OBJECTIVES

As the number of patients with multiple chronic conditions (MCC) increases, the complexity of care increases as well, with a larger number of physicians with different specialties involved in patient care. As a first step of understanding the composition of provider specialties in treating patients, we constructed a national (weighted) specialty network using Medicare claims, and applied social network analysis to detect communities of highly connected specialists.

METHODS

Our cohort consisted of over 65 years old subjects with > 1 chronic conditions, continuously enrolled for Medicare part A and B for 2012-2013, never dually eligible for Medicaid, and without ESRD or disability. Using outpatient Medicare claims, we analyzed patient-provider specialty relationships and constructed a patient-sharing specialty network. We detected specialty communities of the network based on multi-level modularity optimization and studied their composition.

RESULTS

The study cohort included 6,921,692 patients who met study criteria. Patients on average were treated by approximately three different types of specialists. Primary care providers (PCPs) saw 1,979,669 patients, while other specialists treated 6,572,347 patients. A specialty (56 in total) was in average connected to approximately 52 others, with the In average, PCPs were connected with higher strength to other specialties. We detected 3 communities of specialties of sizes 34, 14, and 8. All PCP specialties were included in the largest community. More interestingly, the other two communities primarily contained specialties related to oncology, and to pain/addiction/psychiatry, respectively.

CONCLUSIONS

A specialty network was constructed based on patient visits to providers. PCPs have a central position in this network. The specialty network was meaningfully partitioned in 3 communities. Yet, how patients’ chronic conditions influence the formation of these specialty networks and their community composition remains to be further explored.