Back to 2015 Annual Meeting Posters
A Model For Predicting Prolonged Length Of Stay In Patients Undergoing Anatomic Lung Resection - A National Surgical Quality Improvement Program (NSQIP) Database Study
*Matthew R DeLuzio1, *Hari B Keshava2, *Zuoheng Wang3, *John A Federico4, *Daniel J Boffa4, *Frank C Detterbeck4, *Anthony W Kim4
1Yale-new Haven Hospital, New Haven, CT;2Cleveland Clinic Foundation, Cleveland, OH;3Yale School of Public Health, New Haven, CT;4Yale School of Medicine, New Haven, CT
Objective: To develop a risk adjusted scoring model to predict prolonged length of stay in patients undergoing anatomic lung resection.
Design: The National Surgical Quality Improvement Program database (2005-2013) was culled for patients undergoing anatomic lung resections. Univariate logistic regression analysis was used to identify pre- and perioperative variables which contributed to a prolonged length of stay > 14 days. Variables reaching a p-value <0.05 were then analyzed using stepwise multivariate logistic regression. Statistically weighted scores were determined for each significant variable, and cumulative scores for each patient were calculated. The probability of a prolonged length of stay was determined and a receiver operating characteristic curve was generated to determine the adequacy of the prediction model.
Setting: Hospitalized care
Patients: Thoracic surgery patients
Interventions: Lobectomy or pneumonectomy
Main Outcome Measure: Length of stay
Results: 9797 patients were included in the analysis. 2450 of these patients were randomly selected to be used for internal validation. Fourteen variables were found to be significant for prolonged length of stay; six were preoperative co-morbidities (Age > 65, p=0.003; COPD, p<0.001; Hyponatremia, p=0.016; ASA class 3, p=0.001; ASA class 4/5, p<0.001; open procedure, p<0.001) and eight were perioperative complications (pneumonia, p<0.001; re-intubation, p<0.001; ventilator > 48 hours, p<0.001; UTI, p<0.001; transfusion, p=0.003; DVT, p<0.001; Sepsis, p=0.032; return to the operating room, p<0.001). The C-index of the resulting ROC curve was 0.853.
Conclusions: Utilizing the NSQIP database, general variables predictive of a prolonged length of stay have been identified for patients undergoing anatomic lung resection. A scoring system has been developed based on these variables to aid the practitioner in predicting which patients will experience a prolonged length of stay postoperatively.
Back to 2015 Annual Meeting Posters
|
|
