The LACE+ index fails to predict 30–90 day readmission for supratentorial craniotomy patients: A retrospective series of 238 surgical procedures

•Discharge predictive tool.•Predicting 30–90 day hospital readmission.•LACE + index not viable in brain tumor population. Objective: The LACE + index (Length of stay, Acuity of admission, Charlson Comorbidity Index (CCI) score, and Emergency department visits in the past 6 months) is a tool utilized to predict 30–90 day readmission and other secondary outcomes. We sought to examine the effectiveness of this predictive tool in patients undergoing brain tumor surgery. Patients and methods: Admissions and readmissions for patients undergoing craniotomy for supratentorial neoplasm at a single, multi-hospital, academic medical center, were analyzed. Key data was prospectively collected with the Neurosurgery Quality Improvement Initiative (NQII)-EpiLog tool. This included all supratentorial craniotomy cases for which the patient was alive at 90 days after surgery (n = 238). Simple logistic regression analyses were used to assess the ability of the LACE + index and subsequent single variables to accurately predict the outcome measures of 30–90 day readmission, 30–90 day emergency department (ED) visit, and 30–90 day reoperation. Analysis of the model’s or variable’s discrimination was determined by the receiver operating characteristic curve as represented by the C-statistic. Results: The sample included admissions for craniotomy for supratentorial neoplasm (n = 238) from 227 patients, of which 50.00% were female (n = 119). The average LACE + index score was 53.48 ± 16.69 (Range 9–83). The LACE + index did not accurately predict 30–90 day readmissions (P = 0.127), 30–90 day ED visits (P = 0.308), nor reoperations (P = 0.644). ROC confirmed that the LACE + index was little better than random chance at predicting these events in this population (C-statistic = 0.51−0.58). However, a single unit increase in LACE + leads to a 0.97 times reduction in the odds of being discharged home with fair predictive accuracy (P