A multivariate model to determine prognostic factors in gastrointestinal fistulas

Background: Some studies have identified and selected factors that were associated with prognosis in patients with gastrointestinal fistulas, but a multivariate analysis to determine their relative importance and independent predictive value has not been done. The aim of this study was to determine independent prognostic factors for fistula closure and death in patients with gastrointestinal fistulas using a multivariate model. Study Design: Several variables were assessed related to spontaneous closure, surgical closure, and mortality in 188 patients with digestive fistulas (duodenal 22.3%, jejunoileal 28.7%, colonic 23.9%, biliopancreatic 25%). Selection of the variables was done through a forward stepwise logistic regression procedure; the final models were used to estimate the probability of closure, either spontaneous or surgical, and the probability of death. Results: Variables significant for spontaneous closure were: cause of the fistula (p = 0.027), fistula output (p = 0.037), institutional origin of the patient (p = 0.026), and occurrence of complications (p < 0.001). Organ of origin of the fistula was only marginally significant (p = 0.068). Successful surgical closure was significantly associated with the presence of complications (p = 0.001) and was marginally associated with age (p = 0.069). Variables significant for death were fistula output (p = 0.009) and the presence of complications (p < 0.001). Conclusions: We conclude that the likelihood of spontaneous fistula closure is higher for fistulas with surgical causes, low output, and with no complications. Mortality is higher in patients with complications and with high-output fistulas. (J Am Coll Surg 1999;188:483-490. (C) 1999 by the American College of Surgeons).