Associations Between Person-Centered Primary Care Measure Score, Potential Case-Mix Adjustment Variables, and Other Patient Experience Survey Items

We surveyed BCBSMA members in June–August 2021 in two mail-only waves, using a Massachusetts Health Quality Partners (MHQP) patient experience survey instrument based on the CG-CAHPS Visit Survey 4.0, with embedded PCPCM items. The survey sample included 2400 BCBSMA members who were attributed to PCPs (2000 to 20 physicians, 400 to 29 nurse practitioners) using HMO member self-designation or, in absence of self-designation (most PPO members), the PCP with most recent evaluation and management claim and who had an in-person or virtual primary care visit between September 1, 2020, and February 28, 2021. We stratified the survey sample by length of relationship with attributed PCP (< 1 year, 1–4 years, > 4 years) and by medical complexity based on claims (high, medium, low; DxCG/Cotiviti). For each returned complete survey, we calculated PCPCM scores (0 to 100), scores on two global ratings–based MHQP survey items—overall rating of most recent visit (0 to 10), likelihood to recommend provider (0 to 10)—and seven report-based MHQP composites (i.e., those assessing specific activities or events, all scored 0 to 100): Communication, Access, Integration of Care, Knowledge of Patient, Self-Management Support, Office Staff, and Adult Behavioral Health. Detailed information on these composites is available at MHQP. Using individual patient responses as units of analysis, we fit regression models that predicted each outcome score as a function of multiple potential case-mix adjustment variables (treating each as a linear predictor): self-identified age, gender, educational attainment, general health, mental health, self-identified length of relationship with PCP, whether most recent visit was in-person, claims-based medical complexity, and claims-based length of relationship with PCP. We also computed associations between PCPCM scores and the other outcome scores. We used inverse probability of response weights to account for survey non-response and stratified sample design. We performed a complete case analysis, without imputing item-level non-responses. We considered p -values < 0.05 statistically significant. This research was exempted from review by the WCG IRB. We received 381 survey responses (15.9% response rate). PCPCM scores were statistically significantly associated with self-reported general health, self-reported mental health, claims-based medical complexity, length of relationship with the PCP (both self-reported and claims-derived), and whether the most recent visit was in-person (Table 1 ). Overall rating of most recent visit was associated with self-reported general health. Likelihood to recommend provider was associated with self-reported general health, claims-based medical complexity, and whether the most recent visit was in-person. Of the report-based composites, only Knowledge of Patient was associated with more than one case-mix adjustment variable. PCPCM scores had Pearson’s correlations of 0.74 with overall visit rating, 0.72 with likelihood to recommend, 0.70 with Knowledge of Patient, and correlations < 0.6 with the remaining report-based composites (all statistically significant at p < 0.001).