Validated Model for Predicting Postoperative Respiratory Failure

Abstract

Abdominal wall reconstruction can be associated with significant rates of respiratory events. In this current study, the authors aim to characterize perioperative risk factors associated with postoperative respiratory failure and derive a model with which to predict postoperative respiratory failure. The authors reviewed the 2005 to 2010 American College of Surgeons National Surgical Quality Improvement Program databases, identifying encounters for Current Procedural Terminology codes for both hernia repair (49560, 49561, 49565, 49566, and 49568) and component separation (15734). A predictive model of postoperative respiratory failure was developed using logistic regression analyses and validated using a bootstrap technique. Of 1706 patients undergoing complex abdominal reconstructions in the study period, 102 (6.0 percent) experienced postoperative respiratory failure. Patients experiencing postoperative respiratory failure had longer admissions (21.0±18.5 versus 5.9±5.5 days, p<0.001) and a higher mortality rate (14.7 percent versus 0.1 percent, p<0.001). Multivariate logistic regression revealed eight variables significantly associated with postoperative respiratory failure. A history of chronic obstructive pulmonary disease (p<0.001), dyspnea at rest (p=0.032), dependent functional status (p=0.032), malnutrition (p<0.001), recurrent incarcerated hernia (p=0.006), concurrent intraabdominal procedure (p=0.041), American Society of Anesthesiologists score greater than 3 (p<0.001), and prolonged operative time (p<0.001) were independently associated with higher rates of postoperative respiratory failure. The multivariate model was internally validated using a bootstrap technique and had good discrimination (c statistic=0.78). A validated predictive model and clinical risk-assessment tool of postoperative respiratory failure following abdominal wall reconstruction is presented. Respiratory complications were associated with significantly longer hospital stays and higher rates of mortality. Data derived from this large cohort can be used to risk-stratify patients and to enhance perioperative decision-making. Risk, III.