Association of Team-Based Primary Care with Health Care Utilization and Costs Among Chronically Ill Patients

Research

JAMA Internal Medicine | Original Investigation Association of Team-Based Primary Care With Health Care Utilization and Costs Among Chronically Ill Patients

David J. Meyers, MPH; Alyna T. Chien, MD, MS; Kevin H. Nguyen, MS; Zhonghe Li, MS; Sara J. Singer, MBA, PhD; Meredith B. Rosenthal, PhD

Invited Commentary page 61

IMPORTANCE Empirical study findings to date are mixed on the association between Supplemental content team-based primary care initiatives and health care use and costs for Medicaid and commercially insured patients, especially those with multiple chronic conditions.

OBJECTIVE To evaluate the association of establishing team-based primary care with patient health care use and costs.

DESIGN, SETTING, AND PARTICIPANTS We used difference-in-differences to compare preutilization and postutilization rates between intervention and comparison practices with inverse probability weighting to balance observable differences. We fit a linear model using generalized estimating equations to adjust for clustering at 18 academically affiliated primary care practices in the Boston, Massachusetts, area between 2011 and 2015. The study included 83 953 patients accounting for 138 113 patient-years across 18 intervention practices and 238 455 patients accounting for 401 573 patient-years across 76 comparison practices. Data were analyzed between April and August 2018.

EXPOSURES Practices participated in a 4-year learning collaborative that created and supported team-based primary care.

MAIN OUTCOMES AND MEASURES Outpatient visits, hospitalizations, emergency department visits, ambulatory care–sensitive hospitalizations, ambulatory care–sensitive emergency department visits, and total costs of care.

RESULTS Of 322 408 participants, 176 259 (54.7%) were female; 64 030 (19.9%) were younger than 18 years and 258 378 (80.1%) were age 19 to 64 years. Intervention practices had fewer participants, with 2 or more chronic conditions (n = 51 155 [37.0%] vs n = 186 954 [46.6%]), more participants younger than 18 years (n = 337 931 [27.5%] vs n = 74 691 [18.6%]), higher Medicaid enrollment (n = 39 541 [28.6%] vs n = 81 417 [20.3%]), and similar sex distributions (75 023 women [54.4%] vs 220 097 women [54.8%]); however, after inverse probability weighting, observable patient characteristics were well balanced. Intervention practices had higher utilization in the preperiod. Patients in intervention practices experienced a 7.4% increase in annual outpatient visits relative to baseline (95% CI, Author Affiliations: Department of Health Services, Policy, and Practice, 3.5%-11.3%; P < .001) after adjusting for patient age, sex, comorbidity, zip code level Brown University School of Public sociodemographic characteristics, clinician characteristics, and plan fixed effects. In a Health, Providence, Rhode Island subsample of patients with 2 or more chronic conditions, there was a statistically significant (Meyers, Nguyen); Department of 18.6% reduction in hospitalizations (95% CI, 1.5%-33.0%; P = .03), 25.2% reduction in Pediatrics, Harvard Medical School, Division of General Pediatrics, emergency department visits (95% CI, 6.6%-44.0%; P = .007), and a 36.7% reduction in Department of Medicine, Boston ambulatory care–sensitive emergency department visits (95% CI, 9.2%-64.0%; P = .009). Children’s Hospital, Boston, Among patients with less than 2 comorbidities, there was an increase in outpatient visits Massachusetts (Chien); Department (9.2%; 95% CI, 5.10%-13.10%; P < .001), hospitalizations (36.2%; 95% CI, 12.2-566.6; of Health Policy and Management, Harvard T.H. Chan School of Public P = .003), and ambulatory care–sensitive hospitalizations (50.6%; 95% CI, 7.1%-329.2%; Health, Boston, Massachusetts (Li, P = .02). Rosenthal); Department of Medicine, Stanford University School of CONCLUSIONS AND RELEVANCE While establishing team-based care was not associated with Medicine, Stanford, California (Singer). differences in the full patient sample, there were substantial reductions in utilization among a Corresponding Author: David J. subset of chronically ill patients. Team-based care practice transformation in primary care Meyers, MPH, Department of Health settings may be a valuable tool in improving the care of sicker patients, thereby reducing Services, Policy, and Practice, Brown avoidable use; however, it may lead to greater use among healthier patients. University School of Public Health, 121 S Main St, Box G-S121-3, JAMA Intern Med. 2019;179(1):54-61. doi:10.1001/jamainternmed.2018.5118 Providence, RI 02912 (david_meyers Published online November 26, 2018. @brown.edu).

54 (Reprinted) jamainternalmedicine.com

Downloaded From: https://jamanetwork.com/ on 09/23/2021 Association of Team-Based Primary Care With Health Care Utilization and Costs Among Chronically Ill Patients Original Investigation Research

here is increasing recognition that creating health care teams is critical to improving health care quality and Key Points value. Teaming refers to the dynamic activities, includ- T Question What is the association of a team-based primary care ing coordination and collaboration, that allow individuals to transformation collaborative initiative with patient health care work together to deliver shared goals.1 Some elements of highly utilization and costs? functioning primary care teams and teaming have been de- Findings In this study, among chronically ill patients in 18 scribed in isolated parts of the American health care system, practices who were exposed to team-based care, there was an 2,3 yet team-based care is not the norm. 18% reduction in hospitalizations, a 25% reduction in emergency Past work has found teaming can be effective at improv- department visits, and a 36% reduction in ambulatory ing clinical care and outcomes.4 Introducing teams in hospi- care–sensitive emergency department visits relative to 76 tal settings has reduced mortality and length of stay,4 and care comparison practices. Among healthier patients, there was an delivered by geriatric teams improved elderly patients’ func- increase in outpatient visits and hospitalizations. tional status, mental health, and independence compared with Meaning Team-based approaches to primary care transformation control groups.4-6 Teams based in primary initiatives have may benefit patients with chronic illness by reducing the use of found modest gains, although findings are mixed.7-10 Gaps in acute care; however, it may lead to higher use among healthier the literature remain. More work is needed to understand how patients. academic medical practices’ use of team-based care, how team- based care affects safety net practices, and which patients ben- Required team activities included daily 15-minute “huddles” efit most from these interventions. and implementation of population management systems (eg, sys- In 2012, Harvard Medical School launched the Academic tematic identification and follow-up with patients who required Innovations Collaborative (AIC), a multiyear, multisite care screenings). Team members were required to attend thrice-yearly learning collaborative aimed at establishing team-based care 1.5-day learning sessions and regular interactive webinars to con- at its affiliated primary care practices.11 We examine the asso- nect with other practices and share strategies for improving clini- ciation of this intervention with health care utilization and cal quality through team-based care. To support these efforts, costs. We make several important contributions to the litera- practices received an unrestricted $3 per member per month ture. First, we used an All Payer Claims Database (APCD) that payment to support the transformation process during the provided us with all commercial and Medicaid claims for the first 2 years, and a $0.50 to $1 payment for the latter 2 years of practices in our study, granting us a more complete view of pa- the initiative. These funds could be used by the practice for tient care than single-payer studies. Second, we used a differ- any purpose and did not need to be directly linked to the AIC ence-in-difference approach with inverse probability weight- intervention. ing to isolate the effects of the intervention on patient To date, examinations of the AIC have demonstrated that utilization and costs. Third, our study includes intervention transitions to team-based care involved changing practice con- practices across 6 different academic medical centers, whereas figurations (ie, who worked with whom) and composition past work has typically focused on changes within just (ie, ratios of certain types of personnel to physicians) more than 1 medical center. it changed the overall size of practices (ie, total number of staff within practices).12-17 They have also shown that better team dynamics (eg, team members’ ability to understand each oth- Methods er’s roles) were associated with greater satisfaction with clinical work among primary care clinicians and with more posi- Intervention tive perceptions regarding patient safety among all staff. This study was approved by the Harvard T.H. Chan School of Pub- Qualitative studies have illustrated how the establishment of lic Health institutional review board. A waiver of patient consent primary care teams can provide important scaffolding for resi- was granted because the study used administrative data. The AIC dents when they feel stressed and inadequate during conti- initiative was inspired by the need to control increasing health nuity clinic and with greater job satisfaction among medical care costs through more efficient care delivery and aimed to im- assistants despite a higher workload.12-17 prove care through team-based management of chronic illness.11 Before the AIC, practice-level care delivery was largely decen- Data Sources tralized because most of the included practices had not estab- Our primary data source was the Massachusetts APCD from 2011 lished patient care teams, and participation in multisite collab- to 2015.18 The APCD contains monthly plan enrollment and 100% orative initiatives was limited. medical claims on all residents of Massachusetts, collected across The AIC intervention involved substantial reorganizing of ex- all commercial payers and Medicaid. These files underwent sub- isting clinicians into care teams, empaneling patients to primary stance abuse redaction19 but predate the March 2016 Supreme care clinicians, and introducing a series of activities related to Court decision that allowed self-insured plans to opt out of sub- teamwork. Patients who sought care at the affiliated practices mitting claims to APCDs.20 Our sample includes almost all medi- were assigned to a care team, and the care teams were encour- cal claims data available for patients in Massachusetts younger aged to take ownership of all aspects of the patient’s care. Prac- than 65 years. tices affiliated with Harvard Medical School could choose to opt We linked National Provider Identification (NPI) numbers into the intervention. available in APCD claims to the Massachusetts Health Quality

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine January 2019 Volume 179, Number 1 55

Downloaded From: https://jamanetwork.com/ on 09/23/2021 Research Original Investigation Association of Team-Based Primary Care With Health Care Utilization and Costs Among Chronically Ill Patients

Partners Provider Database file21 to identify claims associated with tainment rates, race/ethnicity percentages, and poverty rates. AIC patients. Using the Massachusetts Health Quality Partners Using the propensity score, we calculated inverse probability Provider Database, we identified all 18 AIC practices and 76 other of treatment weights (IPTW) to balance selection on observ- academically affiliated primary care practices in the greater able characteristics.23 We formally assessed covariate bal- Boston area as a comparison group. Clinicians can practice in mul- ance using the propensity scores to assess their performance tiple locations, so we included any NPI affiliated with an AIC prac- and visually assess the assumption of common support. tice regardless of their other affiliations. To ensure that we captured a complete list of clinicians at each AIC practice, we used Outcomes of Interest lists provided by each practice to the study team of their Our primary outcomes of interest were patient-level utiliza- affiliated clinicians and looked up individual NPIs if they were tion rates and costs following a framework used to evaluate not included in the Massachusetts Health Quality Partners similar interventions.24 For each patient, we calculated their Provider Database. annual outpatient, hospitalization, and emergency depart- We have limited data on patient characteristics, so we in- ment visit rate. We also calculated their ambulatory care– cluded additional zip code level sociodemographic character- sensitive emergency department utilization rate based on the istics from the American Community Survey. Additionally, we New York University algorithm,25 and their ambulatory care– included details on clinician credentials, sex, and specialty from sensitive hospitalization rate based on the Agency for Health- the national NPI registry. care Research and Quality algorithm.26 We were particularly interested in the ambulatory care–sensitive outcomes be- Patient Attribution cause these are the visits that may have been most prevent- To attribute patients to practices, we limited our sample to en- able through better planned and coordinated primary care. rollees with at least 6 months of enrollment within a year. We In addition to utilization rates, we calculated the total costs also excluded those older than 65 years and those dually eli- of care for each patient annually excluding drug costs. We could gible for Medicare and Medicaid because Medicare claims were not include drug costs because pharmacy benefit managers not available for our analysis. For each year, we identified as a were not required to report data to the APCD. All other medi- patient’s primary clinican the clinician NPI with which pa- cal costs were calculated as the total allowed amounts (ie, the tients had the plurality of their evaluation and management copayment plus the health plan paid amount) for all of a pa- visits that year. Patients could be reattributed each year if they tient’s medical services within a year, regardless of where they appeared to change clinician or practice based on evaluation received those services. We fit separate cost models for Med- and management visit patterns. If a patient did not have any icaid and non-Medicaid enrollees because the allowed amounts evaluation and management visits within a calendar year, then may differ between public and private payers. they were not attributed and therefore they were not included in that year for analysis. Statistical Methods For each patient, we collected from the APCD their total To estimate the effects of the AIC intervention on outcomes, months of enrollment each year, their plan identification, their we used a standard difference-in-difference framework.27 The plan type (eg, health maitenance organization, preferred pro- difference-in-difference approach allows for the calculation of vider organization, or point of service) their age group the mean treatment effect of an intervention by comparing the (≤18 years and 19-64 years; the APCD did not include detailed changes in the preperiod and postperiod times between the age information) and their sex. We also calculated an Elix- intervention and comparison groups. For our analysis, we con- hauser Comorbidity index for each patient each year, and clas- sidered the preperiod to be 2011 and 2012, and the postpe- sified those with 2 or more chronic conditions as our chroni- riod 2014 and 2015. We excluded data from 2013 in our analy- cally ill sample.22 sis as that was the first year of the implementation, and for many practices, the focus was more on planning and self- Comparison Practice Selection assessment than adoption of new approaches to team-based To create an adequate comparison group, we first included all care. We visually and formally checked the assumption of academically affiliated primary care practices in the greater parallel trends in the preperiod for valid difference-in- Boston area. We retained practices with 3 to 100 affiliated phy- difference inference. sicians because this was the range in size for the AIC prac- To account for clustering at the practice level within the tices. We attributed patients to the 76 practices that met this analysis, we fit a linear model using generalized estimating criteria using the same approach as for AIC practices. equations.28,29 Each model included an indicator of AIC prac- Because there were some observable differences in pa- tice, an indicator of being in the postperiod, and their inter- tient characteristics between these practices, we estimated pro- action term that provided the difference-in-difference coef- pensity scores for assignment to an AIC practice. We used a logit ficient. We include additional covariates for sex, months of model with a binary outcome of AIC assignment and covari- enrollment, Medicaid enrollment, a linear time trend, pa- ates for sex, the patient’s age group, Medicaid status, type of tient’s Elixhauser index, zip code sociodemographic charac- plan patient was enrolled in (eg, health maintenance organi- teristics, clinician credential (MD/NP/RN), specialty, and sex. zation, preferred provider organization, or point of service), We included plan fixed effects to adjust for potential differ- months of enrollment in their primary plan that year, Elix- ences between insurance plan benefits and weight the obser- hauser comorbidity index, and zip code level educational at- vations using the IPTWs.

56 JAMA Internal Medicine January 2019 Volume 179, Number 1 (Reprinted) jamainternalmedicine.com

Downloaded From: https://jamanetwork.com/ on 09/23/2021 Association of Team-Based Primary Care With Health Care Utilization and Costs Among Chronically Ill Patients Original Investigation Research

Table 1. Descriptive Characteristics of the Study Sample

Unweighted, No. (%) IPW Weighted,a % P Value Variable Comparison Intervention Difference Comparison Intervention P Value Difference Patient characteristics, No. Unique patients 238 455 83 953 NA 238 455 83 953 NA Patient-years 401 573 138 113 NA 401 573 138 113 NA Practices 76 18 NA 76 18 NA ≥2 Comorbidities 46.6 37.0 <.001b 43.8 43.3 .98 Age ≤18, y 14 353 (18.6) 23 087 (27.5) <.001b 20.6 20.2 .20 Age 19-64, y 194 102 (81.4) 60 866 (72.5) <.001b 79.4 79.8 .20 Female 130 673 (54.8) 45 586 (54.3) <.001b 54.6 54.4 .24 Medicaid 48 406 (20.3) 24 011 (28.6) <.001b 21.1 19.7 <.001b Mean enrollment, mo 11.7 11.7 <.001b 11.7 11.7 <.001b Clinician characteristicsc MD or DO degree 228 201 (95.7) 79 839 (95.1) <.001b 95.6 95.7 <.001b Nurse practitioner or registered nurse 6438 (2.7) 1007 (1.2) <.001b 2.4 2.2 .55 Internal medicine 162 865 (68.3) 52 890 (63.0) <.001b 66.9 68.1 .57 Family medicine 41 014 (17.2) 12 845 (15.3) <.001b 16.7 16.3 .23 Obstetrician-gynecologist 2623 (1.1) 1091 (1.3) <.001b 1.1 1.1 .09 Other 5246 (2.2) 672 (0.8) <.001b 1.8 1.8 .14 Female clinician 142 358 (59.7) 46 762 (55.7) <.001b 58.7 57.6 <.001b Zip code characteristics, %d High school graduate 89.4 87.9 <.001b 89.0 89.1 .19 Disability 10.4 10.7 <.001b 10.5 10.5 .96 US born 75.3 72.2 <.001b 74.5 74.7 .002b Primary language not English 29.2 33.5 <.001b 30.3 30.2 .03b Unemployed 8.6 9.0 <.001b 8.7 8.7 .21 Social security 6.1 6.6 <.001b 6.2 6.2 .16 Receiving SNAP benefits 12.9 14.8 <.001b 11.3 13.2 .04b Public insurance 31.8 33.6 <.001b 32.3 32.2 .45 Uninsured 4.0 4.5 <.001b 4.1 4.1 .19 Poverty 13.1 15.4 <.001b 13.7 13.6 .02b White 71.1 68.4 <.001b 70.4 70.7 <.001b Black 16.6 19.3 <.001b 17.3 17.1 .04 AI/AN 0.7 0.8 <.001b 70.0 70.0 .07 Asian 10.0 10.1 <.001b 10.0 9.9 <.001b Hispanic 10.9 14.1 <.001b 11.7 11.7 .42 Abbreviations: AI/AN, American Indian/Alaska Native; IPW, inverse probability c Clinician characteristics represent counts and proportions of patients who weighting; NA, not applicable; SNAP, Supplemental Nutrition Assistance have a clinician of a given characteristic. Program. d Zip code characteristics are means taken from the American Community a Weighted by inverse probabilities from a logit propensity score model Survey and measured at the zip code level. This information was not available adjusting for all covariates in this table plus plan type. at the patient level for this study. b Indicates statistically significant difference from comparison practices.

We fit additional models stratified by at least 2 chronic conditions and less than 2 chronic conditions to test whether the Results intervention had differential associations among those more chronically ill. Because there was a 14% increase in patient Of all residents in the state who met the eligibility criteria, share of Medicaid enrollees among the AIC practices relative 35.4% were able to be attributed to any primary care clinician to comparison practices, we separated the cost outcomes by (n = 1 519 652). Our final sample included 238 455 unique pa- payer source. We fit additional sensitivity analyses both with tients contributing 401 573 person-years between 2011/2012 and without the plan fixed effects and IPTWs and among Med- and 2014/2015 in comparison practices and 83 953 patients ac- icaid enrollees only. All analysis was conducted in SAS, ver- counting for 138 113 patient-years in AIC practices. Descrip- sion 9.4 (SAS Institute Inc). P values less than .05 were con- tive characteristics of the sample are presented in Table 1, both sidered statistically significant, and all P values were 2-sided. with and without IPTW weighting. Prior to weighting, there

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine January 2019 Volume 179, Number 1 57

Downloaded From: https://jamanetwork.com/ on 09/23/2021 Research Original Investigation Association of Team-Based Primary Care With Health Care Utilization and Costs Among Chronically Ill Patients

were significant differences between the 2 practice types in all member-per-month payment. The APCD data we used also al- characteristics. After weighting, the differences between in- lowed us a more complete view on utilization that may have tervention and comparison practices were largely mitigated; taken place outside of the health systems in which our inter- however, some differences in covariates still remained, nota- vention practices are located. bly in percentage of Medicaid patients and months of enroll- What may have driven these associations? The increase in ment; however, the absolute differences were small. The pro- outpatient visits for the full sample may be a result of im- pensity scores met the common support assumption which is proved care planning and treatment in the AIC practices that displayed visually in eFigure 1 in the Supplement. The full could have encouraged patients to make more frequent pri- propensity score model is also available in eTable 1 in the mary care appointments. One of the earliest components of Supplement. the transition to team-based care was patient empanelment, Across all types of utilization, the AIC practices had which, combined with efforts to target improvement in evi- higher rates in the preperiod than the comparison practices. dence-based care processes (eg, cancer screening and chronic In eFigure 2 in the Supplement, we present figures and illness management) may have increased planning and out- empirical tests of the parallel trends assumption and find reach for preventive care. These additional visits and other non– that parallel trends exist in all variables except for total cost visit-based interactions that team members may initiate of care and outpatient visits in the more than 2 comorbidity (eg, email or telephone consultations) offer increased oppor- sample. Parallel trends were not met in any outcome in the tunities to anticipate and prevent exacerbations of illness, and less than 2 comorbidity subsample. In Table 2, we present while they may initially cost more, there is a chance they would the difference-in-difference models for the full sample of be beneficial for patients in the long run. patients and the chronic condition strata. Results are pre- Our analysis also found statistically significant increases sented per 1000 person-years. Full model output is avail- in outpatient use and hospitalizations among patients with less able in eTable 2 in the Supplement. than 2 comorbidities. Hospitalization rates did decrease among In the full sample, there was a statistically significant 7.4% intervention practices; however, they decreased at a faster rate increase in outpatient visits (95% CI, 3.5%-11.3%; P < .001) for among comparison practices, leading to the large difference patients in intervention practices compared with those in com- in difference. However, it is difficult to interpret these results parisons. There were no other significant differences be- because the parallel trends assumption was not met in these tween AIC and non-AIC practices. In the chronically ill models, indicating that the comparison practices may not have subsample, there was an 18.6% reduction in hospitalizations provided an adequate comparison for healthier patients. Nev- (95% CI, 2.3%-35.6%; P = .03), a 25.2% reduction in emer- ertheless, the AIC intervention may have resulted in higher uti- gency department visits (95% CI, 6.4%-47.6%; P = .007), and lization. We cannot determine whether the increase in hospi- a 36.7% reduction in ambulatory care–sensitive emergency de- talizations among this group are the result of an increase in partment visits (95% CI, 17.9%-64.1%; P = .009). While the dif- patients getting needed care as a result of care coordination. ference was not statistically significant, the total cost of care among the chronically ill subsample trended downwards. Limitations Among patients with less than 2 comorbidities, we de- Our study is subject to several limitations. First, our primary tected statistically significant increases in outpatient visits source of data were administrative claims, which do not include (9.2%; 95% CI, 5.1%-13.1%; P < .001), hospitalizations (36.2%; as detailed clinical information as may be found in electronic 95% CI, 12.2%-566.6%; P < .001) and ambulatory care– health records. However, utilization and cost outcomes are gen- sensitive hospitalizations (50.6%; 95% CI, 7.1%-329.2%; erally well captured by insurance claims. Second, our attribution P = .02). The increases in hospitalization were primarily driven of patients to physicians and physicians to practices was imper- by a larger reduction among comparison practices rather than fect. While we assigned patients according to common practice an actual increase in intervention practices. In sensitivity analy- in the literature, it is possible that some patients in the final ses, we found similar trends only among Medicaid enrollees sample could have been more accurately assigned to a different (eTable 3 in the Supplement), both in the full sample and in practice or simply did not have enough exposure to any practice the 2 stratified samples. for valid assignment. Third, the APCD is limited in the number of patient characteristics available. While the IPTW weighting ap- pears to have adequately balanced our observable covariates, Discussion there may be other differences between AIC and comparison patients that may have led to residual confounding. Fourth, the time In the chronically ill strata, we found clinically relevant re- of the study coincides with large changes to the health care sys- ductions in hospitalizations and emergency department vis- tem under the Affordable Care Act. While the difference-in- its 2 years after a collaborative primary care transformation difference framework is robust to systemic changes as long as effort was initiated, and in a healthier strata, we found statis- they do not affect the intervention and comparison groups dif- tically significant increases in outpatient and hospital utiliza- ferentially, we cannot rule out other changes that may have hap- tion. We are aware of 1 study that found similar results from a pened differentially at AIC practices during this time. Relatedly, team-based care initiative in the Intermountain Health we observe that AIC practices have higher utilization across most System.9 We believe our study makes important additional con- utilization types in the preperiod than the comparison practices. tributions because we found greater effect sizes for a lower per- This may reflect other inherent differences in the types of prac-

58 JAMA Internal Medicine January 2019 Volume 179, Number 1 (Reprinted) jamainternalmedicine.com

Downloaded From: https://jamanetwork.com/ on 09/23/2021 Downloaded From: https://jamanetwork.com/ on 09/23/2021 soito fTa-ae rmr aeWt elhCr tlzto n ot mn hoial l Patients Ill Chronically Among Costs and Utilization Care Health With Care Primary Team-Based of Association jamainternalmedicine.com Table 2. Difference-in-Difference Regression Resultsa

Baseline, Visits per 1000 Patients Post, Visits per 1000 Patients Difference-in-Difference Effect as of Intervention Baseline Relative to Comparisons, Result Comparison AIC Comparison AIC Difference-in-Difference (95% CI) % P Value Full sample Outpatient visits 3499.1 3659.5 3951.4 4382.1 270.4 (128.6 to 412.2) 7.40 .<.001 Inpatient hospitalizations 81.6 140.3 53.3 117.6 5.6 (−13.8 to 24.9) 4.00 .57 Ambulatory-sensitive hospitalizations 21.2 42.2 43 22.7 2 (−8.7 to 12.7) 4.70 .72

08Aeia eia soito.Alrgt reserved. rights All Association. Medical American 2018 © Emergency department visit 417.9 669.7 274.8 536 9.4 (−57.5 to 76.3) 1.40 .78 Ambulatory-sensitive emergency 129.1 212.4 91.4 159.4 −15.4 (−48.8 to 18.1) −7.30 .37 department visit Total cost of care, $ Non-Medicaid 5 903 694.00 7 393 556.00 5 149 373.00 6 148 431.00 24 230.80 (−869 138.00 to 917 599.40) 0.30 .96 Medicaid only 5 261 900.00 26 170 000.00 4 093 167.00 20 948 000.00 −50 358.60 (−6 747 440.00 to 6 646 723.00) −0.20 .99 All 5 602 300.00 11 249 000.00 1 656 860.00 7 776 900.00 473 066.70 (−1 254 160.00 to 2 200 291.00) 4.20 .59b ≥2 Chronic conditions Outpatient visits 5194.8 4820.3 6872.4 6587.7 89.8 (−212.1 to 391.7) 1.90 .56b Inpatient hospitalizations 134.6 260.1 172 249 −48.4 (−92.7 to −4.1) −18.60 .03 Ambulatory-sensitive hospitalizations 62.5 117.6 24.6 59.7 −20.1 (−56 to 15.7) −17.10 .27

(Reprinted) Emergency department visit 421.9 971.2 427.3 732.1 −244.5 (−424.4 to −64.6) −25.20 .008b Ambulatory-sensitive emergency 195.4 398 204.8 261.4 −145.9 (−255 to −36.8) −36.70 .009 department visit Total cost of care, $ b AAItra Medicine Internal JAMA Non-Medicaid 14 276 000.00 18 020 000.00 4 598 660.00 8 402 930.00 60 305.20 (−3 248 930.00 to 3 369 542.00) 0.30 .97 Medicaid only 15 675 510.00 79 615 730.00 13 170 350.00 53 571 140.00 −14 589 900.00 (−31 050 700.00 to 1 870 873.00) −18.30 .08b All 6 662 730.00 20 357 000.00 4 121 210.00 13 325 000.00 −4 490 140.00 (−9 397 380.00 to 417 095.10) −22.10 .07 <2 Chronic conditions Outpatient visits 3352.7 3461.1 3540.3 3967.8 319.2 (175.1 to 463.2) 9 <.001b Inpatient hospitalizations 44.3 70.8 7.5 59.6 25.6 (8.7 to 42.5) 36 .003b Ambulatory-sensitive hospitalizations 4.2 16.8 4.8 27.9 8.5 (1.2 to 15.8) 50 .02b b

aur 09Vlm 7,Nme 1 Number 179, Volume 2019 January Emergency department visit 452.4 601.8 273 503.6 81.1 (21.1 to 141.2) 13 .08 Ambulatory-sensitive emergency 116.9 162 72 132.2 14.8 (−7.8 to 37.5) 9 .20b department visit Investigation Original Total cost of care, $ Non-Medicaid 5 170 100 6 330 850 1 837 880 3 031 320 32 695.6 (−600 834 to 666 225.2) 0.5 .92b Medicaid only 8 027 810 15 750 000 1 568 690 19 856 000 9 309 000 (2 519 087 to 19 294 520) 59.1 .01b All 5 721 860 8 062 730 2 205 650 6 980 170 2 433 652 (870 046.50 to 3 997 257) 30 .002b Abbreviation: AIC, Academic Innovations Collaborative. postperiod means are adjusted from the difference-in-difference model.

a Each row represents a separate regression model. Costs are presented for commercial insurance and Medicaid b Statistically significant difference in trends in the preperiod, potentially indicative of a violation of the parallel Research separately as Medicaid mix changes over time in AIC practices, which may affect cost savings. Baseline and trends assumption. 59 Research Original Investigation Association of Team-Based Primary Care With Health Care Utilization and Costs Among Chronically Ill Patients

tices included in the intervention. Fifth, our data from the Mas- utilization that may or may not be necessary for healthier patients. sachusetts APCD did not include claims for Medicare enrollees, Policymakers and health system leaders may do well to consider limiting our analysis (and its generalizability) to a younger pa- similar team-based care approaches in their health systems as tient population. Finally, our results only reflect changes in uti- part of a portfolio of efforts to bend the cost curve. lization and costs and not the quality of patient care outcomes. More research is needed to determine the extent to which the intervention led to changes in patient health outcomes. Conclusions This study reflects changes implemented by a single network of academically affiliated practices. We observed large reductions In conclusion, we found that a collaborative primary care in use for chronically ill patients and increases in utilization transformation initiative conducted across 6 academic among healthier patients. While the same effects may not gen- medical centers was associated with substantial reductions eralize to all other locations, the magnitude of the effects for in emergency department and hospital utilization among a chronically ill patients suggest that collaborative primary care sample of chronically ill patients and increases in hospital- transformation initiatives may be worthwhile to address avoid- izations and outpatient utilization among healthier patients. able utilization of inpatient and emergency services. However, Similar care transformation initiatives may be valuable for these improvements may come at the cost of increases in some managing primary care in other settings.

ARTICLE INFORMATION Trudy Bearden, Jonathan Sugarman, Cory Sevin, 11. Bitton A, Ellner A, Pabo E, et al; The Harvard Accepted for Publication: August 7, 2018. and Rebecca Steinfield. Medical School Academic Innovations Collaborative. The Harvard Medical School Published Online: November 26, 2018. REFERENCES Academic Innovations Collaborative: transforming doi:10.1001/jamainternmed.2018.5118 1. Edmondson AC. Teaming: How Organizations primary care practice and education. Acad Med. Author Contributions: Dr Meyers had full access to Learn, Innovate, and Compete in the Knowledge 2014;89(9):1239-1244. doi:10.1097/ACM. all of the data in the study and takes responsibility Economy. Hoboken, NJ: John Wiley & Sons; 2012. 0000000000000410 for the integrity of the data and the accuracy of the 2. Grumbach K, Bodenheimer T. Can health care 12. Chien AT, Kyle MA, Peters AS, et al. Establishing data analysis. teams: how does it change practice configuration, Concept and design: All authors. teams improve primary care practice? JAMA. 2004; 291(10):1246-1251. doi:10.1001/jama.291.10.1246 size, and composition? J Ambul Care Manage. 2018; Acquisition, analysis, or interpretation of data: 41(2):146-155. doi:10.1097/JAC. Meyers, Nguyen, Li, Singer, Rosenthal. 3. Committee on Quality of Health Care in America 0000000000000229 Drafting of the manuscript: Meyers, Chien. I of M. Crossing the Quality Chasm. A New Health Critical revision of the manuscript for important System for the 21st Century. Washington, DC: National 13. Song H, Ryan M, Tendulkar S, et al. Team intellectual content: Nguyen, Li, Singer, Rosenthal. Academy Press; 2001. dynamics, clinical work satisfaction, and patient Statistical analysis: Meyers, Li, Rosenthal. care coordination between primary care providers: 4. Lemieux-Charles L, McGuire WL. What do we a mixed methods study. Health Care Manage Rev. Obtained funding: Chien, Singer, Rosenthal. know about health care team effectiveness? a Administrative, technical, or material support: 2017;42(1):28-41. doi:10.1097/HMR. review of the literature. Med Care Res Rev. 2006;63 0000000000000091 Meyers, Chien, Nguyen. (3):263-300. doi:10.1177/1077558706287003 Supervision: Rosenthal. 14. Song H, Chien AT, Fisher J, et al. Development 5. Carter BL, Rogers M, Daly J, Zheng S, James PA. Conflict of Interest Disclosures: None reported. and validation of the primary care team dynamics The potency of team-based care interventions for survey. Health Serv Res. 2015;50(3):897-921. doi: Funding/Support: This study is supported by hypertension: a meta-analysis. Arch Intern Med. 10.1111/1475-6773.12257 funding from the Harvard Medical School Center for 2009;169(19):1748-1755. doi:10.1001/archinternmed. Primary Care and the Controlled Risk Insurance 2009.316 15. Blumenthal KJ, Chien AT, Singer SJ. Relationship among team dynamics, care coordination and Company Risk Management Foundation of the 6. Winkler NS, Damento GM, Khanna SS, Hodge Harvard Medical Institutions. Dr Meyers was perception of safety culture in primary care. Fam DO, Khanna CL. Analysis of a physician-led, Pract. 2018. doi:10.1093/fampra/cmy029 additionally supported by an National Institute of team-based care model for the treatment of Aging T32 training fellowship, and Mr Nguyen was glaucoma. J Glaucoma. 2017;26(8):702-707. doi:10. 16. Sheridan B, Chien AT, Peters AS, Rosenthal MB, additionally supported by an Agency for Healthcare 1097/IJG.0000000000000689 Brooks JV, Singer SJ. Team-based primary care: the Research and Quality T32 training fellowship. medical assistant perspective. Health Care Manage Additional support for data analysis was provided 7. Wen J, Schulman KA. Can team-based care Rev. 2018;43(2):115-125. doi:10.1097/HMR. by Rappaport Institute for Greater Boston. improve patient satisfaction? a systematic review of 0000000000000136 randomized controlled trials. PLoS One. 2014;9(7): Role of the Funder/Support: The funding sources e100603. doi:10.1371/journal.pone.0100603 17. Brooks JV, Singer SJ, Rosenthal M, Chien AT, had no role in the design and conduct of the study; Peters AS. Feeling inadequate: residents’ stress and collection, management, analysis, and 8. Medves J, Godfrey C, Turner C, et al. Systematic learning at primary care clinics in the United States. interpretation of the data; preparation, review, or review of practice guideline dissemination and Med Teach. 2017;0(0):1-8. doi:10.1080/0142159X. approval of the manuscript; and decision to submit implementation strategies for healthcare teams and 2017.1413236 the manuscript for publication. team-based practice. Int J Evid Based Healthc. 2010;8(2):79-89. doi:10.1111/j.1744-1609.2010. 18. The Center for Health Information and Analysis. Additional Contributions: The authors 00166.x The Massachusetts All Payer Claims Database. acknowledge the following individuals who http://www.chiamass.gov/ma-apcd/. Accessed May designed and implemented the AIC intervention: 9. Reiss-Brennan B, Brunisholz KD, Dredge C, et al. 24, 2018. Association of integrated team-based care with Jenny Azzara, Asaf Bitton, Juliana DiLuca, Andrew 19. Frakt AB, Bagley N. Protection or harm? Ellner, Kristen Goodell, Russ Phillips, Gordy Schiff, health care quality, utilization, and cost. JAMA. 2016;316(8):826-834. doi:10.1001/jama.2016.11232 suppressing substance-use data. N Engl J Med. Soma Stout, Jessica Zeidman, Rick Lopez, Kim 2015;372(20):1879-1881. doi:10.1056/NEJMp1501362 Ariyabuddhiphongs, Louise Mackisack, Holly Oh, 10. Misra-Hebert AD, Rabovsky A, Yan C, Hu B, Sam Skura, Joe Froklis, Lori Tishler, David Bor, Rob Rothberg MB. A team-based model of primary care 20. Bland SE, Crowley JS, Gostin LO. Strategies for Chamberlin, Fiona McCaughan, Assaad Sayah, delivery and physician-patient interaction. Am J Med. health system innovation after Gobeille v Liberty Kathleen Conroy, Joanne Cox, Matt Carmody, Linda 2015;128(9):1025-1028. doi:10.1016/j.amjmed.2015. Mutual Insurance Company. JAMA. 2016;316(6): Powers, Valarie Stone, Blair Fosburgh, Peter 03.035 581-582. doi:10.1001/jama.2016.8293 Greenspan, Eric Weil, Carol Keohane, Luke Sato,

60 JAMA Internal Medicine January 2019 Volume 179, Number 1 (Reprinted) jamainternalmedicine.com

Downloaded From: https://jamanetwork.com/ on 09/23/2021 Association of Team-Based Primary Care With Health Care Utilization and Costs Among Chronically Ill Patients Original Investigation Research

21. MHQP. Massachusetts Provider Database. and reduce costs of care? a measurement and 28. Zeger SL, Liang K-Y, Albert PS. Models for http://www.mhqp.org/products_and_tools/? research agenda. Med Care Res Rev. 2010;67(4): longitudinal data: a generalized estimating equation content_item_id=226. Accessed May 12, 2018. 476-484. doi:10.1177/1077558710368412 approach. Biometrics. 1988;44(4):1049-1060. doi: 22. Elixhauser A, Steiner C, Harris DR, Coffey RM. 25. Billings J, Anderson GM, Newman LS. Recent 10.2307/2531734 Comorbidity measures for use with administrative findings on preventable hospitalizations. Health Aff 29. Hubbard AE, Ahern J, Fleischer NL, et al. To data. Med Care. 1998;36(1):8-27. (Millwood). 1996;15(3):239-249. GEE or not to GEE: comparing population average 23. Cole SR, Hernán MA. Constructing inverse 26. Bindman AB, Grumbach K, Osmond D, et al. and mixed models for estimating the associations probability weights for marginal structural models. Preventable hospitalizations and access to health between neighborhood risk factors and health. Am J Epidemiol. 2008;168(6):656-664. doi:10. care. JAMA. 1995;274(4):305-311. Epidemiology. 2010;21(4):467-474. doi:10.1097/EDE. 0b013e3181caeb90 1093/aje/kwn164 27. Dimick JB, Ryan AM. Methods for evaluating 24. Rosenthal MB, Beckman HB, Forrest DD, changes in health care policy: the Huang ES, Landon BE, Lewis S. Will the difference-in-differences approach. JAMA. 2014;312 patient-centered medical home improve efficiency (22):2401-2402. doi:10.1001/jama.2014.16153

Invited Commentary Anatomy and Physiology of Primary Care Teams Thomas Bodenheimer, MD

Health care teams have a structure (anatomy) and culture 4 (agree) on the Likert scale. The share-the-care component (physiology). Team structure has 2 facets: (1) who is on the team of team culture at AIC practices is unstudied. Challenges to and (2) how stable is the team. Primary care teams are often team-based care abound: lack of reimbursement for regis- composed of a core team or teamlet (clinicians working with tered nurses and other team members, scope of work laws, cli- medical assistants) and an ex- nicians lacking trust in the team, and forging a culture change

Related article page 54 tended-care team (regis- from “me” to “we.” tered nurses, pharmacists, so- The article by Meyers et al1 adds to previous evidence on cial workers, and behaviorists) that supports several core teams. primary care team effectiveness. Intermountain Healthcare has Team stability means that members of the team always work a long history of team building, dividing its practices into plan- together and patients on the team’s panel receive all care from ning, adoption, and routinized phases of team building. Rou- their team. tinized practices have at least 6 years of team-based care with Team culture also has 2 components: (1) how team mem- standardized workflows. Patients treated in routinized prac- bers work together and (2) how teams share the care, ie, dis- tices, compared with those in nonteam practices, have higher tribute patient care functions among team members. How team rates of diabetes control (blood pressure lower than 140/90 mm

members work together can be assessed with such instru- Hg, hemoglobin A1c lower than 8% [to convert to proportion ments as the Primary Care Team Dynamics Survey1 or the Team of total hemoglobin, multiply by 0.01], and low-density lipo- Culture Scale.2 Examples of sharing the care include training protein cholesterol lower than 100 mg/dL [to convert to mil- medical assistants to independently identify and close care limoles per liter, multiply by 0.0259]), and lower hospital ad- gaps (overdue cancer screenings, immunizations, or routine mission and emergency department use.4 diabetes services) and empowering RNs and pharmacists to in- A 2012 survey of 231 clinicians and 280 staff at 16 primary dependently care for patients with uncomplicated diabetes or care clinics in San Francisco looked at core team structure and hypertension, including titrating medication doses within burnout. Core teams were described in 3 categories: clini- standing orders.3 Sharing the care can make team members’ cians almost always working with the same medical assistant jobs more interesting while reducing clinicians’ work that does (teamlet), clinicians working with a small group of medical as- not require their level of education. sistants, and no stable core team at all. Team culture was mea- The Meyers et al1 article in this issue is a study of 18 Aca- sured with an 8-item Team Culture Scale. For clinicians, the demic Innovations Collaborative (AIC) practices that have ex- emotional exhaustion component of burnout was high when perienced a 4-year journey toward team-based care. Patients team culture was low. When team culture was high, emo- with 2 or more chronic conditions in AIC practices had signifi- tional exhaustion was significantly lower for clinicians work- cantly lower hospitalizations and emergency department vis- ing with the same medical assistant compared with clinicians its than those in comparison practices.1 without a stable team.2 Primary care teams vary widely in the stability of their team To examine the share-the-care aspect of team culture, a structure and their adoption of a collaborative and share-the- 2013 survey was administered to 326 clinicians and 142 staff care team culture. How do the 18 AIC practices studied by in 19 San Francisco primary care clinics. Share-the-care was Meyers et al1 fare? From surveys referenced in Meyers et al,1 measured by asking whether clinicians had confidence that we know that AIC team members only slightly agreed (3.58 on medical assistants (MAs) could independently assume respon- a 5-point Likert scale) that their teams were stable. State- sibility for panel management, ie, identify care gaps, discuss ments about a positive team culture were scored by the prac- the gaps with patients, and use standing orders to close the gaps tices’ clinicians between 3 (neither agree nor disagree) and for cancer screenings, immunizations, and routine diabetes ser-

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine January 2019 Volume 179, Number 1 61

Downloaded From: https://jamanetwork.com/ on 09/23/2021