Prediction models for same-day discharge following benign minimally invasive hysterectomy

Abstract

To develop prediction models for same-day discharge (SDD) following minimally invasive hysterectomy (MIH) using both clinical and nonclinical attributes and to compare model concordance of individual attribute groups. We performed a retrospective study of patients who underwent elective MIH for benign gynecologic indications at 69 hospitals in a statewide quality improvement collaborative between 2012 and 2019. Potential predictors of SDD were determined a priori and placed into attribute groups (Figure 1). To account for clustering of SDD practices among surgeons and within hospitals, hierarchical multivariable logistic regression models were fitted using predictors from each attribute group individually and all attribute groups in a composite model. Receiver operating characteristic (ROC) curves were generated for each model. To compare the concordance of each attribute group within the composite model, the area under the ROC curve (AUC) of the composite model was compared to that of a model from which a single attribute group was removed. The Hanley-McNeil test was used for comparisons, 95% confidence intervals (CI) for the AUCs were calculated, and a p-value of <0.05 was considered significant. Of the 25,770 patients in our study, 5,411 (21.0%) underwent same-day discharge. ROC curves are presented in Figure 2. The composite model had an AUC of 0.777 (95% CI 0.770-0.784). Among models using factors from individual attribute groups, the model using intraoperative attributes had the highest concordance for SDD (AUC 0.715, 95% CI 0.707-0.722). Removal of intraoperative attributes from the composite model was associated with the largest decrease in the composite model AUC (Table 1). Models using surgeon and hospital attributes were second and third most concordant, respectively (AUC 0.678, 95% CI 0.670-0.685; AUC 0.659, 95% CI 0.650-0.667). Models using surgical timing and patient clinical, socioeconomic, and geographic attributes groups were poor (all AUCs <0.6). Even so, factors from each attribute group contributed incrementally to the concordance of the composite model, with the exception of patient geographic attributes. Clinical and nonclinical attributes contributed to a composite prediction model with good discrimination in predicting SDD following MIH. Factors related to intraoperative, hospital, and surgeon attributes produced models with the strongest concordance. Attention to these attributes may aid efforts to improve utilization of SDD following MIH.View Large Image Figure ViewerDownload Hi-res image Download (PPT)View Large Image Figure ViewerDownload Hi-res image Download (PPT)