Triad Census and Subgroup Analysis of Patient-Sharing Physician Collaborations

Abstract

The collaboration among physicians during episodes of care for hospitalized patients makes a significant contribution to effective health outcomes. Although physician collaborations are frequently analyzed to explore their impact on healthcare outcomes, the impact of the grouping structure of such collaborations is still unknown. The main purpose of this paper is to improve health outcomes by analyzing the attributes of patient-sharing physician collaboration networks. This paper explores the impact of different attributes of patient-sharing physician collaboration networks (PCNs) on hospitalization cost, length of stay and readmission rate. We use an electronic health insurance claim dataset to construct and explore PCNs. A PCN is categorized as either ‘low’ or ‘high’ in terms of hospitalization cost, length of stay, and readmission rate. Isomorphic classes of triad census, and clique and clan concepts of subgroup analysis are used to analyze PCNs. The results show that the clique and clan of physician collaborations affect only hospitalization cost and length of stay. Two isomorphism classes of triad census (i.e., closed triad and open triad) impact hospitalization cost, length of stay, and readmission rate. Physician collaborations in larger groups, instead of smaller groups, is related to lower hospitalization cost and shorter length of stay. The findings and insights from this paper can potentially help healthcare stakeholders to formulate better policies, which will eventually improve the quality of care while reducing cost.