Date of Award
Doctor of Philosophy (PhD)
Babur De los Santos
The primary objective of this dissertation is to provide insights for healthcare practitioners to leverage the granularity of their healthcare data. In particular, leveraging the granularity of healthcare data using data analytics helps practitioners to manage operating room scheduling for productivity and nurse retention. This dissertation addresses the practical challenges of operating room (OR) scheduling by combining the existing insights from the prior literature through various tools in data analytics. In doing so, this dissertation consists of three chapters that operationally quantify the operational characteristics of the operating room and surgical team scheduling to improve operating room outcomes, including OR planning and OR nurse retention. This dissertation contributes to healthcare operations research and practice by emphasizing the importance of using granular information from hospitals’ electronic health records. While the prior research suggests that different team compositions affect OR productivity and OR time prediction, the empirical insights on how the team composition information can be utilized in practice are limited. We fill this gap by presenting data-driven approaches to use this information for OR time prediction and nurse retention. The first and third chapters deal with OR time prediction with the granular procedure, patient, and detailed team information to improve the OR scheduling. The second chapter deals with the OR nurse retention problem under OR nurses’ unique work scheduling environment. The first chapter, which is a joint work with Ahmet Colak, Lawrence Fredendall, and Robert Allen, examines drivers of OR time and their impact on OR time allocation mismatches (i.e., deviations of scheduled OR time from the realized OR time). Building on contemporary health care and empirical methodologies, the chapter identifies two mechanisms that spur scheduling mismatches: (i) OR time allocations that take place before team selections and (ii) OR time allocations that do not incorporate granular team and case data inputs. Using a two-stage estimation framework, the chapter shows how under- and over-allocation of OR times could be mitigated in a newsvendor ii setting using improved OR time predictions for the mean and variance estimates. The chapter’s empirical findings indicate that scheduling methods and the resulting scheduling mismatches have a significant impact on team performance, and deploying granular data inputs about teams—such as dyadic team experience, workload, and back-to-back case assignments—and updating OR times at the time of team selection improve OR time predictions significantly. In particular, the chapter estimates a 32% reduction in absolute mismatch times and a more than 20% reduction in OR costs. The second chapter, which is a joint work with Ahmet Colak and Lawrence Fredendall, addresses the turnover of OR nurses who work with various partners to perform various surgical procedures. Using an instrumental variable approach, the chapter identifies the causal relationship between OR nurses’ work scheduling and their turnover. To quantify the work scheduling characteristics—procedure, partner, and workload assignments, the chapter leverages the granularity of the OR nurse work scheduling data. Because unobserved personal reasons of OR nurses may lead to a potential endogeneity of schedule characteristics, the chapter instruments for the schedule characteristics using nurse peers’ average characteristics. The results suggest that there are significant connections between nurse departure probability and how procedures, partners, and workload are configured in nurses’ schedules. Nurses’ propensity to quit increases with high exposure and diversity to new procedures and partners and with high workload volatility while decreasing with the workload in their schedules. Furthermore, these effects are significantly moderated by the seniority of nurses in the hospital. The chapter also offers several explanations of what might drive these results. The chapter provides strategic reasoning for why hospitals must pay attention to designing the procedure, partner, and workload assignments in nurse scheduling to increase the retention rate in the ongoing nursing shortage and high nurse turnover in the U.S. The third chapter, which is a joint work with Ahmet Colak, Lawrence Fredendall, Babur De los Santos, and Benjamin Grant, systematically reviews the literature to gain more insights into addressing the challenges in OR scheduling to utilize granular team information for OR time prediction. Research in OR scheduling—allocating time to surgical procedures—is entering a new phase of research direction. Recent studies indicate that utilizing team information in OR scheduling can significantly improve the prediction accuracy of OR time, reducing the total cost of idle time and overtime. Despite the importance, utilizing granular team information is challenging due to the multiple decision-makers in surgical team scheduling and the presence of hierarchical structure in surgical teams. Some studies provide some insights on the relative influence of team members, which iii partly helps address these challenges, but there are still limited insights on which decision-maker has the greatest influence on OR time prediction and how hierarchy is aligned with the relative impact of surgical team members. In its findings, the chapter confirms that there are limited empirical insights in the existing literature. Based on the prior insights and the most recent development in this domain, this chapter proposes several empirical strategies that would help address these challenges and determine the key decision-makers to use granular team information of the most importance.
Kim, Jaeyoung, "Leveraging the Granularity of Healthcare Data: Essays on Operating Room Scheduling for Productivity and Nurse Retention" (2023). All Dissertations. 3387.