This article examines the association between a large-scale primary care redesign—the Comprehensive Primary Care Plus Initiative—and ambulatory care patterns of Medicare beneficiaries with highly fragmented care.
Objectives: To determine associations between a large-scale primary care redesign—the Comprehensive Primary Care Plus (CPC+) Initiative—and the extent of continuity or fragmentation of ambulatory care for Medicare fee-for-service beneficiaries during the first 3 years of CPC+.
Study Design: We used a difference-in-differences framework with a comparison group of practices that were similar to CPC+ practices at baseline (eg, practice size, demographics, Medicare spending). Regressions controlled for clustering, baseline patient characteristics, and practice fixed effects. Our study covered January 2016 through December 2019 and included 1,085,707 beneficiaries attributed to 2883 CPC+ practices and 2,274,068 beneficiaries attributed to 6912 comparison practices.
Methods: We focused on beneficiaries with highly fragmented care at baseline because they may have changed the most in response to CPC+. Key outcome measures were the numbers of ambulatory visits and unique practitioners, reported by specialty category; the percentage of visits with the usual provider of care (measuring continuity); and the reversed Bice-Boxerman Index (rBBI; measuring fragmentation).
Results: Medicare beneficiaries with high fragmentation (rBBI ≥ 0.85) at baseline (40% of the sample) had a mean of 13 ambulatory visits across 7 practitioners; the most frequent provider of care accounted for only 28% of visits. By contrast, the remaining beneficiaries had a mean of 10 visits across 4 practitioners, with the most frequent provider accounting for 54% of visits. There were no differences in continuity or fragmentation of care for CPC+ vs comparison beneficiaries.
Conclusions: We find no evidence that CPC+ increased continuity or decreased fragmentation of care.
Am J Manag Care. 2022;28(3):e103-e112. https://doi.org/10.37765/ajmc.2022.88843
Many primary care initiatives are being tested, including the Comprehensive Primary Care Plus (CPC+) Initiative. CPC+’s requirements might reduce fragmentation of care.
In January 2017, CMS launched the 5-year Comprehensive Primary Care Plus (CPC+) model in collaboration with 79 private and public payers. CPC+ is a new approach to payment for and delivery of primary care in 3070 diverse primary care practices in 18 regions, and it is intended to improve quality and reduce costs. CPC+ has 2 tracks and supports practices in both tracks with data feedback, learning activities, health information technology vendor supports, and enhanced payments (a median boost of 10% or 15% of annual practice revenue, depending on the track).1 Track 2 practices receive additional financial support, shift more from fee-for-service (FFS) toward population-based payment, and provide more enhanced care delivery approaches to better support patients with complex needs.
One of the 5 key comprehensive primary care functions targeted by CPC+ is “access and continuity.” Interpersonal continuity of care, defined as recurrent visits with the same individual practitioner over time, is a function of primary care because it is considered critical for high-quality and efficient patient care.2-5 CPC+ requires practices to have at least 95% of their patients empaneled to a practitioner or care team and to measure continuity of care for empaneled patients. Despite having a similar goal, the Comprehensive Primary Care (CPC Classic) model was not found to increase continuity of care.6,7 The CPC+ model, however, has more specific targets for continuity of care and provides more substantial payments for overall primary care transformation than CPC Classic. Further, study findings suggest that examining the inverse of continuity—fragmentation, defined as the receipt of care from multiple ambulatory practitioners with no single practitioner accounting for a substantial proportion of visits—can yield more nuanced results.8 Two patients can have the same amount of continuity (eg, 50% of visits with the most frequently seen provider) but different amounts of fragmentation (eg, the remaining 50% of visits spread across 2 providers vs 4 providers).9 To date, measures of care fragmentation have not yet been applied to evaluations of large-scale interventions. Fragmented ambulatory care has been linked to undesirable consequences, such as increased hospitalizations and emergency department use, unnecessary testing, and increased medical costs.6,7,10-14 It is unclear how an intervention designed in part to improve continuity of care will affect fragmentation.
This study examines interpersonal continuity and fragmentation of care over time for Medicare FFS beneficiaries with highly fragmented care at baseline in CPC+ and comparison practices. Because CPC+ practices had to provide continuity of care, we hypothesize that CPC+ would improve care more for beneficiaries in CPC+ practices than for those in comparison practices. We focus on the subset of beneficiaries with the most fragmented care at baseline and follow them over time because we expect this population to have changed the most in response to the intervention. We also present results for the remaining beneficiaries to illustrate the clinical distinction between high and lower fragmentation of care. This study expands the evidence base on how large-scale primary care interventions can affect ambulatory care patterns for a policy-relevant population.
We compared Medicare beneficiaries attributed to 2883 practices that started CPC+ in 2017 with those attributed to 6912 comparison practices. We attributed beneficiaries to the practice that delivered the largest share of their primary care visits over the prior 2 years. Using an intent-to-treat design, we kept beneficiaries in the analysis for the baseline or intervention period after they were attributed in that period (see eAppendix A [eAppendices available at ajmc.com]).
We selected comparison practices separately by track using propensity score matching of practice, market, and beneficiary characteristics. By design, comparison and CPC+ practices had very similar observable characteristics before CPC+, such as practice size and electronic health record use; attributed Medicare beneficiary demographics, spending, and service use; and county’s median income and number of hospital beds, as well as whether there is a shortage of primary care providers.1
Data and Study Population
To construct outcomes, we analyzed Medicare claims from the CMS Virtual Research Data Center over the baseline period (2016) and the first 3 program years (January 2017 through December 2019). We restricted the analysis to practices that joined CPC+ in 2017 and their comparison practices to ensure a 3-year follow-up period.
Following previous studies of care fragmentation, we imposed a series of sample restrictions (Figure 1).6,7 We included beneficiaries attributed to CPC+ and comparison practices without any age restriction. We required that beneficiaries be observed in the same track for baseline and at least 1 follow-up year and be continuously enrolled in Medicare parts A and B. We used a modified version of the National Committee for Quality Assurance’s definition of ambulatory visits to identify beneficiaries with office or other outpatient visits (such as to rural health clinics and critical access hospitals) for evaluation and management; ophthalmological services for medical examination; and evaluation, new enrollee, or annual wellness visits.15,16 (eAppendix A provides details.) We required beneficiaries to have at least 4 ambulatory visits in the baseline year for inclusion in the analysis because measures of continuity may not be reliable if based on fewer visits.8 We relaxed this criterion in the intervention period, in which we required beneficiaries to have at least 1 ambulatory visit, to allow for the possibility that CPC+ affected the number of visits.8
Our final sample at baseline consisted of 490,514 and 595,193 Medicare beneficiaries attributed to 1370 and 1513 Track 1 and Track 2 CPC+ practices, respectively, and 1,635,669 and 1,390,924 beneficiaries attributed to 5236 and 3780 comparison practices in Tracks 1 and 2, respectively.
With our focus on interpersonal continuity and fragmentation of care, we examined visits to different individual practitioners, rather than different practices, for our main analysis. We studied 4 key outcomes measured in each calendar year (eAppendix B provides details).17,18
Number of qualifying ambulatory visits. We summed the number of ambulatory visits over each year at the beneficiary level. In addition to the total number of visits, we also report visits by practitioner type: primary care physicians, specialists, and nurse practitioners (NPs)/physician assistants (PAs)/clinical nurse specialists (CNSs). We treated NPs/PAs/CNSs as primary care providers for attribution but studied the number of visits to these nonphysician practitioners separately from visits to primary care physicians to better understand patient care patterns.
Number of unique practitioners. This represents the number of unique practitioners the beneficiary saw over the year for the ambulatory visits identified above. We also report this measure by practitioner type.
Percentage of visits with the usual provider of care (UPC). This measures the percentage of visits with the most frequently seen ambulatory practitioner, who could be a primary care practitioner or specialist.
Reversed Bice-Boxerman Continuity of Care Index. The Bice-Boxerman Index (BBI) is based on the number of visits, the number of practitioners, and the distribution of visits across practitioners. It captures the spread of visits across practitioners and the relative shares of each practitioner.6,19 We reversed raw BBI scores, calculating 1 minus BBI, so that higher reversed BBI (rBBI) scores reflect more fragmentation, with scores ranging from 0 (least fragmentation) to 1 (most fragmentation).11 Fragmentation scores are inherently skewed toward 1 because there are more permutations of visits and practitioners that yield fragmented care than concentrated care.12 We used a cut point of 0.85 or higher to designate highly fragmented care because it maximized discrimination between the cluster of scores at the skewed end of the scale and the rest. Further, values above this threshold have previously been associated with more hospitalizations, even after adjusting for beneficiaries’ clinical and demographic characteristics.20,21
We used a difference-in-differences (DID) framework and compared the changes in mean fragmentation and continuity measures for Medicare beneficiaries in CPC+ practices between the 12 months before CPC+ (baseline) and the combined first 3 years of CPC+ with changes among beneficiaries in the comparison practices over the same period. We estimated DID models separately by track, reflecting the differences in care delivery requirements and payments. To net out prior observable differences between CPC+ and comparison beneficiaries not fully eliminated by matching, the regression models controlled for practice fixed effects and beneficiaries’ characteristics at baseline, such as demographics; the original reason for Medicare eligibility; the Hierarchical Condition Category (HCC) score, a comprehensive summary of demographic and clinical factors measuring risk for subsequent Medicare expenditures; and multiple individual chronic conditions. (eAppendix C lists the full set of control variables.) P values were 2-sided and considered statistically significant at P < .05. We did not adjust P values for multiple comparisons but did attempt to avoid false positives by examining the magnitude of the estimates, the patterns of findings across time periods, and estimates of related outcomes. We performed data analyses using Stata version 16.1 (StataCorp). Regressions accounted for clustering of the standard error at the practice level.
Characteristics of Beneficiaries With Highly Fragmented Care
More than 40% of beneficiaries in CPC+ and comparison practices had highly fragmented care (rBBI ≥ 0.85) at baseline. Beneficiaries in CPC+ and comparison practices were similar (Table 1 [part A and part B]). Compared with the remaining beneficiaries, individuals with highly fragmented care were more likely to be in practice sites that were larger, were part of a hospital or health system, and had previously participated in primary care transformation initiatives. They were less likely to be dually eligible for Medicaid or 85 years and older. There were no meaningful differences in mean HCC scores across the 2 fragmentation groups. (For the full set of beneficiary characteristics that we studied, see eAppendix Table 5 in eAppendix D.)
Fragmentation and Continuity of Care in the Baseline Period
In Track 1, at baseline, beneficiaries with high fragmentation attributed to CPC+ practices had a mean of 13 ambulatory visits with 7 unique practitioners, 28% of visits with the usual provider of care, and a fragmentation score of 0.91. By contrast, the remaining beneficiaries had a mean of 10 ambulatory visits with 4 unique practitioners, 54% of visits with the usual provider of care, and a fragmentation score of 0.68 (Table 2).
Visits to specialists drove patterns of care among those with high fragmentation. Beneficiaries with high fragmentation had a mean of 8 visits to 5 specialists and a mean of 3 visits to 2 primary care physicians. Beneficiaries with less fragmented care had more visits with primary care physicians (a mean of 4 visits with 1 practitioner) and saw fewer specialists (5 visits with 2 practitioners). The means and medians of our measures were similar (eAppendix Table 6 in eAppendix D). Results were similar for beneficiaries in comparison practices and across both tracks.
Fragmentation and Continuity of Care Over Time
Beneficiaries with highly fragmented care at baseline had increased continuity of care and reduced fragmentation by the intervention period in CPC+ and comparison practices (Figure 2). The mean percentage of visits with the UPC increased from 28% at baseline to 36% in program year (PY) 3 for Track 1 CPC+ beneficiaries, and the mean rBBI decreased from 0.91 to 0.85 over the same period. This was due to little change in the mean number of ambulatory visits (approximately 13 visits at baseline and in PY 3) but a small decline in the number of unique practitioners seen (7.3 practitioners at baseline and 6.7 in PY 3). Similar patterns can be seen for beneficiaries in Track 1 comparison practices and for beneficiaries in Track 2 CPC+ and comparison practices. Among beneficiaries not in the high-fragmentation group, we observe small declines in continuity of care and increases in fragmentation between baseline and the intervention period (eAppendix Figure 1 in eAppendix D).
The Effect of CPC+ on Patterns of Ambulatory Care
CPC+ and number of ambulatory visits. For beneficiaries with highly fragmented care at baseline, regression-adjusted DID estimates show that CPC+ did not affect the number of qualifying ambulatory visits over time in CPC+ practices relative to those in comparison practices. Track 1 and Track 2 practices had similar changes in the mean number of visits for CPC+ and comparison practices (Table 3). Although some DID estimates are statistically significant, their magnitude is small, indicating no substantive change in the number of visits.
CPC+ and the number of unique practitioners. We did not find meaningful reductions in the number of unique practitioners seen by beneficiaries with highly fragmented care in CPC+ practices relative to those in comparison practices. The regression-adjusted DID estimates are small for Track 1 (0.03; P = .05) and Track 2 (0.02; P = .12). We find similar results when we examine the number of unique practitioners by type.
CPC+ and continuity of care. We find no discernable difference between CPC+ and comparison practices in the change in the percentage of visits with the UPC measure over time. Although the regression-adjusted percentage increased over the course of CPC+—from 28% to 36% for Track 1 and 2 CPC+ practice beneficiaries—there were similar increases for beneficiaries in comparison practices (Table 3).
CPC+ and fragmentation of care. Regression-adjusted DID estimates also show that CPC+ did not affect fragmentation of care as measured by the rBBI for beneficiaries with highly fragmented care at baseline in CPC+ relative to comparison practices (Table 3).
CPC+ was not associated with changes in outcomes when we examined year-by-year DID estimates (eAppendix Table 7 in eAppendix D) instead of the combined 3-year estimates. When we derived alternative constructions of our continuity and fragmentation measures that count primary care physicians and NPs/PAs/CNSs in the attributed practice as a single provider to credit team-based care, we again found no effects of CPC+ (eAppendix Tables 8 and 9).
When we estimated DID models separately for the remaining beneficiaries with less fragmented care, we found no effect of CPC+ on continuity and fragmentation (eAppendix Table 10).
Similarly, we found no discernible effects when we analyzed subgroups of patients based on their HCC risk scores (eAppendix Tables 11 and 12) or based on whether the practice participated in an accountable care organization in the Medicare Shared Savings Program (which affects the type of performance-based payments the practice receives) (eAppendix Table 13). We found similar results when limiting the sample to beneficiaries 65 years and older (eAppendix Table 14).
Medicare beneficiaries with high fragmentation at baseline (40% of the sample) had a mean of 13 ambulatory visits across 7 practitioners (5 of them specialists), with the medians being similar to the means. On average, 28% of their visits were with their UPC—in sharp contrast to the remaining beneficiaries, for whom this was 54%. Patterns were similar across tracks and CPC+ vs comparison practices. Although we did observe improvements in continuity of care over time for beneficiaries in the high-fragmentation group—likely driven by a reversion to the mean given our categorization of beneficiaries at baseline—these improvements did not differ between CPC+ and comparison practices.
Despite previous research showing improvements in care management approaches,1 our findings suggest that CPC+ did not meaningfully change continuity or fragmentation of care for Medicare beneficiaries. At the end of the third year, beneficiaries with highly fragmented care at baseline continued to seek care from many practitioners who delivered small proportions of their ambulatory visits. We also found no evidence that CPC+ affected the number of ambulatory visits, which is particularly noteworthy for Track 2 practices, which CMS has increasingly shifted toward non–visit-based payments.
There are a few possible reasons why these findings were not more favorable. First, practices may need more time to fully implement changes in care delivery. Practices can decide which care delivery requirements to implement first, how to implement them, and which staff should be involved.1 We look only at the first 3 years of the model, but deeper practice transformation and more familiarity with the CPC+ payment structure may translate into changes in key outcomes in the last 2 years. Second, practices may require stronger, tailored learning supports or greater rewards-based payments to directly incentivize them to strengthen continuity of care as opposed to the existing performance-based payments based on measures of patients’ experience, clinical quality measures, and utilization (acute hospitalizations and emergency department visits).22 Third, we primarily calculated fragmentation at the practitioner level, which may overstate true fragmentation when there is team-based care, but we obtained the same findings in our sensitivity analyses when we combined practitioners in the same primary care practice. Related, CPC+ aimed to expand access to care at the practice level, but because NPs/PAs/CNSs often handle same-day appointments, this might worsen fragmentation. Finally, CPC+ incentivized primary care practitioners but not specialists. Although care coordination between primary and specialist providers is necessary for effective care management (and fragmentation of care is sometimes appropriate), specialists may require incentives to engage in behaviors that increase continuity with primary care and decrease fragmentation, such as referring patients back to primary care providers, avoiding specialist-to-specialist referrals, and minimizing the number of visits for problems that could be effectively managed by primary care.
This study has some important limitations. First, practices were not randomly assigned to CPC+ vs comparison group status. Therefore, differences in unobservable characteristics between CPC+ and comparison practices could influence outcomes. Second, high fragmentation does not mean the absence of communication among practitioners. Our claims data do not capture the extent to which practitioners coordinate beneficiaries’ care, exchange information, or develop a shared care plan. In addition, claims do not capture emails, texts, calls, and, before the pandemic, many video or phone visits. If such visits are concentrated among few practitioners, fragmentation measured using billable office visits may overstate gaps in communication across practitioners. Further, we cannot determine from our data the clinical appropriateness of visits with different practitioners. Finally, the generalizability of our findings to other large-scale initiatives may be limited because CPC+ was tested in the regions, payers, and practices that volunteered to participate and were selected by CMS. Given the flexibility of the CPC+ model, another set of practices might have transformed care differently, leading to different results.
We did not find evidence that CPC+ improved fragmentation or continuity of care among Medicare beneficiaries any more than in secular trends. Future interventions that target fragmentation more directly may be more successful in changing that outcome.
The authors thank Mr Tyler Rose and Ms Sandi Nelson for excellent programming. They also thank Mr Timothy J. Day (of CMS) and Dr Deborah Peikes (of Mathematica) for thoughtful input and Dr Leah Hackleman-Good, Mr Donovan Griffin, Ms Jamie John, and Ms Stephanie Barna (of Mathematica) for their work in editing and producing this article.
Author Affiliations: Mathematica, Chicago, IL (LT), Washington, DC (CU, ER), and Princeton, NJ (AG); Weill Cornell Medicine (LMK), New York, NY.
Source of Funding: Funded by HHS, CMS under contracts HHSM-500-2014-00034I/HHSM-500-T0010 and HHSM-500-2014-00034I/75FCMC19F0005. The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of HHS or any of its agencies.
Author Disclosures: Dr Kern is a consultant to Mathematica and receives grant funding from the National Heart, Lung, and Blood Institute. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (LT, CU, LMK, AG, ER); acquisition of data (AG); analysis and interpretation of data (LT, CU, LMK, AG, ER); drafting of the manuscript (LT, CU, AG, ER); critical revision of the manuscript for important intellectual content (LT, CU, LMK, AG, ER); statistical analysis (LT, CU, AG); and supervision (LT, AG).
Address Correspondence to: Lori Timmins, PhD, Mathematica, 111 E Wacker Dr, Ste 3000, Chicago, IL 60601. Email: firstname.lastname@example.org.
1. Anglin G, Peikes D, Petersen A, et al. Independent Evaluation of the Comprehensive Primary Care Plus Initiative: Second Annual Report. Center for Medicare and Medicaid Innovation; 2020.
2. O’Malley AS, Rich EC, Maccarone A, DesRoches CM, Reid RJ. Disentangling the linkage of primary care features to patient outcomes: a review of current literature, data sources, and measurement needs. J Gen Intern Med. 2015;30(suppl 3):S576-S585. doi:10.1007/s11606-015-3311-9
3. Comprehensive Primary Care Plus (CPC+) request for applications. Center for Medicare and Medicaid Innovation. January 6, 2017. Accessed December 10, 2019. https://innovation.cms.gov/Files/x/cpcplus-rfa.pdf
4. Wasson JH, Sauvigne AE, Mogielnicki RP, et al. Continuity of outpatient medical care in elderly men. a randomized trial. JAMA. 1984;252(17):2413-2417. doi:10.1001/jama.1984.03350170015011
5. Lofgren RP, Gottlieb D, Williams RA, Rich EC. Post-call transfer of resident responsibility: its effect on patient care. J Gen Intern Med. 1990;5(6):501-505. doi:10.1007/BF02600880
6. Comprehensive Primary Care Initiative. Center for Medicare and Medicaid Innovation. Accessed December 10, 2019. https://innovation.cms.gov/initiatives/comprehensive-primary-care-initiative
7. Peikes D, Dale S, Ghosh A, et al. The Comprehensive Primary Care Initiative: effects on spending, quality, patients, and physicians. Health Aff (Millwood). 2018;37(6):890-899. doi:10.1377/hlthaff.2017.1678
8. Kern LM, Seirup JK, Rajan M, Jawahar R, Stuard SS. Fragmented ambulatory care and subsequent healthcare utilization among Medicare beneficiaries. Am J Manag Care. 2018;24(9):e278-e284.
9. Kern LM, Seirup JK, Rajan M, Jawahar R, Stuard SS. Fragmented ambulatory care and subsequent emergency department visits and hospital admissions among Medicaid beneficiaries. Am J Manag Care. 2019;25(3):107-112.
10. Nyweide DJ, Anthony DL, Bynum JPW, et al. Continuity of care and the risk of preventable hospitalization in older adults. JAMA Intern Med. 2013;173(20):1879-1885. doi:10.1001/jamainternmed.2013.10059
11. Romano MJ, Segal JB, Pollack CE. The association between continuity of care and overuse of medical procedures. JAMA Intern Med. 2015;175(7):1148-1154. doi:10.1001/jamainternmed.2015.1340
12. Pham HH, Schrag D, O’Malley AS, Wu B, Bach PB. Care patterns in Medicare and their implications for pay for performance. N Engl J Med. 2007;356(11):1130-1139. doi:10.1056/NEJMsa063979
13. Frandsen BR, Joynt KE, Rebitzer JB, Jha AK. Care fragmentation, quality, and costs, among chronically ill patients. Am J Manag Care. 2015;21(5):355-362.
14. Liu CW, Einstadter D, Cebul RD. Care fragmentation and emergency department use among complex patients with diabetes. Am J Manag Care. 2010;16(6):413-420.
15. Kern LM, Seirup JK, Casalino LP, Safford MM. Healthcare fragmentation and the frequency of radiology and other diagnostic tests. J Gen Intern Med. 2017;32(2):175-181. doi:10.1007/s11606-016-3883-z
16. HEDIS Volume 2: Technical Specifications. National Committee for Quality Assurance; 2015.
17. Breslau N, Reeb KG. Continuity of care in a university-based practice. J Med Educ. 1975;50(10):965-969. doi:10.1097/00001888-197510000-00006
18. Pollack CE, Hussey PS, Rudin RS, Fox DS, Lai J, Schneider EC. Measuring care continuity: a comparison of claims-based methods. Med Care. 2016;54(5):e30-e34. doi:10.1097/MLR.0000000000000018
19. Bice TW, Boxerman SB. A quantitative measure of continuity of care. Med Care. 1977;15(4):347-349. doi:10.1097/00005650-197704000-00010
20. Kern LM, Rajan M, Ringel JB, et al. Healthcare fragmentation and incident acute coronary heart disease events: a cohort study. J Gen Intern Med. 2021;36(2):422-429. doi:10.1007/s11606-020-06305-z
21. Kern LM, Ringel JB, Rajan M, et al. Ambulatory care fragmentation and subsequent hospitalization: evidence from the REGARDS study. Med Care. 2021;59(4):334-340. doi:10.1097/MLR.0000000000001470
22. Peikes D, Fries Taylor E, O’Malley AS, Rich EC. The changing landscape of primary care: effects of the ACA and other efforts over the past decade. Health Aff (Millwood). 2020;9(3):421-428. doi:10.1377/hlthaff.2019.01430