Improvement of physician cooperation in an integrated health care network: A social network analysis.
Nicolas Larrain, Sophie Wang, Oliver GröneBackground: Gesundes Kinzigtal is an integrated care system that started in 2005 in the south-west of Germany. Guided by the ‘triple aim’, one of the principles of the initiative is to improve physician cooperation. The premise is that providers who cooperate more promote more efficient use of health services while improving health outcomes.
It has been documented that cooperation among physicians happens while actively sharing clinical information among themselves, through informal discussions that occur in the process of providing care to shared patients. The existence of shared patients can be used to identify information sharing relations among physicians and, hence, can be used to construct physician professional networks of cooperation.
Objectives: To describe the changes in cooperation of the physician’s network in Gesundes Kinzigtal and to study the effect on cooperation of being part of the integrated care initiative.
Methods: We constructed the professional network of physicians based on their shared patients using claims data for the years 2004 to 2017 from the largest insurance company in the region. Using social network analysis methods to obtain indicators related to cooperation, and entropy balancing for the creation of a control group, we studied the effect on the evolution of cooperation among integrated physicians with a panel time-series difference-in-difference model. Using exponential random graph models, we studied changes in the structural properties of both the whole network of physicians and the integrated care only network of physicians with a time-series linear regression.
Results: Indicator changes suggest increasing cooperation in Gesundes Kinzigtal doctors over time. Being part of the integrated network did not affect significantly the number of cooperation ties per year or information bottleneck potential (betweenness). However, the centrality of the integrated physicians increased, measured by page-rank, with a mean increase(p-value) of 72% (0.001) per year. Likewise, the strength of the cooperation ties increased significantly at 1.2%(<0.001) in the mean per year. The edge density and transitivity of both the whole and integrated network of physicians showed small, insignificant changes in time. The mean distance between nodes compared to the randomly generated graphs showed signs of a tighter network by decreasing in 11.5%(<0.001) in the mean per year for the whole physician network, while no significant changes in the integrated physicians-only network were found. Sensitivity analysis over the threshold number of patients to identify a cooperation tie showed similar results.
Discussion: Preliminary results suggest that the advocacy for cooperation instilled in Gesundes Kinzigtal had a positive impact. However, this can only be supported by proof of greater influence (centrality) in the network and stronger ties by the integrated physicians. Spillover effects over the number of ties and network characteristics might influence the insignificance of the effects found over these indicators. Further analysis is proposed to provide more conclusive evidence over Gesundes Kinzigtal’s impact on physician cooperation.