Validation of the LACE readmission and mortality prediction model in a large surgical cohort: Comparison of performance at preoperative assessment and discharge time points

Study objective: The LACE index (Length of stay, admission Acuity, Charlson comorbidity index, and Emergency department visits within 6 months of current admission) is a practical tool designed to predict the risk of readmission or mortality within 30 days of hospital discharge. We sought to validate and examine its performance in a large surgical population at both the preoperative assessment and discharge time points. Design: Retrospective cohort study. Setting: We identified all admissions with a surgery or procedure at Vanderbilt University Medical Center (VUMC) between 2010 and 2015. Patients: A total of 192,670 admissions (age >= 18) were included in the study. Interventions: None. Measurements: LACE scores were calculated and analyzed with multivariable logistic regression. Discrimination was assessed with the c-statistic, calibration was assessed with calibration plots, and overall performance evaluated with the Brier score. Four models were created: admissions with any surgery or procedure, surgical admissions using actual length of stay (ALOS), surgical admissions using estimated length of stay (ELOS) and non-surgical procedural admissions. Main results: 192,670 admissions were included. The all admissions model c-statistic was 0.77 with a Brier score of 0.13. Surgical admissions with ALOS and ELOS had a c-statistic of 0.80, 0.82 and a Brier score of 0.10, 0.08 respectively. Non-surgical procedural admissions had a c-statistic of 0.76 and a Brier score of 0.14. Calibration for all models was adequate. Conclusions: The LACE model for surgical and procedural admissions had good discrimination and adequate calibration. Analysis of the model applied to surgical admissions using ELOS demonstrated slightly better overall performance than ALOS, suggesting that LACE could be utilized for readmission risk stratification at the time of preoperative assessment.