Geographic proximity to primary care providers as a risk-assessment criterion for quality performance measures

Abstract

Importance: Previous studies have found a mixed association between Patient-Centered Medical Home (PCMH) designation and improvements in primary care quality indicators, including avoidable pediatric emergency department (ED) encounters. Whether these associations persist after accounting for the geographic locations of providers relative to where patients reside is unknown.

Objective: To examine the association between geographic proximity to primary care providers versus hospitals and risk of avoidable and potentially avoidable ED visits among children with pre-existing diagnosis of attention-deficit/hyperactivity disorder or asthma.

Methods: Retrospective cohort study of a panel of pediatric Medicaid claims data from the South Carolina from 2016-2018 for 2,959 beneficiaries having a pre-existing diagnosis of attention-deficit/hyperactivity disorder (ADD, ages 6-12) and 6,390 beneficiaries with asthma (MMA, ages 5-18), as defined using Healthcare Effectiveness Data and Information Set (HEDIS) performance measures. We calculated differences in avoidable and potentially avoidable ED visits by the beneficiary's PCMH attribution type and in relation to differences in proximity to their primary care providers versus hospitals. Outcomes were defined using the New York University Emergency Department Algorithm (NYU-EDA). Differences in ED visit risk were assessed using generalized estimation equations and compared using marginal effects models.

Results: The 2.4 percentage point reduction in risk of avoidable ED visits among children in the ADD cohort who attended a PCMH versus those who did not increased to 3.9 to 7.2 percentage points as relative proximity to primary care providers versus hospitals improved (p < 0.01). Children in the ADD and MMA cohorts that were enrolled in a medical home, but did not attend one for primary care services exhibited a 5.4 and 3.0 percentage point increase in avoidable ED visit compared to children who were unenrolled and did not attend medical homes (p < 0.05), but these differences were only observed when geographic proximity to hospitals was more convenient than primary care providers. Mixed findings were observed for potentially avoidable visits.

Conclusions: In several health care performance evaluations, patient-centered medical homes have not been found to reduce differences in hospital utilization for conditions that are treatable in primary care settings among children with chronic illnesses. Analytical approaches that also consider geographic proximity to health care services can identify performance benefits of medical homes. Expanding risk-adjustment models to also include geographic data would benefit ongoing quality improvement initiatives.

Fig 1. Bivariate choropleth map of county proportion of medical homes relative to all primary care providers and the proportion of county population enrolled in Medicaid.

Figure created by the authors using the University of South Carolina’s site license for ArcMap. Data to recreate the map is publicly available through the American Community Survey, the NPI Certification files distributed by the Centers for Medicare and Medicaid Services, and the NCQA data feed file (NCQA data available through subscription).