Implementation and Evaluation: A Population Health Guide for Primary Care Models

October 2012 Care Continuum Alliance 701 Pennsylvania Ave. N.W., Suite 700 Washington, D.C. 20004-2694 (202) 737-5980 [email protected] www.carecontinuumalliance.org

© Copyright 2012 by Care Continuum Alliance, Inc. All Rights Reserved. No part of the material protected by this copyright notice may be reported or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without written permission from the copyright owner.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 2 Executive Summary

Why Was This Guide Developed? This Implementation and Evaluation Guide (I&E Guide) was developed by the Care Continuum Alliance to inform and guide the implementation of key components of population health and specific strategies and suggestions for primary care-centered models to embed the components into their practice. In addition, this Guide offers suggestions and resources on measuring the impact of these efforts from both a cost and a quality perspective. The Guide also offers recommendations for population health implementation for a variety of models and recognizes that models vary widely by the resources available, the culture of the practice, organization or group of organizations working together, and their level of health information technology sophistication. Ultimately, any health care delivery model that is centered around primary care can benefit from the information delivered in this Guide.

What Are the Goals of This Guide? The goal of this Guide is to offer education and guidance on the development and measurement of population health strategies embedded into the framework of a primary care-centered models. This Guide focuses on the overall value of population health strategies for primary care and how these strategies could be both implemented and measured based on the level of sophistication of the model. This Guide is intended as a resource for primary care-centered models regardless of where they are in the transformation process and offers suggestions and insight into specific tactics that can be utilized by any practice at both the clinician level as well as the organization level.

Who Is This Guide For? This Guide is for any health care entity working towards a patient-centered population health model of care. It can also be useful for individual primary care and multispecialty practices that are transforming into a model of care that is whole-patient, whole-population focused. Models that may find the information and considerations in this Guide especially useful would include: • Integrated delivery systems, • Accountable care organizations, • Patient-centered medical homes, • Primary care practices, • Multispecialty practices, • community health collaboratives, • State health exchanges, and • Large hospital systems.

At the end of this Guide is a reference section with tools and resources that offer additional detail on several of the topics discussed within the Guide itself. In addition, we have included general resources in this section that readers will also find useful.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 3 How to Use This Guide

As a resource and tool for primary care practices interested in implementing population health, this Guide can be read in its entirety for an indepth overview of the value and benefits of population health. Each section can also be a stand-alone resource on very specific pieces of population health, including the value of the process, implementation, and evaluation. The following table lists specific topics that each section covers.

Section Selected Topics Page Number

Population Health What are the key components of population health? 9 Overview As a clinician or practice manager, what are the 14 objectives and the benefits of population health? What are the key benefits of population health for my 15 patients? How can I implement population health based on my own 18 needs and resources?

Areas of Impact What kinds of impacts can population health have on my 19 practice or model of care? What is the value proposition for each of the components 21 of population health? What types of data should I consider if I am assessing the 24 health of my patient population? Why should I go through the process of risk stratifying my 21 patient population? What are some strategies that I can use to engage my 27 patients in their care? Can population health help me to better coordinate the 28 care that patients receive? What should I think about when I am trying to measure 33 savings of my population health efforts? What is a comparison group, and why is it important in an 36 evaluation process? What are leading and lagging indicators, and how will they 39 help me improve quality for my patients?

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 4 Table of Contents

Foreword...... 6 Acknowledgments...... 7 Population Health Overview...... 9 best Practices Framework...... 13 Areas of Impact...... 19 Impacts Model...... 19 the Value Proposition...... 21 Drivers of Change and Patient Engagement...... 25 care Coordination...... 28 Measuring Savings...... 33 Appendix: Special Topics...... 43 Medicaid and Underserved Populations...... Release Date: December 2012 Oncology...... Release Date: December 2012 Reference A – Health Information Technology Framework...... 44 Reference B – Population Health Management Program Evaluation...... 46 Methodological Considerations Reference C – Evaluation Study Design Considerations...... 54 Reference D – Methods to Define Outliers...... 55 Reference E – Evaluation Considerations for Small Populations...... 56 Reference F – Utilization Measures...... 59 Reference G – Self Management Measures...... 61 Reference H – Medication Adherence Measures...... 63 Reference I – Productivity Measure...... 74 Reference J – Selection Criteria Considerations...... 76 Reference K – Additional Resources...... 81 References...... 82

Figures and Tables Figure 1, Population Health Conceptual Framework...... 9 Figure 2, Population Health Process Model...... 12 Figure 3, Population Health Impacts Model...... 20 Figure 4, Population Levers for Change...... 26 Figure 5, Engagement Strategies Wheel...... 27 Figure 6, PHM Impacts on Care Coordination...... 29 Figure 7, Disease Progression Chart...... 39 Figure 8, Leading and Lagging Indicators...... 42 table 1, Population Health Objectives...... 14 table 2, Population Health Benefits...... 15 table 3, Population Health Components – Best Practice Implementation Levels for Primary Care Clinicians...... 18 table 4, Data Sources Value...... 24 table 5, PHM Drivers of Change for Primary Care...... 25 table 6, Areas for Assessing Savings...... 35 table 7, Comparison Group Options...... 37 table 8, External Comparison Sources...... 38 table 9, Utilization Measure Options...... 40

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 5 Foreword

Amid the backdrop of ongoing political debate about its merits, health care reform and all that it entails is quickly being implemented in every state. New models of care with primary care-based population health at the center are coming into focus as they rapidly propagate through the health care landscape.

Population health is a priority because of the financial and outcomes pressures inherent in reform. Not only do providers need to concern themselves with patients who seek care, they also now must engage whole populations in order to meet expectations. A population-driven, patient-centered model of care can meet the needs of all consumers regardless of where those consumers are on the continuum of health. With primary care at the center of a model surrounded by support that includes a combination of health information technology, the care team and ancillary providers, diverse care needs can be met, quality can be improved, and cost will be sustainably impacted.

Embedding population health into these new models and assessing its impact can be challenging for models already in the midst of transformation in so many other ways. The Care Continuum Alliance represents the population health industry and has developed the following Implementation and Evaluation Guide as a resource for primary care-centered models that are transitioning to population health.

The foundation for the I&E Guide is the CCA Population Health Conceptual Framework (see Figure 1). The Conceptual Framework, released in 2010, outlines the key components necessary to deliver population health to any defined population and in any setting. This Guide builds upon each of the components in the framework, offering insight into the essential purpose of each component as well as how to implement and evaluate a broad population health strategy. The Guide also incorporates several years of Care Continuum Alliance efforts that explore appropriate program evaluation criteria for population health management programs.

Many industry experts and partner organizations worked together to develop and offer comments and feedback on the Guide, and we are grateful to all who supported this important work.

Jason Cooper, MS, and David Veroff, MPP Co-Chairs, CCA Quality & Research Committee

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 6 Acknowledgments

Quality & Research Committee Kelly Shreve, Capital Blue Cross Co-Chairs: Earl Thompson, HealthFitness Jason G. Cooper, MS Barry Zajac, MHSA, Blue Cross Blue Shield of David Veroff, MPP, Health Dialog, Inc. Louisiana

Reviewers: EVALUATION DESIGN IMPACT Jason G. Cooper, MS R. Allen Frommelt, PhD, Nurtur Donald W. Fisher, PhD, CAE, American Medical Andre Gibrail, AxisMed Gestao Preventiva da Group Association Saude S.A. Helene Forte, RN, MS, PAHM, Aetna Gary Persinger, National Pharmaceutical Council, Inc. Sue Frechette, BSN, MS, MBA, Northfield Associates LLC Tina Ross-Knapp, CCP, APS Healthcare, Inc. Cindy Hochart, RN, MBA, PMP, Heartland Health David Veroff, MPP, Health Dialog, Inc. Marcia Nielsen, PhD, MPH, Patient Centered Kimberly Westrich, National Pharmaceutical Primary Care Collaborative Council, Inc. Mary Jane Osmick, MD, American Specialty Health DRIVERS OF PATIENT & PROVIDER CHANGE David Veroff, MPP, Health Dialog, Inc. Felicia Brown, RN, Blue Cross Blue Shield Association Work Groups: Helene Forte, RN, MS, PAHM, Aetna PHM PRIMARY CARE BEST PRACTICES Cynthia Hallam, RN, MBA, Blue Cross Blue Shield FRAMEWORK of Louisiana Mary Jane Osmick, MD, American Specialty Cindy Hochart, RN, MBA, PMP, Heartland Health Health Tina Ross-Knapp, CCP, APS Healthcare, Inc. Christobel E. Selecky, ZIA Healthcare Kelly Shreve, Capital Blue Cross Consultants Cindy Worrix, RN, CCP, Aetna Susan Weber, RN, CCM, MHP, StayWell Health Management CARE COORDINATION MEASURES Nancy Wilson-Ramon, IdealHealthIT Marybeth Farquhar, PhD, MSN, RN, URAC Betsy Farrell, RN, Aetna VALUE PROPOSITION FRAMEWORK Helene Forte, RN, MS, PAHM, Aetna Felicia Brown, RN, Blue Cross Blue Shield Association Andre Gibrail, AxisMed Gestao Preventiva da Saude S.A. Steven Burch, RPh, PhD, GlaxoSmithKline Garry Goddette, RPh, MBA, Alere Sue Frechette, BSN, MS, MBA, Northfield Associates LLC Diane M. Hedler, RN, MS, CHIE, Kaiser Permanente R. Allen Frommelt, PhD, Nurtur Cindy Hochart, RN, MBA, PMP, Heartland Health Thomas L. Knabel, MD, Ingenix Inc. Suzanne Janczak, Health Integrated, Inc. Jennifer Pitts, PhD, Edington Associates Peter J. Kapolas, RN, MBA, CPHQ, Healthways Tatiana Shnaiden, MD, ActiveHealth Management, Inc. Erik Lesneski, AllOne Health Cynthia O’Neill, URAC Mary Jane Osmick, MD, American Specialty Health

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 7 Urvashi Patel, PhD, Horizon Blue Cross Blue Medicaid and Underserved Populations Shield of New Jersey Jason G. Cooper, MS Gary Persinger, National Pharmaceutical Council, R. Allen Frommelt, PhD, Nurtur Inc. Carl Garrett, Centene Corporation Lisa Saheba, MPH, URAC Toni Miller, CareSource Management Group Chris Tourville, RN, MSHM, FAHM, Cigna Arnold Ari Wegh, ActiveHealth Management, Inc. SPECIAL TOPICS – SHARED DECISION-MAKING Jason G. Cooper, MS TOTAL COST SAVINGS Andrea Fong, Health Dialog David Aronoff, Nurtur Natalie Heidrich, Ethicon Endo-Surgery Jean Ann Cherry, BSN, MBA, OptumHealth Paul C. Mendelowitz, MD, MPH, ActiveHealth Natalie Heidrich, Ethicon Endo-Surgery Management, Inc. Cindy Hochart, RN, MBA, PMP, Heartland Health Julie Slezak, MS, Silverlink Communications Iver Juster, MD, ActiveHealth Management, Inc. Arnold Ari Wegh, ActiveHealth Management, Inc. Diana Potts, APS Healthcare, Inc. Carrie Wolbert, APS Healthcare, Inc. Julie Slezak, MS, Silverlink Communications David Veroff, MPP, Health Dialog, Inc.

SPECIAL TOPICS – ONCOLOGY Courtney Cantrell, RN, Aetna Jason G. Cooper, MS R. Allen Frommelt, PhD, Nurtur Jody Garey, PharmD, US Oncology Deb Harrison, US Oncology Jad Hayes, MS, ASA, MAAA, McKesson Specialty Health

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 8 Population Health Overview

The Care Continuum Alliance has developed As mentioned, there are two specific models frameworks to illustrate, both conceptually or frameworks used in this Guide that will be and operationally, the process and activities referenced: the Population Health Conceptual associated with population health. These Framework (Figure 1), which will be referred frameworks have been developed as a guide to as the “Conceptual Framework”, and the for care delivery models seeking to integrate Population Health Process Model (Figure 2), and implement population health strategies, which will be referred to as the “Process Model.” components, and processes. The population The intent of the Conceptual Framework is to health framework can be embedded into a identify the general components of population primary care integrated system in a variety health and how they relate to one another. of different ways. For example, primary care- The Conceptual Framework depicts the centered delivery models such as integrated identification, assessment and stratification of delivery systems and accountable care patients. The core of the model (central blue organizations, as well as in patient-centered box) includes the continuum of care, as well medical home practices, can adopt the as patient-centered interventions. The patient processes and key components outlined in these is central in the model, and is surrounded by frameworks to assess their own capabilities and various overlapping sources of influence on of his to guide the development of expanded and or her health. This can include, but is not limited integrated care delivery models. to, organizational interventions, Figure 1. Population Health Conceptual Framework

Patient & Provider

Primary Care

Care Continuum Alliance 9 provider interventions and family and interventions in a continuous cycle of quality community resources. Operational measures are improvement and improved patient experience. represented as are the core outcome domains. Finally, the cycle of quality improvement In addition, this process can offer information based on process learnings and outcomes is that will be extremely helpful in a clinician’s prominently depicted by the large curved green efforts to engage with patients in the patient’s arrows. plan of care. It is becoming increasingly evident that effective enrollment and engagement is key The intent of the Population Health Process to impacting the health of a patient population. Model is to help improve our understanding of the essential and detailed elements of Risk Stratification population health. This Process Model outlines The next step in the population health process the process flow associated with delivering is to stratify patients into meaningful categories the key components of population health, for patient-centered intervention targeting, using beginning with monitoring the population and information collected in the health assessments. identifying patients who are appropriate for an This process yields information that the system activity or intervention. It also includes a health can use to divide the patient population into assessment stage, followed by risk stratification, different levels to ensure ROI based on resources the application of engagement strategies, allowed. Stratification should include categories the availability of multiple communication that represent the continuum of care in the and delivery modalities, patient-centered patient population. While some organizations interventions across the care continuum use complicated and proprietary mathematical and the process of evaluating the impact of algorithms to predict risk, others use a simple these efforts in multiple domains. Finally, it count of risks to classify individuals. It is not includes a feedback loop that reflects the need our intent to prescribe how risk stratification to incorporate process and quality-related should be conducted, rather to emphasize the improvements based on learnings from the importance of having some type of stratification impact evaluation. The sections below provide in place to help align patients with appropriate a detailed description of the components of the intervention approaches, thereby maximizing the Process Model. health improvement impact of care. This process is designed to aid both the organizations and Health Assessment clinicians by helping them focus appropriate The Health Assessment section of the Process resources on those patients and segments of the Model represents the effort to assess the health population with greatest need. Furthermore, the of a specific population (i.e., patient panel, care team will be better equipped to identify diabetic population, etc.). This assessment opportunities to impact a patient’s health either typically “triangulates” by drawing on available by addressing gaps in care or by offering new types of information, including self-reported evidence-based interventions determined by a health questionnaires, health insurance claims, new diagnosis or newly discovered risk factor. laboratory and pharmacy data and clinician- documented information. Analytics and the Patient-Centered Interventions ability to combine and analyze this data is a key Whenever possible, the components of part of this process. It also is important to point population health can and should be offered out that, while there is an initial assessment, through a variety of communications and repeated measures over time are necessary to interventions in order to maximize the clinician’s demonstrate changes in health status of patients resources and reach and to accommodate and populations over time. This monitoring of the preferences and technological abilities of results in a continuous feedback loop for the patients with the ultimate goal of increased care team facilitates documenting the progress patient engagement and support for self of any population-based care over time, management. For example, some patients, establishing new baselines and adjusting care perhaps those with low risk, may prefer to

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 10 receive everything through the mail, while the conceptualization of the overall strategy others might want to participate through an and specific intervention approaches. Careful on-line program geared toward education and consideration of the chain of effects that will information sharing. Some interventions are eventually lead to the ultimate goal or outcome, best delivered directly by the provider during and inclusion of those outcomes in the impact a standard office visit, while other interventions evaluation framework, can help clinicians to and care plans may be offered through a identify the components needed to impact combination of intervention modalities. The those outcomes. Additionally, because there are Process Model includes social media as a many that contribute to the financial impact of delivery modality to reflect the increasing an intervention, explicitly outlining the predicted popularity and promise of this type of health short-, intermediate- and long-term outcomes education and support. Matching intervention can help primary care-centered models modalities to the preferences of patients likely understand the full range of impacts and the will lead to an increased level of participation expected time frame for ultimately generating and engagement, and ultimately to improved cost savings. Finally, a well-constructed patient health. conceptual outcomes framework can help with interpretation of outcomes and shed light on Impact Evaluation the practical implications of evaluation findings. To maximize the health impact of a patient- Demonstrating that short- and moderate- centered intervention or activity, it is important term outcomes are occurring as expected can to consider the environment of patients and, provide early evidence to clinicians that efforts whenever possible, to employ interventions are benefitting patients. Conversely, if early designed to create a supportive environment outcomes are contrary to expectations, early and organizational culture for patients. The link in reporting allows for midcourse corrections to the the outcomes framework between environment activities. and the actual tailored interventions represents the implicit hypothesis that population health Quality Improvement Process will impact psychosocial variables that will then Lastly, Quality Improvement Process is also drive changes in health behaviors, including represented in the both the Conceptual self-management and the use of screening and Framework and the Process Model. The cycle preventive services. Improvements in these of quality improvement includes changes to behaviors will, in turn, have a positive impact on both interventions and processes (including patient health and clinical outcomes. In addition, assessment, stratification and engagement/ the Impact Evaluation section of the Process enrollment strategies) based on process Model represents the ultimate impact on service learnings from operational measures, as well utilization, provider and patient satisfaction, and as outcomes. This process also highlights the financial outcomes derived from improvements patient's voice through data collection that will in health behaviors, health and clinical outcomes lead to an enhanced patient experience. and productivity. Health information technologies (HIT) continue Outlining a framework for an intervention’s to increase in their importance to population associated outcomes can have several practical health. CCA developed the HIT Framework to applications. It can help systematize the help identify the key components necessary design and implementation, as well as the to fully operationalize population health. evaluation processes. Whether the evaluation Reference A includes a full discussion of the HIT framework is created before or parallel to Framework, first released in Volume 5 of the the intervention deployment, it can help with Outcomes Guidelines Report.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 11 Figure 2 – Population Health Process Model

Population Monitoring / Identifi cation

Health Assessment 1 HRA Medical Claims Lab Data Other Incentives & Rewards

Risk Stratifi cation2 Incentive Reward Enrollment/ Participation Healthy Health/Emotional Chronic Illness End Of Life Engagement Outcomes Risk

Enrollment / Engagement Strategies

Communication and Intervention Delivery Modalities1,2

Mail E-mail Telephone Internet/Intranet Social Media Face-to-Face Visits

Patient-Centered Interventions1

Health Continuum Organizational Interventions • Program Referrals (External/Internal) Culture/Enviornment • Integrated/Coordinated Components

Health Promotion, Health Risk Care Coordination/ Disease/ Wellness, Management Advocacy Case Management Preventive Services Quality Improvements Based on Process Learnings and Outcomes Tailored Interventions2

Operational Measures Impact Evaluation Program Outcomes Health Status and Clinical Outcomes Psychosocial Drivers Health Behaviors

Quality of Self-Management Productivity Life Satisfaction Patient/Provider Screening /Preventive Services Service Utilization

Financial Outcomes

Time frame for Impact

1 Represents example components for each essential element. Does not necessarily reflect the universe of components. 2 Communication may utilize one or more touch points within the delivery system.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 12 Best Practices Framework staff may implement population health in a very different way. The best practices framework Population health is a framework that can section has been developed to help each model be implemented in a variety of settings and and practice understand the various options for many different populations. In addition, available for implementing population health the strategy can be implemented in varying and specifically at a tactical level what those degrees or levels based on resources, options look like. technology sophistication and the practice’s current stage of transformation. Even basic The section begins with detail on the basic differences in practices will very likely play a objectives and benefits of each population role in how population health is implemented. health component for the organization as well as For example, a small practice of primary care for the clinician, and for the patient. Following physicians, who have an electronic health these grids is a framework that offers steps to record and disease registry in place as well as a population health implementation at a tactical care coordinator, may be able to implement a level specifically for the clinician. Additional population health strategy at a very high level, frameworks will be added for the other levels at while a rural, integrated delivery system with a later date. few technology resources in place and limited

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 13 Table 1. Population Health Objectives Mary Jane Osmick, MD

Population Health Domain Organization Clinician Patient Patient Population Identification Use eligibility/administrative data to push Become aware of all patients in managed Link self to medical home and organization updated ”population list” to clinicians population Health Assessment Assess customer base demographics, values Use validated tools to assess patient health • Increase awareness of health risks and and special needs risks, preferences, activation and values within conditions defined patient panel • Increase understanding of health risks and conditions Risk Stratification • Identify cost drivers, at-risk individuals in • Prioritize at-risk patients and intervene to • Understand condition severity patient population decrease both acute and long-term risks • Understand how behaviors affect risks and • Prioritize at-risk patients for clinicians • Offer appropriate patient support based conditions • Identify and offer tailored interventions on risk and segment for segments Engagement • Support engagement of patient Offer patient-specific care plans and ancillary • Participate in defining customized care plan population interventions based on identified patient • Receive information and support tools to • Help patients access care and needs, preferences, activation, values, become activated in care interventions appropriately capabilities Patient-Centered Interventions Direct resources toward the areas of greatest Assure every at-risk patient receives timely Learn how to implement self-care plan to population risk and opportunities for health care and has access to resources to help improve/stabilize health improvement manage acute and chronic health needs Impact Evaluation • Use analytics to understand and improve • Access ”scorecard” to understand and Improve health risks and control of conditions population health interventions impact improve performance relative to others • Push “scorecard” to individual clinicians • Identify areas for care improvement

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 14 Table 2. Population Health Benefits

Population Health Domain Organization Clinician Patient Patient Population Identification Understands make-up of assigned population Focuses defined resources on identified Has medical home and trusted organization patients Health Assessment Drives organizational strategy and allocation Defines and directs staff/ancillary resources • Creates individual patient base line of resources to support identified population required to meet needs of identified • Provides opportunity for more meaningful individuals clinician encounters Risk Stratification • Identifies cost drivers, patients at risk • Provides more efficient encounter for • Provides appropriate level of care based on • Helps define interventions required to patients/clinicians condition severity support population and segments • Enables proactive interventions to maximize • Offers resources specific to identified needs outcomes and P4P payments Engagement • Reduces out-of-network utilization Enhances practice efficiency (seeing patients • Provides customized care experience • Promotes outcomes-driven use of the appropriately) while being comfortable that • Promotes partnership with clinician system the entire patient population’s needs are being met Patient-Centered Interventions • Optimizes population engagement • Enhances practice efficiency (seeing • Promotes improved likelihood of patient/ consistent with preferences, values patients appropriately) while being family participation in care plan • Focuses resources on appropriate comfortable that the entire patient • Promotes improved adherence to evidence- population cohorts population’s needs are being met based interventions • Optimizes outcomes and P4P payments Impact Evaluation • Identifies improvement opportunities • Improves health of clinician population • Provides feedback, motivation • Identifies savings opportunities • Increases revenue through quality and P4P • Promotes self-care management payments

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 15 Various primary care-centered models are In Table 3, the six population health components likely to implement population health and its are arrayed across the page from left to right: individual components in a variety of ways. How, 1. Identification, as well as how completely, the components are 2. Health Assessment, implemented will depend largely on the specific characteristics of the health care practice 3. Risk Stratification, or organization, the resources available to 4. Engagement, support the effort, and the collaborations and 5. Patient-centered Interventions, and partnerships that exist within the matrix of the organization. Although implementations may 6. Impact Evaluation. vary widely based on how organizations learn and grow, best practices will certainly emerge Each of the six components are broken down over time. One can assume that organizations into five “Population Health Best Practice will take a phased-in approach, and demonstrate Levels” (from Level I at the bottom through iterative improvement as they become more Level V at the top). In each of the five cells under sophisticated in defining their own delivery the six population health components, a brief model and responding to the need to produce description of the clinician function at each level favorable outcomes. is presented. The goal of presenting Levels I to V is to demonstrate progression towards clinician In Table 3, we present a clinician-specific best practice in each of the six components. framework which highlights how the role Moving upward in any of the six components of the clinician must change based on (from Level I to V) demonstrates enhanced the components of population health. (In integration among clinicians, improved data subsequent publications, the framework will be access and connective technology, use of valid expanded and also focus on the changing role of measurement and decision-support tools – all of organizations, as well as the patient.) which strengthen the medical home model.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 16 Each of the five best-practice levels is described knowledge of other practitioner interventions below: becomes easier. In addition, at this level all clinicians and facilities identify the concept • Level I represents (mostly) a “manual” system, and need for a patient medical home, and with no or rudimentary connection to wider are working with each other and technology systems of care. Here, the clinician (or group to make this happen. Often in this level, of clinicians) works individually with a patient, text-based, non-searchable documents exist, generally becoming aware of need only disallowing true integration of longitudinal when the patient presents for care. At this patient data. Clinicians may begin to level, the clinician tends to be reactive, and communicate with patients electronically in “waits” for individuals to identify themselves a secure and HIPAA-compliant environment. with specific health care needs. Information Clinicians begin to receive outcomes data is limited to what is shared between patient from the larger health care system, and and clinician at point of care and is refreshed performance targets are set. Clinicians may as the patient presents to the clinician time have ability to share personal health records over time. The clinician is required to function with patients. as the integrator of information – patient and practice-specific. Longitudinal patient data is • Level V is characterized by the existence difficult to identify. of valid, frequently refreshed data and information represented in a dashboard- • Level II demonstrates that clinician and staff type format to enhance the patient-clinician have an awareness of the patient population, relationship. At a high level, infrastructure, but may lack connectivity. The clinician information, and incentives are all aligned continues in “manual mode”, although some and in place for fully-coordinated patient functions may be accomplished electronically care across applicable care settings. More (i.e., billing). They may identify and focus on specifically, decision support tools flag specific diagnoses (such as diabetes, etc.) and opportunities for error reduction/patient individual complex patients who frequently safety, enhanced outcomes, etc. Here, there present for care. is full viewing of all medical information in a • Level III begins the transition toward HIPAA- compliant way for all clinicians and population health, as the practice shifts patients. Patients decide what and how much to electronic venues for some patient information they choose to have available. In interactions. A registry of specific health addition, two-way ongoing communication conditions and risks may be available to the occurs through all available electronic and clinician and staff. Proactive outreach to face-to-face modalities. Peer support is individuals identified with high risk become available for patients who choose this method possible to prevent avoidable hospitalizations of self-management. A team that supports and ED visits. At this level the clinician is still the patient population is also clearly identified reactive, but this is the earliest form of an at this level. Finally, a patient/family/support automatic “push" of patient information to the structure is in full collaboration with the clinician. clinician and coordinated care team (who have • Level IV includes the assumption that all the patient information needed to play electronic connectivity exists within the their role). practice with some ability to connect to the larger system of care. In this setting,

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 17 Table 3. Population Health Components - Best Practice Implementation Levels for Primary Care Clinicians

1 2 3 4 5 6

Patient Population Health Assessment Risk Stratification Engagement Patient-centered Impact Evaluation Identification Interventions

Clinician receives Clinician auto-notified Valid tools auto- stratify “Medical home”; Clinician/Patient Real-time feedback; Level V real-time, patient & of new or conflicting patients & population clinician monitors, collaborative care plan; outcomes meet & population specific data info requiring resolution across all clinicians; optimizes care plan & 1°, 2°, 3° prevention exceed patient , peer, at point of care gaps flagged for action care team across all focus; coordinated team population goals settings

Patient information Patient health, values, Stratification lists Clinician engages with Clinician aware of & Clinician receives Level IV available from all preferences assessed; available based on patient in “medical responds to patient patient outcome info; clinicians - ID, risks, clinician receives info claims, HA, labs, home,” coordinates needs/preferences performance goals set condition control for consideration screening info across connected focus on 1°, 2°, 3° in peer organization settings prevention

Clinician registry – key Clinician evaluates New health risks Clinician engages with Clinician focuses on Clinician unaware of Level III diagnoses, tests, Hx, health risks based identified through patient focusing on 1°, 2°, 3° prevention; patient outcome unless and condition control on year-over-year health assessments and both past and newly strategies for risks directly involved in care comparing assessments via registry lists identified risks identified PHM Best Practice Level Clinician has patient list Clinician asks patients Risk based on “frequent Clinician engages with Intervention based on Clinician unaware of Level II with diagnoses for baseline health flier” status & clinician patient episodically at current patient need patient outcome unless assessment; assesses lists with diagnoses patient presentation and known health risk(s) directly involved in care patient at the visit

Clinician identifies Clinician assesses Clinician aware of high- Clinician engages with Intervention based on Clinician unaware of patient through direct patient at the visit risk patients based on patient episodically at current patient need patient outcome unless Level I interaction and hard- “frequent flier” status patient presentation and known health risk(s) directly involved in care copy records

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 18 Areas of Impact

Once embedded in a primary care-centered The relationship between the patient and the model of care, the process of delivering clinician can have a strong impact on patient population health as outlined in the Conceptual engagement in the care process, as well as on Framework and Process Model (see pages 9 the patient’s treatment adherence, engagement and 12) can impact that model in a variety of in tailored population health interventions, self different ways. The Population Health Impacts management, and a healthy lifestyle. Model (Figure 3) offers a high level overview of the purpose, value and clinician-related impact An important feature of population health is that of each of the components of population health. it can have positive impact on both the patient Following the model are four subsections that and the clinician. As depicted in the model, specifically discuss the impact of population impacts on the clinician include, but are not health on 1) primary care, 2) drivers of change limited to, more comprehensive understanding and patient engagement, 3) care coordination, of patient health risks, more efficient and and 4) measuring savings. effective use of resources, better quality care, increased overall satisfaction, and ultimately, Impacts Model more positive patient outcomes. These patient outcomes include, but are not limited to, better The Population Health Impacts Model represents awareness and self-efficacy (psychosocial the primary elements of the Conceptual impacts), improved health behaviors, enhanced Framework (health risk assessment, risk health status and quality of life, and more stratification, engagement, patient-centered appropriate service utilization. interventions, and impact evaluation). In addition, the model represents the purpose, A final feature of the Impacts Model is the value proposition, and clinician impact for each quality improvement process that can be of these areas, as well as the patient impact in facilitated by the ongoing evaluation of impact. several important domains. Information from the impact evaluation can be used to enhance and refine the health Like the Conceptual Framework, the Population assessment process, risk stratification, the Health Impacts Model includes patient-centered intervention process and content of the interventions as the core, and the patient is interventions, and ultimately, the relationship central. But unlike the Conceptual Framework, between the patient and clinician. the patient is not alone in the center of the model. Here, the patient-clinician interaction is More detailed information about the value central. Health assessment and risk stratification proposition for each of the Model components give the clinician important information that can be found in the sections that follow. brings richness and value to the patient- For further discussion on self management clinician conversation. The patient-centered measures see Reference G, and for medication interventions give the clinician valuable tools adherence measures see Reference H. to offer patients across the health continuum.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 19 Figure 3. Population Health Impacts Model Jennifer Pitts, PhD

Health Assessment Clinician Impact Purpose Value Proposition • C omprehensive understanding of patient health/risk • C ollects important information • P rovides a comprehensive view of • E nhanced care plan about patient health risks and health health status and individual risks in • S tronger patient engagement behaviors clinician’s practice panel communication • Increased clinician work satisfaction

Risk Stratification Value Proposition Clinician Impact Purpose • Improves clinician understanding of Quality Indicators • S tratify patients into meaningful how to guide and support patient • Efficient and effective use of resources categories for personalized efforts to maintain health and/or • Quality of the care plan for individual intervention targeting reduce risks patients

Engagement in Patient- Centered Interventions PATIENT- clinician INTERACTION • Optimal use of time with patient Health Promotion • Targeted communication and education Disease Management Preventive Services • Quality of communication Case Management

• Engage in shared decision-making Continuous Quality Improvement

Population Health Across the Health Continuum

Value Proposition Purpose Clinician Impact • A ssure every at risk patient receives • P rovide resources for patients across • Improved patient health status timely care and has access to resources the health continuum to support the • Improved patient health management to help manage acute and chronic needs of the entire patient population • Improved quality and cost outcomes health needs

Impact Evaluation Patient Outcomes Clinician Impact Healthy Behaviors Quality of Life Service Utilization • Better understanding Psychosocial Drivers Clinical/Health Status of opportunity to • S elf-Management Improved • In- and Out-patient • Awareness enhance patient care • S creening & • Health Status communication and Visits • Readiness • Knowledge to self Prevention • BMI, BP, Labs relationship with • ER Visits • Self-efficacy assess and improve as • Treatment Adherence clinician • P harmacy a clinician

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 20 The Value Proposition in that particular band of the continuum. For example, providing nutrition education to all Sue Frechette, BSN, MS, MBA, R. Allen Frommelt, patients may promote behavior change for PhD, Thomas L. Knabel, MD, Tatiana Shnaiden, MD, Kelly Shreve, Earl Thompson, and Barry Zajac, MHSA some. However, targeting specific patients who are at-risk for diabetes and/or are obese As health care continues to transform, based on their risk status would be more population health is often designated as a key impactful. part of the process. The Conceptual Framework identifies the six core component to the • At the organization level, risk stratification process of delivering population health. This yields information that can be used to section reviews the value proposition for each effectively and efficiently allocate resources component as well as the ultimate impact of and lead to the greatest health impact. population health overall. Without a clear picture of the risk of a patient population, decisions regarding what type Health Assessment Value and to whom an intervention should be delivered can be imprecise and unfocused. Assessing the health of a patient benefits the For example, if a practice finds through a risk primary care-centered model for both the stratification process that its patient panel clinician as well as the organization by enhancing consists of a high percentage of healthy the available knowledge of the overall health of people and people at low health/emotional a patient and/or a group of patients. There are risk, then resources could be allocated for many types of data and data sources available interventions that focus on prevention and for this process, each adding its own value to wellness. However, if risk stratification reveals the assessment. Table 4 identifies both the data a higher percentage of patients with chronic source and the value of each. illness, then the practice may decide to invest resources in chronic care and complex case Bringing together individual level data from management. multiple sources provides value to the primary care team. For example, an ACO affiliated with Engagement Value a payer could understand how accessing claims data would be relatively easy, while an ACO in Engagement requires an alignment of personal the Boston area—where there are a relatively and program goals in the overall context of large number of smaller payers—would see that intrinsic motivation and is different from a same process as requiring a greater investment. patient’s general participation. Two relevant uses An ACO affiliated with a hospital system that from Merriam-Webster’s dictionary apply here: has implemented and enjoys a high adoption (1) emotional involvement or commitment and (2) rate of electronic health records (EHRs) would the state of being in gear.1 In short, engagement make different investment decisions than one is (1) a psychological state which (2) manifests that doesn’t, and the presence of an advanced in positive behavior change. As such, it consists regional health exchange would also affect that of self-determined participation in intervention- decision. directed activities in alignment with patient goals to which the patient is dedicated. Engaging Risk Stratification Value patients in their own health improvement from a clinician perspective includes patients Risk stratifying a patient population offers two and families engaging with their primary care key values: practice to improve health care delivery and • For the individual clinician, risk stratification patients and families engaging in the health of gives the information they need to match their communities. Engagement requires several patients to the most appropriate intervention. psychological and environmental conditions that This matching depends on where the patient must be present to some degree. The seven lands on a stratification continuum and the precursors to positive behavior change are listed nature of the factors that place the patient on the next page.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 21 The value of engagement from a patient Patient-Centered Interventions perspective is in having the capability to make The value of having a broad range of behavior change, maintain recommended organizational and tailored population health behavior, or self manage health. From the interventions is the ability to provide the best clinician and organization perspective, the (or most appropriate) intervention from the perceived value is having realistic expectations right source and delivered in the right way of the largest superset of the patient population for each patient, depending upon where they that could be impacted by an intervention, are on the health continuum, as well as to thereby improving health and lowering overall enable a measurable change in behavior with cost. corresponding measurable change in health status (or outcome). Tailored interventions will Patients and their primary care team are partners vary based on both the availability of those in patient-centered models of care. Population interventions and the current reimbursement health management requires both prevention model. In addition, the most appropriate and treatment of disease and a focus on wellness interventions can only be determined once and quality of life. The primary care practice the health of the patient population has been engages with the patient to support improved assessed and stratified by risk. Clinicians may health behaviors (e.g., medication management, initially focus on patients in the higher risk glucose monitoring, etc.) and self-management categories but ultimately will deliver a broad of chronic conditions. range of patient-centered interventions to all patients. The lack of ability and information Engagement begins with a clear understanding necessary to tailor interventions based on risk by the care team of the patient’s health and and patient need could result in ineffective behavior change goals which are documented and inefficient use of limited health care in the patient's care plan. Engagement can then resources. In addition, resources could be used be measured by assessing specific behavior unnecessarily, resulting in an increase in health changes through self or other administered care consumption without improvement in either assessment. There are several standardized tools health or cost. Examples of high-risk tailored available to accomplish this, including the Patient interventions would include: Activation Measure. In addition, an indirect • For the patient with congestive heart failure measure can be taken by monitoring behavioral (CHF) at high risk for ER use: progress toward the goals required. Examples of indirect or process measures include: • c HF clinics (typically sponsored by regular communication on progress, refills hospitals) of medications, office visits, activity logging, • Home care visits appropriate screenings performed, etc. • Home monitoring equipment (BP, HR, weight) • case and chronic care management • c aregiver and community engagement

• For patients with diabetes: • Diabetic educators/nutritionists Clinician Checklist: Precursors to Behavior Change • Medication management Sense of necessity for change. Willingness to experience anxiety or difficulty. • Self management programs (Several Awareness of the problem. national programs are being adopted by confronting the problem. hospitals.) effort toward change. • Diabetes support groups Hoping for a positive change. Social support for change.2

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 22 Clinician Checklist: Intervention Evaluation Considerations • What will be evaluated? • What parts of the program will be considered in the evaluation? • What will be considered a success? • What evidence will be used as a benchmark? • Do the conclusions make sense? Are they feasible? • How will the lessons learned be used?

• For patients who are morbidly obese: Impact evaluation determines the level of impact • b ariatric surgery centers from the intervention on these domains so that clinicians and the organization overall can • Other weight loss resources (e.g., Weight ascertain the value of their efforts. In addition, Watchers) ongoing evaluation can provide information that will inform quality improvement efforts Impact Evaluation Value and identify opportunities for additional efforts Evaluating the impact of an intervention can be moving forward. accomplished by addressing three inter-related domains: merit (quality), worth (or value, i.e. cost For more information, please refer to References effectiveness) and significance (importance). B and C.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 23 Table 4. Data Sources Value

Data Type (Data Sources) Patient-reported (HRA Claims (commercial Clinical (EHRs, HIEs, Billing (practice Health management (care programs, patient payers, TPAs, Medicare, lab results, pathology management systems, management programs, portals, other) Medicaid) and radiology reports, clearinghouses) employers, wellness pharmacy registry data)) programs)

Risk stratification value Augments and enhances Provides the broadest Provides clinical detail May provide more timely Augments and enhances the the risk assessment and detailed financial and that enhances risk and information with value risk assessment similarly to stratification process, utilization information stratification, such as similar to claims data in self- reported data, often in especially in key areas of about health care abnormal lab values that circumstances in which a more rigorous, clinically risk behavior, health and services used by a reflect disease or contribute claims data may not be objective and structured way. functional status, well- patient across clinicians; to risk and tumor specific available or timeliness is being, social support, and is the foundation of most information that improves important. psychosocial and economic common risk assessment the assessment of cancer factors. and stratification risk. methodologies.

Resource allocation/ Readiness to change and These data are used in Enhances the direction of Same as claims data. Documents health population management engagement assessments algorithms to quantify cost resources for interventions management program inputs, value help direct appropriate risk and identify specific by increasing the accuracy such as communications and initial resources; data levers of risk mitigation, of the risk assessment and other interventions, verifies about health habits such including gaps in care, to quantifying the magnitude that the program is operated as diet, exercise and risky inform resource allocation of desired clinical as planned and establishes behavior, and family history and intervention selection. improvement goals. process milestones, behavioral help guide interventions goal achievement and other and are useful in indicators of effectiveness. assessing receptivity to health interventions, as well as helping match communication channel and style with patient’s needs and preferences.

Outcomes evaluation value Data about health behavior Used to measure the Complementary to the Same as claims data. Provides evidence of specific habits and compliance, impact of population health value of claims and self- interventions and their health and functional interventions on cost, reported data; allows relationship to individuals status, and pain levels health service utilization, evaluation of achievement and their motivation or measured longitudinally risk and compliance with of clinical outcomes and/ likelihood of responding to can assess the impact of evidence-based care or improvement in clinical interventions; establishes a interventions. guidelines status. causal link between specific interventions and outcomes.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 24 Drivers of Change and population, they must reach beyond individual Patient Engagement level interventions and consider strategies that support the health of their entire patient panel. Felicia Brown, RN; Helene Forte, RN, MS, PAHM; The American Medical Group Association and Cindy Worrix, RN, CCP (AMGA) recently published a brief describing Primary care practices (or advanced primary how physician practices must change to care practices) seek to improve the health of effectively manage their populations.3 The their patients. As clinicians become increasingly brief outlines the key changes that will occur in responsible for the health of a defined Table 5.

Table 5. PHM Drivers of Change for Primary Care

Current State: Volume-Based Reimbursement Future State: Risk-Based Reimbursement (ACO/ (Fee-for-Service) Shared Savings/ Capitation and Quality-Oriented) Low financial accountability for cost of care High accountability for cost of care Defines population as patients who present at the Defines population as every patient in the provider doctor's office organization panel, regardless of whether they present at the doctor's office Minimal infrastructure (technology, staff, data, etc.) Must have infrastructure to manage the entire to manage more than the sickest/most complex population patients Culture rewards volume and operational efficiency Culture rewards optimization of cost and quality

Source: AMGA. ACOs and Population Health Management: How Physician Practices Must Change to Effectively Manage Patient Populations.

Driving change for an entire population is a identified on the engagement wheel. These key component of population health and is include identification, technology solution enabled through a variety of engagement efforts support, incentives, and interventions and throughout the entire process. In addition, these communications. The Identification section engagement efforts must include a variety of of the wheel includes strategies that may be modalities and must be tailored to meet the deployed in one of the earlier components of specific needs of patients. Figure 4 offers a view the Conceptual Framework and Process Model, of the continuum of patient health status as well including the population identification, health as potential levers that improve health based on assessment, and risk stratification components. where they fall along the continuum. During this phase of delivering population health, the clinician, while working toward The population health management industry is identifying and assessing the health of a patient leading the effort in innovative and successful panel, can be simultaneously engaging patients engagement for both consumers and patients. as well. The second section of the wheel focuses Recent research by Deloitte and the Care on technology solutions. These include tactics Continuum Alliance identified engagement like social networking, mobile technology and components embedded in many programs. electronic monitoring and are often used at Figure 5 displays these components grouped a variety of stages in the population health by the part of the process with which they process including the identification and patient- might be closely connected. For example, most centered intervention stages. The third section patient engagement begins with the process of of engagement tactics represented in the identification. The engagement diagram groups engagement wheel focuses on those deployed the key engagement components for this part of during an incentives effort. These could include the process at the top of the wheel. both internal and external motivation such Four groups of engagement strategies are as gaming motivation and standard incentive

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 25 Figure 4. Population Levers for Change

80% Members 20% Members

Healthy At Risk Acute Chronic Catastrophic End of Life No Disease Obesity Physician Visits Diabetes Head Injury Terminal Illness Smoking ER Visits Multiple Sclerosis Transplant

– Prevention – Disease/ – Health risk measures Condition – Palliative assessment – Health – C oordination Management – C omplex Case – Hospice – Lifestyle education of Care – Drug Therapy Management management – Advanced – Wellness Management Directives – Health coaching incentives – Monitoring

• Medication Management (i.e., Adherence) • resolve gaps in care

evers • early symptom ID and personal mgt • exercise – activity level • Diet • regular clinical evaluation and follow up • Weight management • Home environment issues • Stress management • Work environment issues

ey Activityey L • co-morbid complication coordination K

Source: Deloitte Consulting LLP. Consumer Engagement Across the Health Care Spectrum. Presentation at CCA Forum 11 on September 7-9, 2011, in San Francisco, CA.

programs. Incentives are often used during These efforts include health coaching, the use the intervention part of the population health of patient portals for communication between process but have also been used in the health the clinician and the patient, as well as all other assessment and outcomes phases as well. clinician-directed interactions. All of these tactics The fourth section of the wheel focuses on together comprise a comprehensive menu of engagement strategies deployed during the options available to primary care-centered intervention and communications efforts of the models to engage their patients as they direct clinicians and other members of the care team. their population health efforts.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 26 Figure 5. Engagement Strategies Wheel

1. Patient Population Identification

Predictive Modeling Health Risk Receptivity & On-line Assessment Willingness Portal, / Biometric Virtual Tools Screening 2. Technology Solutions Support Employer Based Social On-site Networking Programs ommunications

C Patient Telephonic/ Engagement Mobile Virtual Technologies Health

4. Interventions/4. Provider Monitoring Based Devices Programs

Incentive Gaming Programs Motivation

3. Incentives

Source: Deloitte Consulting LLP. Consumer Engagement Across the Health Care Spectrum. Presentation at CCA Forum 11 on September 7-9, 2011, in San Francisco, CA.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 27 Care Coordination • Identification of team members responsible for coordination; Betsy Farrell, RN, Garry Goddette, RPh, MBA, Suzanne Janczak, Erik Lesneski, and Chris Tourville, • Information exchange across care interfaces; RN, MSHM, FAHM • Interventions that support care coordination; As stated by the National Center for • Monitoring and adjustment of care; and Biotechnology Information (NCBI) 4, there are many definitions and concepts of care • evaluation of outcomes, including coordination, but they all share the key concept identification of care coordination issues. that the overall intent of care coordination is to aid in the delivery of health care at the right The fishbone diagram (Figure 6) displays the time, in the right order, in the right setting, core components of population health and their and to the right people. Population health core alignment with both AHRQ’s definition and components can help with this goal through the components of care coordination. Each of the health assessment, risk stratification and delivery components and its benefits to the process of of individualized and tailored interventions coordinating care are detailed below. for an identified patient population. Agency for Healthcare Research and Quality’s (AHRQ) Health Assessments and Risk Stratification comprehensive definitions of care coordination Assessing the health of a group of patients and states: using that information to better understand the risk and health needs of that population is an Care coordination is the deliberate important first step to organizing, planning, and organization of patient care activities coordinating overall care. The AHRQ definition between two or more participants (including of care coordination points out that ongoing the patient) involved in a patient’s care to care involves “marshalling of personnel and facilitate the appropriate delivery of health other resources needed to carry out all required care services. Organizing care involves the patient activities.”4 Understanding what each marshaling of personnel and other resources patient requires and who must work with that needed to carry out all required patient patient to meet those requirements can only be care activities, and is often managed by the accomplished through the health assessment exchange of information among participants and risk stratification process outlined in the responsible for different aspects of care.4 population health framework. In addition, In addition, AHRQ developed a comprehensive one of AHRQ’s core components specifically list of the key components of care coordination: points out that an assessment of a patient’s • e ssential care tasks and responsibilities; care coordination needs must be undertaken– another goal that can be met through the health • Assessment of a patient’s care coordination assessment and risk stratification component of needs; population health. • Development of a coordinated care plan; Engagement Engaging a patient or a group of patients in care can be accomplished through a variety AHRQ Definition of Care Coordination: Care coordination is of ways. Patient-centered engagement varies the deliberate organization of patient care activities between two depending on individual patient needs identified or more participants (including the patient) involved in a patient’s through the health assessment phase described care to facilitate the appropriate delivery of health care services. above. One part of care coordination is the Organizing care involves the marshaling of personnel and other identification of loops and methods to close the resources needed to carry out all required patient care activities, loops discovered through the health assessment and is often managed by the exchange of information among process. Ultimately, different patients will have and need different levels of engagement, and participants responsible for different aspects of care. the definition of good engagement for each patient may be different as well. Depending on

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 28

Figure 6. PHM Impacts on Care Coordination

Health Patient- Assessment & Centered Risk Interventions Stratification

Assess education Contribute to and health literacy increased patient engagement Identify risk Contribute to Assess increased patient technology and physician needs interaction

Assess resource Improve patient needs satisfaction with care delivered Begin care plan development Ensure that the care plan is successfully Track population implemented trends AHRQ Definition and Core Components of Involves patient Facilitates Care Coordination in care plan information process sharing

Helps to identify Increases patient gaps and other care motivation and related issues activation Allows for Improves continuous quality communication improvement between clinician and patient Allows for the evaluation of the impact on the process

Impact Engagement Evaluation

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 29 that support care coordination. Utilizing the AHRQ Core Components of Care Coordination: information obtained from the health assessment • essential care tasks and responsibilities and risk stratification components of population health, a clinician could deliver interventions • Assessment of a patient’s care coordination needs that were patient-centered and focused on the • Development of a coordinated care plan specific needs of the patient so that care would • Identification of team members responsible for coordination be meaningful and well coordinated. In addition, tailored and patient-centered intervention • Information exchange across care interfaces would ensure that the care plan was successfully • Interventions that support care coordination implemented and would contribute to higher levels of engagement and communication • Monitoring and adjustment of care between the clinician and the patient. • evaluation of outcomes, including identification of care coordination issues To support a population-based healthcare model, tailored clinician interventions should be applied in a standardized, evidence-based health status, different levels of engagement will manner for quality driven, cost effective results. be required. A plan of care can be developed The interventions should be managed as specific as the beginning of a long path of engagement treatment plans or care plans that can be between the patient and the clinician. The applied consistently based on chronic condition/ following examples depict different engagement diagnosis and risk. strategies depending on individual health needs:

1. Improving diabetes control – improved patient Through the health assessment and risk self-management, with reports (telephonic, stratification process, both the individual email or in-person) to the care team every clinician and organization as a whole needs three months. to determine, with a population-based 2. Improving allergy management – improved approach, what patient group(s) provides the patient self-management with on-line best opportunities for improvement in clinical assistance, seeing nurse as needed, receiving outcomes, patient adherence, healthcare prescribed allergy shots, and taking utilization and medical costs. The patient prescribed medications. population can also be prioritized based on identified risk over a specific time period such as 3. Well person with some health risks only – HgA1c level, emergency department utilization, needs support from health coach. hospitalizations, biometrics, presence of multiple chronic conditions/diagnoses, and follow-up visit Engagement can also occur in many different compliance. ways—face-to-face, on the telephone, on-line— or by multiple different modalities (e.g., person- For example, a clinician with a large number to-person, virtual, etc.). This is an important of patients with diabetes may find it helpful to consideration for primary care-centered models apply a standard assessment template as part that must coordinate care and engage all of each condition visit. A standard template patients, including those who may not come in could be added to the patient’s health record to the clinician's office setting. electronically or as a hard copy that allows multiple visit dates and the corresponding data Patient-Centered Interventions to be displayed together. The clinician could AHRQ’s definition and core components of care then use the information gathered to work with coordination point out the need to organize the patient on opportunities to improve health all patient activities and deliver interventions and self management.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 30 Components to be captured would include over time. With the initial visit and each follow- those grounded in evidence-based medicine, up visit, the clinician will need to incorporate such as: ongoing interventions, based on engagement • HgA1c, activities, and new/ongoing measures/ treatments to be addressed. Improvement • blood pressure (BP), of health care outcomes relies on educating • body mass index (BMI), patients, promoting engagement and ownership • Dental exam, of their condition, and ongoing self-care. • Dilated retinal exam, It is important for the clinician to be • Foot exam, and knowledgeable about the patient’s healthcare • Lipids. benefits and their family/social network. Part of Medication management and compliance also the treatment plan should contain information should be addressed, along with any over-the- regarding the patient’s home situation that may counter medications, vitamins or herbals the impact their ability to become/stay compliant patient may be taking. In addition, the template with their self-care management. Items for (based on the patient’s gender and age) can the clinician’s consideration may include include the appropriate preventative care inquiries regarding the patient’s financial status, measures that should be addressed, such as: constraints that could lead to medication compliance issues, transportation resource • colorectal screening, challenges that could cause difficulty with follow- • cervical cancer screening, up visits to the clinician office, and lack of family/ • breast cancer screening, and caregiver’s engagement/assistance with dietary needs or weight management. Along with • Flu and pneumovax immunizations. family/social network considerations, the clinician should consider other resources that can help Lifestyle management components also should align and coordinate the patient’s care needs. be captured as part of the patient’s overall Multiple resources may exist, including: health status and as an opportunity to address positive behavioral changes when appropriate, • c ommunity resources (e.g., dieticians, weight including: management programs, etc.), • Weight management, • educational materials (e.g., flyers, handouts, web-based resources, etc.), • Stress management, • Healthcare benefits under the patient’s • tobacco use, medical plan (e.g., case management, lifestyle • Diet/nutrition, management programs such as smoking • Physical activity/exercise, and cessation, etc.), and • Family/social network. • Pharmacy components such as home delivery services and brand-to-generic alternatives. When applying the standard template components during the initial visit, the Impact Evaluation establishment of the identified clinical and/or Among the AHRQ list of core components behavioral improvements should be aligned for care coordination is an impact evaluation with the defined clinical treatment guidelines. process, the last component identified in the Patient goals should be set that drive towards Conceptual Framework. Impact evaluation could appropriate clinical improvements and positive, contribute to the ability of a primary care- individualized behavioral goals. For example, a centered model to exchange information across patient diagnosed with diabetes and an elevated care interfaces, monitor and adjust care, and BMI can have personal goals addressing diet and evaluate outcomes. exercise. These goals will align with the desired clinical results such as HgA1c improvements

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 31 Care coordination is a planned process of A section of that data information must be care that is often initiated by the collection of process information obtained from the patient. data and information. AHRQ describes care One such patient survey has been developed coordination as a deliberate organization of by the National Transitions of Care Coalition patient activities. In order to be deliberate, the (NTOCC)5 to help the patient/caregiver to track primary care-centered model must obtain the health information and potentially prevent information necessary to make decisions around other health problems. The guide is to be the care needed and the opportunities that exist filled out “each time [patient ] meet[s] with a to improve that care. health care clinician (such as a doctor, nurse, pharmacist, or social worker); visit[s] a hospital, The overall plan of care must have in place nursing center, or other health care facility; or impact indicators of measurable goals or receive[s] care in [patient’s] home.”6 By sharing outcomes that specify the impact desired. all of this information among the care team and These indicators would be developed by the the patient, a continuous quality program of analysis of information collected at both the care would be facilitated. The planned closing point of care and practice levels of the model. of gaps could be facilitated, monitored and An example from diabetes, “… (1) the degree measured. to which clinicians adhere to the guidelines; (2) the extent of glycemic, lipid, and blood pressure Once an episode has been concluded, or for an control in patients with diabetes; and (3) the ongoing program of care for a chronic condition, roles of organizational and patient population an evaluation assessment could be more easily characteristics in affecting both clinician conducted. If shared amongst all parties, this adherence and patient outcome measures for impact evaluation would enable a thorough diabetes.”5 review of the efficacy of the planned care program. All clinicians, but especially primary Once the plan has been activated, feedback care physicians, play a key role in coordinated loops must be established to assess the progress care for their patients. The "Clinician Checklist" and evaluate the process. An assessment of below is a simple dashboard clinicians could impact would measure to what extent the plan incorporate into practice information to control has achieved the expected outcomes measured and monitor the status of care coordination at an against the impact indicators. In addition, an individual patient level. articulated data gathering and data delivery process should be well-established at the outset to assure the appropriate exchange of information.

Clinician Checklist: Care Coordination Considerations ✓ Health Assessment & Risk Stratification ✓ Engagement Identify risk Involves patient in care plan process Assess technology needs Increases patient motivation and activation Assess resource needs Improves communication between clinician and begin care plan development patient track population trends ✓ Impact Evaluation ✓ Patient-Centered Interventions Facilitates information sharing contribute to increased patient engagement Helps to identify gaps and other care related issues contribute to increased patient and clinician Allows for continuous quality improvement interaction Allows for the evaluation of the impact on the Improve patient satisfaction with care delivered process ensure that the care plan is successfully implemented

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 32 Measuring Savings specific to program selection, the concepts and considerations are very similar to those a primary David Aronoff, Cindy Hochart, RN, MBA, PMP, care-centered model would face in their own Iver Juster, MD, Diana Potts, Julie Slezak, MS, and David Veroff, MPP efforts to identify the correct patient population for cost savings analysis. It is important to understand the impact of any intervention or activity for both quality What to Measure improvement and resource allocation decisions. Evaluating the savings impact of population All organized medical practice systems strive health is an important part of the overall process to beneficially impact the cost of care while and should be considered during the planning improving or maintaining quality and the stages of the effort. The Measuring Savings experience (satisfaction) of their patients. From section offers insight into five key areas of the perspective of impact on cost of care, what evaluating population health efforts. These key measures are most appropriate? The answer areas include: depends on the practice model, how risk is shared, and the data sources available for • Identifying the Population analysis. • What to Measure • Using a Comparison Group When selecting a measure of cost impact, consideration of the analytic intent is necessary • The Role Time Plays in Evaluation to make sure the measure is accurate and • Leading and Lagging Indicators complete for its purpose. For example, from  a population health perspective, a review of Population Identification cost should include (in addition to the cost of Identifying a population is important for both paid claims) the cost from the perspective of the deployment of interventions and resources the patient—known as out-of-pocket (OOP) as well as for the process of evaluating those cost—which includes deductibles, coinsurance, interventions. For cost savings evaluations, and copays. This perspective is called allowed, population identification is the first step in covered, or reimbursed cost.I An ACO that is an accurate evaluation of savings. Although analyzing cost to the system would likely be the majority of primary care-centered models interested in only costs paid by the payer. have a clearly defined set of patients, they are held accountable for determining which set of Because allowed cost neutralizes the impact of patients received which intervention and for year-to-year changes in the fraction of cost of what time period. This accountability plays an care that is borne by the payer, it should be used important role in determining the cost savings (if claims data are available through a partnership achieved through these activities. The process with a third-party payer) to evaluate the extent for identifying patients to be included in a to which specific interventions, or the program cost savings evaluation can be done in several or practice as a whole, impacted financial different ways. Tools common to primary care- outcomes. Allowed cost is net of contractual centered models can be helpful in this process. differences and benefit design variability. These include accessing information through However, allowed costs differ across payers. This an electronic health record, disease registry, factor should be considered if multiple payer care management system or patient portal. data sets are included in the analysis. The Care Continuum Alliance also developed However, billed charges should not be used to selection criteria that offers specific insight evaluate cost outcomes because of their often and suggestions for selecting a population remote connection to what is allowed or paid. specifically for evaluation and this work is All cost evaluations based on claims data are included in Reference J. Although this work is inherently impacted by the variance of the

l In some settings, Allowed, Covered, and Reimbursed costs have different definitions; here we use a simple definition: the cost specified in a contract with the payer, which—barring volume and other special discounts — is the same as the total amount reimbursed to the service provider.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 33 claims payment process such as claims lag, but if measure for a service line in a continuous claims are the data source, allowed cost is often improvement process where the goal is to drive the best measure of services provided across inefficiencies out of a process. A comparison of employer plans and populations. net revenue to the fixed and variable costs of Many ACOs have available to them, financial providing care will be appropriate for an ACO data from the collection of fees for clinical that is operating under a capitated model. services rendered which will include payments As this discussion illustrates, the best selection from commercial and government payers as well of the measure of cost needs to be made as from the collection of the out-of-pocket costs with a thoughtful review of the purpose of the to the consumer. The measure these entities analysis, the data sources available and the will use is reflective of their role in providing perspective for which audience, the analysis is care. Net revenue is, for a clinician, the most intended. ACOs and other organizations that appropriate measure for global analysis of share financial risks can use a combination of system impact related to the cost of care to metrics to measure cost savings. These entities an ACO that combines the interests of several can use a hybrid of prevention-based initiatives constituents such as a hospital and physician in conjunction with health maintenance to offset partnerships. current and future costs. These are savings which can be compared to previous periods and similar The true cost of providing care can be broken population groups who did not receive the same down into fixed and variable costs within a care initiatives. system or practice assuming the availability of a robust cost accounting methodology. This level See Table 6 below. For more information, please of cost detail is used as the most proximate refer to References D and E.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 34 Table 6. Areas for Assessing Savings

Measure Description Pro Con Total Billed Charges Amount billed by clinician Does not reflect benefit design difference and contracted payment amounts. The only patients who pay billed charges are self-pay who do not negotiate a settlement. Net Charges Amount remaining after applying This measure allows inclusion Is not reflective of benefit negotiated third party payer of the consumer OOP portion design differences. contracts. of the charge. Allowed Charge Charge after contractual Use of allowed charge levels Claims based – requires arrangements and benefit design the playing field for analysis claims availability and factors are applied. across employer and product typically has a 2-3 month design. time lag between service and availability for analysis. Total Cost Direct and indirect costs of providing Clinicians will use this measure From financial /billing service. to assess profitability of a systems. May include costs department or service line as a not relevant to the direct part of the enterprise. provision of the health care service. Total Revenue Amount of dollars brought into the Includes pass-through organization. dollars that do not reflect the organization’s actual available revenue such as room and board costs for care of hospice patients in long-term care facilities. Net Revenue Portion of total revenue that is Reflects the amount of income Does not reflect the actual available for ACO operations. to ACO before removing cost of care. actual cost to deliver service. Direct Costs Portion of total cost that is directly Measure that is most proximal Does not include related to the provision of service. to the actual service before administrative and other dividing into fixed and variable indirect costs incurred by costs. the organization Fixed Cost Portion of direct costs that stays Used when trying to evaluate Difficult to make changes constant across all services provided. reductions in cost for the to fixed costs. overhead portion of costs such Only reflects a portion of as building used for service. the cost to provide service Variable Cost Portion of direct costs that change Used when trying to adjust Only reflects a portion of with the number of services provided the unit cost of a service (i.e., the cost to provide service. (i.e., staff costs). the number of nursing hrs/ procedure).

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 35 Using a Comparison Group the insured population that doesn’t participate in In order to assess the effectiveness of primary the current activity or visit the same primary care care-centered model activities, a form of practice and appropriately match to deal with comparison should be incorporated into the identified confounders. If a comparison group is overall evaluation. A comparison will help to difficult to find there are other methods as well. determine the true impact of the intervention (See References B and C). Table 7 outlines the vs. other confounding factors that may have viable comparison group methods that could impacted the desired outcome during the be used for a wellness/prevention program and same time that the intervention was taking gives additional insight into the considerations place. An example of this could be evaluating for each method. the effectiveness of a weight management intervention for a patient population that also Without a legitimate population-based was exposed to other activities focused on comparison population, benchmarks or industry weight management, such as a workplace statistics could be used as simple comparisons. wellness program. We could look at a Table 8 provides some examples of sources for comparison patient population from a sample of these comparisons.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 36 Table C1. Comparison Group Options

METHOD TO DEVELOP Randomized concurrent Matched control Non-participating Historical control A COMPARISON control individuals comparison (as defined by CCA GROUP guidelines) General description of Intervened population Intervened population Intervened participating Intervened population method compared with compared with individuals compared compared with similarly individuals randomly individuals from non- with intervened non- identified population in a selected to have services participating groups participating individuals baseline period (with costs withheld matched to have similar intervened individuals characteristics as trended forward) intervened individuals Comparison time frame Concurrent to Concurrent to Concurrent to Prior period intervention intervention intervention Population selection bias None Somewhat significant Significant None Source of comparison Population for Population for Population for Population for group whom program was whom program was whom program was whom program was implemented, randomly implemented, purchaser implemented, individual implemented, in prior selected group withheld decision to not decision to not period from program participate participate Trend factor Not required Not required Not required Individuals without measured conditions in population for whom program was implemented “Regression to mean” None None None Low issue Credibility of causal Extremely strong (gold Strong (with proper Poor Moderate statements standard) design) Applicability to all types Strong Strong Strong Uncertain of programs/program designs Program sponsor Strong Moderate (confidentiality None Low resistance to approach of non-participating groups can be an issue) Ease of implementation Very difficult Difficult Easy Somewhat difficult Clarity of method to lay Very clear Very unclear Clear Somewhat unclear audiences Multiyear application vs. Much harder About the same About the same Somewhat harder single-year application Method availability Rarely possible Occasionally possible Frequently possible Usually possible Bleed of interventions to Control group likely Control group likely Control group likely Control group unlikely to comparison population to get provider-based to get provider-based to get provider-based get any interventions interventions interventions interventions and softer member-based interventions Key strengths Generally seen as gold Potential for a concurrent Easy to implement Relatively universally standard evaluation comparison population Method is easy to available method method with many of the same understand Application of well- Ease/strength of characteristics as the accepted actuarial interpretation intervened population processes Credibility of casual statements can be very strong Key problems/biases Sponsor resistance to Difficult to get access Very significant bias Difficult to ensure implementation to non•participating associated with equivalence between Difficulty of multiyear groups of sufficient size differences in motivation baseline and intervention assessment in same geographic between participants year (particularly in area, with similar benefit Likely not possible in populations with many structure “opt-out” program shifts in size,composition, “Black box” approach models and/or benefit structure) difficult to understand Difficulty in deriving credible trend factor

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 37 Table 8. External Comparison Sources

Identified measures of interest Potential industry comparisons Outreach Disease management outreach Enrollment Disease management Health and wellness programs Health plan enrollment Loyalty (is patient continuing with same Health plan loyalty measures clinicians) Engagement Engagement company statistics: • Wellness programs • Chronic care management programs • Health plan information Behavior change Historical self-reported patient information Research benchmarks Adherence Pharmacy adherence stats DM adherence to treatment plan stats Moving closer to clinical targets HEDIS Morbidity Disease management, Research literature Mortality Standard statistics Functional status Research Quality of life Research Productivity performance • Research • See also Reference I.

The Role Time Plays in Evaluation What role does it play and why is it important? Primary care-based population health programs will be oriented to meet cost savings objectives within a defined time horizon. For example, shared savings models defined in the Medicare Pioneer ACO program set savings objectives that increase every year. Other programs define savings explicitly as single year objectives. Clinicians short- and/or long-term savings objectives in mind. Population health solutions may differ as a result of these different time frames. For example, services to improve care coordination, reduce readmissions, help with appropriate specialist selection, improve short- term treatment decision-making, and improve self-care skills would be more likely to drive

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 38 short-term cost savings than services supporting What are key takeaways and recommendations? tobacco cessation, weight loss, improved • Short term savings targets will likely focus the physical activity, and improve screening rates. population health interventions and will result Even for short-term focused programs, there in less attention to longer term preventive may be an accumulation of cost impact that activities and changing health behaviors. would be worthwhile to assess, since benefits • Savings accumulate over time with any should grow over time. program that has sticky behavior or health utilization impact. In addition, as the time horizon for measurement increases, more data accumulates and statistical • Short-term medical cost impact is harder to variance in costs reduces. Longer time horizons detect than longer term medical cost impact may therefore enable more certainty in cost because of statistical variance in medical cost savings assessments. Proxy measures that are data. In short-term cost assessments, it may less highly variable than actual costs, such as be necessary to estimate cost savings using utilization rates, may be necessary in shorter indicators that are less variable than medical term cost savings assessments. costs (such as utilization rates). Table 9 reviews several options for utilization measures. Reference F provides additional insight into utilization measures.

Leading and Lagging Indicators Safely reducing the per-capita cost of healthcare is one of the three “triple aims” promoted by the Institute of Healthcare Improvement as required to improve the U.S. healthcare system. The medical cost of healthcare is driven by several components including: the underlying risk and

Figure 7. Disease Progression Chart

progression chronic AT-Risk pre-chronic End of Life Disease complications

Indentify Risks and Opportunities to Improve Health

Lifestyle-related Risks Risks for Progression Risks for Complications Risks to Quality of Life

Enroll: Get Started • Participate: Make Healthy Changes Sustain: Maintain Healthy Habits Lifelong, Activated Engagement with Wellness

Through effective population health management, the population can be “moved to the left” by preventing or slowing individuals’ progress to the right.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 39 Table 9. Utilization Measures Options

Measure Utility/Comments All-cause utilization of diseased population • relatively sensitive for major cost driver impacted by chronic care management programs • Measures impact on comorbidities as well as primary conditions of interest • Does not test accuracy of identification of diseased population Condition-specific utilization of diseased population • More sensitive for expected key impact of chronic care manage- ment programs but does not measure chronic care management impact on comorbidities • Does not test accuracy of identification of diseased population Condition-specific utilization of entire popula- • Specific for one expected impact of chronic care management tion – known as “plausibility indicator” by Disease programs but insensitive to other possible program impacts Management Purchasing Consortium International • Does not measure chronic care management impact on (DMPC) (Variant 1 – one measure) comorbidities • Sensitive to condition prevalence changes, so prevalence adjustment required • May serve as “end to end” test of chronic care management identification, outreach, enrollment, engagement, impact, retention, etc. Condition-specific utilization of entire population – • Same as for Variant 1 known as “plausibility indicators” by DMPC (Variant • Adds ED visits, which reflect additional chronic care management 2 – two separate measures) impact beyond admissions reductions Condition-specific utilization of entire population – • Same as for Variant 1 (Variant 3 – one combined measure) • Adds ED visits to admissions in a single weighted aggregate measure that may reflect additional chronic care management impacts beyond just admissions reductions All-cause utilization of entire population • the most inclusive measure of utilization but with attribution problems • May be insensitive to chronic care management program impact, as many other influences impact total utilization changes over time Utilization x unit cost to build up total cost estimate • relatively sensitive for major cost driver impacted by chronic care (surrogate for measured cost approach) management programs • Measures impact on comorbidities as well as primary conditions of interest • Does not test accuracy of identification of diseased population

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 40 demographic characteristics of the population, important to take into account the expected benefit design (which may induce changes in time course over which savings can be expected demand), utilization (quantity and distribution) of to occur (discussed in more detail above). The services and pricing of services, equipment, and relationship between markers of engagement medications. and savings varies widely depending on clinical or risk scenario, intervention, and a patient’s While cost trend may be influenced in the short baseline risk. As a general rule the further term by changes in benefits design and pricing to the right on the continuum of health, the (and to some extent redistribution of service shorter the time between engagement and utilization), sustained impact on per capita cost impact. For example, the impact time course depends on long-term reduction in risk-for-age, of screening mammography in a population of and avoidable utilization of services such as ER 50-year-old women on the incidence of breast visits, hospitalizations, and procedures that are cancer is measured in decades, while the impact sensitive to the preferences of clinicians and their of starting beta-blockers after heart attack is patients (which results in unwarranted variations measured in days. in the use of such procedures). Understanding how markers of engagement with The ultimate aim in sustainably lowering health causes the ultimately desired outcomes healthcare costs is to prevent them through of utilization and cost savings supports the reducing the population’s risk for developing development of a framework of leading chronic or complex acute disease, and through (early markers) and lagging (later utilization preventing or slowing the rate of progression and monetary results) indicators, similar to and exacerbations of chronic disease. This the often-used framework of leading and is expressed by the concept of “moving the lagging economic indicators as a framework or population to the left” (Figure 7) through initial, dashboard of a nation’s economic health. sustained, and ultimately lifelong engagement with one’s health-improvement and risk- Think of lagging indicators—utilization and reduction opportunities as they evolve over time. cost savings—as the result of processes of The population health perspective promotes health care, care management, and patients’ engagement with opportunities to improve engagement with their health—the leading health and prevent or mitigate risk as key to indicators. When leading indicators are strong, slowing or preventing the progression “to the the (highly valued) lagging indicators will follow. right” at the individual level, and to “moving to We consider leading indicators to include the left” at the population level. voluntary personal behaviors (e.g. tobacco use, medication persistence); healthcare processes When implementing a framework of measures by (screening for subclinical disease or monitoring which to gauge the financial effectiveness (i.e., for exacerbations, worsening, or complications cost savings of population health embedded of established disease); healthcare management into primary care-centered models) it is processes (engaging patients or members in

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 41 Figure 8. Leading and Lagging Indicators

Leading INDICATORS: Forecasts outcomes we care about

• Identification, stratification, and methods of offering engagement • e nrollment (initial engagement) • Sustained and Activated engagement ... life-long • behavior change ... maintenance • Processes of care (taking a blood pressure, monitoring HbA1c, prescribing a drug) • Adherence to treatment.... medications • Achieving clinical targets (e.g., blood pressure, HbA1c, LDL cholesterol, BMI)

LAGGING INDICATORS: Outcomes we care about

• Morbidity (chronic conditions, exacerbations, disease progression) • Mortality • Functional status • Quality of life and well-being • Productivity (absenteeism, disability, and presenteeism) and performance • Healthcare claims cost

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 42 Appendix: Special Topics

The components of population health discussed and the considerations that a primary care- in this Guide can be applied to many different centered model may need to address specific populations, but there are populations that to this population. Future additions will include require special considerations based on the a section focused on an oncology population characteristics of that population. The Special as well as a section focused on dual eligibles— Topics section of this Guide is devoted to persons who qualify for both Medicare and these populations. This section will include a Medicaid coverage. discussion regarding a Medicaid population

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 43 Reference A

Health Information Technology Committee and the PHM Outcomes Workgroup, Framework attempts to identify the key components of both health information and health Population health relies increasingly on data- technology necessary to fully operationalize driven, HIT-supported interventions. Numerous population health management programs. federal government initiatives are underway These technologies extend beyond electronic to expand the adoption and implementation medical records to encompass a wide of health information technologies (HIT). spectrum of innovative technology devices and Appropriately, these initiatives focus on applications. As such, the HIT Framework is establishing standards, defining effective use, intended to complement and expand upon the and measuring quality improvement achieved Population Health Conceptual Framework by via the role of HIT. As technology continues first identifying the upper-most levels of data to evolve, so will its ability to enhance the exchange, envisioned by the Health Information health consumer’s experience and enable Technology for Economic and Clinical Health communications and information sharing across (HITECH) Act (a part of the American Recovery all health care settings. and Reinvestment Act, Public Law 111-5) as “health information exchanges,”7 through Care Continuum Alliance members develop, broad systems and information hubs, such as utilize and support numerous health person-level databases, continuing to identify technologies and components. The HIT components through to end-user, or consumer- Framework, developed jointly by the HIT enable, devices.

Figure A1 – HIT Framework

Regional Data Liquidity

Systems- and Person-Level Databases

Infrastructure and Services ACOs, HIEs, RHIOs EHR, Lab Communication and Claims Rules Engines, Enabling Devices Processing Decision Support Systems* Tools, Intervention- Home Health End-User Level Databases* Hubs, PHR, Medical Devices Monitoring Cell Phones, Smart Phones, Devices* IVR, iPads, Personal Computers, Digital TVs*

* Examples only; not meant to be all-inclusive

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 44 Regional Data Liquidity patients. This infrastructure combined with the This component of the HIT Framework circles complementary services enables the enrollment the entire framework and refers to regional and engagement process as well as the process data hubs that compile and distribute data of communication and intervention delivery. for a population that exists within a specific Examples of infrastructure tools and the services geographic region. Data can be collected from included within them are rules engines, decision a variety of health delivery sources, including support tools and intervention-level databases. hospitals and providers, as well as non- traditional sources, such as community centers Communication and Enabling Devices and public health agencies. Examples of these This component is focused on devices that allow regional or high-level data hubs are envisioned and enhance communication between and through accountable care organizations, health among health care providers and health care information exchanges and regional health consumers. These devices also enhance the information organizations. ability for providers and consumers to exchange and share information and contribute to most Systems- and Person-Level Databases of the processes outlined in the Population This component of the HIT Framework refers Health Process Model, including the process of to the databases and systems used to identify, enrolling and engaging, program delivery and assess, stratify, and enroll a population. Both measurement of outcomes. Examples of these systems and databases enable the assessment, devices include home health hubs, personal stratification and engagement processes health records and monitoring devices. identified in the Population Health Process Model, as well as ability to measure the program End-User Medical Devices outcomes. This section includes both the data This component of the HIT Framework includes and technology needed to perform these devices used by health care consumers to functions. Examples of these databases and communicate and exchange information with systems include electronic health records, as well health care providers, including, but not limited as lab and claims processing systems. to physicians. These devices contribute to the process of successfully communicating and Infrastructure and Services delivering program components. Advancements The third component of the HIT Framework in technology have produced a variety of such is focused on the information and technology devices, including personal computers, cell needed to support health providers in their phones and iPads. efforts to enhance the services they provide to

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 45 Reference B

Methodological Considerations • Establish framework to judge relative strengths of evaluation approaches. The When conceiving of guidelines for PHM health services research and medical care program evaluation, the workgroup started literature are replete with examples of study with some general goals and objectives. It was designs that utilize various sources of data to recognized that the environment in which PHM evaluate the impacts of different interventions measurement occurs is sufficiently different, on populations. This diversity is useful to less well-developed and more varied from meet the needs of organizations that might chronic care management; and that it would have limited access to certain types of data be unreasonable to imagine that there would but plentiful access to other types. In this be consensus around a given methodology or context, and because there has not been sufficient evidence to support the preference of a convergence of methodologies in PHM one well-designed and executed approach over evaluation as in chronic care management, another. it is useful to remain open-minded and have a means to assess each methodology in an The objectives are to move the industry forward objective and systematic way. In addition, in the following areas: the opportunity for thought leaders to • Establish minimum requirements for differentiate their products based on rigor of attribution. As will be discussed later, evaluation design is likely to raise the quality evaluating programs designed and intended of these evaluations for the industry. to affect an entire population challenges • Define a research agenda to encourage many of the assumptions embedded in the studies to benchmark methodology and disease management evaluation guidelines establish empirical standards for PHM presented in Volume 4. For example, evaluations. Before identifying even minimal the trend methodology, along with risk standards and eventually establishing a adjustment, helps establish estimates of rigorous benchmark for PHM program what would have occurred if the disease evaluations, significant research is needed on management program had not intervened, methodology. A “comparative effectiveness” all things being equal (ceteris paribus), which approach to various experimental and quasi- helped justify attribution – the ability to claim experimental study designs is needed to that changes in costs were caused by the assess their validity and potential weaknesses. program’s interventions. In the PHM world, Since behavioral change is a key focus where there is a potential for all strata of of PHM, social intervention evaluation a population to be receiving interventions approaches 8,9 should be further studied and of some kind, there is no reasonable way applied. These guidelines can help frame the to estimate what the “underlying” trend is research questions that will best advance the and therefore no good way to establish a PHM industry. similar ceteris paribus assumption. Setting standards for identifying what is required to Client/Customer Expectations make a reasonable claim of causality, without A PHM strategy includes coordinated program conducting a randomized controlled trial, is initiatives for the continuum of needs across an important step. a population and creates an expectation for • Set minimum standards for adjustment. outcomes measurement with equivalent span. The populations targeted for PHM are Unlike stand-alone chronic care management, more varied and opportunities for selection there is no intent to leave significant portions bias and non-comparability of study and of the eligible population unaffected by the comparison groups are expanded. There is a programs. need to match adjustment approaches to the study design to minimize the impact of bias and confounding.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 46 This comprehensiveness in intervention consistent with the intent of the measurement approach creates two specific evaluation under specific circumstances. The needs: a consolidated methodology assessing measurement can be reasonably expected performance across all program components; and to represent what it is intended to measure a component-level assessment to differentiate – for example, a difference in the utilization the effectiveness of the individual programs and of services from what would otherwise have interventions within the overall approach. been expected for this population. 3. External validity: the results of the For the purpose of focusing on a definable and evaluation can be generalized beyond the common level of comprehensiveness and an specific analysis, i.e., the measurement can achievable scope, these initial PHM evaluation be reasonably expected to represent the guidelines focus on the combination of common impact of the intervention across instances chronic disease and wellness programs. and settings. As such, measures and study outcomes are comparable broadly across Methodology Framework: programs and/or organizations – a difference Study Design in the utilization of services from what The objective of measurement is to render would otherwise be expected for any similar a quantitative or qualitative description. To population, for example. describe a program outcome, one can describe a state or a change that, to be valuable, must be The following framework outlines the prevailing valid. Validity has many flavors. For this purpose, approaches that are being or could be used one can imagine a hierarchy of validity types: to assess program effectiveness. As previous 1. Accuracy: measurements correctly capture Outcomes Guidelines Reports have noted, the data and compute metrics. This definition is more rigorous a method’s ability to produce independent of the meaning of the results. conclusive or valid results, the more impractical it is likely to be for routine, periodic reporting. 2. Internal validity: metrics and study design These are noted in both the complexity and key are constructed and executed in a way issues rows.

Table B1 – Methodology Framework: Study Design

DESIGN Randomized Non-Experimental Concurrent Trial Historic Control Pre/Post Industry Trend

Source of Population for Population for Composite of Expected Trend Population for Comparison Group whom program was whom program was populations for (typically observed whom program was implemented, randomly not implemented whom program was trend of individuals implemented, in selected group withheld (not contracted), implemented without measured prior year from program or, population for conditions in whom program population for was implemented, whom program but individuals who was implemented) were not contacted applied to or chose not to population for participate. whom program was implemented

Threats to Validity Very low, internal and Low (controlling for Moderate, internal High, external High, internal external appropriate selection and external and external bias), internal and external

Complexity Very high High Currently high Low Very low

Key Issues Not suited for routine Cost, availability of High upfront Questionable Regression to mean, reporting due to cost, comparison group and investment cost validity without risk selection bias timing, and limited complete comparison adjustment generalizability group data

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 47 Randomized Controlled Trial (RCT): A • Non-contacted individuals: The comparison randomized controlled trial is a scientific group consists of individuals who were experiment to evaluate the effectiveness of eligible for the intervention but not contacted an intervention (or a product). Subjects are by the chronic care management organization. randomly allocated to the treatment group that Examples are individuals with incomplete receives the intervention and the control group addresses or missing phone numbers. that does not. The random assignment is meant • Non-participants: The comparison group to maximize the probability that both groups consists of individuals who were identified for are fully comparable at baseline so that later a chronic care management intervention but differences can be attributed to the intervention. chose not to participate. Thus, the RCT is considered the gold standard of study design. Its disadvantages are the Concurrent Industry Trend: This trend would complexity of organizing such a trial, concerns be estimated by combining longitudinal data about the limited generalizability of the findings from a large number of care management because of the tight control over the intervention organizations that reflect the overall industry under trial conditions, and the reluctance of reasonably well. The trend would then become assigning subjects to a control group that may the equivalent of the S&P 500 for the chronic not receive treatment. care management industry, i.e., a national benchmark against which individual programs Study designs with non-experimental would be compared. For a fair comparison, the comparison groups: A strong non-experimental trend would have to be adjusted to account for design requires the comparison group to be differences in region, case-mix, etc., between drawn from comparable populations or to use the national peer group data and the individual individuals matched to the intervention group program. Of note, a program’s effect would not members. Comparability can also be achieved reflect the impact compared to no treatment but with statistical adjustment. Below are some compared to the rest of the industry. examples of control group opportunities: • Non-contracted group: This can be referred Historic control: This design compares the to as natural experiment. In this design, trend in the sub-population equivalent to the researchers try to find a naturally occurring intervention population prior to start of the control group that is equivalent to the intervention and afterward to an expected trend treatment group in all other aspects but based on analysis of the rest of the population. exposure to the intervention. Thus, the design Any difference in observed findings to the is similar to an RCT but without the random expected results is credited to the intervention. assignment of individuals. For example, if a health plan offers a care management Pre-post comparison: This design compares intervention to the fully insured but not to the baseline to post-intervention data. The self-insured business lines in the same region, difference is counted as program effect. the members in the self-insured products could be investigated as an equivalent, non- Most industry activity has been focused in the randomly assigned comparison group. historic control method. Comparison group • Non-qualified individuals:T he comparison methods, which are derived from within the group consists of individuals who were not same population in which the program was identified for the chronic care management implemented, such as non-participants, are intervention. This group can include currently emerging both in terms of population individuals with no chronic conditions, with and individual (matched control) comparisons. chronic conditions that are not part of the Limited opportunities arise to conduct studies, chronic care management contract, or with in which the comparison group is identified chronic conditions that are part of the chronic from outside the program population, such as care management contract but fail to meet a different employer group and region of the program inclusion criteria. health plan. Even rarer is the situation to conduct

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 48 a prospective randomized control design which outlier adjustments. Linear regression modeling is best suited for special studies and research. simultaneously adjusts and assesses the association of multiple variables to the outcome Methodology Framework: measure. When using a matched-control Adjustment design to assess outcome effect, propensity Measurement adjustment is another framework score weights can be used to determine the dimension, in addition to study design or appropriateness of the match between the attribution. Measurement comparability individual participant and non-participants requires appropriate adjustments for factors based on the appropriate covariates. These that influence the measure but are not related scores are a result of regression modeling. to the program intervention. For instance, This approach requires careful attention due adjustments for patient demographics (age and to the various parameter estimates involved. gender) are commonly applied to utilization Otherwise, the appearance of rigor may mask rates. Adjustment can be considered along a invalid matches. continuum similar to study design approach related to complexity and validity, except Attribution Adequacy – that the adjustment methodology continuum Illustration & Examples is effectively cumulative; each successively Attribution Adequacy – Illustration of Framework ”higher” level of adjustment is intended to In an effort to illustrate adequacy of various actually or effectively include the methodologies approaches for valid attribution to the below it. Outcomes Guidelines Report, Vol. 4 interventions or program provided, the following (see Volume 4, page 60), documents whether schematic combines study design and risk and how to approach risk adjustment for a adjustment dimensions along a “strength” particular program, which is dependent upon continuum of low to high. The two dimensions the outcome being measured and whether that create an area that has been further delineated measure is: as “adequate” and “not adequate” for • impacted only by exogenous confounders routine reporting of valid program outcomes to and not the interventions, where there is no purchasers of population health management concern that program impacts will be altered programs. by the adjustment (e.g., outlier exclusions, non-chronic trend), or Approaches with lower strengths of attribution, • impacted by exogenous confounders as well such as pre-post and historic control, may be as by program interventions that potentially adequate when certain levels of adjustment may be inappropriately distorted or are included. Though the adequacy concept is discounted by the adjustment (e.g., condition new, the schematic and the shaded areas are prevalence or severity, case mix). in line with previous CCA outcomes guideline recommendations. Control group approaches may require statistical adjustments, in addition to stratification and

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 49 Chart B1 – Attribution Adequacy Illustration of Framework

Strength of Adjustment HIGH

Statistical modeling (incorporating HIGH below components) ty

Stratification and complexi and outlier restrictions

ethodology cost M Outlier No restrictions ade t only qua te LOW None LOW HIGH Pre-post Historic Control Concurrent Non-experimental RCT Industry trend Comparison Group Strength of attribution approach

Attribution Adequacy Examples EXAMPLE C – A participant/non-participant Examples have been provided to clarify how comparison at the population level. The two this framework can be used as a guideline in comparison groups are stratified by population assessing validity among various study designs demographics and condition. To further control and measurement adjustments. An example self-selection bias, each of the sub-stratums is with insufficient adjustment is provided. normalized/adjusted for risk differences. Adequate examples include previously endorsed approaches, as well as emerging methods. EXAMPLE D – A program eligible/non-eligible member-level comparison. Covariates in a linear EXAMPLE A – A historical control (pre/ regression model adjust for variables associated post) comparison of program participants with the outcome measure but not directly excluding for high-cost outliers. Factors that are attributed to the program intervention. considered inadequate include: participants as their own control group vs. eligible population; EXAMPLE E – A program eligible/non-eligible using their own experience for historic trend; not member-level comparison. Eligible individuals stratifying by chronic condition. are matched to non-eligible members based on similar propensity scores. Outcome changes EXAMPLE B - A historical control (pre/post) between baseline and program time periods are comparison of program eligible members. then assessed. This is commonly referred to as This example outlines a CCA recommended “difference in differences” analysis. approacha incorporating non-chronic trend adjustments, condition stratification and outlier exclusions.

a See “Trend.” Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010, pp. 71-73.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 50 Chart B2 – Attribution Adequacy Examples

Example A: Pre-Post Comparison Strength of Adjustment HIGH Intervention group: Members Statistical actively participating in chronic care modeling management program (incorporating HIGH below Comparison strategy: Historic components) cost trend based on 3 years of pre- A ty intervention data Level of analysis: Aggregated Stratification and complexi and outlier data restrictions Adjustment: Truncation of outlier ethodology cost cost at $100K/PMPY M Outlier No restrictions ade t only qua te LOW None LOW HIGH Pre-post Historic Control Concurrent Non-experimental RCT Industry trend Comparison Group

Strength of attribution approach

Example B: Pre-Post CCA Recommended Method Strength of Adjustment HIGH Intervention group: Members eligible for chronic care management Statistical program and identified with chronic modeling (incorporating HIGH condition below components) Comparison strategy: Historic B y cost trend based on 2 years of pre- t intervention data for the non-chronic population Stratification and complexi and outlier Level of analysis: Aggregated restrictions data ethodology cost M Adjustment: Truncation of Outlier No outlier cost at $100K/PMPY; restrictions adeq t pre-intervention trend relativity only ua between chronic and non-chronic; te stratification by condition and co- morbidity LOW None LOW HIGH Pre-post Historic Control Concurrent Non-experimental RCT Industry trend Comparison Group

Strength of attribution approach

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 51 Example C: Non-Participating Group as Comparison Population Strength of Adjustment HIGH Intervention group: Members Statistical participating in chronic care modeling management program (incorporating HIGH below Comparison strategy: Non- components) participating members, 2 years of C ty baseline and 2 years of intervention data Stratification and complexi and outlier Level of analysis: Aggregated restrictions data

ethodology cost Adjustment: Truncation of outlier N M cost at $100K/PMPY; stratification Outlier o by disease, sex and age; risk score restrictions t ade normalization only qua te LOW None LOW HIGH Pre-post Historic Control Concurrent Non-experimental RCT Industry trend Comparison Group

Strength of attribution approach

Example D: Non-Contracted Group as Comparison Population Strength of Adjustment HIGH Intervention group: Members Statistical eligible for cardiac chronic care modeling management program (incorporating HIGH below Comparison strategy: Members components) not meeting program eligibility D ty criteria, 2 years of baseline and 2 years of intervention data Stratification and complexi and outlier Level of analysis: Member-level restrictions data

ethodology cost Adjustment: Truncation of outlier N M cost at $100K/PMPY; multivariate Outlier o regression restrictions ade t only qua te LOW None LOW HIGH Pre-post Historic Control Concurrent Non-experimental RCT Industry trend Comparison Group

Strength of attribution approach

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 52 Example E: Natural Experiment Strength of Adjustment HIGH Intervention group: Fully Statistical insured members for health plan, modeling eligible for chronic care management HIGH (incorporating program below components) Comparison strategy: Self- E ty insured members to whom program was not available, 2 years of baseline Stratification and 2 years and complexi and outlier of intervention data restrictions Level of analysis: Member-level ethodology cost data N M Outlier o Adjustment: Truncation of outlier restrictions t ade cost at $100K/PMPY: Propensity only qua te score matching; Difference in differences analysis LOW None LOW HIGH Pre-post Historic Control Concurrent Non-experimental RCT Industry trend Comparison Group

Strength of attribution approach

Methodology Caveats 4. Control Group – As programs increase Selecting an approach within the acceptable engagement rates, the availability of non- continuum of attribution and adjustment does participants as the comparison group source not guarantee valid results. The following key will be limited. considerations should be kept in mind. 5. Specific Application – The methodologies 1. Population Size – Previous CCA work a has described may be more suitable to chronic illustrated the impact of population size on care, wellness and other true “population”- the variability of financial outcome measures. based programs than to case management Measurements of populations less than and other low-prevalence programs. 15,000 may have confidence intervals that 6. Limited Empirical Evidence – There is span zero, even when using more advanced general recognition that evaluations using methodologies. comparison groups, such as match control, 2. Acceptance – When using sophisticated are more rigorous than population-based statistical methods and population analyses previously recommended. However, comparisons, the presentation of results must empirical comparisons with randomized account for limited customer understanding. control designs have yet to be published. Also, these approaches are more challenging to replicate for audit purposes. 3. Implementation – As was previously pointed out, methods which improve measurement validity result in higher costs as the reporting process requires assessment by a sophisticated evaluator and can’t be fully automated.

a See “Small Populations.” Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010, p. 82. Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 53 Reference C

Table C1. Comparison Group Options

METHOD TO DEVELOP Randomized concurrent Matched control Non-participating Historical control A COMPARISON control individuals comparison (as defined by CCA GROUP guidelines) General description of Intervened population Intervened population Intervened participating Intervened population method compared with individuals compared with individuals compared compared with similarly randomly selected to individuals from non- with intervened non- identified population in a have services withheld participating groups participating individuals baseline period (with costs matched to have similar intervened individuals characteristics as trended forward) intervened individuals Comparison time frame Concurrent to Concurrent to Concurrent to Prior period intervention intervention intervention Population selection bias None Somewhat significant Significant None Source of comparison Population for Population for Population for Population for group whom program was whom program was whom program was whom program was implemented, randomly implemented, purchaser implemented, individual implemented, in prior selected group withheld decision to not decision to not period from program participate participate Trend factor Not required Not required Not required Individuals without measured conditions in population for whom program was implemented “Regression to mean” None None None Low issue Credibility of causal Extremely strong (gold Strong (with proper Poor Moderate statements standard) design) Applicability to all types Strong Strong Strong Uncertain of programs/program designs Program sponsor Strong Moderate (confidentiality None Low resistance to approach of non-participating groups can be an issue) Ease of implementation Very difficult Difficult Easy Somewhat difficult Clarity of method to lay Very clear Very unclear Clear Somewhat unclear audiences Multiyear application vs. Much harder About the same About the same Somewhat harder single-year application Method availability Rarely possible Occasionally possible Frequently possible Usually possible Bleed of interventions to Control group likely Control group likely Control group likely Control group unlikely to comparison population to get provider-based to get provider-based to get provider-based get any interventions interventions interventions interventions and softer member-based interventions Key strengths Generally seen as gold Potential for a concurrent Easy to implement Relatively universally standard evaluation comparison population Method is easy to available method method with many of the same understand Application of well- Ease/strength of characteristics as the accepted actuarial interpretation intervened population processes Credibility of casual statements can be very strong Key problems/biases Sponsor resistance to Difficult to get access to Very significant bias Difficult to ensure implementation non•participating groups associated with equivalence between Difficulty of multiyear of sufficient size in same differences in motivation baseline and intervention assessment geographic area, with between participants year (particularly in similar benefit structure Likely not possible in populations with many “Black box” approach “opt-out” program shifts in size,composition, difficult to understand models and/or benefit structure) Difficulty in deriving credible trend factor

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 54 Reference D

Methods to Define Outliers during the year. This approach is preferable to excluding the participant’s entire experience Patients can incur extraordinarily high costs for because it enables the inclusion of a greater numerous reasons. These costs often are for proportion of the managed population in events randomly distributed over a population measurement. Moreover, it does not create a and unrelated to a chronic care management distortion if the program is involved in shifting a program—accidental trauma, for example. member above or below the stop-loss threshold. High costs create substantial volatility to claims There are several methods commonly used cost trends and can distort financial savings across population health management to identify calculations, particularly for smaller populations. and mitigate these outlier costs. These methods A stop-loss approach excludes, for the purpose are outlined in Table D1. CCA recommends a of measurement, claims costs for individual review of the information in the table prior to members in excess of the stop-loss threshold selection of a method.

Table D1 – Methods to Define Outliers

METHOD Stop-Loss Method Percentile Distribution Standard Deviation (SD) Method Method General description of A threshold value is determined (e.g., A threshold value is A threshold value is determined method $100K) and costs above threshold are determined based on the based on X (e.g., 3SD) standard excluded and are indexed to grow X percentile of claims costs deviations from population mean. in future years concurrent with an (e.g., 99.5%); costs above this Costs above this threshold are appropriate trend. threshold are excluded. excluded. Trend factor Should be used to account for Not required Not required medical cost trends year to year. Applicability to all types of May require lower threshold to offset N/A May not be appropriate for programs/program designs variability in small populations. populations with non-normal distributions. Ease of implementation Very Easy Easy More difficult Clarity of method to lay Very Clear Clear Requires knowledge of simple audiences statistics. Multiyear application vs. Trend adjustment should be Easy to apply Should recalculate SD for each year. single-year application considered for multiyear assessment. Key strengths Better for evaluations where two Better for evaluations where Better for evaluations where two groups are expected to have two groups are expected to groups are expected to have unit different frequency cost distributions have unit cost variance rather cost variance rather than different rather than unit cost variance. than different frequency cost frequency cost distributions. Ease of application and most distributions. Can be used without adjustment on sensitive to varying rates of Can be used without small populations. catastrophic claims. adjustment on small populations. Key limitations Stop-loss threshold is arbitrary and Threshold is also arbitrary Threshold is also arbitrary but data- not data-sensitive. but data-sensitive to varying sensitive, and is only somewhat Threshold selection for smaller percentages of shock-loss sensitive to a varied percentage of populations is by nature more claims. claims categorized as shock-loss. subjective than other methods listed. Handles variant cost The calculations are the most May not appropriately filter random distributions with variant involved, but involve standard drivers of variance in small groups thresholds, which may not be simple statistics. Using standard without lowering the stop-loss desirable with variant rates of concepts of significance for amount. catastrophic claims. selecting a standard deviation Lack of threshold trend adjustment threshold may not apply to the will deflate overall trended results proper selection of shock-loss proportional to the presence of claims. catastrophic cases. Handles variant cost distributions with variant thresholds, which may not be desirable with variant rates of catastrophic claims.

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 55 Reference E

Evaluation Considerations Medical cost data are highly variable in small for Small Populations populations. Average costs for individuals with many of the common chronic care management The value and risks of applying the methods conditions can show severe variation with high recommendations articulated in these Guidelines standard deviations. Even a few participants to small populations is an important concern. with high costs can have an impact on averages The ultimate end users of outcomes results, calculated for PMPM costs and result in wide typically, are organizations representing groups confidence intervals around estimates of of people (such as employers, state and federal these measures. This is, in fact, one reason agencies, health plans and provider groups). why medical management professionals were Given the importance of employer-based health attracted to these conditions in the first place: care management, the number of potential The elimination of unnecessary variation was users of outcomes information through our considered one of the key goals of chronic employer groups alone is extremely large. These care management interventions. In larger groups can vary widely in size. Employer groups populations, the impact of a few cost outliers ranging in size from 50 people to more than on the measure variance does not have as 250,000 people are actively engaged in chronic significant an impact. care management and have an increasingly active interest in understanding outcomes Table E1 demonstrates the significance of from these services. This section is intended to the impact of this variability on medical cost provide important contextual information for assessments. The information presented in the understanding outcomes measures for groups table is meant to be an example of the range of individuals aggregated into relatively small of differences that can be seen in a sample numbers. In fact, the information below provides population and is not intended to represent relative information for groups that have a wide results that would be seen for all populations range of sizes, up to 50,000 individuals. of this size. The table was created by starting with a large population of individuals who

Table E1 – Small Populations, Part I

Population with Chronic Total Population Total Population Conditions PMPM Medical Cost Savings* (95% Confidence Intervals) 30 500 -$17.29 to $29.20 60 1,000 -$21.79 to $35.50 120 2,000 -$11.07 to $24.02 180 3,000 -$7.02 to $17.77 240 4,000 -$3.91 to $15.24 300 5,000 -$3.74 to $14.71 600 10,000 -$2.88 to $13.26 900 15,000 $0.24 to $10.24 1,200 20,000 $0.74 to $10.00 1,500 25,000 $1.43 to $9.04 1,800 30,000 $1.98 to $8.86 2,400 40,000 $2.06 to $8.36 3,000 50,000 $2.53 to $8.09

*Savings for the population with chronic conditions is divided by the total population member months

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 56 participate in a chronic care management • this process was then repeated for different program, then repeatedly taking various sample size samples, up to 3,000 members, and sizes and computing the economic impact of the with chronic conditions, a procedure intervention for each sample. called “bootstrapping” in mathematical • the table shows values for the upper and programming circles. lower confidence intervals on first year PMPM medical cost savings estimates in a population Clearly, caution is advised in producing medical with a robust chronic condition management cost savings measures in subpopulations program in place. with small numbers of members receiving management for chronic medical conditions. • the savings estimates were derived from High variability will frequently result in methods compliant with the CCA Outcomes conflicting, misleading and/or grossly inaccurate Guidelines Report. indications of program-related impact. • the variance was estimated using repeated Looking at this table slightly differently, the samples from a large commercial plan upper and lower confidence interval limits population for which a chronic condition were graphed vs. total sample size to get an management program was implemented. appreciation of the direction and magnitude of • For example, in the first line, 500 samples this effect. of members were pulled from the entire population, generating, in each case, 30 Of interest, several items should be noted. members with chronic conditions. The • As the sample size increases, the upper medical cost savings algorithms applied to and lower confidence intervals converge on each of these samples and the variation in the PMPM savings that a large population these results are measured to produce the likely would recognize from this chronic care confidence interval cited. management program.

Chart E1 – Small Populations, Part II

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 57 • For small numbers of participants, the range Typically, this large number of participants is of measured PMPM savings can be striking— not commonly available in modest size or small from as high as $35 PMPM to as low as -$20 companies. Consider that in diabetes, where the PMPM for population sample sizes in the prevalence rate is approximately 5 percent, you 1,000 to 2,000 range. In other words, a small would need an employer with 50,000 employees company or group being serviced by this to identify 2,500 diabetics—a number that, on program could show PMPM impacts ranging this chart, still is in a range characterized by wide from -$20 to $35 just on chance alone. variation. • Also of interest, the lower curve rises gradually, crossing $0 savings at a total Ideally, the owner of a chronic care management population of 15,000 members/900 with the program would calculate the number of condition. One might argue that to avoid individuals necessary in a program to ensure misleading clients or coming to the incorrect statistical significance of results, a process called conclusion that there were no cost savings a “power calculation” by statisticians, before or even a loss, a minimum number of 15,000 embarking on a program. This would prepare population size should be included in a them for the level of certainty they would have calculation before PMPM savings calculations later in estimating outcomes. are made for this program.

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 58 Reference F

Utilization Measures • Measures impact on comorbidities, as well as primary conditions of interest. Within population health management, there • Does not test accuracy of identification is widespread use of utilization measures as algorithm for diseased population. complements to the primary financial outcomes measure for understanding and validating • condition-specific admission rate for the program savings. Multiple utilization measures entire insured or covered population (using are currently used for this purpose, with principal diagnosis only). different utilization measures revealing different • Specific for one expected impact of information about chronic care management chronic care management programs, but program performance. insensitive to other possible program impacts. Hospital admission measures (typically, • Does not measure chronic care admission rate expressed as the number of management impact on comorbidities. hospitalizations per thousand members per year) • Sensitive to condition prevalence changes, and emergency department utilization measures so prevalence adjustment required. (typically, ED visit rate expressed as the number of visits per thousand members per year) are the • May serve as “end to end” test of utilization measures chronic care management identification, outreach, enrollment, programs most directly impact. These measures, engagement, impact, retention, etc. derived from medical claims, are suitable for pre-post comparison, as well as year-on-year Other utilization measures that might be tracking for programs beyond their baseline year, of interest to purchasers and suppliers of to complement and corroborate the primary population health management programs may financial measures. be collected and reported at the option and agreement of the parties. Table F1, next page, Two of the most commonly used admissions elaborates on several of these measures. measures are: • All-cause admission rate for the diseased or eligible population. • r elatively sensitive for major cost driver impacted by chronic care management programs.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 59 Table F1. Utilization Measures Options

Measure Utility/Comments All-cause utilization of diseased population • relatively sensitive for major cost driver impacted by chronic care management programs • Measures impact on comorbidities as well as primary conditions of interest • Does not test accuracy of identification of diseased population Condition-specific utilization of diseased population • More sensitive for expected key impact of chronic care manage- ment programs but does not measure chronic care management impact on comorbidities • Does not test accuracy of identification of diseased population Condition-specific utilization of entire popula- • Specific for one expected impact of chronic care management tion – known as “plausibility indicator” by Disease programs but insensitive to other possible program impacts Management Purchasing Consortium International • Does not measure chronic care management impact on (DMPC) (Variant 1 – one measure) comorbidities • Sensitive to condition prevalence changes, so prevalence adjustment required • May serve as “end to end” test of chronic care management identification, outreach, enrollment, engagement, impact, retention, etc. Condition-specific utilization of entire population – • Same as for Variant 1 known as “plausibility indicators” by DMPC (Variant • Adds ED visits, which reflect additional chronic care management 2 – two separate measures) impact beyond admissions reductions Condition-specific utilization of entire population – • Same as for Variant 1 (Variant 3 – one combined measure) • Adds ED visits to admissions in a single weighted aggregate measure that may reflect additional chronic care management impacts beyond just admissions reductions All-cause utilization of entire population • the most inclusive measure of utilization but with attribution problems • May be insensitive to chronic care management program impact, as many other influences impact total utilization changes over time Utilization x unit cost to build up total cost estimate • relatively sensitive for major cost driver impacted by chronic care (surrogate for measured cost approach) management programs • Measures impact on comorbidities as well as primary conditions of interest • Does not test accuracy of identification of diseased population

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 60 REFERENCE G

Self Management Measures problem solving and active collaboration among participants, family/caregivers and CCA recognizes the critical role individuals play others on the health care team. in managing their health on a daily basis. Within • A ssessment of an individual’s self the context of managing a chronic disease— management capabilities relies on behavioral diabetes, for example—this role becomes more measures that include self-efficacy, health complex with the daily need to self-administer beliefs and readiness to change, knowledge and manage multiple medications, self-monitor of the condition and its treatment and self- and manage blood sugar levels and respond care skills required to manage the condition. appropriately, implement and follow dietary recommendations and incorporate healthful Criteria for Self Management Metrics lifestyle behaviors, such as daily exercise. • M etric can be influenced by a chronic care Successful self management of a chronic disease management program. can slow disease progression and improve overall quality of life. • M etric assesses an issue or problem that has a substantive impact on the participant cohor t Chronic care management programs, therefore, over time. should incorporate self management assessment • T here is an evidence base for designing or and education to increase awareness of and selecting the metric. compliance with treatment guidelines, facilitate • T he metric is routinely measured or is problem solving skills, support and motivate measurable by use of a validated tool or individuals in making healthful behavioral method. changes and promote open communications with providers. • The resulting information is useable to assess and refine the intervention to lead to This section includes: improved patient outcomes. • Definition of self management. Self Management Metrics • Criteria for selecting and prioritizing the development of (self management) metrics. Eight possible metrics applicable to the development and implementation of disease self • S pecification of metrics to assess self management education programs have been management both on the individual and identified. program level. • T he knowledge of condition/health literacy. CCA defines self management as: • Of condition/issue. Self management consists of the ongoing • Of solution/intervention. processes and actions taken to manage/ • R eadiness to change on applicable behaviors control one’s own condition, with the goal (both generic and condition-specific). The of improving clinical outcomes, health status stages in the process that individuals may and quality of life. go through to engage in and fully adopt • C ore components of the self management behaviors. process include incorporating the needs, • S elf-efficacy (both generic and condition- goals and life experiences of the individual, in specific). addition to being guided by evidence-based standards. • Individual’s belief about his/her ability to produce desired effects. • T he objectives of self management interventions are to support informed • Related constructs. decisionmaking, improve and promote use of • Confidence. self-care skills and behaviors and encourage • Perceived control.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 61 • the use of devices and tools designed to support self management. • the presence of collaborative goal-setting activities. • the use and content of assessments of participant self management skills. • the presence of individual action plans designed to guide self management. • the presence and frequency of use of specific self-monitoring activities (both generic and condition-specific).

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

Care Continuum Alliance • Implementation and Evaluation: A Population Health Guide for Primary Care Models 62 REFERENCE H

Medication Adherence Measures medication adherence support over time. To define the level of the Framework, the Organizational Approach to Medication workgroup first envisioned a fully functional Adherence Best Practice system and infrastructure that promoted The workgroup goal of developing a best medication adherence in every person with practice framework for medication adherence whom it interacted. Various common elements was achieved through the creation of the appeared to be required, and include: Organizational Best Practice Framework • s urveys/tools and training to raise individual (Figure H1). The Framework includes a step- practitioner awareness of potential patient wise set of systematic interventions that will medication non-adherence; enable organizations to promote best practice • organization-practitioner alignment that in supporting patients to achieve high rates medication non-adherence is a critical of medication adherence. Through review element to achieve positive patient outcomes; of the current literature and discussions, the workgroup identified a variety of approaches • a defined organization-specific suite of used by providers and organizations aimed at practitioner/organization interventions to improving medication adherence. A discussion lessen the risk of individual patient non- that identifies surveys and tools used in self- adherence; reported medication adherence, based on the • a method to aggregate adherence/non- literature review, is included in this section (Table adherence data across populations; H2). Less information, however, was available • goal-setting among practitioners and on how organizations can systematically change the organization to achieve high rates of medication adherence rates in their patient adherence across populations; populations; more research appears to be needed. • incentives/reimbursement to practitioners/ providers and organizations based on best While no two organizational approaches to practice outcomes; the problem of medication adherence are • r eal-time medication adherence data and alike, there are common areas that define a decision support for providers at point of potentially generalizable medication adherence care; improvement continuum. The best practice • organizational value-chain alignment in continuum, as currently defined, is based upon support of medication adherence; and available literature, as well as expert input from • o verall organization success linked to high the workgroup, representing multiple areas of rates of population medication adherence. the health delivery system. The Framework is not meant to be comprehensive in scope, but rather These elements were gauged as more or less offers a starting point from which organizations systematic in approach, and arrayed across a may begin to assess themselves and their grid (left to right). Of particular note, the grid focus on improving medication adherence. takes into account changes and interventions Theoretically, as more organizations work to required from both individual providers and improve medication adherence systematically, organizations to best support medication the current Framework will evolve to include adherence. On the grid, the term “HCP” or new research and additional key interventions “Health Care Provider” refers to any provider of that support best patient and population service, including physicians, pharmacists, case outcomes. The purpose of defining the level or disease managers and ancillary providers. of best practice, therefore, is to begin to assist Also, the term “Organization” refers to any organizations to: 1) evaluate their current state organization that plays a role in health care, of function; and 2) work in a step-wise manner including those that directly deliver care to to systematically improve their approach to

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 63 patients (e.g., health delivery systems, individual The five levels include: or group providers, pharmacies, home care • Level I – Person assessment; agencies, long term care facilities), as well as • Level II – Identification of at-risk individuals; those removed from direct patient care (e.g., health plans, population health vendors, other • L evel III – Basic intervention to improve health care-related organizations). medication adherence; • L evel IV – Advanced intervention – goals- The elements are ordered into a stepped based management to improve medication continuum, arraying the less systematic adherence; and interventions to the left of the continuum and • L evel V – Fully functional system in support of the more systematic approaches to the right. medication adherence for the individual and Related HCP and organization interventions the population. are “stacked.” The model defines five levels of systematic approach—assigned Levels I through The visual representation of the five levels V—with organizations progressing along the defines the steps to completion of each level, best practices continuum from Level I to Level V. based on progression from lower to higher “Systematic Approach” (see Figure H1).

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 64 Figure H1 – Organizational Best Practice Framework Level V – Fully Building a Best-practice Approach to Improve Medication Adherence in Organizations* Functional Systematic Interventions

Organization’s clinical strategy links improved medica- tion adherence to overall success Level IV – Advanced across all service Intervention (Goals- lines Based Management)

Organization collectively sets Organization actively targets & goals to works to align its improve organiza- value-chain to tional performance improve medication in medication non- adherence across all Level III – Basic adherence in the populations Intervention population(s) served

Organization inte- Organization builds grates medication Organization formal database to adherence data base aggregates popula- track & report trends into daily provider tion data to identify in medication non- functions, provides medication non-ad- adherence in the decision support herence trends Level II – Identify population(s) served for the HCP point of At-Risk Individuals care

Organization identi- Organization offers fies critical points of suite of targeted Organization identi- Organization is care & interventions assessments and fies low HCP per- aligned to assess to overcome barriers interventions to formers & challenges individual risk of of medication non- lessen individual them to improve medication non- adherence in at-risk risk for medication medication non- adherence Level I – individuals non-adherence in adherence outcomes Person Assessment & populations population(s) served

HCPs: A) receive/ HCPs are aware of HCPs receive incen- access data for own HCPs** have access At point of care & assess individuals tives, increased population; B) are to validated surveys/ delivery, HCPs use at critical points of quality payments trained to identify/ tools to assess indi- the validated sur- care to identify and or reimbursement close gaps; C) set vidual medication veys/tools to identify overcome medica- based on medication goals to decrease non-adherence at-risk individuals tion non-adherence adherence medication non- barriers performance adherence

Lower Systematic Approach Higher

* Definition of “Organization” is “any organization” in the health care space—could be health care delivery systems, pharmacies, analytics companies, population health vendors, etc. ** HCPs are any provider of any type, including physicians, case managers, pharmacists, ancillary staff, etc. Note: This Organizational Best Practice Framework has been updated from the original framework published in 2010 in the CCA Outcomes Guidelines Report, vol. 5.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 65 MEASURE OF MEDICATION • Whom to Report: Reported by condition POSSESSION RATIO and by drug classes applicable to that CCA recommends the measures of condition. Individuals with multiple conditions Medication Possession Ratio and Persistence (e.g., CAD and diabetes) will be counted when assessing the outcomes of a program. for all conditions and for all appropriate Medication possession ratio (MPR) is an drug classes. See Table H1, next page, for a operational definition of adherence: a representative drug class list, per condition. retrospective assessment of days supplied over • How to Report: Each appropriate an evaluation period.10 A recommended method condition/drug class combination, per the for calculating MPR, one aspect of the total representative drug classes list, would have an adherence equation, is described in the section MPR reported. For example, for beta blockers below. in CAD, this measure would include the sum of all days (for all members with CAD taking Methodology beta blockers) in the denominator and the MPR is a population-based measure reported as sum of all days supplied (for all members with a percentage: CAD taking beta blockers) in the numerator. To calculate disease-specific (columns in drug • Data sources: Administrative pharmacy class table) or drug class-specific (rows in drug claims and eligibility data. class table) MPR totals, one must sum all days • Evaluation Period: A fixed calendar length: and days supply for appropriate column/row 12 months (annual). One month run-out will and then perform the division to calculate be allowed for claims lag; therefore, the percentage. measure can be calculated at the end of month 13. Inclusion/Exclusion Criteria • Enrollment Criteria: A continuous evaluation • I ntended for more prevalent common chronic period with no more than a 45-day gap in conditions (persistent asthma, CAD, CHF, pharmacy benefits coverage. diabetes, hypertension and hyperlipidemia). • Denominator: The duration from first (index) • I ntended for oral medications only (CAD, CHF, prescription to the end of the evaluation diabetes, hypertension and hyperlipidemia). period. • I ntended for oral medications and inhaled • Numerator: The days supplied over the same medications only (persistent asthma). period. • Excludes liquid-form medications. • Numerator Limit: To control for potential • I ndex prescription must occur within the first confounders of concomitant therapies and six (6) months of the evaluation period. overlap (due to medication switches within drug class, lost medications, etc.), prior • A minimum of two claims, per member, for to disease-specific or drug class-specific, a specific drug class must be incurred to population-based MPR roll-ups (see “Whom include the member in the calculation. to Report,” below), each individual’s MPR • E xcludes “carry in” from a prior evaluation numerator/denominator set will be limited period. For example, a 30-day supply filled on such that the numerator is less than or equal Dec. 15, 2006, would not be considered for to the denominator (i.e., days supply <= days; the 2007 evaluation period. individual MPR cannot exceed 100 percent). • E xcludes “carry out,” when medication • What to Report: MPR as a percentage and supply goes beyond the evaluation period. by quartiles (e.g., box plot, box-and-whisker Using the above example, the same 30-day diagram). supply is filled on Dec. 15, 2006. For the 2006 evaluation period, only the days supply from Dec. 15 to Dec. 31 (17 days) would count in the numerator.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 66 Table H1 - Representative Drug Classes

DRUG CLASS AST CAD CHF DIAB HYPL HTN angiotensin II converting enzyme inhibitors O O O O angiotensin II receptor antagonists O O O O anti-inflammatory agents (cromolyn sodium, I nedocromil) beta blockers O O O biguanides O bile acid sequestrants O O calcium channel blockers O O cardiac glycosides O corticosteroids I diuretics O O fibric acid derivatives O O hydroxymethylglutaryl reductase O O inhibitor combos hydroxymethylglutaryl reductase inhibitors O O leukotriene modifiers O long-acting beta agonists I sulfonylureas O thiazolidinediones O xanthines O

AST - Persistent Asthma Included for condition/drug class O = Oral CAD - Coronary Artery Disease Not included for condition/drug class I = Inhaled CHF - Congestive Heart Failure DIAB - Diabetes Note: Combination drugs (e.g. ACE + Diuretic; ICS + LABA) will be HYPL - Hyperlipidemia measured in all conditions and all classes to which that medication HTN - Hypertension belongs.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 67 • Potential confounders of contraindications, • Whom to Report: Reported by condition samples and clinical utilization (inpatient and by drug classes applicable to that admissions) may impact MPR. condition. Individuals with multiple conditions (e.g., CAD and diabetes) will be counted for MEASURE OF MEDICATION PERSISTENCE all conditions and for all appropriate drug Medication persistence can be defined as the classes. See Table H1 for a representative “amount of time that an individual remains drug class list, per condition. on chronic drug therapy.”11 A recommended • How to Report: Each appropriate method for calculating persistence, another condition/drug class combination, per the aspect of the total adherence equation, is representative drug classes list, would have a described in the section below. time series persistence reported. This can be visualized as a survivability graph (much like Methodology a Kaplan-Meier curve). The x-axis consists of Persistence is a population-based measure the ETP periods and the y-axis consists of the reported as a percentage over time: persistence percentage at each PET period. • Data sources: Administrative pharmacy Inclusion/Exclusion Criteria claims data. • I ntended for more prevalent common chronic • Permissible Refill Gap: 60 days. conditions (persistent asthma, CAD, CHF, • Annual Evaluation Period: A fixed calendar diabetes, hypertension and hyperlipidemia). length: 12 months (annual). One month run- • I ntended for oral medications only (CAD, CHF, out will be allowed for claims lag. diabetes, hypertension and hyperlipidemia). • Evaluation Time Periods (ETP): The time • I ntended for oral medications and inhaled from last fill’s run-out date (i.e., date of fill + medications only (persistent asthma). days supply) + permissible refill gap (not to exceed the end of the calendar year). • Excludes liquid-form medications. • Denominators: The number of eligible • E xcludes “carry in” from prior evaluation members, in each ETP period, from first period. For example, a 30-day supply filled on (therapy initiation) prescription to the end of Dec. 15, 2006, would not be considered for the evaluation period. the 2007 evaluation period. • Numerators: The number of eligible • E xcludes “carry out,” when medication members, in each ETP period, from first supply goes beyond the evaluation period. (therapy initiation) prescription to the end of Using the above example, the same 30-day the evaluation period who do not exceed the supply is filled on Dec. 15, 2006. For the 2006 permissible refill gap. evaluation period, only the days supply from Dec. 15 to Dec. 31 (17 days) would count in • What to Report: Persistence as a percentage the numerator. per ETP period. • Potential confounders of contraindications, lost medications, samples and clinical utilization (inpatient admissions) may impact persistence.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 68 Examples

Chart H1 – Diabetes Persistence Example

Time 0 Time 1 Time 2 Time 3 Time 4 Time 5 Time 6

1 1 1 1 1 1 1 Eligible at year start; Mbr1 100% persistent

Eligible at year start; Mbr2 1 1 1 1 0 0 0 50% persistent

Eligible mid-year; Mbr3 1 1 1 1 – – – 100% persistent

Eligible mid-year; 33% Mbr4 1 1 0 0 – – – persistent

P% 100% 100% 75% 75% 50% 50% 50% Time 0 = Initiation of Therapy, regardless of calendar date @ start – = “Null” entry (i.e. non-measurable data point due to start date

Figure H2 – Diabetes Persistence Curve

100%

90%

80%

70%

60%

50%

40% % Persistence 30%

20%

10%

0% Time 0 Time 1 Time 2 Time 3 Time 4 Time 5 Time 6

Biguanides

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 69 Frequently Asked Questions Regarding MPR variance – so, that might explain periods of and Measures of Persistence adherence versus non-adherence (when a Q: What are medication possession ratio (MPR) person is “off-season”). and persistence meant to measure? Q: When might be considered a “triggering A: MPR and persistence measurements are point” for interventions regarding MPR and complementary. MPR attempts to highlight persistence? the proportion of filled doses while on a therapy regimen. Persistence generally A: Many of the effectiveness studies performed captures those individuals who have halted by the pharmaceutical industries have (dropped off) therapy. The intervention focused on MPR ≥ 80 percent. Therefore, strategies to close the gaps for such many presume MPR rates below that heterogeneous groups can differ greatly. to be suboptimal. An optimal MPR threshold has not been causally proved, Q: What are commonly recurring barriers to via RCTs originally designed for such, and optimal MPR and persistence? experience reasons that such optimal MPR thresholds might differ based on multiple A: Forgetfulness, dose/schedule alteration and/ factors (e.g., population demographics, or cost are often cited as reasons for poor socioeconomic factors, side effect profiles, MPR outcomes. Whereas side effects, cost etc.). However, a recent study by Hansen et and/or administration issues might be linked al. in the March 2009 issue of The Annals of to persistence failure. These are examples Pharmacotherapy suggests that 80 percent taken from published literature; however, the might represent a point to determine reasons for each can significantly overlap. adherent/non-adherent cohorts, while Since both MPR and persistence are affected retaining parity between sensitivity and by differential response, based on an specificity.12 For persistence, once a person individual’s diagnosis and the therapeutic has missed an allowable refill gap, they are drug class utilized, multiple elements should considered to be non-adherent. Triggering be considered when determining effective points will also depend on how one views clinical strategies for such populations. the data: by book of business, by individual, by lines of business/markets or other Q: Why is measurement more variable with dimensions. The MPR/persistence methods inhaled medications (e.g., asthma)? are meant to standardize outcomes reporting A: An example of differential response, and a in this domain. In either case, differential challenge of administrative pharmacy claims response and each individual’s motivation data, lies in the domain of pharmacotherapy and behavioral skills will play significant roles measurements for persistent asthma. in determining appropriate intervention This form of asthma is chronic with acute triggers and strategies. exacerbations. Days supply for inhaled medications is also more difficult to perfect Q: Although the recommendations call for – individual lung volume and inhalation annual measurement with appropriate habits are inconsistent, and may change claims run-out, could my organization with age, circumstance or other variable. In calculate on an ongoing basis or quarterly? addition, the accuracy of the days supply A: A method for routine measurement that field is variable, depending on factors such minimally impacts inclusion/exclusion criteria as prescriber, prescribing system, filling and remains statistically less variant would adjudication system, etc. For these reasons, be a rolling 12 months view. One would it is not uncommon to see MPR rates for merely adjust the “book ends” of the analysis asthmatics well below that of other chronic time period to be 12 months. Regardless conditions. And, for many people, asthma of measurement frequency (monthly, severity can be influenced by seasonal

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 70 quarterly, annually), 12 months of claims to offer and expand low-cost discount drug data will minimize impact of some known lists, prescriptions associated with such confounders, such as plan design changes, programs might not appear in claims data. mail versus retail fills and medication This occurs primarily due to cash purchases guidelines changes. in which a benefit card is not presented and an administrative claim is not generated. Q: Why did you decide to exclude liquid form medications? Q: How does the drug class-level MPR A: Liquid medications have similar challenges methodology account for concomitant as inhalers, with days supply fields being therapy and medication switches? subject to considerable variation and data A: If an individual has overlapping days supply collection issues. For example, blood glucose for two medications within the same drug control in insulin-dependent diabetics class, such as when a patient is switched can be driven by several factors – current from one statin to another, then a drug class- weight or prior blood glucose readings, for level MPR calculation would result in that example. The adherence workgroup, from a person’s days supply exceeding the days in complexity reduction standpoint, decided to the evaluation period. As such, those days approach the more straightforward forms to supply (numerator) would be capped so that vet these standardized methods. they could be equal to, but not greater than, the days (denominator). If a person switches Q: How do you define drug class? medications between two drug classes (i.e., discontinues therapy in drug class x and A: The recommended therapeutic drug classes begins new therapy in drug class y), the MPR were chosen with clinical practice in mind for drug class x would be deflated, whereas (i.e., the way the medications are employed drug class y would remain a truer reflection to treat specific conditions). A drug class of that individual’s ongoing adherence. encompasses medications with similar modes of action that would not, under Q: How does the condition-level MPR the majority of circumstances, be used as methodology account for concomitant concomitant therapy. therapy and medication switches?

Q: How did you arrive at the drug classes for A: If an individual has claims for two drug each condition? classes for a condition, the calculation of condition-level MPR must make an A: The list is recommended as a general assumption regarding whether the situation starting point, but it is by no means all- involves concomitant therapy or a switch inclusive. The conditions were chosen due in therapy. If the methodology assumes to their higher overall prevalence rates. The concomitant therapy, it will accumulate medications had to be commonly used and each drug’s days supply in the numerator, relevant in the medication management and each drug’s potential days of therapy of the designated condition (guidelines- in the denominator. If the methodology based), be a chronic medication (not just assumes switching, it will accumulate each taken at acute phases or used with tapering drug’s days supply in the numerator, but will regimens) and have relatively consistent only accumulate potential days once in the claims data availability. denominator. Thus, the first approach can deflate MPR, and the second approach can Q: How do discount drug programs impact inflate MPR. Neither is necessarily right or these measurements? wrong, since there is currently no way for A: The growth of discounted prescription a claims-based analysis to ever truly know drug programs might have an effect on whether these situations involve concomitant measurement accuracy. As companies begin therapy or switches. Therefore, it is essential

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 71 for condition-level MPR reports to clearly Ogedegbe, et al. in the Journal of Clinical explain which assumption and methodology Epidemiology also outlines a process to develop is chosen. a medication self-efficacy scale.14 Other known methods are to apply validation through the use Measures of Self-reported of pharmacy claims comparisons or comparisons Medication Adherence to other known survey instruments. Self-report surveys for measuring medication adherence vary in structure and purpose. The primary goal for the self-reported To identify empirically validated self-report instruments table was to provide those that medication adherence surveys, a literature could be used for multiple chronic conditions. search was conducted from 1980 to present. This Disease-specific instruments may be the subject literature search yielded instruments that were of future work. Key words for inclusion in the general as well as disease-specific. Specifically, literature search were: medication adherence, adherence surveys have been developed for self-report, validation and risk assessment. therapeutic areas such as hypertension, HIV and References from the development and depression, among others. These surveys may validation papers of each of the instruments include different elements, such as general self were also useful in locating additional published report, attitudes, compliance or identification of instruments. Table H2 offers self-reported barriers to taking medicines. Other instruments instruments that have been used to evaluate focused on the predictability of those individuals various aspects of medication adherence. The that would be non-adherent. Typically, the types of instruments include scales to assess surveys consisted of various questions grouped low medication adherence, barriers to taking together (common factors) to help discuss medicines, treatment satisfaction and risk for known factors that impact the beliefs of taking non-adherence. medicines. Use of the instruments From an item development perspective, there is The type of self-report medication instrument no standard methodology available. The survey selected for a program or study may depend on instruments were typically developed using the desired outcome. Considerations include standard methodologies, such as those used to use in a clinical setting where an immediate develop quality of life instruments.13 At minimum, response can take place, or use in a pre-post items for the development and validation study that might evaluate behavior change. process may include the question development Other instruments may be used in the prediction process, statistical analysis demonstrating of adherence and, based on what they measure, the instrument is psychometrically sound (i.e., will require different interventions (or actions).” reliable and valid) and interpretation (statistical Last, the length of the survey and the ability and clinical significance).13 to interpret the results have an impact on the selection and use of the desired survey Additional measures may provide comparisons instrument. to self-reported medication-taking behaviors.

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 72 Table H2 – Medication Adherence Self-Report Instruments

Survey or Scale Purpose Number of Items Citation and Permissions Morisky Scale identifies low medication 4 or 8 Morisky DE, et al. Med Care 1986 Jan; 24:67-74; Medication adherence in clinical settings Krousel-Wood M et al. Am J Manag Care 2009; Adherence Scale 15(1):59-66. (MMAS) no special permission Brief Adherence Assessment of medication 3 questions and visual Byerly M, et al. Schizophrenia Research 100(1):60-69. Rating Scale adherence of outpatients analogue scale no special permission (BARS) Adherence Assessment of taking medicines 12 Kripalani S, et al. Value Health 2009. to Refills and as prescribed and refilling awaiting permission information Medications Scale medicines on schedule (ARMS) Adherence Starts Assessment of barriers 20 or 12 Matza LS, et al. Ann Pharmacother 2—0; 43(10): with Knowledge to medication adherence 1621-1630. Hahn S, et al. Curr Med Res Opin 2008; (ASK 20, ASK 12) in subscales including 24(7):2127-2138. Matza LS, et al. Curr Med Res Opin inconvenience/forgetfulness, 2008; 24(11):3197-3206. treatment beliefs and behavior permission required to use in a clinical study only Brief Medication Screen for adherence and 9 Svarstad BI, et al. Pt Ed and Conseling 1999; 37:113- Questionnaire barriers to adherence – regimen, 124. (BMQ) belief and recall Treatment Assessment of treatment 9 Atkinson ML, et al. Value Health 2005 Nov-Dec; Satisfaction satisfaction in which there is 8 Suppl 1:S9-S24. Bharmal M. Health Qual Life Questionnaire a potential for side effects to Outcomes 2009 Apr 27;7:36. for Medication interfere with routine clinical care www.quintiles.com/tsqm for permission (Version II and TSQM – 9) Adherence Score serves as a low, medium 3 McHorney CA. Curr Med Res Opin 2009; 25(1):215- Estimator and high risk for non-adherence 238. Contact Merck for user agreement Medication Items to determine the 5 Thompson K, et al. Schizophrenia Research 2000; Adherence Report willingness to take medications 42:241-247. Scale (MARS) everyday Copyright Robert Horne, 1998 for conditions of use – [email protected]

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 73 REFERENCE I

Productivity Measure report. Several different approaches have been developed as a result. These approaches Health-related, on-the-job productivity losses include: assessment of perceived impairment; typically arise from two sources: absenteeism comparison of one’s own performance and and presenteeism. Work absenteeism and its productivity to that of others without health related costs have been measured using self- related-problems; and an estimation of reporting of days missed for health-related unproductive time spent at work. All methods reasons, as well as data from administrative attempt to measure the same construct, loss records of absences, short- or long-term of productivity, by asking the respondent to disability, workers’ compensation and Family evaluate his or her own work performance as a Medical Leave. Challenges arise in using function of time not on task and the quality and administrative data to assess the impact of quantity of work produced as a result. Numerous wellness or chronic care management programs survey instruments have been developed on workplace productivity, as it might be and assessed for validity, consistency and difficult to differentiate days lost to illness from reproducibility. absenteeism associated with non-health-related causes, such as child care, personal days and No matter which questionnaire is used, the vacations. In addition, policy differences in following factors should be considered in disability and workers’ compensation may vary selecting an instrument that is appropriate for by state or by employer, further complicating the setting, the population and the program: comparisons using these data sources across • Instrument reliability and validity. multiple program sites. • Applicability across industries and Presenteeism – a Definition occupations (appropriate for the target population). Variations in the definition of presenteeism result from differences in focus – i.e., health- • A pplicability across the health care continuum related, work-related or more general/life- (appropriate for the target population and related. Moreover, presenteeism has been program). used in both a positive sense (being fully • R epresentativeness of the behavior recall present and productive while at work)15 and period used by the survey to the study in a negative sense (being unproductive while period. 16 being present at work) in more recent studies. • User-friendliness, available languages and For the purposes of our work, the definition of reading level (appropriate for the target presenteeism will reflect the prevailing attitude population). of researchers and employers and therefore will retain the more pejorative connotation and the • M echanism of administration. focus on both health and the workplace. • L ength of survey. • A bility to integrate into other program Presenteeism definition: Presenteeism is processes. decreased on-the-job productivity associated • L icensing and cost requirements. with health concerns or problems.

Monetization Measuring Presenteeism An additional consideration in selecting a Presenteeism losses are challenging to both presenteeism instrument is whether the findings quantify and measure. They do not show up in can be converted to some measure of economic claims data or appear on time cards. Productivity impact. The most common method is to use losses are also intrinsically difficult to measure, salary information about the study population since much work output is not quantifiable or the industry to estimate costs. Given that and many measures must rely solely on self- there are many presenteeism instruments, as

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 74 well as several methods for monetization, it is It is not necessary to monetize presenteeism not surprising that there is no clear consensus losses to evaluate program effectiveness. on how best to quantify presenteeism-related Relative changes in presenteeism are in productivity costs. As one reviewer noted, themselves meaningful outcomes. Careful “the greatest impediment to estimating the consideration of the general and disease- cost of productivity lost to illness is the lack specific presenteeism measures and judicious of established and validated methods for choice of the most appropriate instrument for monetization.”17 the study population and program will help to ensure the validity and reliability of the resulting productivity data.

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 75 REFERENCE J

Selection Criteria Volumes 2 and 3 of the CCA Outcome Guidelines Report developed a philosophical Background and Prior Work framework to allow construction and evaluation The work to develop standardized selection of identification algorithms for the five core criteria began in 2008 with the recognition chronic conditions: diabetes, asthma, chronic that achieving consensus on how to select obstructive pulmonary disease, coronary populations for the evaluation of chronic care artery disease and congestive heart failure. management programs would contribute to The resulting selection criteria are the improving standardized program evaluation, standardized characteristics used to identify as well as help facilitate rigorous performance people for inclusion in the measurement pools comparisons and increase transparency for (denominators) of outcomes metrics. purchasers. In 2009, the work focused on testing the The term “selection criteria” refers to denominators specified in Volume 3 using standardized characteristics (observed in data varying identification time frames and minimum sets) used to identify people for inclusion in the eligibility time frames for their suitability for measurement pools (denominators) of outcomes program comparison. Organizations used their metrics. The project’s goal was to develop a own data to test whether CCA’s denominator fundamental approach to guide denominator specifications produced: prevalence rates that specifications for comparison of chronic care were consistent with their own experience; management programs for the five chronic reasonable specificity without unduly conditions. Specification of selection criteria sacrificing sensitivity when tested over time; requires: identifying data sources to be used; and acceptable overlap between individuals specifying the algorithm to be used to query the identified using CCA selection criteria and those data; and selecting the diagnostic, procedural using proprietary criteria. and other codes for use in the algorithm. 2010 Scope of Work It’s important to emphasize that selection criteria Building on the work in 2009’s Outcomes represent an identification algorithm’s intent to Guidelines Report Volume 4, the Selection accurately identify people who have a disease Criteria Workgroup tested selection criteria while not (falsely) identifying those who do (denominators) for the five common chronic not have the disease. It is well-recognized that conditions for their suitability for comparing administrative data cannot completely succeed the performance of chronic care management at both tasks—that algorithms that identify all programsa on clinical, utilization and financial (or nearly all) people who have a disease will outcomes measures. include some false-positive identifications, and vice versa. In previous volumes of the Outcomes Consistent with our prior work, we defined Guidelines, we discussed the nature of this “suitability” to mean that measures derived “sensitivity-specificity balance.” An additional from the selection criteria are fair and issue with using selection criteria—discussed representative of program experience and are in Volume 4—is that standardized selection acceptable to most programs for the purpose of criteria will produce denominators that may not outcomes comparisons. completely overlap with those produced by programs. This is important to understand for While these criteria may also be appropriately those who use standardized program evaluation used for internal program evaluation and reports to compare programs. improvement, it is not our expectation that they will be; however, we recognize that programs

a These denominators are appropriate for measures in any population.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 76 will not be willing to be compared unless the questions. It is important to bear in mind that CCA selection criteria identify a relevant set of the purpose of constructing these metrics was individuals and is capable of showing the impact not to define actual detailed measures but to of their programs. test the denominator suitability over a variety of conditions and outcomes types. To assess suitability, the workgroup tested the five chronic condition denominators’ General Measure Criteria performance with a basket of clinical, utilization Each measure embodies the following elements: and financial measures. Good performance was – Identification frame (incurred begin/end defined as yielding stable results over time and dates for claims): 24 months with a 3 months’ across vendors (testers). In 2009—as reported paid-through run out past the end of the in Volume 4—the first criterion was tested identification frame. This is the timeframe (overlap); in 2010, the focus was on the second based on date of service, used to determine and third. The fourth and fifth criteria were not whether a member qualifies as “having” the explicitly tested (measure overlap and overlap denominator’s disease. stability over time), but the expectation is that – Measurement frame (incurred begin/end programs will do so using the published criteria: dates for claims): last 12 months of the 24 1. (Reported in Outcomes Guidelines Report, month identification frame Volume 4) The identification overlap between – Minimum eligibility (during the the tester (vendor) and CCA selection criteria measurement frame) is adequate (i.e., most of the people identified by the tester are also found by CCA, and • C linical measures: “HEDIS continuous” - CCA does not require tester to report on too eligible for the entire measurement frame many people who don’t fit their identification with a single allowable gap of up to 45 criteria). days 2. In a given measurement year, the • Utilization and financial measures: Any measurements using the CCA criteria six months, not necessarily contiguous. correlate sufficiently well across testers (i.e., While testing was done with this criterion the measure gives stable results across (to conform to that specified in previous testers). editions of the Guidelines), it is recognized that some organizations may wish to use 3. The results from using CCA selection criteria the same denominators for all outcomes over multiple, consecutive measurement years measures. It is strongly recommended is plausibly stable (i.e., doesn’t vary more than that reporting organizations specify which expected—the measure gives stable results eligibility criterion was used. across time).

4. (Not explicitly tested) In a given measurement Clinical Test Measures year, the measurements using the CCA The test measures were simplified from formal and tester’s own selection criteria correlate comparison measures in that clinical exclusions sufficiently well. were not allowed, and simplified, calendar- 5. (Not explicitly tested) In a series of based identification timeframes were used. measurement years, the results from the CCA The denominator definitions may be found in selection criteria are, on average, directionally the CCA Outcomes Guidelines Report, vol. 5.b consistent with that found by using the The following measures were tested (numerator testers’ (vendors’) own criteria. evidence as per 2010 HEDIS Technical Specifications): We devised clinical, utilization, and financial – D iabetes: A1c test during the measurement metrics to address the second and third year.

b See "Appendix." Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010, p. 121.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 77 – D iabetes: Nephropathy screen or at least consistent with HEDIS at that time. HEDIS now one fill for ACE inhibitor or ARB during the goes through age 50 for asthma. measurement year. – A sthma (ages 5-17): At least one fill for an Financial Measures asthma controller medication during the The following general measure specification measurement year. criteria were used: – A sthma (ages 18-56): At least one fill for an – O utcomes: Year 1 $pmpm (paid) and Year 1 – asthma controller medication during the Year 2 trend for the chronic populations with measurement year. CAD, CHF, diabetes, asthma (separate for population age 5-17 and 18-56), and COPD. – C OPD: At least one fill for a (short- or long- acting) beta-agonist. – M inimum eligibility: at least six months in the measurement year. – C HF: At least one fill for an ACE inhibitor or ARB during the measurement year. Results and Commentary – C HF: At least one fill for a beta-blocker during Several CCA members participated in the testing the measurement year. process. The following results are reported – C AD: LDL-cholesterol testing during the for the four populations on whom all tests measurement year. were performed. These populations are large, geographically diverse, commercial populations Utilization Measures with chronic care management programs. The following general measure specification criteria were used: It was concluded that the criterion of suitability is met because the test results represent – O utcome: All-cause (except maternity and both plausible results and a reasonable range perinatal) emergency department (ED) considering that the four populations measured visits and hospitalizations per 1,000 chronic might have different demographics, coinsurance (by disease) members per year, reported and deductible levels, physician practice separately for members with each of CAD, patterns, population risk and years of disease CHF, diabetes, asthma, and COPD. Also and population health management. report for all diseases combined (eliminating double-counting). Table J1 displays the average condition-specific – D enominators: One for each population with prevalence rates across the four populations, as CAD, CHF, diabetes, asthma (separate for well as the average prevalence rates reported in population age 5-17 and 18-56), and COPD Volume 4. For the 2010 data, both the average who were eligible at least six months in the prevalence and the +- 95th percentiles for measurement year. Year 1 and Year 2 populations are presented. – N umerators: Number of all-cause (except Measured prevalence rates for the five maternity and prenatal) ED visits or chronic conditions (with asthma split into two hospitalizations for members in each age groups) for Year 1 and Year 2 data met denominator. expectations: They were consistent with rates expected by the testers and the workgroup – A hospitalization was counted as occurring members, based on their experience; they were on its date of admission (not its date of consistent across testers; and, with the exception discharge). of CAD, in line with prevalence rates found in – H ospital transfers were not counted as the 2009 tests. Note that in the 2009 tests, it separate hospitalizations. was not possible to present separate averages for juvenile and adult asthma. Finally, prevalence Note: For clinical test measures and utilization rates were found to be consistent between Years measures, asthma was tested through age 56, 1 and 2. to align with the original specifications that were

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 78 Table J1 – Prevalence Rates Percent of Population with Diabetes Asthma Asthma COPD CAD CHF (5-17) (18-56)

V4 CCA Outcome Rate 4.50% 3.90% 3.90% 0.72% 1.47% 0.39% Report (DOR) V5 Testers: Year 1 Rate 5.39% 5.72% 3.20% 0.65% 0.79% 0.33% 95% CI (+/-) 0.34% 2.00% 0.79% 0.03% 0.18% 0.07% V5 Testers: Year 2 Rate 5.49% 5.99% 3.29% 0.66% 0.77% 0.31% 95% CI (+/-) 0.33% 2.20% 0.88% 0.03% 0.18% 0.06%

Table J2 presents condition-specific average rates and +- 95th percentile for Year 1 clinical outcomes across the four populations. Measured clinical outcome rates for Year 1 data for the chronic conditions likewise met expectations: They were consistent across testers. Table J2 – Clinical Measures Percentage with Specified Test or Treatment Diabetes Asthma Asthma COPD CAD CHF (5-17) (18-56)

Annual A1c Nephropathy >= 1 fill for >= 1 fill for >= 1 fill for >= 1 fill for controller LDL Screen Test screen beta-agonist ACE or ARB beta-blocker

Rate 80% 74% 78% 77% 37% 76% 65% 67%

95% Cl (+/-) 8% 2% 4% 5% 8% 12% 4% 5%

Table J3 presents condition-specific rates and the +- 95 percentiles for Year 1 utilization measures. Measured utilization (ED and hospitalization) rates for Year 1 data for the chronic conditions were plausible according to testers’ expectations based on their experience, though showed fairly wide inter-tester variation (not unexpected given the composition differences among populations and the lack of risk adjustment in our results).

Table J3 – Utilization Measures

Diabetes Asthma Asthma COPD CAD CHF (5-17) (18-56)

Admits/1,000 183 38 92 400 530 860

95% Cl (+/-) 20 5 24 69 165 158

ED Visits/1,000 272 376 396 438 444 558

95% Cl (+/-) 65 80 115 148 161 131

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 79 Finally, Table J4 presents the average, condition- $PDMPM and trend) rates for the chronic specific, incurred per diseased member per conditions were plausible according to testers’ month (PDMPM) paid amounts and +- 95 expectations based on their experience, though percentiles for the four populations. Also as with utilization showed fairly wide inter-tester included is the +- 95 percentile for the Year variation. 1/Year 2 trend. Measured financial (Year 1

Table J4 – Financial Measures

Diabetes Asthma Asthma COPD CAD CHF All (5-17) (18-56) Chronic

PMPM (for members with identified $907 $268 $608 $1,330 $1,884 $2,684 $834 conditions)

95% CI (+/-) $158 $50 $30 $213 $134 $834 $165

95% CI (+/-) on PMPM Trend 4% 11% 4% 8% 1% 5% 4%

It was concluded that the results of all tests CCA recommends that the five condition are consistent with meeting the criterion of selection criteria be used as denominators suitability. Based on this conclusion, for clinical utilization and financial measures specifically for program comparison.

Source: CCA. Outcomes Guidelines Report, vol. 5. Washington, DC: Care Continuum Alliance, 2010.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 80 REFERENCE K

I and E Guide Resources

AHRQ Guide – Developing and Running a Primary Care Practice Facilitation Program: A How to Guide pcmh.ahrq.gov/portal/server.pt/gateway/PTARGS_0_11787_956233

AHRQ – Care Coordination Atlas http://www.ahrq.gov/qual/careatlas/

AMGA Guide – ACO’s and Population Health Management http://www.amga.org/AboutAMGA/ACO/Articles/CaseStudy_final.pdf

CDC Guide – A Framework for Patient Centered Health Risk Assessments http://www.cdc.gov/policy/opth/hra/

Health Plans of American Guide – Medicaid Accountable Care http://www.mhpa.org/_upload/Medicaid_Accountable_Care_FINAL1.pdf

National Transitions of Care Coalition (NTOCC) http://www.ntocc.org

PCPCC Guide – Benefits of Implementing PCMH http://www.pcpcc.net/resources

PCPCC Guide – Care Coordination Guide http://www.pcpcc.net/resources

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 81 REFERENCES

1. engagement. 2012. Merriam-Webster.com. Retrieved September 27, 2012, from http://www. merriam-webster.com/dictionary/engagement?show=0&t=1348800229

2. Hanna, F. (2008). Therapy with Difficult Clients: Using the Precursors Model to Awaken Change. Washington D.C.: American Psychological Association.

3. AMGA. ACOs and Population Health Management: How Physician Practices Must Change to Effectively Manage Patient Populations. Available from: http://www.amga.org/AboutAMGA/ ACO/Articles/CaseStudy_final.pdf

4. McDonald KM, Sundaram V, Bravata DM, et al. Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies (Vol. 7: Care Coordination). Rockville (MD): Agency for Healthcare Research and Quality (US); 2007 Jun. (Technical Reviews, No. 9.7.) 3, Definitions of Care Coordination and Related Terms. Available from: http://www.ncbi.nlm.nih.gov/books/ NBK44012/

5. Ward MM, Yankey JW, Vaughn TE, BootsMiller BJ, Flach SD, Welke KF, Pendergast JF, Perlin J, Doebbeling BN. Physician process and patient outcome measures for diabetes care: relationships to organizational characteristics. Med Care. 2004 Sep;42(9):840-50.

6. National Transitions of Care Coalition. Taking Care of MY Health Care. Available from: http:// www.ntocc.org/Portals/0/PDF/Resources/Taking_Care_Of_My_Health_Care.pdf

7. American Recovery and Reinvestment Act, Pub. L. No. 111-5, 123 Stat. 115 (2009).

8. Norman G. Disease management outcomes: are we asking the right questions yet? Population Health Management. 2008; 11(4): 185-188.

9. Berwick DM. The science of improvement. JAMA. 2008 Mar; 299: 1182-1184.

10. Cramer JA, Roy A, Burell A, et al. Medication compliance and persistence: terminology and definitions. Value Health. 2008; 2(1): 44-47.

11. Sikka R, Xia F, Aubert RE. Estimating medication persistency using administrative claims data. Am J Manag Care. 2005 Jul; 11(7): 449-456.

12. Hansen R, et al. Comparison of methods to assess medication adherence and classify nonadherence. The Annals of Pharmacotherapy. 2009; 43(3): 413-422.

13. Leidy NK, Revicki DA, Geneste B. Recommendations for evaluating the validity of quality of life claims for labeling and promotion. Value in Health. 1999; 2(2): 113-127.

14. Ogedegbe G, Mancuso CA, Allegrante JP, Charlson ME. Development and evaluation of a medication adherence self-efficacy scale in hypertensive African-American patients. J Clin Epid. 2003; 56(6): 520-529.

15. Koopman C, et al. Stanford Presenteeism scale: health status and employee productivity. JOEM. 2002; 44(1): 14-20.

16. For example, see Schultz AB and Edington DW. Employee health and presenteeism: a systematic review. Journal of Occupational Rehabilitation. 2007; 17(3): 547-579.

17. Mattke S, Balakrishnan A, Bergamo G, Newberry SJ. A review of methods to measure health- related productivity loss. Am J Manag Care. 2007; 13(4): 211-217.

CARE CONTINUUM ALLIANCE • Implementation and Evaluation: A Population Health Guide for Primary Care Models 82