Effects of observed variations in primary care practice on Medicare patients' hospital utilization

Hospitalization is a sentinel event for older adults. Emerging evidence indicates that health system changes emphasizing patient-centered care, such as the patient-centered medical home (PCMH), can reduce hospitalizations; however, our previous system-wide analysis failed to detect a significant longitudinal change in hospitalizations. Empirical observations indicate that providers vary in how they apply patient-centered practices. Therefore, we assess the impact of Care by Design (CBD), the University of Utah Community Clinics' (UUCC) PCMH, on hospitalizations at the level of primary care practice cluster (grouping providers who practice similarly). First, ANOVA explored the relationship between a provider's clinic/team and their CBD implementation (30 elements categorized into one of three components: Appropriate Access (AA) , Care Teams (CT), and Planned Care (PC); 0=not implemented and 4=fully implemented; data collected in 2011). Statistically significant variation exists in AA, CT, and PC across UUCC clinics (p-value<0.000, <0.000, =0.038 respectively) and provider teams (p-value=0.024, <0.000, =0.009 respectively). Second, primary care practice clusters were formed using k-means cluster analysis on providers' CBD implementation. ANOVA assessed cluster quality. 6 distinct provider clusters were formed (all dimensions' p-values<0.000). Third, ANOVA is performed to assess the impact of each cluster on hospitalizations (2011 CMS Medicare data), controlling for patient demographics. We hypothesize that Medicare patients will experience different levels of hospitalization based on their provider's primary care cluster (analysis in process). This poster presents hospital utilization and analysis of variance for primary care clusters, within the UUCC system, ranging in PCMH implementation.