Social network analysis of hospitals admitting ventilatordependent patients in Taiwan

Objectives: The integrated delivery system (IDS) for ventilator-dependent patients (VDPs) has been in implementation for more than 18 years. This study clarified structural changes and differences in characteristics among hospital networks admitting VDPs. Methods: Social network analysis (SNA) was used to analyze the health insurance data of VDPs during 2000-2013 from all hospitals a part of the national health insurance administration, Central Division. Patient transfer and interaction between hospitals were analyzed; thereafter, degree centrality index was used to combine graph theory and geographical information for cluster structure analysis. Finally, analysis of variance (ANOVA) was used to test the differences in various characteristics between hospital networks. Results: According to the number of VDPs, three phases were classified: growing phase, stable phase, and decline phase. The top three hospital clusters comprised the networks in which medical centers operated as core hospitals, showing a clear concentration and geographical boundary over time. Medical expenditures of each cluster network, the average number of patient transfers between hospitals, and the average number of transfers per patient showed a consistent pattern of changes across the networks. However, intensive care unit (ICU) return rate and average number of ICU returns per patient did not decrease significantly during the three phases in each cluster. The average of degree centrality of networks A01 and A03 showed a significant increase. The findings imply that hospitals within and between the cluster networks exhibit competitive and interdependent relationships. In response to the declining number of VDPs, which reduced the frequency of patient transfers, the increased ICU return rate seemed to be a common strategy among hospitals. Conclusions: Employing SNA can broaden the understanding of structures and characteristic changes of hospital networks, which not only helps interpret hospital network relationships and business operating patterns but also can be used as a reference for future IDS policy development. The use of SNA in the IDS study also obtains a reference example for the implementation of other health policies and research into hospital behavior.