J. Martin1, Y. Li3, V. A. Ferraris1, S. P. Saha1, E. Q. Ballert1, R. Freyberg4, J. W. Harris1, P. L. Almenoff5 1University Of Kentucky,Department Of Surgery,Lexington, KY, USA 3University Of Washington,VHA OAR, OIA,Seattle, WA, USA 4University Of Cincinnati,VHA OAR, OIA,Cincinnati, OH, USA 5University Of Missouri,VA Center Of Innovation,Columbia, MO, USA
Introduction: Serious postoperative complications (SPC) nearly always precede operative mortality. Early identification of patients likely to have SPC provides an opportunity to intervene and to limit morbidity and mortality. Furthermore, failure to rescue patients having SPC from postoperative mortality (FTR) occurs may be a quality indicator.
Methods: We evaluated surgical admissions to 110 Veterans Affairs (VA) medical centers from 10/2012-9/2013. We used VA administrative databases to identify risk factors for SPC and to measure SPC by the presence of five serious treatable complications (pneumonia, PE/DVT, sepsis, shock/cardiac arrest, GI hemorrhage/acute ulcer) developed by the AHRQ. To predict the risk of developing SPC during a hospitalization, we conducted hierarchical logistic regressions with a random intercept at the hospital level on the development sample and validated the models using the validation sample.
Results: Of 38,840 surgical admissions, 2.73% of patients developed SPC. Among these, 17.42% died, compared to 1.22% for patients who had no SPC. The risk prediction model for SPC had a c-statistics of 0.91 (95%CI=0.89-0.92). Increased risk is associated with being 65 years and older, direct admission to ICUs, having pulmonary circulation, peripheral vascular, and fluid/electrolyte disorders, and the proportion of patients who had the same DRG and developed a SPC. Ninety percent of the patients with FTR were in the highest risk quintile for SPC.
Conclusions: Our risk model predicts SPC with good accuracy. Our model provides a framework to test usefulness of therapeutic interventions to decrease morbidity and mortality and to limit FTR in high risk populations.
