A clinical prediction rule for perioperative mortality and major morbidity after laparoscopic giant paraesophageal hernia repair.

ABSTRACT

OBJECTIVE: In the current era, giant paraesophageal hernia repair by experienced minimally invasive surgeons has excellent perioperative outcomes when performed electively. However, nonelective repair is associated with significantly greater morbidity and mortality, even when performed laparoscopically. We hypothesized that clinical prediction tools using pretreatment variables could be developed that would predict patient-specific risk of postoperative morbidity and mortality. METHODS: We assessed 980 patients who underwent giant paraesophageal hernia repair (1997-2010; 80% elective and 97% laparoscopic). We assessed the association between clinical predictor covariates, including demographics, comorbidity, and urgency of operation, and risk for in-hospital or 30-day mortality and major morbidity. By using forward stepwise logistic regression, clinical prediction models for mortality and major morbidity were developed. RESULTS: Urgency of operation was a significant predictor of mortality (elective 1.1% [9/778] vs nonelective 8% [16/199]; P < .001) and major morbidity (elective 18% [143/781] vs nonelective 41% [81/199]; P < .001). The most common adverse outcomes were pulmonary complications (n = 199; 20%). A 4-covariate prediction model consisting of age 80 years or more, urgency of operation, and 2 Charlson comorbidity index variables (congestive heart failure and pulmonary disease) provided discriminatory accuracy for postoperative mortality of 88%. A 5-covariate model (sex, age by decade, urgency of operation, congestive heart failure, and pulmonary disease) for major postoperative morbidity was 68% predictive. CONCLUSIONS: Predictive models using pretreatment patient characteristics can accurately predict mortality and major morbidity after giant paraesophageal hernia repair. After prospective validation, these models could provide patient-specific risk prediction, tailored for individual patient characteristics, and contribute to decision-making regarding surgical intervention.