1915. Predicting Real-Time Risk of Complications in the Postoperative Setting With Temperature as a Single Variable

BACKGROUND: No real-time postoperative risk stratification model exists to predict complications following surgery. The aim of this work is to understand if we can successfully risk stratify patients across three distinct surgeries using group-based trajectory modeling (GBTM) with only a single variable, temperature. METHODS: We performed a retrospective study of adults undergoing elective total knee arthroplasty (TKA), total hip arthroplasty (THA), colectomy, and pancreatectomy at an academic medical center from October 2014 to February 2018. Clinical data were abstracted using definitions from the National Surgical Quality Improvement Program (NSQIP) and temperature data were extracted from the Database Warehouse. GBTM was used to identify distinct clusters of patients with similar temperature trajectories. We calculated rates of complications and combined all NSQIP infectious and inflammatory complications into a single metric hence forth labeled inflammatory complications. Chi-square test was used to compare categorical variables. RESULTS: We identified 815 independent surgical patients: 307 TKA/THA, 195 pancreatectomy, and 313 colectomy patients. Rates of all NSQIP complications were 1.6% for TKA/THA, 35.4% for pancreatectomy and 10.2% for colectomy at 30 days after surgery. Pancreatectomy patients clustered into two temperature trajectories and both TKA/THA and colectomy patients (Figure 1) clustered into three groups. Inflammatory complication frequencies were significantly different in colectomy and trended toward significance for TKA/THA and pancreatectomy (Table 1). CONCLUSION: Temperature trajectory modeling may help identify postoperative patients at higher risk for surgical complication after surgery. While risk stratification seems to work better in high complication surgeries or models with more patients, the promise of this modeling technique relies on the ability to identify high-risk patients with a single variable. DISCLOSURES: All authors: No reported disclosures.