9 History of falls 85 (1.9) 1593 (2.2) –0.4 –2.5 HIV infection 6 (0.1) 96 (0.1) 0 –0.1 Hyperlipidemia 2735 (60.4) 44653 (62.4) –2.1 –4.3 Hyperparathyroidism 48 (1.1) 786 (1.1) 0 –0.4 Hypertension* 2798 (61.8) 47063 (65.8) –4.1 –8.5 Hyperthyroidism 99 (2.2) 1515 (2.1) 0.1 0.5 Inflammatory bowel disease (IBD)* 67 (1.5) 1005 (1.4) 0.1 0.6 Ischemic heart disease* 482 (10.6) 8886 (12.4) –1.8 –5.6 Liver disease* 311 (6.9) 5299 (7.4) –0.5 –2.1 Metastatic cancer* 38 (0.8) 688 (1) –0.1 –1.3 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1496 (33) 24431 (34.2) –1.1 –2.4 Other fractures* — — — — Palpitations 346 (7.6) 5576 (7.8) –0.2 –0.6 Parkinson’s disease 28 (0.6) 532 (0.7) –0.1 –1.5

351

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 51 (1.1) 1005 (1.4) –0.3 –2.5 coronary syndromes (MI/ACS)* Preventive care 3162 (69.8) 48818 (68.3) 1.5 3.3 Prior MI 19 (0.4) 327 (0.5) 0 -0.6 Prior stroke 9 (0.2) 198 (0.3) –0.1 –1.6 Peripheral vascular disease (PVD) or PVD 207 (4.6) 3834 (5.4) –0.8 –3.7 surgery Rheumatoid arthritis* 237 (5.2) 4068 (5.7) –0.5 –2 Recent MI 1 (0) 17 (0) 0 –0.1 Recent stroke 4 (0.1) 48 (0.1) 0 0.8 Renal dysfunction 399 (8.8) 7680 (10.7) –1.9 –6.5 Schizophrenia 15 (0.3) 354 (0.5) –0.2 –2.6 Transient ischemic attack* 69 (1.5) 1340 (1.9) –0.4 –2.7 Upper GI diseases* 1378 (30.4) 22604 (31.6) –1.2 –2.6 Urinary tract infection 809 (17.9) 13727 (19.2) –1.3 –3.5 Baseline medication use, n (%) Systemic steroid* 303 (6.7) 5457 (7.6) –0.9 –3.7 Bile acid sequestrants 66 (1.5) 1207 (1.7) –0.2 –1.9 IV osteoclast inhibitors* 19 (0.4) 506 (0.7) –0.3 –3.9 352

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 407 (9) 7162 (10) –1 –3.5 Lipid-lowering agents 1718 (37.9) 29943 (41.9) –4 –8.1 Niacin and fibrates 149 (3.3) 2958 (4.1) –0.8 –4.5 Angiotensin-converting enzyme 1198 (26.4) 21203 (29.7) –3.2 –7.2 inhibitors (ACEIs)* Agents for dementia 56 (1.2) 1163 (1.6) –0.4 –3.3 Antidiabetics* 722 (15.9) 14635 (20.5) –4.5 –11.8 Antiparkinson agents* 164 (3.6) 2849 (4) –0.4 –1.9 Antiplatelets* 2 (0) 33 (0) 0 -0.1 Angiotensin II receptor blockers (ARBs)* 951 (21) 17187 (24) –3 –7.3 Aromatase inhibitors 84 (1.9) 1309 (1.8) 0 0.2 Atypical antipsychotics 169 (3.7) 3180 (4.4) –0.7 –3.6 Beta-blockers* 1323 (29.2) 23160 (32.4) –3.2 –6.9 Biologic disease-modifying 30 (0.7) 502 (0.7) 0 -0.5 antirheumatic drugs (DMARDs)* Calcium channel blockers* 979 (21.6) 18003 (25.2) –3.6 –8.4 Digoxin* 67 (1.5) 1250 (1.7) –0.3 –2.1 Heparin and low-molecular-weight 33 (0.7) 587 (0.8) –0.1 –1.1 heparin*

353

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Lithium 18 (0.4) 333 (0.5) –0.1 –1 Loop diuretics* 424 (9.4) 7992 (11.2) –1.8 –6 Monoamine oxidase inhibitors (MOAs) 0 (0) 0 (0) 0 -0.2 Nitrates* 174 (3.8) 3417 (4.8) –0.9 –4.6 Nonbiologic DMARDs 214 (4.7) 3594 (5) –0.3 –1.4

Nonselective nonsteroidal anti- 1270 (28) 20901 (29.2) –1.2 –2.7 inflammatory drugs (NSAIDs)* Opioids* 2097 (46.3) 34935 (48.9) –2.6 –5.2 Oral anticoagulants* 133 (2.9) 2538 (3.5) –0.6 –3.5 Other anticoagulants* 2 (0) 19 (0) 0 1 Other antihypertensives 163 (3.6) 3377 (4.7) –1.1 –5.6 Other newer and atypical 564 (12.4) 9560 (13.4) –0.9 –2.7 antidepressants* Potassium-sparing diuretics 401 (8.9) 7486 (10.5) –1.6 –5.5 Renin inhibitor 22 (0.5) 398 (0.6) –0.1 –1 Selective COX-2 inhibitors* 209 (4.6) 3449 (4.8) –0.2 –1 Selective estrogen receptor blockers 126 (2.8) 1652 (2.3) 0.5 3 Selective serotonin-norepinephrine 464 (10.2) 8058 (11.3) –1 –3.3 reuptake inhibitors (SNRIs)* 354

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 1759 (38.8) 28056 (39.2) –0.4 –0.8 (SSRIs)* Tricyclic antidepressants (TCAs)* 280 (6.2) 4900 (6.9) –0.7 –2.7 Thiazide diuretics* 1481 (32.7) 26696 (37.3) –4.6 –9.8 Typical antipsychotics* 17 (0.4) 401 (0.6) –0.2 –2.7 No lipid-lowering drug despite diagnosis 1239 (27.3) 18733 (26.2) 1.1 2.6 of hyperlipidemia* Long-term use of opioids* 1005 (22.2) 16918 (23.7) –1.5 –3.5 Long-term current use of steroids* 25 (0.6) 552 (0.8) –0.2 –2.7 Lack of 2nd RX among all the interested 389 (8.6) 4897 (6.8) 1.7 6.5 drugs groups* Opioids baseline days’ supply ≥ 90* 593 (13.1) 11310 (15.8) –2.7 –7.8 index days’ supply* 35 (15.9) 35 (18.5) –0.7 –4.3 Use of preventive services Colonoscopy*, n (%) 520 (11.5) 8528 (11.9) –0.4 –1.4 Fecal occult blood test*, n (%) 683 (15.1) 10712 (15) 0.1 0.3 Mammography*, n (% of females) 1703 (37.8) 25613 (36.3) 1.5 3.7 Any prior lab test, n (%) 0 (0.5) 0 (0.5) 0 3.5 Bone mineral density test*, n (%) 200 (4.4) 2881 (4) 0.4 1.9

355

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 1933 (42.7) 32170 (45) –2.3 –4.7 INR test, n (%) 0 (2) 0 (2.1) –0.1 –3 Use of wheelchair, walker, crutches, or 79 (1.7) 1495 (2.1) –0.3 –2.5 cane, n (%) Use of oxygen, n (%) 67 (1.5) 1375 (1.9) –0.4 –3.4 Lipid test, n (%) 2887 (63.7) 46291 (64.7) –1 –2.1 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (2.7) 2 (2.9) –0.3 –10 diagnoses Total days in hospital in prior year* 1 (3.6) 1 (4.1) –0.2 –5 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –12.6 Copayment 617 (640.5) 689 (752.1) –71.7 –10.3 Number of office visits 18 (17.9) 20 (21.1) –1.2 –6.4 Number of unique drugs of interest (by 5 (2.9) 5 (4.6) –0.7 –18.7 NDC) in prior year or on index date Number of office visits with a 4 (7) 5 (9.9) –0.7 –7.7 cardiovascular diagnosis Number of drugs of interest 15 (12.1) 17 (14.9) –2.1 –15.7 dispensations in 365 days in prior year or on index date*

356

Appendix 45. Characteristics of raloxifene vs bisphosphonate initiators in adherence prediction score tertile 1 after propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (0.5) 0 (0.6) 0 –5.1 year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –13.3 (non-mail)* Number of cardiovascular 0 (0.4) 0 (0.4) 0 –4.8 hospitalizations Index copayment 39 (38) 43 (71.1) –3.2 –5.6 Vaccinations* 0 (0.7) 0 (0.7) 0 –2.7 Physician visits* 10 (7.1) 10 (8.5) –0.6 –7.6 Number of drugs of interest 3 (1.6) 3 (2.4) –0.4 –17.3 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 4 (2.6) 5 (2.7) –0.5 –20.3 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 4 (2.7) 5 (2.9) –0.6 –21 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 14 (8.6) 16 (9.9) –1.7 –18.3 days before or on index date* Number of any drug dispensations in 33 (25.2) 38 (31.2) –4.5 –16 365 days before or on index date*

357

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 66 (10.3) 66 (10.8) –0.6 –5.9 Sex: female* 4298 (99.3) 69638 (96.9) 2.4 17.9 Combined comorbidity score* 0 (1.5) 1 (1.8) –0.2 –13.7 Commercial plan type, n (%) 2290 (52.9) 35885 (49.9) 3 6 Benefit plan types, n (%) EPO* 281 (6.5) 4324 (6) 0.5 2 HMO* 1287 (29.7) 22631 (31.5) –1.7 –3.8 IND* 132 (3) 2341 (3.3) –0.2 –1.2 OTH* 853 (19.7) 14352 (20) –0.3 –0.6 POS* 1420 (32.8) 22156 (30.8) 2 4.3 Comorbidities, n (%) Acute coronary syndrome, with or without 130 (3) 2859 (4) –1 –5.3 revascularization Acute coronary syndrome, with 32 (0.7) 689 (1) –0.2 –2.4 revascularization Atrial fibrillation 23 (0.5) 641 (0.9) –0.4 –4.3

358

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Alzheimer’s disease* 91 (2.1) 2119 (2.9) –0.8 –5.4 Angina 117 (2.7) 2504 (3.5) –0.8 –4.5 Coronary atherosclerosis* 329 (7.6) 7287 (10.1) –2.5 –8.9 Coronary artery bypass graft (CABG), new 1 (0) 72 (0.1) –0.1 –3.1 CABG, old 15 (0.3) 545 (0.8) –0.4 –5.6 Any cancer 964 (22.3) 16137 (22.4) –0.2 –0.4 Any malignancy, including lymphoma and 891 (20.6) 15008 (20.9) –0.3 –0.7 leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 464 (10.7) 8600 (12) –1.2 –3.9 Chest pain 659 (15.2) 12649 (17.6) –2.4 –6.4 Heart failure* 160 (3.7) 3954 (5.5) –1.8 –8.6 Heart failure hospitalization 5 (0.1) 230 (0.3) –0.2 –4.4 Conduction disorders 61 (1.4) 1268 (1.8) –0.4 –2.8 Chronic obstructive pulmonary disease 429 (9.9) 8179 (11.4) –1.5 –4.8 (COPD) Depression* 601 (13.9) 11121 (15.5) –1.6 –4.5 Diabetes* 818 (18.9) 17929 (24.9) –6 –14.6 Drug-induced osteoporosis* 65 (1.5) 1043 (1.5) 0.1 0.4

359

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Falls* 63 (1.5) 1260 (1.8) –0.3 –2.4 History of falls 59 (1.4) 1142 (1.6) –0.2 –1.9 HIV infection 2 (0) 48 (0.1) 0 –0.9 Hyperlipidemia 2571 (59.4) 44493 (61.9) –2.5 –5.1 Hyperparathyroidism 40 (0.9) 716 (1) –0.1 –0.7 Hypertension* 2762 (63.8) 49074 (68.3) –4.4 –9.4 Hyperthyroidism 98 (2.3) 1765 (2.5) –0.2 –1.3 Inflammatory bowel disease (IBD)* 47 (1.1) 878 (1.2) –0.1 –1.3 Ischemic heart disease* 353 (8.2) 7800 (10.8) –2.7 –9.2 Liver disease* 174 (4) 3295 (4.6) –0.6 –2.8 Metastatic cancer* 35 (0.8) 705 (1) –0.2 –1.8 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1563 (36.1) 25978 (36.1) 0 0 Other fractures* — — — — Palpitations 256 (5.9) 4534 (6.3) –0.4 –1.6 Parkinson’s 27 (0.6) 683 (1) –0.3 –3.7

360

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute coronary 35 (0.8) 987 (1.4) –0.6 –5.4 syndromes (MI/ACS)* Preventive care 3071 (71) 49585 (69) 2 4.3 Prior MI 8 (0.2) 240 (0.3) –0.1 –2.9 Prior stroke 1 (0) 174 (0.2) –0.2 –6 Peripheral vascular disease (PVD) or PVD 163 (3.8) 3545 (4.9) –1.2 –5.7 surgery Rheumatoid arthritis* 127 (2.9) 2692 (3.7) –0.8 –4.5 Recent MI 0 (0) 4 (0) 0 -1 Recent stroke 0 (0) 5 (0) 0 –1.2 Renal dysfunction 292 (6.7) 7018 (9.8) –3 –11 Schizophrenia 12 (0.3) 182 (0.3) 0 0.5 Transient ischemic attack* 60 (1.4) 1270 (1.8) –0.4 –3.1 Upper GI diseases* 1013 (23.4) 18139 (25.2) –1.8 –4.3 Urinary tract infection 667 (15.4) 11937 (16.6) –1.2 –3.3 Baseline medication use, n (%) Systemic steroid* 181 (4.2) 4718 (6.6) –2.4 –10.6 Bile acid sequestrants 47 (1.1) 938 (1.3) –0.2 –2 IV osteoclast inhibitors* 32 (0.7) 825 (1.1) –0.4 –4.2 361

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 252 (5.8) 5065 (7) –1.2 –5 Lipid-lowering agents 2022 (46.7) 36262 (50.4) –3.7 –7.4 Niacin and fibrates 198 (4.6) 4114 (5.7) –1.1 –5.2 Angiotensin-converting enzyme inhibitors 1268 (29.3) 23874 (33.2) –3.9 –8.4 (ACEIs)* Agents for dementia 60 (1.4) 1379 (1.9) –0.5 –4.2 Antidiabetics* 617 (14.3) 14362 (20) –5.7 –15.2 Antiparkinson agents* 103 (2.4) 2184 (3) –0.7 –4.1 Antiplatelets* 1 (0) 57 (0.1) –0.1 –2.5 Angiotensin II receptor blockers (ARBs)* 959 (22.2) 18549 (25.8) –3.6 –8.5 Aromatase inhibitors 82 (1.9) 1322 (1.8) 0.1 0.4 Atypical antipsychotics 107 (2.5) 2297 (3.2) –0.7 –4.4 Beta-blockers* 1362 (31.5) 25229 (35.1) –3.6 –7.7 Biologic disease-modifying antirheumatic 16 (0.4) 336 (0.5) –0.1 –1.5 drugs (DMARDs)* Calcium channel blockers* 1148 (26.5) 21262 (29.6) –3 –6.8 Digoxin* 82 (1.9) 1807 (2.5) –0.6 –4.2 Heparin and low-molecular-weight 21 (0.5) 474 (0.7) –0.2 –2.3 heparin*

362

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Lithium 11 (0.3) 239 (0.3) –0.1 –1.5 Loop diuretics* 320 (7.4) 7216 (10) –2.6 –9.4 Monoamine oxidase inhibitors (MOAs) 1 (0) 0 (0) 0 2.1 Nitrates* 131 (3) 3050 (4.2) –1.2 –6.5 Nonbiologic DMARDs 143 (3.3) 2918 (4.1) –0.8 –4

Nonselective nonsteroidal anti- 968 (22.4) 17446 (24.3) –1.9 –4.5 inflammatory drugs (NSAIDs)* Opioids* 1512 (34.9) 27797 (38.7) –3.7 –7.7 Oral anticoagulants* 129 (3) 3238 (4.5) –1.5 –8 Other anticoagulants* 2 (0) 28 (0) 0 0.4 Other antihypertensives 138 (3.2) 3181 (4.4) –1.2 –6.5 Other newer and atypical 327 (7.6) 6201 (8.6) –1.1 –3.9 antidepressants* Potassium-sparing diuretics 431 (10) 8747 (12.2) –2.2 –7 Renin inhibitor 11 (0.3) 184 (0.3) 0 0 Selective COX-2 inhibitors* 180 (4.2) 3279 (4.6) –0.4 –2 Selective estrogen receptor blockers 90 (2.1) 1519 (2.1) 0 –0.2 Selective serotonin-norepinephrine 202 (4.7) 3971 (5.5) –0.9 –3.9 reuptake inhibitors (SNRIs)* 363

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Selective serotonin reuptake inhibitors 1141 (26.4) 19698 (27.4) –1 –2.3 (SSRIs)* Tricyclic antidepressants (TCAs)* 211 (4.9) 3835 (5.3) –0.5 –2.1 Thiazide diuretics* 1571 (36.3) 30154 (41.9) –5.6 –11.6 Typical antipsychotics* 6 (0.1) 134 (0.2) 0 –1.2 No lipid-lowering drug despite diagnosis 835 (19.3) 13506 (18.8) 0.5 1.3 of hyperlipidemia* Long-term use of opioids* 868 (20.1) 15335 (21.3) –1.3 –3.1 Long-term current use of steroids* 13 (0.3) 493 (0.7) –0.4 –5.5 Lack of 2nd RX among all the interested 53 (1.2) 683 (0.9) 0.3 2.7 drugs groups* Opioids baseline days’ supply ≥ 90* 274 (6.3) 6089 (8.5) –2.1 –8.2 Index days’ supply* 43 (24.5) 44 (25.6) –1.4 –5.7 Use of preventive services Colonoscopy*, n (%) 514 (11.9) 8593 (12) –0.1 –0.2 Fecal occult blood test*, n (%) 813 (18.8) 12824 (17.8) 0.9 2.5 Mammography*, n (% of females) 1956 (45.5) 30683 (44.1) 1.4 5.1 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 2.4 Bone mineral density test*, n (%) 497 (11.5) 7255 (10.1) 1.4 4.5

364

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Electrocardiography, n (%) 1695 (39.2) 29989 (41.7) –2.5 –5.2 INR test, n (%) 0 (1.7) 0 (2.5) –0.2 –7.4 Use of wheelchair, walker, crutches, or 73 (1.7) 1537 (2.1) –0.5 –3.3 cane, n (%) Use of oxygen, n (%) 36 (0.8) 1007 (1.4) –0.6 –5.4 Lipid test, n (%) 2778 (64.2) 46389 (64.5) –0.3 –0.7 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (2.2) 2 (2.6) –0.4 –15.6 diagnoses Total days in hospital in prior year* 1 (2.4) 1 (3.9) –0.3 –8.3 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –14 Copayment 652 (645.2) 756 (778.4) –104.6 –14.6 Number of office visits 15 (15) 17 (17.5) –1.7 –10.2 Number of unique drugs of interest (by 5 (2.7) 6 (4.2) –0.9 –24.3 NDC) in prior year or on index date Number of office visits with a 4 (5.7) 4 (7.5) –0.8 –12.6 cardiovascular diagnosis Number of drugs of interest dispensations 19 (13.5) 21 (16.5) –2.7 –17.6 in 365 days in prior year or on index date*

365

Appendix 46. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 2 after propensity score fine stratification and weighting in Aim 2 Number of hospitalizations in the prior 0 (0.4) 0 (0.6) 0 –7.6 year Medication synchronization metrics (non- 0 (0.2) 0 (0.2) 0 –16.1 mail)* Number of cardiovascular hospitalizations 0 (0.3) 0 (0.4) 0 –8 Index copayment 44 (48.2) 46 (75.8) –2.8 –4.4 Vaccinations* 0 (0.7) 0 (0.8) 0 –4.2 Physician visits* 8 (6) 9 (6.9) –0.7 –11.5 Number of drugs of interest dispensations 3 (1.7) 4 (2.5) –0.5 –23.5 with days’ supply that overlap index date Number unique (by NDC) drugs in 365 5 (2.5) 6 (3) –0.8 –28.8 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 5 (2.7) 6 (3.2) –0.9 –30.7 period with days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 12 (6.9) 14 (8.4) –2 –26.3 days before or on index date* Number of any drug dispensations in 365 34 (25.2) 40 (32.8) –5.6 –19.2 days before or on index date*

366

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 66 (9.7) 66 (10.2) –0.5 –5.2 Sex: female* 4528 (99.4) 70623 (98.6) 0.8 7.9 Combined comorbidity score* 0 (1.5) 0 (1.6) –0.1 –5 Commercial plan type, n (%) 2828 (62.1) 43127 (60.2) 1.9 3.8 Benefit plan types, n (%) EPO* 283 (6.2) 4472 (6.2) 0 -0.1 HMO* 1373 (30.1) 22824 (31.9) –1.7 –3.7 IND* 400 (8.8) 6505 (9.1) -0.3 –1.1 OTH* 609 (13.4) 9534 (13.3) 0.1 0.2 POS* 1444 (31.7) 21473 (30) 1.7 3.7 Comorbidities, n (%) Acute coronary syndrome, with or 102 (2.2) 1703 (2.4) –0.1 –0.9 without revascularization Acute coronary syndrome, with 34 (0.7) 555 (0.8) 0 -0.3 revascularization Atrial fibrillation 18 (0.4) 325 (0.5) –0.1 –0.9 Alzheimer’s disease* 108 (2.4) 1925 (2.7) –0.3 -2

367

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Angina 90 (2) 1501 (2.1) –0.1 –0.9 Coronary atherosclerosis* 277 (6.1) 4962 (6.9) –0.8 –3.4 Coronary artery bypass graft (CABG), 2 (0) 17 (0) 0 1.1 new CABG, old 25 (0.5) 459 (0.6) –0.1 –1.2 Any cancer 981 (21.5) 15368 (21.4) 0.1 0.2 Any malignancy, including lymphoma 903 (19.8) 14232 (19.9) 0 -0.1 and leukemia, except malignant neoplasm of skin* Cardiovascular system symptom 418 (9.2) 6529 (9.1) 0.1 0.2 Chest pain 619 (13.6) 10160 (14.2) –0.6 –1.7 Heart failure* 145 (3.2) 2660 (3.7) –0.5 –2.9 Heart failure hospitalization 2 (0) 57 (0.1) 0 –1.5 Conduction disorders 53 (1.2) 931 (1.3) –0.1 –1.2 Chronic obstructive pulmonary disease 405 (8.9) 6789 (9.5) –0.6 –2 (COPD) Depression* 430 (9.4) 7085 (9.9) –0.5 –1.5 Diabetes* 718 (15.8) 13060 (18.2) –2.5 –6.6 Drug-induced osteoporosis* 63 (1.4) 959 (1.3) 0 0.4

368

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Falls* 53 (1.2) 828 (1.2) 0 0.1 History of falls 58 (1.3) 930 (1.3) 0 -0.2 HIV infection 1 (0) 12 (0) 0 0.4 Hyperlipidemia 2582 (56.7) 41234 (57.5) –0.9 –1.8 Hyperparathyroidism 39 (0.9) 676 (0.9) –0.1 –0.9 Hypertension* 2675 (58.7) 43924 (61.3) –2.6 –5.3 Hyperthyroidism 95 (2.1) 1540 (2.1) –0.1 –0.4 Inflammatory bowel disease (IBD)* 34 (0.7) 506 (0.7) 0 0.5 Ischemic heart disease* 303 (6.6) 5375 (7.5) –0.9 –3.3 Liver disease* 128 (2.8) 2237 (3.1) –0.3 –1.8 Metastatic cancer* 55 (1.2) 969 (1.4) –0.1 –1.3 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 1471 (32.3) 23358 (32.6) –0.3 –0.7 Other fractures* — — — — Palpitations 210 (4.6) 3149 (4.4) 0.2 1 Parkinson’s 30 (0.7) 602 (0.8) –0.2 –2.1

369

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Post-myocardial infarction/acute 33 (0.7) 578 (0.8) –0.1 –0.9 coronary syndromes (MI/ACS)* Preventive care 3153 (69.2) 49025 (68.4) 0.8 1.7 Prior MI 7 (0.2) 95 (0.1) 0 0.6 Prior stroke 9 (0.2) 162 (0.2) 0 –0.6 Peripheral vascular disease (PVD) or 139 (3.1) 2443 (3.4) –0.4 –2 PVD surgery Rheumatoid arthritis* 127 (2.8) 2184 (3) –0.3 –1.5 Recent MI 0 (0) 1 (0) 0 -0.5 Recent stroke 2 (0) 18 (0) 0 1 Renal dysfunction 259 (5.7) 4863 (6.8) –1.1 –4.6 Schizophrenia 5 (0.1) 96 (0.1) 0 –0.7 Transient ischemic attack* 45 (1) 764 (1.1) –0.1 –0.8 Upper GI diseases* 875 (19.2) 14150 (19.7) –0.5 –1.4 Urinary tract infection 574 (12.6) 9229 (12.9) –0.3 –0.9 Baseline medication use, n (%) Systemic steroid* 184 (4) 3457 (4.8) –0.8 –3.8 Bile acid sequestrants 47 (1) 736 (1) 0 0 IV osteoclast inhibitors* 26 (0.6) 482 (0.7) –0.1 –1.3 370

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Inhaled glucocorticoid* 266 (5.8) 4875 (6.8) –1 –4 Lipid-lowering agents 2355 (51.7) 38148 (53.2) –1.6 –3.1 Niacin and fibrates 300 (6.6) 5155 (7.2) –0.6 –2.4 Angiotensin-converting enzyme 1465 (32.1) 24556 (34.3) –2.1 –4.5 inhibitors (ACEIs)* Agents for dementia 80 (1.8) 1448 (2) –0.3 –2 Antidiabetics* 575 (12.6) 10880 (15.2) –2.6 –7.4 Antiparkinson agents* 89 (2) 1580 (2.2) –0.3 –1.8 Antiplatelets* 2 (0) 44 (0.1) 0 –0.8 Angiotensin II receptor blockers 914 (20.1) 15850 (22.1) –2.1 –5.1 (ARBs)* Aromatase inhibitors 72 (1.6) 1093 (1.5) 0.1 0.4 Atypical antipsychotics 90 (2) 1663 (2.3) –0.3 –2.4 Beta-blockers* 1482 (32.5) 24719 (34.5) –2 –4.2 Biologic disease-modifying 14 (0.3) 239 (0.3) 0 –0.5 antirheumatic drugs (DMARDs)* Calcium channel blockers* 1173 (25.7) 20045 (28) –2.2 –5 Digoxin* 88 (1.9) 1663 (2.3) –0.4 –2.7

371

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Heparin and low-molecular-weight 23 (0.5) 426 (0.6) –0.1 –1.2 heparin* Lithium 10 (0.2) 165 (0.2) 0 –0.2 Loop diuretics* 360 (7.9) 6859 (9.6) –1.7 –5.9 Monoamine oxidase inhibitors (MOAs) 1 (0) 12 (0) 0 0.3 Nitrates* 133 (2.9) 2489 (3.5) –0.6 –3.2 Nonbiologic DMARDs 176 (3.9) 3008 (4.2) –0.3 –1.7

Nonselective nonsteroidal anti- 862 (18.9) 14011 (19.6) –0.6 –1.6 inflammatory drugs (NSAIDs)* Opioids* 1251 (27.5) 20911 (29.2) –1.7 –3.8 Oral anticoagulants* 164 (3.6) 3189 (4.5) –0.9 –4.3 Other anticoagulants* 3 (0.1) 39 (0.1) 0 0.4 Other antihypertensives 194 (4.3) 3515 (4.9) –0.6 –3.1 Other newer and atypical 261 (5.7) 4497 (6.3) –0.5 –2.3 antidepressants* Potassium-sparing diuretics 566 (12.4) 9897 (13.8) –1.4 –4.1 Renin inhibitor 11 (0.2) 202 (0.3) 0 –0.8 Selective COX-2 inhibitors* 236 (5.2) 3890 (5.4) –0.3 –1.1 Selective estrogen receptor blockers 70 (1.5) 963 (1.3) 0.2 1.6

372

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Selective serotonin-norepinephrine 148 (3.2) 2549 (3.6) –0.3 –1.7 reuptake inhibitors (SNRIs)* Selective serotonin reuptake inhibitors 934 (20.5) 15012 (21) –0.5 –1.1 (SSRIs)* Tricyclic antidepressants (TCAs)* 228 (5) 3640 (5.1) –0.1 –0.4 Thiazide diuretics* 1810 (39.7) 30735 (42.9) –3.2 –6.5 Typical antipsychotics* 18 (0.4) 259 (0.4) 0 0.5 No lipid-lowering drug despite 652 (14.3) 10204 (14.2) 0.1 0.2 diagnosis of hyperlipidemia* Long-term use of opioids* 830 (18.2) 13338 (18.6) –0.4 –1 Long-term current use of steroids* 19 (0.4) 366 (0.5) –0.1 –1.4 Lack of 2nd RX among all the 14 (0.3) 210 (0.3) 0 0.3 interested drugs groups* Opioids baseline days’ supply ≥ 90* 209 (4.6) 3938 (5.5) –0.9 –4.2 Index days’ supply* 58 (29.9) 59 (28.9) –0.6 –2 Use of preventive services Colonoscopy*, n (%) 587 (12.9) 9043 (12.6) 0.3 0.8 Fecal occult blood test*, n (%) 932 (20.5) 14328 (20) 0.5 1.1 Mammography*, n (% of females) 2359 (52.1) 36104 (51.1) 1.0 2.8

373

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Any prior lab test, n (%) 0 (0.4) 0 (0.4) 0 1.6 Bone mineral density test*, n (%) 1441 (31.6) 21520 (30) 1.6 3.4 Electrocardiography, n (%) 1509 (33.1) 24563 (34.3) –1.2 –2.5 INR test, n (%) 0 (1.7) 0 (2.1) –0.1 –3.7 Use of wheelchair, walker, crutches, or 42 (0.9) 859 (1.2) –0.3 –2.7 cane, n (%) Use of oxygen, n (%) 55 (1.2) 989 (1.4) –0.2 –1.5 Lipid test, n (%) 2569 (56.4) 40305 (56.2) 0.1 0.3 Health care utilization, mean (SD) Number of distinct cardiovascular 2 (1.9) 2 (2.1) –0.1 –6 diagnoses Total days in hospital in prior year* 0 (2.6) 1 (3.2) –0.1 –3.1 Medication synchronization metrics 1 (0.3) 1 (0.3) 0 –6.7 Copayment 724 (695.5) 777 (827.3) –52.9 –6.9 Number of office visits 13 (12.8) 14 (14.4) –0.6 –4.5 Number of unique drugs of interest 5 (2.5) 5 (3.8) –0.4 –13.7 (by NDC) in prior year or on index date Number of office visits with a 3 (4.1) 3 (5.1) –0.3 –5.9 cardiovascular diagnosis

374

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Number of drugs of interest 19 (15) 21 (17.1) –1.6 –9.9 dispensations in 365 days in prior year or on index date* Number of hospitalizations in the prior 0 (0.4) 0 (0.4) 0 –3.4 year Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –9.4 (non-mail)* Number of cardiovascular 0 (0.3) 0 (0.3) 0 –3.3 hospitalizations Index copayment 48 (47.4) 48 (60.9) –0.3 –0.6 Vaccinations* 0 (0.7) 0 (0.7) 0 –0.9 Physician visits* 7 (5.4) 7 (5.9) –0.3 –5.2 Number of drugs of interest 4 (2) 4 (3) –0.4 –15.8 dispensations with days’ supply that overlap index date Number unique (by NDC) drugs in 365 6 (2.9) 6 (3.4) –0.5 –15.2 days before or on index date with days’ supply that overlap index date Number of dispensations in baseline 6 (3.4) 7 (3.8) –0.6 –16.8 period with days’ supply that overlap index date*

375

Appendix 47. Characteristics of raloxifene vs bisphosphonates initiators in adherence prediction score tertile 3 after propensity score fine stratification and weighting in Aim 2 Number of unique (by NDC) drugs in 11 (6.7) 12 (7.5) –0.9 –13.2 365 days before or on index date* Number of any drug dispensations in 35 (27.6) 38 (34.8) –2.9 –9.1 365 days before or on index date*

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Raloxifene Bisphosphonates Absolute Standardized Difference Differences Age, mean (SD)* 65 (10.4) 65 (10.8) –0.6 –5.7 Sex: female* 13330 212019 (98.5) 0.8 8.2 (99.3) Combined comorbidity score* 0 (1.6) 1 (1.7) –0.1 –6.4 Commercial plan type, n (%) 7777 (58) 119751 (55.6) 2.3 4.7 Benefit plan types, n (%) EPO* 960 (7.2) 15027 (7) 0.2 0.7 HMO* 3812 (28.4) 64705 (30.1) –1.7 –3.6 IND* 575 (4.3) 9891 (4.6) –0.3 –1.5 OTH* 2255 (16.8) 36262 (16.8) 0 –0.1 POS* 4696 (35) 71024 (33) 2 4.2

376

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Comorbidities, n (%) Acute coronary syndrome, with or without 447 (3.3) 8083 (3.8) –0.4 –2.3 revascularization Acute coronary syndrome, with revascularization 139 (1) 2319 (1.1) 0 –0.4 Atrial fibrillation 86 (0.6) 1754 (0.8) –0.2 –2 Alzheimer’s disease* 282 (2.1) 5390 (2.5) –0.4 –2.7 Angina 399 (3) 7115 (3.3) –0.3 –1.9 Coronary atherosclerosis* 1061 (7.9) 19539 (9.1) –1.2 –4.2 Coronary artery bypass graft (CABG), new 9 (0.1) 182 (0.1) 0 –0.6 CABG, old 73 (0.5) 1425 (0.7) –0.1 –1.5 Any cancer 2986 (22.3) 47575 (22.1) 0.2 0.4 Any malignancy, including lymphoma and leukemia, 2763 (20.6) 44186 (20.5) 0.1 0.2 except malignant neoplasm of skin* Cardiovascular system symptom 1523 (11.4) 25134 (11.7) –0.3 –1 Chest pain 2229 (16.6) 38014 (17.7) –1 –2.8 Heart failure* 519 (3.9) 10206 (4.7) –0.9 –4.3 Heart failure hospitalization 25 (0.2) 633 (0.3) –0.1 –2.2 Conduction disorders 199 (1.5) 3500 (1.6) –0.1 –1.2 Chronic obstructive pulmonary disease (COPD) 1392 (10.4) 24180 (11.2) –0.9 –2.8 377

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Depression* 2094 (15.6) 34754 (16.1) –0.5 –1.5 Diabetes* 2504 (18.7) 47509 (22.1) –3.4 –8.5 Drug-induced osteoporosis* 217 (1.6) 3492 (1.6) 0 0 Falls* 209 (1.6) 3718 (1.7) –0.2 –1.3 History of falls 202 (1.5) 3499 (1.6) –0.1 –1 HIV infection 9 (0.1) 129 (0.1) 0 0.3 Hyperlipidemia 7889 (58.8) 129768 (60.3) –1.5 –3 Hyperparathyroidism 127 (0.9) 2170 (1) –0.1 –0.6 Hypertension* 8236 (61.4) 138591 (64.4) –3 –6.2 Hyperthyroidism 292 (2.2) 4748 (2.2) 0 –0.2 Inflammatory bowel disease (IBD)* 148 (1.1) 2418 (1.1) 0 –0.2 Ischemic heart disease* 1138 (8.5) 20866 (9.7) –1.2 –4.2 Liver disease* 613 (4.6) 10599 (4.9) –0.4 –1.7 Metastatic cancer* 128 (1) 2275 (1.1) –0.1 –1 Osteoporotic fracture: non-vertebral* — — — — Osteoporotic fracture: vertebral* — — — — Osteoporosis* 4531 (33.8) 73615 (34.2) –0.4 –0.9 Other fractures* — — — —

378

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Palpitations 811 (6) 12879 (6) 0.1 0.3 Parkinson’s disease 85 (0.6) 1820 (0.8) –0.2 –2.5 Post-myocardial infarction/acute coronary syndromes 119 (0.9) 2381 (1.1) -0.2 –2.2 (MI/ACS)* Preventive care 9388 (70) 147994 (68.8) 1.2 2.6 Prior MI 34 (0.3) 627 (0.3) 0 –0.7 Prior stroke 19 (0.1) 433 (0.2) –0.1 –1.4 Peripheral vascular disease (PVD) or PVD surgery 509 (3.8) 9294 (4.3) –0.5 –2.7 Rheumatoid arthritis* 491 (3.7) 8487 (3.9) –0.3 –1.5 Recent MI 1 (0) 19 (0) 0 –0.2 Recent stroke 6 (0) 94 (0) 0 0 Renal dysfunction 950 (7.1) 18264 (8.5) –1.4 –5.2 Schizophrenia 32 (0.2) 597 (0.3) 0 –0.8 Transient ischemic attack* 174 (1.3) 3148 (1.5) –0.2 –1.4 Upper GI diseases* 3268 (24.4) 54272 (25.2) –0.9 –2 Urinary tract infection 2050 (15.3) 34532 (16) –0.8 –2.1 Baseline medication use, n (%) Systemic steroid* 668 (5) 12422 (5.8) –0.8 –3.5

379

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Bile acid sequestrants 160 (1.2) 2765 (1.3) –0.1 –0.8 IV osteoclast inhibitors* 77 (0.6) 1780 (0.8) –0.3 –3 Inhaled glucocorticoid* 926 (6.9) 16809 (7.8) –0.9 –3.5 Lipid-lowering agents 6094 (45.4) 103866 (48.3) –2.8 –5.7 Niacin and fibrates 647 (4.8) 12033 (5.6) –0.8 –3.5 Angiotensin-converting enzyme inhibitors (ACEIs)* 3931 (29.3) 68844 (32) –2.7 –5.8

Agents for dementia 197 (1.5) 3771 (1.8) –0.3 –2.3 Antidiabetics* 1914 (14.3) 38434 (17.9) –3.6 –9.8 Antiparkinson agents* 356 (2.7) 6449 (3) –0.3 –2.1 Antiplatelets* 5 (0) 109 (0.1) 0 –0.6 Angiotensin II receptor blockers (ARBs)* 2824 (21) 50841 (23.6) –2.6 –6.2 Aromatase inhibitors 238 (1.8) 3722 (1.7) 0 0.3 Atypical antipsychotics 367 (2.7) 7013 (3.3) –0.5 –3.1 Beta-blockers* 4168 (31.1) 72049 (33.5) –2.4 –5.2 Biologic disease-modifying antirheumatic drugs 60 (0.4) 1034 (0.5) 0 –0.5 (DMARDs)* Calcium channel blockers* 3300 (24.6) 58586 (27.2) –2.6 –6 Digoxin* 237 (1.8) 4589 (2.1) –0.4 –2.6

380

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Heparin and low-molecular-weight heparin* 77 (0.6) 1386 (0.6) –0.1 –0.9 Lithium 39 (0.3) 755 (0.4) –0.1 –1.1 Loop diuretics* 1106 (8.2) 21110 (9.8) –1.6 –5.5 Monoamine oxidase inhibitors (MOAs) 2 (0) 14 (0) 0 0.8 Nitrates* 438 (3.3) 8390 (3.9) –0.6 –3.4 Nonbiologic DMARDs 533 (4) 9230 (4.3) –0.3 –1.6

Nonselective nonsteroidal anti-inflammatory drugs 3102 (23.1) 51790 (24.1) –0.9 –2.2 (NSAIDs)* Opioids* 4862 (36.2) 82086 (38.1) –1.9 –3.9 Oral anticoagulants* 427 (3.2) 8471 (3.9) –0.8 –4.1 Other anticoagulants* 7 (0.1) 82 (0) 0 0.7 Other antihypertensives 495 (3.7) 9587 (4.5) –0.8 –3.9 Other newer and atypical antidepressants* 1152 (8.6) 19852 (9.2) -0.6 –2.2 Potassium-sparing diuretics 1398 (10.4) 25700 (11.9) –1.5 –4.8 Renin inhibitor 44 (0.3) 774 (0.4) 0 –0.5 Selective COX-2 inhibitors* 625 (4.7) 10531 (4.9) –0.2 –1.1 Selective estrogen receptor blockers 286 (2.1) 3956 (1.8) 0.3 2.1

381

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Selective serotonin-norepinephrine reuptake 815 (6.1) 14368 (6.7) –0.6 –2.5 inhibitors (SNRIs)* Selective serotonin reuptake inhibitors (SSRIs)* 3835 (28.6) 61907 (28.8) –0.2 –0.4 Tricyclic antidepressants (TCAs)* 719 (5.4) 12174 (5.7) –0.3 –1.3 Thiazide diuretics* 4861 (36.2) 86867 (40.4) –4.1 –8.5 Typical antipsychotics* 41 (0.3) 755 (0.4) 0 –0.8 No lipid-lowering drug despite diagnosis of 2728 (20.3) 42424 (19.7) 0.6 1.6 hyperlipidemia* Long-term use of opioids* 2702 (20.1) 44772 (20.8) –0.7 –1.6 Long-term current use of steroids* 57 (0.4) 1197 (0.6) –0.1 –1.9 Lack of 2nd RX among all the interested drugs 456 (3.4) 5854 (2.7) 0.7 3.9 groups* Opioids baseline days’ supply ≥ 90* 1077 (8) 20537 (9.5) –1.5 –5.4 Index days’ supply* 45 (26.1) 46 (26.6) –1.1 –4.2 Use of preventive services Colonoscopy*, n (%) 1622 (12.1) 26507 (12.3) –0.2 –0.7 Fecal occult blood test*, n (%) 2428 (18.2) 37777 (17.8) 0.4 1.4 Mammography*, n (%) 6019 (44.9) 93469 (43.4) 1.4 2.9 Any prior lab test, n (%) 0 (0.5) 0 (0.5) 0 2.4

382

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Bone mineral density test*, n (%) 2138 (15.9) 32720 (15.2) 0.7 2 Electrocardiography, n (%) 5137 (38.3) 85159 (39.6) –1.3 –2.6 INR test, n (%) 0 (1.8) 0 (2.1) –0.1 –3.5 Use of wheelchair, walker, crutches, or cane, n (%) 194 (1.4) 3544 (1.6) –0.2 –1.6 Use of oxygen, n (%) 159 (1.2) 3117 (1.4) –0.3 –2.3 Lipid test, n (%) 8235 (61.4) 132268 (61.4) –0.1 –0.2 Health care utilization, mean (SD) Number of distinct cardiovascular diagnoses 2 (2.3) 2 (2.5) –0.2 –7.7 Total days in hospital in prior year* 1 (2.9) 1 (3.4) –0.1 –3.8 Medication synchronization metrics 0 (0.2) 0 (0.2) 0 –10 Copayment 665 (662.6) 729 (769.5) –64.5 –9 Number of office visits 16 (15.5) 16 (17.6) –0.8 –4.9 Number of unique drugs of interest (by NDC) in prior 5 (2.7) 5 (4.1) –0.6 –16.7 year or on index date Number of office visits with a cardiovascular 4 (5.8) 4 (7.5) –0.4 –6.2 diagnosis Number of drugs of interest dispensations in 365 17 (13.8) 19 (16.3) –1.9 –12.3 days in prior year or on index date* Number of hospitalizations in the prior year 0 (0.4) 0 (0.5) 0 –4.2

383

Appendix 48. Characteristics of raloxifene vs bisphosphonates initiators overall across all tertiles after propensity score fine stratification and weighting in Aim 2 Medication synchronization metrics (non-mail)* 0 (0.2) 0 (0.2) 0 –11.4 Number of cardiovascular hospitalizations 0 (0.3) 0 (0.4) 0 –3.9 Index copayment 44 (44.3) 45 (70.8) –1.7 –3 Vaccinations* 0 (0.7) 0 (0.7) 0 –2.1 Physician visits* 8 (6.3) 8 (7.2) –0.4 –5.8 Number of drugs of interest dispensations with days’ 3 (1.9) 4 (2.8) –0.4 –16.9 supply that overlap index date Number unique (by NDC) drugs in 365 days before or 5 (2.8) 5 (3.1) –0.5 –18.2 on index date with days’ supply that overlap index date Number of dispensations in baseline period with 5 (3.1) 6 (3.4) –0.6 –19.2 days’ supply that overlap index date* Number of unique (by NDC) drugs in 365 days before 12 (7.5) 14 (8.5) –1.3 –15.7 or on index date* Number of any drug dispensations in 365 days before or on index date*

384

Appendix 49. Patient characteristics at baseline and prior adherence measures for statin initiators in Aim 3 Age, mean (SD) 62.5 (13.1) Female, n (%) 380,982 (54.7) Index statins, n (%) Atorvastatin 187,170 (26.9) Fluvastatin 553 (0.1) Lovastatin 38,234 (5.5) Pitavastatin 5,718 (0.8) Pravastatin 135,617 (19.5) Rosuvastatin 75,649 (10.9) Simvastatin 252,218 (36.3) Days’ supply of index statins, median (IQR) 30 (30-60) Medical cost burden Copayment for index statins, median (IQR) 10 (4-18) Benefit plan type, n (%) POS 266,058 (38.3) EPO 52,254 (7.5) HMO 177,723 (25.6) IND 12,158 (1.8) PPO 45,020 (6.5) Other 141,946 (20.4) Copayment for any drugs, median (IQR) 313 (140-636) Resource utilization Number of hospitalizations in the prior year, mean (SD) 0.3 (0.7) Total days in hospital in prior year, mean (SD) 1.6 (6.1)

385

Appendix 49. Patient characteristics at baseline and prior adherence measures for statin initiators in Aim 3 Number of physician visits, mean (SD) 7.8 (6.2) Number of INR tests, mean (SD) 0.6 (3.0) Number of vaccinations, mean (SD) 0.4 (0.7) Fecal occult blood test, n (%) 65,607 (9.4) Colonoscopy, n (%) 63,589 (9.2) Mammography, n (% of females) 131,661 (34.6) Comorbidities, n (%) Previous fracture 3,295 (0.5) HIV infection 1,813 (0.3) Recent MI 13,718 (2.0) Prior MI 19,246 (2.8) Recent stroke 8,635 (1.2) Prior stroke 13,195 (1.9) Ischemic heart disease 137,186 (19.7) Transient ischemic attack 26,873 (3.9) Hypertension 513,705 (73.9) Diabetes 250,802 (36.1) Hyperlipidemia 556,472 (80.0) Heart failure 60,667 (8.7) Peripheral vascular disease 52,508 (7.6) Parkinson’s disease 4,977 (0.7) Alzheimer’s disease or dementia 30,976 (4.5) Depression 109,291 (15.7) Schizophrenia 3,498 (0.5) Rheumatoid arthritis 14,844 (2.1)

386

Appendix 49. Patient characteristics at baseline and prior adherence measures for statin initiators in Aim 3 Cancer 116,116 (16.7) Liver disease 35,625 (5.1) Renal dysfunction 95,061(13.7) Combined comorbidity score 0.8 (2.2) Other medication use, n (%) ACEIs 285,907 (41.1) ARBs 141,794 (20.4) Renin inhibitor 3,077 (0.4) Beta-blockers 250,743 (36.1) Calcium channel blockers 188,766 (27.2) Thiazide diuretics 234,828 (33.8) Loop diuretics 77,995 (11.2) Potassium-sparing diuretics 47,883 (6.9) Other antihypertensives 47,698 (6.9) Oral anticoagulants 46,449 (6.7) Heparin and low-molecular-weight heparin 8,641 (1.2) Other anticoagulants 529 (0.1) Antiplatelets 67,771 (9.6) Antianginal agents 45,909 (6.6) Digoxin 14,281 (2.1) Other lipid-lowering agents 85,005 (12.2) Antidiabetics 203,689 (29.3) Antiparkinson agents 16,936 (2.4) SSRIs 145,642 (21.0) SSNRIs 37,256 (5.4)

387

Appendix 49. Patient characteristics at baseline and prior adherence measures for statin initiators in Aim 3 TCAs 25,139 (3.6) MAO inhibitors 36 (0.0) Typical antipsychotics 2,977 (0.4) Atypical antipsychotics 22,039 (3.2) Other newer and atypical antidepressants 54,753 (7.9) Lithium 2,145 (0.3) Nonbiologic DMARDs 13,202 (1.9) Biologic DMARDs 2,674 (0.4) NSAIDs 157,688 (22.7) Selective COX-2 NSAIDs 17,345 (2.5) Agents for dementia 14,634 (2.1) Opioids 263,450 (37.9) Medication burden Number of drug dispensations in prior year, median 25 (15-42) (IQR) Number of unique drugs dispensed in prior year, mean 9.9 (5.7) (SD) Number of concurrent unique drugs dispensed on the 5.2 (3.1) index date, mean (SD) Prior adherence measures for other chronically used medications, mean (SD) Mean PDC 76.1 (24.8) Median PDC 77.8 (26.1) Maximum PDC 84.9 (24.1) Minimum PDC 64.2 (32.2) No of medication classes occurring discontinuation, 2.1 (1.6) mean (SD)

388

Disclaimer: The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.

Acknowledgment: Research reported in this report was [partially] funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#ME-1309- 06274) Further information available at: https://www.pcori.org/research-results/2014/comparing-statistical-models-predict-if-patients-will-take-new-medicine

389