Primary Care Physician Organization Affiliations with Health Systems between 2014 and 2019 Were Associated with Modest Improvements in Quality of Care

Abstract

We examined whether affiliation of independent primary care physician organizations (POs) with vertically-integrated health systems improved quality of care or service utilization for Original Medicare enrollees overall and for groups of POs expected to benefit from affiliation. Using national Medicare claims, PECOS, and IRS-990 data from 2013 to 2019, we estimated changes in 13 quality and utilization measures following affiliation. We found that affiliation was not consistently associated with meaningful improvements in preventive care, chronic disease management, or utilization. Modest gains were observed in diabetic eye exams and organization-level care continuity, whereas primary care visits and specialist visits declined. Small POs exhibited additional gains in breast cancer screening relative to large POs, while low-quality POs and rural POs experienced lower relative performance on at least one other measure. These results suggest that health system affiliation alone does not yield broad improvements in quality across diverse types of primary care POs.

INTRODUCTION

Vertically integrated health systems (hereafter, “health systems”) bring together hospitals and physician organizations (POs) under common ownership to achieve economies of scale and scope and, potentially, to improve care coordination through operational alignment (Burns et al., 2013). Vertical integration has expanded rapidly over the past decade: the proportion of physicians employed by hospitals or health systems grew from roughly one-quarter in 2012 to over 55 percent in 2024 (Avalere Health, 2024). This consolidation has been driven by health systems seeking greater leverage with payers, stronger control of referrals for specialty and hospital services, and enhanced ability to capture savings under alternative payment models by managing care across the continuum (Burns et al., 2000; O'Malley et al., 2011).

For many physicians, employment within a health system appears attractive. Salary-based payment, job stability, predictable hours, and access to advanced technology and staffing support are frequently cited advantages (Christianson et al., 2014; Lee et al., 2005). Health systems also claim to deliver more “integrated” care by facilitating specialist access and improving communication between primary care and specialty clinicians. Common electronic health record (EHR) platforms can provide real-time access to patient information and recent utilization to support tailored care and smoother transitions. For these reasons, many newly trained physicians now favor hospital employment over independent practice by nearly two-to-one across most specialties, including primary care (Neprash et al., 2020).

However, affiliation also introduces challenges. Surveys of employed physicians reveal concerns about reduced autonomy, productivity pressures, and weakened patient relationships. In one 2023 national survey of physicians, a majority of hospital or health system-employed physicians reported that ownership changes worsened patient care, reflecting reduced physician autonomy in clinical decisionmaking, curtailed communication and shorter visits (Physicians Advocacy Institute, 2023). Centralized management can further limit clinicians’ influence over front-office operations, while patients often describe fragmented communication and increased reliance on call centers (Chuang et al., 2022). Expectations to see more patients or meet extensive documentation requirements may heighten administrative burden, calling into question whether system employment reliably improves the quality or continuity of ambulatory care (Reid et al., 2022).

Despite vigorous debate, empirical evidence on how provider consolidation affects care quality has several limitations. Existing studies are predominantly cross-sectional or focus on hospitals, and the few longitudinal analyses of PO ownership transitions show improvements in structural capabilities—such as health information technology and quality-improvement infrastructure—but not necessarily in outcomes (Bishop et al., 2016; Friedberg et al., 2009). Some research has found better cancer screening or lower emergency department use following affiliation, yet these studies examined single regions and may not generalize nationally (Carlin et al., 2015; Crespin et al., 2016). To date, no work has assessed the longer-term quality or utilization effects of health system affiliation among a large, diverse sample of primary care POs or examined heterogeneity in effects by PO characteristics. In particular, small, rural, and lower-quality POs are especially likely to encounter operational constraints and infrastructure limitations that limit their ability to provide accessible, comprehensive, and coordinated care (Lee et al., 2005; Rittenhouse et al., 2011). Health system affiliation could provide access to shared technology, managerial support, and care coordination resources that are often unavailable in resource-limited or geographically isolated settings (Harvey et al., 2020; Ridgely et al., 2020). Similarly, lower-quality POs may particularly benefit from the standardized processes and quality-improvement capabilities of larger health systems (Weiner, 2009).

To address this gap, we conducted a nationwide longitudinal analysis of primary care POs to evaluate the impact of health system affiliation on quality of care and service use for Medicare beneficiaries. We ask two related questions: (1) How does health system affiliation affect quality and utilization among primary care POs? and (2) Do these effects differ for small, rural, or lower-quality POs that may stand to benefit most from health system affiliation? We hypothesize that health system infrastructure—such as data platforms, quality-improvement processes, and operational support—offers the greatest potential advantage for these organizations.

NEW CONTRIBUTION

This study provides one of the most comprehensive national, longitudinal assessments to date of how affiliation with vertically integrated health systems by independent POs affects quality of care and service use among Medicare beneficiaries. Uniquely, we focus on primary care physician organizations—the front line of patient care and the setting where coordination, continuity, and population health management are most directly influenced by system integration. Prior studies have relied largely on cross-sectional or hospital-based data, often limited to a single state or sub-state region, which limits generalizability given the substantial variation in health system structure, market dynamics, payment policy, and patient populations across the U.S. (Furukawa et al., 2020). In contrast, our analysis incorporates POs from across the nation, capturing the diversity of U.S. health system environments and allowing for findings that are broadly applicable to policymakers. Using linked PECOS and IRS data, we follow POs for up to five years post-affiliation to assess a broad set of preventive, chronic-care, medication-adherence, and utilization outcomes. Furthermore, beyond examining overall effects, we assess heterogeneity in impacts by three PO characteristics—size, rurality, and baseline quality—to provide a more nuanced view of why health system affiliation may lead to different outcomes in different contexts. By exploring these differential effects, our study helps shed light on possible mechanisms that may explain changes in care delivery following affiliation and highlighting when and where integration is most likely to improve quality or require greater support from health systems.

METHODS

Identifying physician organizations.

We used the Medicare Data on Provider Practice and Specialty (MDPPAS) to identify all taxpayer identification numbers (TINs) for which at least one physician billed a plurality of Medicare Part B services using the TIN between 2013 and 2019. For 265 of these TINs that were associated with academic health systems, we used crosswalks derived by Welch (2016) and updated by our team to combine TINs from the same system, because academic health systems often used multiple TINs corresponding to specialty-specific departments. We used the resulting set of single and combined TINs to represent unique “physician organizations” (POs). To focus the analysis on primary care POs, we restricted the sample to POs with at least one primary care physician and at least two total physicians in the baseline year (2013). Primary care was defined by a specialty of general internal medicine, family medicine, geriatric medicine, or general practice. We also required POs to bill Medicare in each year and have at least 30 attributed beneficiaries (see Medicare beneficiary attribution below) eligible for a quality or utilization measure in each year (Appendix Table 1). This “balanced panel” design minimizes bias from compositional changes due to the entry and exit of POs from the sample, while the denominator size thresholds minimize statistical noise.

Measuring PO affiliations with health systems.

We used data from the Medicare Provider Enrollment, Chain, and Ownership System (PECOS) and Internal Revenue Service (IRS) Form 990 Schedule R from 2013 to 2019 to measure PO affiliations with health systems. We define health systems as entities composed of at least one short-term acute care hospital and at least one PO that is affiliated through shared ownership or contracting relationships for payment and service delivery. Using PECOS, we defined affiliated POs as those that were either owned or managed by a health system using relationship codes in the data: (1) a 5 percent or greater direct or indirect ownership interest in the PO or (2) operational or managerial control over the PO whether or not there is an ownership interest by the system. For health systems with one or more non-profit hospitals, we supplemented the PECOS data with IRS Form 990 data by identifying all POs that were listed on a health system’s tax form. For 11 POs that were identified as affiliated in two non-consecutive years but not listed as affiliated in the intervening year(s), we assumed that the PO was affiliated in each intervening year.

Identifying newly affiliating POs.

Identifying new health system affiliations is challenging when using TINs as a PO identifier because affiliating POs commonly adopt the TIN of their health system leading to a censoring problem for legacy TINs in Medicare FFS claims data. To address this issue, we first identified POs for which the majority of physicians billing with a TIN in one calendar year begin using a different TIN in the next year and for which the original TIN never appears in subsequent years in the MDPPAS. These TIN changes could reflect either a consolidation between organizations or simply a change in the use of a TIN without an actual consolidation. Newly affiliating POs were identified as those with both a change of TIN between two consecutive years and a change from no affiliation to affiliation, as measured in the PECOS/IRS 990 data, between the two years. For these POs, we refer to the PO’s original TIN as its “legacy” TIN and the new TIN as the “surviving” TIN, and we updated the TIN listed for each provider in the MDPPAS (i.e., surviving TIN) with the provider’s legacy TIN. We refer to this version of the MDPPAS file as the “updated MDPPAS,” and we carried forward the legacy TIN in each subsequent year of the updated MDPPAS so long as the provider continued to bill Medicare Part B services most commonly using the surviving TIN listed in the original MDPPAS.

Medicare beneficiary attribution.

Our sample comprised all Medicare beneficiaries enrolled in Original Medicare aged 65 and older whose usual source of care was an eligible PO. We attributed beneficiaries to the PO that provided a plurality of evaluation and management (E&M) visits to primary care providers (including physicians, nurse practitioners, and physician assistants) in the calendar year. For beneficiaries without visits to primary care providers, we attributed beneficiaries to the PO that provided a plurality of E&M visits to internal medicine subspecialists (See Appendix Table 2 (Specialist visits) for a list of these specialties). We used both the “original MDPPAS” and “updated MDPPAS” to attribute beneficiaries to POs. We first attributed beneficiaries to POs based on TINs listed in the original MDPPAS file. We then used the updated MDPPAS to update PO attributions for beneficiaries attributed to a surviving TIN to determine whether to attribute the beneficiary to the legacy TIN or surviving TIN based on the plurality of E&M visits to providers associated with each TIN.

Measuring PO characteristics.

We used MDPPAS to measure the number of unique physicians practicing at each PO and each PO’s mix of physician specialties (i.e., primary care only, primary care and other specialties, or only specialties other than primary care). We used Medicare enrollment files to measure the characteristics of beneficiaries attributed to each PO, including their age, gender, race, ethnicity, dual-eligibility, and original entitlement due to disability. To serve as a proxy for beneficiary comorbidities, we used Hierarchical Condition Category (HCC) scores. Rurality was measured using categorizations of Rural-Urban Commuting Area (RUCA) codes (U.S. Department of Agriculture Economic Research Service, 2023). Each beneficiary’s neighborhood socioeconomic deprivation was measured using a ZIP code-level index composed of six items from the American Community Survey (Bird et al., 2010).

Quality of care and utilization measures.

We used 100% Medicare fee-for-service (FFS) claims files to derive eight quality measures that reflect delivery of preventive services (breast cancer screening), chronic condition management (diabetes eye exams and statin prescribing for patients with cardiovascular disease), medication adherence (adherence to blood pressure medications), care coordination and transitions (timely follow-up after acute events and all-cause hospital readmissions), and primary care continuity at both the physician and PO levels. We also used five utilization measures: primary care clinician visits, internal medicine specialist visits, ophthalmologist or optometrist visits, Emergency Department (ED) visits, and a composite measure of ambulatory care sensitive hospitalizations and ED visits. Appendix Table 2 provides detailed specifications for each measure.

Statistical analysis.

We used linear mixed models with a time-varying indicator of affiliation status that was equal to one in the first full calendar year following affiliation with a health system, fixed effects for year, and random effects for PO. This approach yields effect estimates, captured by the coefficient on affiliation status, that averages the impact of affiliation across anywhere from one to five post-affiliation years depending on the year of affiliation for each PO (Timbie et al., 2025). All analyses used a random sample of beneficiaries in both affiliating and non-affiliating POs for computational efficiency. Appendix Table 3 reports full samples and analytic sample sizes for each measure in the baseline year. The model adjusted for beneficiary age, gender, original Medicare entitlement due to disability, race and ethnicity, HCC score, RUCA category, and area-level socioeconomic deprivation. All models included adjustment for the following time-varying PO characteristics: number of attributed beneficiaries, number of physicians, specialty mix (primary care only vs. primary care and other specialties), mean beneficiary age, mean risk score, mean area-level deprivation, percent disabled, percent female, percent of beneficiaries in each RUCA category, and percent of beneficiaries by race and ethnicity. Inclusion of these characteristics helps to control for changes in patient mix and PO attributes unrelated to affiliation that may be associated with changes in quality and utilization and is consistent with our assumption of conditional parallel trends.

We used an event-study design to confirm the presence of conditional parallel trends in the pre-affiliation period between POs that were never affiliated and POs that became affiliated later in the study period for each measure included in the analysis (Bailey & Goodman-Bacon, 2015) (Appendix Table 4). Specifically, we included leads of affiliation switching (e.g., indicators for one, two, three, and four or more years prior to affiliation) and used the p-values and coefficient estimates to determine whether affiliated and unaffiliated POs exhibited different trends prior to the former becoming affiliated. If they are close to zero and not significant, the event study model does not refute the parallel trends assumption. Appendix Table 5 includes samples sizes and standard errors for all estimates. Adjustment for multiple comparisons using the Benjamini Hochberg procedure did not change the statistical significance of any of these results.

Subgroup analyses.

We examined the consistency of findings from our main analysis across three types of PO subgroups defined by baseline quality, size, and rurality. Low-quality POs were defined as POs in the lower quartile of performance in 2013 (the baseline year) whereas high-quality POs were defined as those in the upper quartile of performance. Each group was defined separately for each measure. Thus, the same PO might appear in the low-quality subgroup for one measure but the high-quality subgroup for another measure. Small and large POs were defined as POs with less than 10 clinicians versus 10 or more clinicians, respectively. Rural POs were defined as those for whom more than 10 percent of the Medicare beneficiaries attributed to the PO lived in rural or small-town zip codes in 2013 based on RUCA codes; non-rural POs were defined as all other POs. For each PO subgroup, we fit the same linear mixed models described previously separately. We then derived confidence intervals for the difference in affiliation impacts between subgroups to determine the statistical significance of subgroup differences. Appendix Tables 6–8 include samples sizes and standard errors for all subgroup analyses. Given their exploratory nature, no adjustments were made for multiple comparisons.

RESULTS

A total of 7410 primary care POs and over 6 million Medicare FFS beneficiaries attributed to these POs per year met all eligibility criteria and were included in the analysis. POs that affiliated with health systems during the study period were larger than non-affiliating POs in terms of both the number of physicians (18 vs. 10) and attributed beneficiaries (1,515 vs. 876) in the baseline year (2013) (Table 1). Affiliating POs were also more likely than non-affiliating POs to include both primary care physicians and physicians with other specialties as opposed to primary care physicians only (43 percent vs 26 percent). Beneficiaries attributed to affiliating POs in the baseline year were less likely to be dually eligible for Medicaid (8.9 percent vs 11.6 percent) and more likely to be White (89.3 percent vs 85.0 percent) and less likely to be Asian/Pacific Islander, Black, Hispanic, or some other race. Beneficiaries attributed to the two types of POs were similar with respect to rurality, HCC scores, and area-level socioeconomic indicators.

Table 1.

Beneficiary and PO characteristics at baseline (2013)

	POs that Affiliate with Health Systems between 2014 and 2018	POs that do not Affiliate with Health Systems between 2014 and 2018
Beneficiary-Level Characteristics	N=796,118 beneficiaries	N=5,500,355 beneficiaries
Age, mean (SD)	75.7 (7.6)	75.5 (7.5)
Female, %	59.5	59.2
Dual eligibility, %	8.9	11.6
Disabled, %	8.2	8.7
Race/ethnicity, %
Asian/Pacific Islander	1.2	1.9
Black	5.5	6.5
Hispanic	2.6	4.5
White	89.3	85.0
Other	0.8	1.4
Rurality, %
Metropolitan	79.2	79.9
Micropolitan	11.3	11.1
Small Town	5.8	5.3
Rural	3.6	3.6
HCC Score, mean (SD)	1.1 (1.0)	1.1 (1.0)
Area-level SES Index, mean (SD)	0.2 (0.9)	0.1 (0.9)
Organization-Level Characteristics	N=573 POs	N=6,837 POs
Number of beneficiaries, mean (SD)	1,514.9 (2,389.8)	875.7 (1,700.6)
Number of physicians, mean (SD)	18.0 (47.3)	10.3 (71.7)
Specialty mix, n (%)
Primary care only	56.9	73.7
Primary care and other specialties	43.1	26.3

Note. Characteristics of 7,410 POs meeting all inclusion criteria for the analysis are displayed in the table. All characteristics are measured in the baseline year (2013). All differences between affiliating POs and non-affiliating POs are statistically significant (p<0.001) except for HCC Score. SD=Standard deviation; HCC=Hierarchical Condition Category; SES=Socioeconomic Status

Baseline quality and utilization metrics differed across POs and between PO subgroups (Table 2). In subgroups defined by quality of care in the baseline year, we observed the largest differences, including a 30-percentage point difference in rates of breast cancer screening between high- and low-quality POs and a similar difference for diabetic eye exams (28 percentage points). Differences in rates of follow-up care after acute events were also notably higher (22 percentage points) in high-quality POs compared with low-quality POs. Differences in quality between rural POs and urban POs were much smaller across all measures. The largest baseline differences included a 4-percentage point lower rate of breast cancer screening and a 3-percentage point lower rate of diabetic eye exams among beneficiaries who received care from rural POs compared with urban POs. We observed few differences between small and large POs across the eight quality measures.

Table 2.

Baseline PO Performance, by Measure

Measure	All POs	Low-Quality or Low-Utilization POs	High-Quality or High-Utilization POs	Small POs	Large POs	Rural POs	Urban POs
Quality Measures, %
All-Cause Hospital Readmissions	14	9	18	13	14	14	14
Breast Cancer Screening	72	52	82	71	73	69	73
Continuity of Care – PCC, mean	77	65	87	78	76	79	77
Continuity of Care – PO, mean	87	75	92	87	86	88	86
Diabetes Eye Exams	60	44	72	59	60	58	60
Follow-up after Acute Events	68	56	78	67	68	67	68
Hypertension Medication Adherence	83	74	87	83	83	83	83
Statin Prescription for CVD	68	57	76	67	68	66	68
Utilization Measures, Visits per 100 Beneficiaries
Primary care clinician visits	442	304	636	454	416	456	438
Internal medicine specialist visits	200	105	358	194	214	142	218
Ophthalmologist/optometrist visits	114	81	144	113	114	103	117
ED visits	37	23	57	36	40	42	35
Ambulatory care-sensitive ED visits/hospitalizations	8	4	14	8	8	9	7

Notes: Performance estimates are unadjusted and are measured in the baseline year, 2013. PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease, ED=emergency department.

Baseline differences in rates of visits to selected specialists were notable for small and rural POs, both of which exhibited similar patterns compared to large and urban POs. Beneficiaries receiving care from small POs had higher rates of primary care visits in the baseline year (38 more visits per 100 beneficiaries than beneficiaries receiving care from large POs) (Table 2). Similarly, beneficiaries in rural POs had 17 more visits per 100 beneficiaries than beneficiaries in urban POs. However, beneficiaries in small and rural POs had lower rates of specialist visits in the baseline year (21 fewer visits per 100 beneficiaries for beneficiaries in small POs compared to large POs and 76 fewer visits per 100 beneficiaries for beneficiaries in rural POs than urban POs).

Over a follow-up period of up to 5 years following a PO’s affiliation with a health system, we found few statistically significant and clinically meaningful improvements in quality for affiliating POs relative to non-affiliating POs (Table 3). The largest improvements observed were a 1.5-percentage point relative increase in continuity of care with a beneficiary’s attributed PO (p<0.001) and a 0.9-percentage point relative improvement in diabetic eye exams (p<0.001). Meanwhile, continuity of care with a beneficiary’s usual primary care clinician decreased by 1.6 percentage points relative to beneficiaries in non-affiliating POs (p<0.001). We found no other meaningful improvements in quality of care for other process, medication adherence, or readmission measures. Only clinician-level primary care continuity was associated with differential trends between affiliating and non-affiliating POs prior to affiliation (Appendix Table 4) and may explain the relative reductions for affiliating POs observed for that measure.

Table 3.

Estimated impact of health system affiliation on quality of care

	All POs		Low-quality POs*		High-quality POs*		Small POs		Large POs		Rural POs		Urban POs
Measure	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value
All-cause Hospital Readmissions	−0.1	0.624	−0.1	0.704	0.3	0.287	−0.2	0.435	0.0	0.800	0.5**	0.069	−0.2	0.113
Breast Cancer Screening	0.2	0.350	0.6	0.368	−0.1	0.838	1.0**	0.008	0.0	0.825	0.3	0.454	0.1	0.466
Continuity of Care – PCC	−1.6	0.000	−0.8**	0.000	−2.1	0.000	−2.2**	0.000	−0.5	0.000	−1.4	0.000	−1.7	0.000
Continuity of Care – PO	1.5	0.000	2.8**	0.000	1.6	0.000	1.2**	0.000	1.6	0.000	1.6	0.000	1.5	0.000
Diabetes Eye Exams	0.9	0.000	0.9	0.296	−0.4	0.423	1.7	0.003	0.6	0.043	−0.4**	0.406	1.3	0.000
Follow-up after Acute Events	0.6	0.124	−5.3**	0.000	−1.0	0.391	−0.4	0.687	0.8	0.111	2.3	0.011	0.4	0.428
Hypertension Med. Adherence	−0.3	0.018	0.2	0.708	−0.2	0.305	0.0	0.938	−0.5	0.003	−0.2	0.478	−0.3	0.021
Statin Prescription for CVD	0.3	0.267	0.4	0.686	1.4	0.026	0.0	0.985	0.1	0.709	0.0	0.935	0.4	0.221

PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease. Linear mixed models were used to estimate the impact of health system affiliation, and all measures use a scale from 0–100. Thus, all estimates represent the average change in each measure, in percentage points, following affiliation.

^*“Low quality” was defined as the lowest quartile of PO performance for each measure except for All-cause Hospital Readmissions, for which “low quality” was defined as performance in the upper quartile.

^**Statistically significant difference between subgroups.

Subgroup analyses defined according to a PO’s quality of care, size, and rurality in the baseline year revealed heterogeneous changes in quality for selected measures for some subgroups but largely reflected the overall findings (Table 3). For example, small POs that affiliated with health systems experienced modest improvements in breast cancer screening (1.0 percentage points) that were not observed for large POs (p<0.05 for all comparisons between subgroups reported here). Lower-quality POs that affiliated with health systems were not associated with relative improvements in any measure except within-PO care continuity and were associated with large reductions in follow-up visits after acute events (5.3 percentage points) despite no such reduction for high-quality POs. Rural POs that affiliated with health systems exhibited modest reductions in diabetic eye exams (−0.4 percentage points) despite improvements for urban POs and increases in readmission rates (0.5 percentage points) relative to no changes for urban POs.

Overall, visits with primary care clinicians decreased by 9 visits per 100 beneficiaries and visits with internal medicine specialists decreased by 4 visits per 100 beneficiaries for POs that affiliated with health systems relative to non-affiliating POs (p<0.001) (Table 4). Despite these changes, affiliating POs were not associated with either relative increases or decreases in either ED visits or a composite measure of avoidable hospitalizations and avoidable ED visits. Changes in utilization for PO subgroups mimicked the overall results in all but a few cases. Most notably, the reductions in primary care clinician visits observed in the main analysis were concentrated within POs that had the highest primary care visit rates at baseline while the reductions in specialist visits were primarily among large and urban POs and those with the lowest specialist visit rates at baseline.

Table 4.

Estimated impact of health system affiliation on utilization measures

	All POs		Low-utilization POs*		High-utilization POs*		Small POs		Large POs		Rural POs		Urban POs
Measure	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value	Estimate	P value
Primary care clinician visits	−9.1	0.000	−3.6	0.142	−30.4	0.000	−9.8	0.000	−7.3	0.000	−13.1	0.000	−8.3	0.000
Internal medicine specialist visits	−3.6	0.000	−4.7**	0.002	11.6	0.001	0.0	0.991	−3.7	0.003	−0.3	0.854	−4.3	0.001
Ophthalmologist/ optometrist visits	1.0	0.103	1.1	0.408	−0.7	0.615	1.2	0.270	1.1	0.122	1.4	0.209	0.8	0.253
ED visits	0.6	0.071	0.0	0.966	0.2	0.748	1.1	0.034	0.5	0.219	1.5	0.030	0.3	0.411
Ambulatory care-sensitive ED visits/ hospitalizations	0.1	0.658	−0.4	0.365	−0.4	0.055	−0.2	0.398	0.1	0.619	0.6	0.057	−0.1	0.593

ED=emergency department. Linear mixed models were used to estimate the impact of health system affiliation. All estimates represent the average change in visits per 100 beneficiaries following affiliation.

^*Low-utilization and high-utilization subgroups are defined as the lowest and highest quartile for each measure, respectively. Although lower utilization is not necessarily “better” for all measures, reductions in ED visits and ambulatory care-sensitive ED visits/ hospitalizations are generally considered to be indicators of improved PO performance.

^**Statistically significant difference between subgroups.

DISCUSSION

In one of the largest national studies to examine how independent, primary care POs perform after affiliating with health systems, we found limited evidence that these affiliations lead to meaningful improvements in the quality of primary care. Across multiple dimensions of care—including preventive services, chronic disease management, medication adherence, and care coordination—performance remained largely unchanged after affiliation. The few areas showing modest relative gains, such as diabetic eye exams and continuity of care at the organization level, were potentially offset by corresponding declines in visits with primary care clinicians and specialists. Results varied only slightly across PO subgroups: small POs exhibited modest improvements in one additional measure, breast cancer screening, whereas low-quality and rural POs showed mixed effects including worse follow-up care for low-quality POs relative to high-quality POs and both higher readmission rates and lower rates of diabetic eye exams for rural POs relative to urban POs. Taken together, our findings suggest that health system affiliation alone is unlikely to produce broad improvements in the quality of care for primary care POs even for subgroups that might be expected to benefit most from affiliation.

The limited effects of health system affiliation for primary care POs are consistent with prior studies that have documented weak or mixed associations with quality of care (Ianni et al., 2024; Liu et al., 2022; Machta et al., 2019). Several factors may explain this pattern. First, meaningful practice transformation may require several years of investment to implement standardized care-management processes and performance feedback mechanisms used by health systems or to optimize electronic health records (Bazzoli et al., 2004; Christianson et al., 2014). Although we followed POs for up to five years, the average duration of follow-up was just over three years and, as a result, we might have captured only the early stages of organizational changes, before such initiatives are fully operational. Additionally, some processes may even deteriorate temporarily after affiliation as practices adapt to modified workflows, reporting requirements, or patient redistribution. In particular, if health system physicians are more constrained in their ability to devote adequate time to patients during visits due to productivity requirements or restrictive visit scheduling protocols, reductions in patient access could blunt improvements in quality (Physicians Advocacy Institute, 2023). Although we observed reductions in both primary care visits and specialist visits following affiliation, it remains unclear whether these findings indicate poorer access or more efficient delivery of services.

Despite examining three subgroups of primary care POs that theoretically stood to gain the most from health system affiliation, we found no evidence of disproportionate benefit for these groups. Small, rural, or lower-performing POs often face significant operational challenges, including limited information technology, insufficient support staff, and weaker quality-improvement infrastructure, all of which hamper their capacity to adopt new care processes or manage population health effectively after joining a larger system. Health system integration could confer advantages by supplying access to formal clinical protocols, advanced data tools, and managerial resources. However, successful uptake of these supports may require organizational readiness and strong local leadership, qualities that low-performing or resource-constrained POs may initially lack (Shaw et al., 2012; Weiner, 2009). Similarly, affiliations might offer small organizations ready access to EHRs, decision support, and dedicated quality-improvement personnel—resources otherwise out of reach (Lee et al., 2005; Rittenhouse et al., 2011; Wolfson et al., 2009)—as well as standardized workflows that could ease administrative burden and improve protocol adherence (Ridgely et al., 2020). Limited improvements for small, rural, or lower-performing POs suggests that the hypothesized advantages of system affiliation are either deployed equally across affiliating PO types or used at insufficient levels to drive quality improvement.

The substantial variation in health system characteristics observed across the United States may also help account for the generally modest or null benefits of affiliation identified in our study. A growing literature has documented that health systems differ widely in structural complexity, organizational relationships, integration strategies, and leadership approaches (Furukawa et al., 2020; Ridgely et al., 2020; Scanlon et al., 2020). While our subgroup analyses focused on size, rurality, and baseline quality, other unmeasured attributes of POs or their health systems may be more determinative of an organization's ability to benefit from affiliation. For example, the degree and type of integration between the PO and the health system (such as shared governance structures, clinical integration, or intensity of management support, financial arrangements, local competitive environment, or the priorities of system leadership in primary care redesign could all play important roles in improving quality outcomes (Harvey et al., 2020; Ridgely et al., 2020; Scanlon et al., 2020). Similarly, differences in physician or staff turnover, patient case mix, or the presence of local “champions” for quality improvement may influence each PO’s post-affiliation trajectory. Future research should seek to elucidate these factors to better identify drivers of effective affiliation processes that produce the largest benefits to patient care.

Another potential explanation for the limited improvements we observed is that motivations for vertical integration may be unrelated to the goals of enhancing primary care quality. Strategic drivers may include increasing market power, negotiating leverage with payers, expanding referrals to affiliated hospitals, or optimizing financial performance rather than direct investments in clinical quality for ambulatory patients. Indeed, previous studies have documented that vertical integration and health system consolidation are frequently associated with increased prices and spending (Baker et al., 2014; Capps et al., 2018), even as quality improvement remains modest or inconsistent. Recent literature also highlights that integrated delivery systems can facilitate greater “upstream” referrals and intensity of coding, boosting system revenues (Post et al., 2022), while the expected returns from primary care redesign or investments in high-value chronic care management may be secondary considerations.

This study has several limitations that should be considered when interpreting the results. Although our analysis uses national data on POs, we excluded single-physician practices and limited the sample to organizations that deliver primary care services, which included both POs that provide primary care exclusively as well as multi-specialty practices, potentially masking heterogeneity in the effects of affiliation. The balanced-panel approach and accompanying size requirements further restrict the applicability of findings to comparatively larger POs. Our data sources also pose inherent limitations. Because no definitive dataset exists for identifying health system affiliation, we drew on both PECOS and IRS filings—a strategy employed in prior research (Cohen et al., 2017). Nonetheless, some independent organizations may have been incorrectly classified as system-affiliated, and the reverse may also occur. Moreover, the analytic sample was confined to POs whose affiliation with a health system was identified by both concurrent changes in PECOS ownership or management relationships and by a new billing TIN. While this stringent rule reduces the likelihood of misclassification, it may fail to identify some affiliations. Moreover, the term “affiliation” can encompass heterogeneous arrangements—ranging from employment models to contractual or other partnership forms—that could influence outcomes differently; however, insufficient detail in available data prevented separate examination of these differences. Future research might consider heterogeneity in ownership percentages reported in sources such as PECOS. Because Medicare Advantage encounter data were unavailable for the early study years, analyses were limited to beneficiaries enrolled in Original Medicare, which may affect generalizability. We also did not examine potential lagged effects of joining a system or effects that might vary by year over the post-affiliation period, which could provide additional insight. Finally, our evaluation relied solely on administrative claims, which restricts assessment to measures observable in billing data. A more comprehensive assessment would include beneficiary-reported experiences and direct clinical indicators—such as hemoglobin A1c and blood pressure control.

CONCLUSION

In a national, longitudinal analysis of primary care POs from 2014 through 2019, health system affiliation was not associated with consistent or clinically meaningful improvements in quality of care or service utilization. Modest increases in selected preventive and coordination measures were offset by potential reductions in primary care and specialist visit rates, indicating that affiliation may yield mixed effects on the organization and delivery of primary care. Subgroup findings further reveal that benefits are modest for small POs and even more mixed for rural and lower-quality POs. These results contribute new evidence on the limits of vertical integration as a mechanism for quality improvement. For policymakers and health system leaders, they suggest that improving outcomes after affiliation likely requires deliberate investments in practice infrastructure, staff, and data systems. Future research should explore the organizational strategies, incentive designs, and contextual factors that enable health system affiliation to translate into measurable gains in patient care.

Technical Appendix

Appendix Table 1.

Derivation of Physician Organization and Medicare Beneficiary Sample

	Physician Organizations		Medicare Beneficiaries
	#	%	#	%
1. POs with >=2 physicians, >=1 primary care physician, >=1 attributed beneficiary, and billed Medicare in all 7 years of study period	11,479	100	13,935,190	100
2. Exclude POs that are always affiliated	9,500	82.8	7,941,848	57.0
3. Exclude POs that only de-affiliate	9,275	80.8	7,467,234	53.6
4. Exclude POs that both affiliate and de-affiliate	9,106	79.3	7,077,064	50.8
5. Exclude POs that have <30 attributed beneficiaries in any year	7,410	64.6	6,296,473	45.2

Notes: PO eligibility in step 1 was based on physician and attributed beneficiary counts in the baseline year, 2013.

Appendix Table 2.

Quality Measure Specifications

Measure	Description
All-Cause Hospital Readmissions	Indicator of an unplanned readmission for any reason within 30 days of discharge from the hospital. We used the National Committee for Quality Assurance’s HEDIS measure specifications for Plan All-Cause Readmissions for measurement years 2013–2019. Beneficiaries are eligible for the measure if they are age 18 and older and are discharged alive from the hospital. The unit of analysis is the hospital discharge. A hospital discharge is identified as having a readmission if the beneficiary is readmitted to a hospital within 30 days, either for the condition associated with the index hospital stay or for a different reason and excludes planned readmissions such as for chemotherapy. The measure specifications include case mix adjustment using a beneficiary’s age, gender, primary discharge diagnosis, comorbidities, and other factors.
Breast Cancer Screening	Indicator for receipt of a mammogram during the measurement year or the preceding 15 months. We used the National Committee for Quality Assurance’s HEDIS measure specifications for Breast Cancer Screening for measurement years 2013–2019. The measure assesses whether female Medicare beneficiaries aged 50–74 receive a mammogram to screen for breast cancer during the measurement year or the preceding 15 months. Women with a diagnosis of bilateral mastectomy in their claims history are excluded from the measure.
Continuity of Care – Primary care cliniciana	Continuous measure of the proportion of primary care visits that were made by a beneficiary to his or her usual source of primary care during the measurement year. A beneficiary’s usual source of primary care was defined as the primary care clinician associated with a plurality of the beneficiary’s primary care visits. Primary care clinicians were defined as those who reported one of the following 8 specialties in the MDPPAS file: family medicine, general internal medicine, general pediatrics, general practice, nurse practitioner, physician assistant, geriatrics, and gynecology. Eligible primary care clinicians were those whose first “primary specialty” listed in MDPPAS was one of the 8 specialties listed above and who either had no second “primary specialty” or whose second “primary specialty” was also included in the list above. Beneficiaries were eligible for this measure if they had three or more primary care visits during the measurement year. The denominator for the measure is the total count of primary care visits for the beneficiary in the measurement year based on the specifications for the measure Primary Care Visits (see below). The numerator was the number of these visits that were made to the beneficiary’s usual source of care.
Continuity of Care – Physician organizationa	Continuous measure of the proportion of primary care visits that were made by a beneficiary to the physician organization that represents the beneficiary’s usual source of primary care during the measurement year. The measure is analogous to Continuity of Care – PCP but aggregates counts of primary care visits to the physician organization level rather than to the level of individual clinicians.
Diabetes Eye Exam	Indicator for receipt of an eye screening for diabetic retinal disease during the measurement year among beneficiaries with Type 1 or Type 2 diabetes. We used the National Committee for Quality Assurance’s HEDIS measure specifications for the measure Comprehensive Diabetes Care for measurement years 2013–2019. The measure assesses whether beneficiaries age 18–75 with Type 1 or Type 2 diabetes received an eye screening for diabetic retinal disease during the measurement year. The eye exam must be performed by an eye care professional (i.e., optometrist or ophthalmologist). Alternatively, an eye exam provided by an eye care professional in the prior year with a negative test result for retinopathy would also qualify as a success for the measure. Beneficiaries who had a diagnosis of gestational diabetes or steroid-induced diabetes during the measurement year are excluded from the measure.
Follow-up after Acute Events	Indicator of timely follow-up after acute events. We use the specifications for the measure Timely Follow-up after Acute Exacerbations of Chronic Conditions, developed by IMPAQ International, LLC.1 This measure assesses timely follow-up care for beneficiaries who have one of 6 chronic conditions: hypertension, asthma, heart failure, coronary artery disease, chronic obstructive pulmonary disease, or diabetes mellitus, and who experience acute events identified by diagnosis codes and utilization in one of three settings: emergency department, observation unit, or inpatient hospital. Follow-up visits are defined as “timely” if they occur within 7 days (hypertension), 14 days (asthma, heart failure, and CAD), or within 30 days (COPD and diabetes).
Hypertension Medication Adherence	Indicator for a beneficiary who adheres to his or her prescribed drug therapy for renin angiotensin system (RAS) antagonists: angiotensin converting enzyme inhibitor (ACEI), angiotensin receptor blocker (ARB), or direct renin inhibitor medications. We used specifications from the Medicare 2017 Part C & D Star Ratings2 which are endorsed by the National Quality Forum (NQF) and the Pharmacy Quality Alliance (PQA). Beneficiaries are eligible for this measure if they are age 18 years and older and filled prescriptions for at least two RAS antagonists on unique dates of service during the measurement year. Adherence is defined as a sufficient number of prescription fills covering 80% or more of the time the beneficiary is supposed to be taking the medication (i.e., proportion of days covered at 80 percent or higher).
Statin Prescription for Cardiovascular Disease	Indicator for dispensing of statin medication during the measurement year for beneficiaries with diabetes. We used the National Committee for Quality Assurance’s HEDIS measure specifications for Statin Therapy for Patients With Cardiovascular Disease for measurement years 2017–2019. We use the indicator: Received Statin Therapy, which identifies beneficiaries who were dispensed at least one high-intensity or moderate-intensity statin medication during the measurement year. Beneficiaries are eligible for the measure if they are males 21–75 years of age and females 40–75 years of age during the measurement year, and who have clinical atherosclerotic cardiovascular disease.
Primary care clinician visitsa	Count of face-to-face encounters with a clinician reporting one of the following 9 specialties in the MDPPAS file: family medicine, general internal medicine, general pediatrics, general practice, nurse practitioner, physician assistant, geriatrics, and gynecology. Eligible primary care clinicians were those whose first “primary specialty” from MD_PPAS was one of the 8 specialties listed above and who either had no second “primary specialty” or whose second “primary specialty” was also included in the list above. Visits were defined using HCPCS codes for evaluation and management services. Facility-based providers (i.e., critical access hospitals, rural health clinics, federally qualified health centers, Electing Teaching Amendment (“ETA”) hospitals) that provide evaluation and management services were identified using specific revenue center codes and facility type fields on outpatient claims.
Internal medicine specialist visitsa	Count of face-to-face encounters with a physician who reported one of the following 11 internal medicine sub-specialties as his or her “primary specialty” in the MDPPAS file: cardiology, critical care medicine, endocrinology, gastroenterology, hematology, immunology, infectious disease, nephrology, oncology, pulmonary disease, and rheumatology. Visits were defined using HCPCS codes for evaluation and management services. Facility-based providers (i.e., critical access hospitals, rural health clinics, federally qualified health centers, Electing Teaching Amendment (“ETA”) hospitals) that provide evaluation and management services were identified using specific revenue center codes and facility type fields on outpatient claims.
Ophthalmologist/optometrist visitsa	Count of visits to ophthalmologists or optometrists. Encounters were identified using evaluation and management procedure codes or codes corresponding to one of 16 eye exam procedures. Eye doctors were identified as those whose “primary specialty” in the MDPPAS file was either ophthalmologist or optometrist.
Emergency department visitsa	Count of visits to the Emergency Department for any reason during the measurement year. Visits are identified by revenue center codes and CPT codes. ED visits include both visits that do and do not lead to admission.
Ambulatory care-sensitive ED visits and hospitalizationsa	Count of ambulatory-care sensitive visits to acute care settings during the measurement year. We used the AHRQ Prevention Quality Indicators3 to identify ambulatory-care sensitive hospitalizations, and we applied these specifications to Emergency Department claims to identify ACS-ED visits following an approach used previously by others.4 The 8 indicators comprising the measure include hospitalizations or ED visits for: diabetes short-term complications (PQI 01); diabetes long-term complications (PQI 03); chronic obstructive pulmonary disease or asthma in older adults (PQI 05); hypertension (PQI 07); heart failure (PQI 08); community-acquired pneumonia (PQI 11); urinary tract infection (PQI 12); and uncontrolled diabetes (PQI 14).

^aBeneficiaries were eligible for these measures if they were age 18 and older, had 12 months of Medicare entitlement in the measurement year, had no months of Medicare Advantage enrollment, were alive at the end of the measurement year, were not entitled to Medicare coverage because of End-Stage Renal Disease (ESRD), and had no hospice use in the measurement year.

Appendix Table 3.

Measure-specific sample sizes

	Full sample				Analytic Sub-sample
Measure	Becomes Affiliated		Always Unaffiliated		Becomes Affiliated		Always Unaffiliated
	# POs	# beneficiaries	# POs	# beneficiaries	# POs	# beneficiaries	# POs	# beneficiaries
All-Cause Hospital Readmissions	480	142,011	5346	943,670	480	101,280	5346	25,000
Breast Cancer Screening	459	152,148	5167	1,070,138	459	130,551	5167	25,000
Continuity of Care – PCC	548	531,217	6657	3,779,519	548	150,000	6657	25,000
Continuity of Care - PO	548	531,217	6657	3,779,519	548	150,000	6657	25,000
Diabetes Eye Exams	343	58,447	3219	396,194	343	58,447	3219	25,000
Follow-up after Acute Events	143	18,291	1017	97,267	143	15,692	1017	25,000
Hypertension Medication Adherence	518	225,267	5857	1,548,101	518	146,401	5857	25,000
Statin Prescription for CVD	255	38,991	2274	250,435	255	38,991	2274	25,000

Notes: All POs included in the analysis met all inclusion criteria defined in Appendix Table 1 and had at least 30 beneficiaries eligible for the measure of interest in each of the 7 study years. Beneficiary counts are measured in the baseline year, 2013. PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease.

Appendix Table 4.

Parallel Trends Assessment

	Year -4 and -5		Year -3		Year -2
Measure	Estimate	P value	Estimate	P value	Estimate	P value
Quality Measures, %
All-cause Hospital Readmissions	0.0	0.886	0.1	0.636	−0.2	0.174
Breast Cancer Screening	−0.4	0.102	−0.4	0.059	0.0	0.878
Continuity of Care – PCC	2.0	0.000	1.4	0.000	0.7	0.000
Continuity of Care – PO	1.5	0.000	1.5	0.000	1.2	0.000
Diabetes Eye Exams	−0.4	0.102	−0.4	0.059	0.0	0.878
Follow-up after Acute Events	−1.1	0.039	−1.0	0.073	−0.1	0.792
Hypertension Medication Adherence	0.1	0.442	0.2	0.284	0.2	0.098
Statin Prescription for CVD	0.0	0.977	−0.3	0.387	0.2	0.596
Utilization Measures, Visits per 100 Beneficiaries
Primary care clinician visits	−5.9	0.002	0.9	0.576	1.5	0.249
Internal medicine specialist visits	−0.9	0.545	0	0.987	0.8	0.474
Ophthalmologist/optometrist visits	−0.3	0.749	−0.5	0.502	−0.4	0.538
ED visits	−0.3	0.460	−0.4	0.307	−0.2	0.453
Ambulatory care-sensitive ED visits/hospitalizations	−0.1	0.412	0	0.794	0.2	0.127

Note: All estimates are difference-in-differences estimates relative to the year prior to a PO’s affiliation with a health system (“Year -1”). Statistically significant estimates indicate a departure from parallel trends. PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease, ED=emergency department.

Appendix Table 5.

Impact of health system affiliation on quality of care

	All POs
Measure	Pre-Affiliation Beneficiary Years*	Post-Affiliation Beneficiary Years*	Impact of Affiliation for Affiliating POs
			Estimate	SE	P value
Quality Measures
All-cause Hospital Readmissions	489,932	213,530	−0.1	0.1	0.624
Breast Cancer Screening	633,459	320,344	0.2	0.2	0.350
Continuity of Care – PCC	700,174	374,861	−1.6	0.1	0.000
Continuity of Care – PO	700,174	374,861	1.5	0.1	0.000
Diabetes Eye Exams	397,226	170,363	0.9	0.2	0.000
Follow-up after Acute Events	226,494	35,126	0.6	0.4	0.124
Hypertension Med. Adherence	692,525	361,480	−0.3	0.1	0.018
Statin Prescription for CVD	337,665	114,480	0.3	0.3	0.267
Utilization Measures **
Primary care clinician visits	700,287	374,781	−9.1	1.3	0.000
Internal medicine specialist visits	700,287	374,781	−3.6	1.0	0.000
Ophthalmologist/optometrist visits	700,287	374,781	1.0	0.6	0.103
ED visits	700,287	374,781	0.6	0.3	0.071
ACS ED visits/hospitalizations	700,287	374,781	0.1	0.1	0.658

^*All values are numbers of beneficiaries across all pre-affiliation years or post-affiliation years, except for All-Cause Hospital Readmissions and Follow-up after Acute Events which are numbers of discharges.

^**All estimates are visits per 100 beneficiaries.

PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease, ACS=ambulatory care-sensitive

Appendix Table 6.

Impact of health system affiliation for low-quality and high-quality POs and low-utilization and high-utilization POs

	Low-Quality or Low-Utilization POs					High-Quality or High-Utilization POs
Measure	Pre-Affiliation Beneficiary Years*	Post-Affiliation Beneficiary Years*	Impact of Affiliation for Affiliating POs			Pre-Affiliation Beneficiary Years*	Post-Affiliation Beneficiary Years*	Impact of Affiliation for Affiliating POs
			Estimate	SE	P value			Estimate	SE	P value
Quality Measures
All-cause Hospital Readmissions	63,742	26,068	−0.1	0.4	0.704	50,339	34,111	0.3	0.3	0.287
Breast Cancer Screening	62,854	21,959	0.6	0.7	0.368	164,329	106,686	−0.1	0.3	0.838
Continuity of Care – PCC	102,153	64,771	−0.8	0.2	0.000	119,754	65,997	−2.1	0.2	0.000
Continuity of Care – PO	77,652	53,986	2.8	0.2	0.000	167,369	89,636	1.6	0.1	0.000
Diabetes Eye Exams	53,393	20,597	0.9	0.9	0.296	82,152	44,341	−0.4	0.5	0.423
Follow-up after Acute Events	36,296	5,371	−5.3	1.4	0.000	42,828	7,326	−1.0	1.1	0.391
Hypertension Med. Adherence	66,155	25,764	0.2	0.5	0.708	156,289	123,245	−0.2	0.2	0.305
Statin Prescription for CVD	50,194	9,319	0.4	1.0	0.686	63,734	21,059	1.4	0.6	0.026
Utilization Measures **
Primary care clinician visits	130,818	105,420	−3.6	2.5	0.142	93,205	56,352	−30.4	4.4	0.000
Internal medicine specialist visits	146,510	96,414	−4.7	1.5	0.002	112,893	80,641	11.6	3.5	0.001
Ophthalmologist/optometrist visits	109,229	71,240	1.1	1.3	0.408	140,662	112,520	−0.7	1.5	0.615
ED visits	124,410	80,658	0.0	1.0	0.966	98,211	73,916	0.2	0.6	0.748
ACS ED visits/hospitalizations	111,127	69,745	−0.4	0.4	0.365	97,447	73,513	−0.4	0.2	0.055

^*All values are numbers of beneficiaries across all pre-affiliation years or post-affiliation years, except for All-Cause Hospital Readmissions and Follow-up after Acute Events which are numbers of discharges.

^**All estimates are visits per 100 beneficiaries.

PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease, ACS=ambulatory care-sensitive

Appendix Table 7.

Impact of health system affiliation for small and large POs

	Small POs					Large POs
Measure	Pre-Affiliation Beneficiary Years*	Post-Affiliation Beneficiary Years*	Impact of Affiliation for Affiliating POs			Pre-Affiliation Beneficiary Years*	Post-Affiliation Beneficiary Years*	Impact of Affiliation for Affiliating POs
			Estimate	SE	P value			Estimate	SE	P value
Quality Measures
All-cause Hospital Readmissions	151,971	26,600	−0.2	0.3	0.435	349,560	186,930	0.0	0.1	0.800
Breast Cancer Screening	189,885	47,926	1.0	0.4	0.008	462,113	272,418	0.0	0.2	0.825
Continuity of Care – PCC	244,996	93,224	−2.2	0.1	0.000	480,178	281,637	−0.5	0.1	0.000
Continuity of Care – PO	244,996	93,224	1.2	0.1	0.000	480,178	281,637	1.6	0.1	0.000
Diabetes Eye Exams	135,786	23,166	1.7	0.6	0.003	269,837	147,197	0.6	0.3	0.043
Follow-up after Acute Events	101,132	5,214	−0.4	1.0	0.687	127,142	29,912	0.8	0.5	0.111
Hypertension Med. Adherence	248,929	85,271	0.0	0.2	0.938	468,596	276,209	−0.5	0.2	0.003
Statin Prescription for CVD	123,969	15,009	0.0	0.6	0.985	219,580	99,471	0.1	0.3	0.709
Utilization Measures **
Primary care clinician visits	259,343	103,457	−9.8	2.2	0.000	465,944	271,324	−7.3	1.5	0.000
Internal medicine specialist visits	259,343	103,457	0.0	1.8	0.991	465,944	271,324	−3.7	1.2	0.003
Ophthalmologist/optometrist visits	259,343	103,457	1.2	1.1	0.270	465,944	271,324	1.1	0.7	0.122
ED visits	259,343	103,457	1.1	0.5	0.034	465,944	271,324	0.5	0.4	0.219
ACS ED visits/hospitalizations	259,350	103,460	−0.2	0.2	0.398	465,937	271,321	0.1	0.2	0.619

^*All values are numbers of beneficiaries across all pre-affiliation years or post-affiliation years, except for All-Cause Hospital Readmissions and Follow-up after Acute Events which are numbers of discharges.

^**All estimates are visits per 100 beneficiaries.

PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease, ACS=ambulatory care-sensitive

Appendix Table 8.

Impact of health system affiliation for rural and urban POs

	Rural POs					Urban POs
Measure	Pre-Affiliation Beneficiary Years*	Post-Affiliation Beneficiary Years*	Impact of Affiliation for Affiliating POs			Pre-Affiliation Beneficiary Years*	Post-Affiliation Beneficiary Years*	Impact of Affiliation for Affiliating POs
			Estimate	SE	P value			Estimate	SE	P value
Quality Measures
All-cause Hospital Readmissions	103,921	40,166	0.5	0.3	0.069	386,011	173,364	−0.2	0.1	0.113
Breast Cancer Screening	138,178	60,122	0.3	0.4	0.454	495,281	260,222	0.1	0.2	0.466
Continuity of Care – PCC	155,803	87,800	−1.4	0.2	0.000	544,371	287,061	−1.7	0.1	0.000
Continuity of Care – PO	155,803	87,800	1.6	0.1	0.000	544,371	287,061	1.5	0.1	0.000
Diabetes Eye Exams	93,502	33,679	−0.4	0.5	0.406	303,724	136,684	1.3	0.3	0.000
Follow-up after Acute Events	59,597	6,199	2.3	0.9	0.011	166,897	28,927	0.4	0.5	0.428
Hypertension Med. Adherence	162,331	81,944	−0.2	0.3	0.478	530,194	279,536	−0.3	0.1	0.021
Statin Prescription for CVD	80,060	19,537	0.0	0.6	0.935	257,605	94,943	0.4	0.3	0.221
Utilization Measures *
Primary care clinician visits	161,823	88,881	−13.1	2.8	0.000	538,464	285,900	−8.3	1.4	0.000
Internal medicine specialist visits	161,823	88,881	−0.3	1.8	0.854	538,464	285,900	−4.3	1.2	0.001
Ophthalmologist/optometrist visits	161,823	88,881	1.4	1.1	0.209	538,464	285,900	0.8	0.7	0.253
ED visits	161,823	88,881	1.5	0.7	0.030	538,464	285,900	0.3	0.3	0.411
ACS ED visits/hospitalizations	161,853	88,899	0.6	0.3	0.057	538,434	285,882	−0.1	0.1	0.593

^*All values are numbers of beneficiaries across all pre-affiliation years or post-affiliation years, except for All-Cause Hospital Readmissions and Follow-up after Acute Events which are numbers of discharges.

^**All estimates are visits per 100 beneficiaries.

PCC=Primary care clinician, PO=physician organization, CVD=cardiovascular disease, ACS=ambulatory care-sensitive