Abstracts

Rationale: The "Hub and Spoke" system offers an organizational model in which advanced subspecialty care centers (hubs) may deliver cost effective care to affiliated clinical partners (spokes). Measuring the model's effectiveness is particularly challenging. Drawing from social network analysis, we developed measures of epilepsy referral networks, using the Veteran Administration Healthcare System (VA) administrative data. Methods: A national network of epileptologists, neurologists, physiatrists, psychiatrists, and primary care physicians were identified within the VA. The number of patients shared in the VA outpatient setting, from fiscal year (FY) 2008-2013, between each epileptologist and any other physician was identified. With epileptologists and providers defined as nodes of the network, a link was created in the network between providers if they shared at least two patients in the given time period. We visualized the national network of provider connections and computed node degree, coreness, strength, and betweenness of epileptologists within the VA medical system. To compute node strength, the link weights were specified to be the number of patients shared. Node betweenness is determined by the link weights and were taken to be the inverse of the number of patients shared, since affinity between nodes most typically corresponds to a small link weight. We also performed the same analysis at the hospital level, grouping all epileptologists at a given station into a single super-provider node. All analyses were performed in R, using the igraph library. Nodes represent individual physicians and are color coded by specialty. Node size is scaled proportional to the number of Veterans with epilepsy evaluated by the station. Results: The network of 130 stations examined included sixteen VA Epilepsy Centers of Excellence (ECOE) which (see Table 1 for characteristics of each center) employed 54 epilepsy specialists. Of the 6,875,758 Veterans seen by the 26,030 physicians in the network, 68,552 Veterans visited epilepsy specialists from FY2008-2013. Network measures of each ECOE are summarized in Table 1. The visualization (Figure 1) of the network provides additional information regarding the patterns and relationships of hubs and spokes. Patterns observed from the network suggest that the proportion of general neurology, physiatry, and psychiatry network connectivity varies across epilepsy centers. Interestingly, epilepsy specialists in Texas, who manage a larger volume of Veterans with epilepsy, are relatively better connected to general neurologists. Surprisingly, many consortium sites do not share any patients with ECOE sites. Conclusions: Social network analysis is a power tool to examine the organization of subspecialty care and referral networks in specialties such as epilepsy. Network analysis provides visual data of health care service coordination that may not be appreciated by traditional quantitative analytical methods.