Predicting Fragmented Care: Beneficiary, Primary Care Physician, and Practice Characteristics

Abstract

Research ObjectiveIdentifying characteristics of beneficiaries, primary care physicians, and primary care practice sites that predict highly fragmented ambulatory care (that is, care spread across multiple providers without a dominant provider) is essential to develop effective interventions targeted at reducing fragmentation. High care fragmentation is associated with unnecessary procedures and testing, increased emergency department visits and hospitalizations, and increased medical costs.Study DesignThis study was conducted in the context of the Comprehensive Primary Care Plus Model (CPC+), a large primary care redesign initiative. We used Medicare claims data from January through December 2018 on Medicare fee‐for‐service (FFS) beneficiaries attributed to primary care practice sites participating in CPC+ and to comparison practices that were similar at baseline. We used hierarchical linear models to predict the likelihood of a beneficiary receiving highly fragmented care, defined as having a fragmentation score (measured by the reversed Bice‐Boxerman Index) ≥ 0.85. We used an extensive set of explanatory variables at each level (74 total variables) and group‐level random intercepts to understand how characteristics at each level help explain variation in fragmentation. We estimated separate models for the two CPC+ transformation/payment tracks.Population Studied3,541,136 Medicare FFS beneficiaries attributed to 26,344 primary care physicians in 9300 primary care practice sites.Principal FindingsThe three sets of explanatory variables (beneficiary, physician, and practice site) together only explained about 5 percent of the variation in the likelihood of high care fragmentation. Unobserved differences between primary care physicians and between primary care practice sites together accounted for only 4 percent of the variation. Instead, more than 91 percent of the variation in fragmentation consisted of unobserved residual variance. We identified several characteristics of beneficiaries (age, reason for original Medicare entitlement, and dual status), physicians (gender and measures of comprehensiveness of care), and practice sites (size, being part of a system/hospital, and census region) that had small associations with high fragmentation. Findings were similar by track.ConclusionsAlthough we identified a number of characteristics that predict high care fragmentation, most of the variation in fragmentation was not explained by observed beneficiary, primary care physician, or primary care practice characteristics. This suggests other providers and beneficiaries' preferences may be important factors.Implications for Policy or PracticeOur findings show that primary care physician and practice site characteristics explain only a small share of variation in care fragmentation. Behaviors of other health care providers not captured by regional controls, as well as unmeasured patient preferences, are likely to be important predictors of high care fragmentation.One implication of these findings is that interventions focused on primary care need to be sizable and targeted to decrease fragmentation. Further, future health care innovations might need to expand their focus beyond primary care to consider how specialists and hospitals work with primary care, ways to modify beneficiaries' self‐referral behavior, and the effect of market factors on the primary care practice environment. In addition, fully understanding the factors that drive fragmentation (and opportunities to reduce it) will require more data on specialists and their practices.Primary Funding SourceCenters for Medicare and Medicaid Services.