Development and validation of a multivariable prediction model in open abdomen patients for entero-atmospheric fistula

Background Laparostomy or Open Abdomen (OA) has matured into an effective strategy in the management of abdominal catastrophe. Single prognostic factors have been identified in a previous systematic review regarding entero-atmospheric fistula (EAF). Unfortunately, no prognostic multivariable model for EAF exist. The aim was to develop and validate a multivariable prediction model from a retrospective cohort study involving three hospital's databases. Methods Fifty-seven variables were evaluated to develop a multivariable model. Univariate and multivariable logistic regression analyses were performed for on a developmental data set from two hospitals. Receiver operator characteristics analysis with area under the curve (AUC) and 95% confidence intervals (CI) were performed on the developmental data set (internal validation) as well as on an additional validation data set from another hospital (external validation). Results Five-hundred and forty-eight patients managed with an OA. Two variables remained in the multivariable prediction model for EAF. The AUC for EAF on internal validation were 0.74 (95% CI: 0.58-0.86) and 0.79 (95% CI: 0.67-0.92) on external validation. Conclusions A multivariable prediction model for EAF was externally validated and an easy-to-use probability nomogram was constructed using the two predictor variables. Level of evidence: III; prognostic.