Applying an implementation science lens to understand physician-level variation in patient length of stay in internal medicine

Abstract Background & objectives Length of Stay (LoS) is a critical quality metric and focus of improvement efforts in healthcare. Successfully managing LoS depends on understanding the drivers of variation amenable to change. This study aims to (1) characterize physician-level variation in LoS; (2) identify physician actions associated with LoS; and (3) explore the individual-, team-, and hospital-level factors influencing this variation to generate hypotheses for further study. Methods This mixed-methods comparative case study approach examined six General Internal Medicine (GIM) departments in Toronto, Ontario. Physician-level variation in LoS was calculated using a random-intercept negative binomial regression model and sensitivity analysis. Semi-structured interviews and ethnographic observations were conducted and analyzed using the AACTT Framework (Action-Actor-Context-Target-Time), the Consolidated Framework for Implementation Research (CFIR), and the Theoretical Domains Frameworks (TDF). Hospitals with the lowest and highest physician-level variation in LoS were compared. Results Physician-level variation in LoS ranged from 1.7 to 7.0%, which—though modest numerically—represents meaningful differences in physician decision-making not explained by patient complexity, and no significant hospital-level effect was observed. Qualitative analysis from 12 observations and 67 interviews (32 GIM physicians and residents, 35 nurses and other health professionals) identified eight discrete physician actions influencing LoS, along with five individual-level factors and five team- and hospital-level factors. The nature of these factors was different when comparing hospitals with the lowest and highest variation. Organizational culture and perceptions of the patient population shaped physician perceptions of their professional role, while GIM departmental culture, structural characteristics, and communication networks informed physician beliefs about team capabilities and consequences of action (or inaction). Conclusion This study highlights the complex interplay between physician actions and factors influencing physician-level variation in LoS. Interventions that target physicians but do not attend to team and hospital factors are likely insufficient to achieve sustained improvements in LoS. Aligning individual-level feedback and environmental restructuring with organizational values and needs of the patient population may offer a more promising approach to sustained improvement.