摘要：Objectives. We applied social network analyses to determine how hospitals within Orange County, California, are interconnected by patient sharing, a system which may have numerous public health implications. Methods. Our analyses considered 2 general patient-sharing networks: uninterrupted patient sharing (UPS; i.e., direct interhospital transfers) and total patient sharing (TPS; i.e., all interhospital patient sharing, including patients with intervening nonhospital stays). We considered these networks at 3 thresholds of patient sharing: at least 1, at least 10, and at least 100 patients shared. Results. Geographically proximate hospitals were somewhat more likely to share patients, but many hospitals shared patients with distant hospitals. Number of patient admissions and percentage of cancer patients were associated with greater connectivity across the system. The TPS network revealed numerous connections not seen in the UPS network, meaning that direct transfers only accounted for a fraction of total patient sharing. Conclusions. Our analysis demonstrated that Orange County's 32 hospitals were highly and heterogeneously interconnected by patient sharing. Different hospital populations had different levels of influence over the patient-sharing network. In many ways, hospitals in a county are analogous to individual people within a social network. Both people and hospitals have individual characteristics (e.g., height, weight, age, and gender for people; bed capacity, facility type, and payer mix for hospitals). People exhibit behaviors; hospitals implement policies and interventions. Just as people are connected by social ties and interactions, hospitals are often connected to each other through sharing patients, because patients discharged from one hospital may be admitted to other hospitals in the same region. Similar to social interactions among people, patient sharing can have significant implications for disease and patient-demographic epidemiology (e.g., hospital-acquired infections) and the impact of disease control measures, patient education and prevention programs, and fiscal policies. 1 – 3 Social network constructs have been applied to individuals in populations to better understand the spread of behaviors, ideas, 4 – 10 and diseases 4 , 11 – 13 and their control measures. 4 , 12 , 7 However, to our knowledge our study is the first to apply social network frameworks and measures to hospitals to understand how hospitals within a county are interconnected by patient sharing. We collected data from all acute care health facilities in a large United States county and constructed a social network model representing their patient-sharing connections. We used social network analyses to determine and characterize: The amount of patient sharing occurring and the heterogeneity of this sharing among different hospitals, Whether certain hospitals had a greater influence over the patient-sharing network and what characteristics may predict that influence, How patient sharing correlated with geographic distance (i.e., whether hospitals that were closer together were more likely to share patients), and The proportion of patient sharing that occurred via direct patient transfers, as opposed to patients being discharged and having an intervening stay in the community before being readmitted to another hospital.
