Back to 2014 Annual Meeting Posters
Predicting Mortality, Length of Stay, and Readmission for Surgical Patients
David E Clark, *Amy E. Haskins, Brad M. Cushing
Maine Medical Center, Portland, ME
Objective: Simultaneous evaluation of factors affecting mortality, hospital length of stay (LOS), and readmission.
Design: Retrospective cohort.
Setting: NSQIP participating hospitals.
Participants: Data from patients who underwent elective general surgical operations were obtained from the 2012 NSQIP Participant Use File (PUF). For each of 30 postoperative days (POD), each patient was considered to be in one of the following compartments: Initial hospitalization, discharged home, discharged but not home, readmitted, or dead. Daily transition rates between compartments were estimated using exponential regression, conditional on patient and operative characteristics, and assuming constant transition rates for specified time periods.
Main Outcome Measures: For each transition, time-varying rates and effects of covariates; for the system as a whole, cumulative LOS, readmissions, and mortality.
Results: Older age, comorbidities, and laboratory abnormalities increased the probabilities of outcomes other than discharge home. The longer a patient had been hospitalized, the lower the rate of discharge home and the higher the rate of discharge to other locations. Readmission was more likely the longer a patient had been hospitalized, but less likely the longer a patient had been discharged. Mortality increased with longer LOS, for both hospitalized and discharged patients; mortality decreased markedly the longer a patient had been discharged home, but decreased only slightly the longer a patient had been discharged to another location. Outcomes simulated using the compartment model closely reproduced actual outcomes, and allowed separation of covariate and time effects on each transition, visualization of time trends in the entire system, and analysis of cumulative days of institutional care. Conclusions: Compartment models enable more detailed evaluation (potentially including interhospital comparisons) of surgical outcomes, including simultaneous prediction of LOS, readmission, and mortality.
Back to 2014 Annual Meeting Posters
