OBJECTIVES

Multiple chronic conditions (MCC), defined as having two or more chronic conditions, is one of the most pressing concerns in care quality and delivery in Medicare. We constructed a national provider network using social network analysis methods to identify closely connected provider communities for patients with MCC and examine their characteristics.

METHODS

Using Medicare claims data (2012-2013), we defined a cohort of subjects ≥ 65 who were continuously enrolled in Medicare Part A and B with at least one claim, never dually eligible for Medicaid, without ESRD or disability, and had at least 2 chronic conditions. A patient-sharing provider network was then constructed with unique provider identifiers. We detected non-overlapping provider communities within the network using Clauset-Newman-Moore algorithm and examined each community’s characteristics in association with patient characteristics.

RESULTS

The eligibility criteria reduced the cohort to 6,916,918 patients treated by 330,243 providers (2012-2013). On average, patients had 4.0 distinct chronic conditions. The provider network had an average degree of 105.6 (standard deviation = 131.3). We identified 130 communities, of which 28 had 3 providers and 10 patients. The largest community was composed of 26,459 providers and 424,252 patients. Across the 28 communities, demographic characteristics varied (female between 52.6% and 60.3%, white between 40.5% and 97.4%) as well as the number of chronic conditions per patient (between 3.8 and 4.2), with hypertension being the most prevalent in all 28 communities. The proportion of degree sum within-community to the total degree sum for the community ranged between .68 and .91. Emergency Medicine physicians as a group had the highest average degree centrality in 11 out of 28 communities among other specialties.

CONCLUSIONS

Medicare patients with MCC are treated in a provider network composed of communities with diverse characteristics. Further research is required to understand the impact of network and community characteristics on health-related outcomes.