Objective model using only gender, age and medication list predicts in-hospital morbidity after elective surgery

Background. Most current surgical risk models contain many variables: Some of which may be esoteric, require a physician's assessment or must be obtained intraoperatively. Early preoperative risk stratification is essential to identify high risk, elective surgical patients for medical optimization and care coordination across the perioperative period. We sought to create a simple, patient-driven scoring system using: Gender, age and list of medications to predict in-hospital postoperative morbidity. We hypothesized that certain medications would elevate risk, as indices of underlying conditions. Methods. Two Logistic regression models were created based on patient's gender, age, and medications: GAMMA (Gender, age and type of medications to predict in-hospital morbidity) and GAMMA-N (Gender, age and number of medications to predict in-hospital morbidity). A logistic regression models predicting in-hospital morbidity based on ASA score alone was also created (ASA-M). The predictive performance of these models was tested in a large surgical patient database. Results. Our GAMMA model predicts postoperative morbidity after perioperative care with high accuracy (c-statistic 0.819, Brier score 0.034). This result is similar to a model using only the ASA score (c-statistic 0.827, Brier score 0.033) and better than our GAMMA-N model (c-statistic 0.795, Brier score 0.050). Conclusions. The combination of a patient's gender, age, and medication list provided reliable prediction of postoperative morbidity. Our model has the added benefit of increased objectivity, can be conducted preoperatively, and is amenable to patient-use as it requires only limited medical knowledge.