Impacts of Care Provider Collaborations on the Service Time for Inpatient Stays: An Analysis Using Electronic Health Record Audit Logs and Dynamic Graphs

Problem definition: Collaboration among care providers is critical to healthcare delivery, yet its impact on inpatient service times—especially at the interpersonnel level—remains unclear. Prior work has focused on team familiarity or shared experience. In contrast, we examine collaboration, which is defined as real-time coactivity during an inpatient stay. This study explores the nonlinear effects of collaboration on service time, focusing on weak initial collaboration, strengthening dynamics, and variability in collaboration strength. Methodology/results: Using electronic health record audit logs from a major U.S. hospital, we develop a temporal data-mining algorithm to construct dynamic graphs that measure collaboration strength among care providers. We then estimate its impact on inpatient length of stay (LOS), controlling for team familiarity and other confounders. We find a curvilinear (inverted U) relationship between collaboration and service time; weak initial collaboration increases LOS, but stronger collaboration—once past a tipping point of approximately two to four prior interactions—leads to efficiency gains, consistent with a learning curve. Importantly, this relationship holds even after accounting for familiarity. Additionally, we find that high variability in collaboration strength among providers leads to longer LOS, with a 5.4%–20.4% increase associated with every 0.5-standard-deviation increase in collaboration variability. Managerial implications: This study contributes to healthcare operations management by emphasizing the importance of strong, consistent collaboration among care providers. It suggests that optimizing collaboration dynamics by fostering more uniform collaboration patterns can enhance service efficiency and reduce inpatient service times. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2022.0176 .