Himani Gupta, Prateek Gupta, and Lee Morrow of Creighton have done us all a favor by mining a national database (the National Surgical Quality Improvement Program) to create and validate a risk calculator for perioperative pulmonary complications, which they unveil in the November CHEST.
Pulmonologists are consulted every day to weigh in on the risk of perioperative complications in patients planned for surgery who have obstructive sleep apnea, chronic obstructive pulmonary disease, asthma, or interstitial lung disease. We do our best, usually coming up with phrases of dubious value like “While this patient’s [insert disease here] places him at elevated risk for complications including postoperative respiratory failure and pneumonia, the benefits of surgery appear to outweigh the possible risks at this time.”
Hey, it’s not our fault. We haven’t had the large data sets needed to create a useful predictive model like the revised cardiac risk index for cardiac complications after surgery, which would permit rational risk stratification and decision-making algorithms like the one in Circulation.
The NSQIP data set is robust: more than 460,000 patients over two years, at 180 hospitals of every kind (academic, community, etc), and surgeries of all types. The authors used 2007 as a training set and 2008 as a validation set. They used logistical regression to create a model requiring only 5 variables (procedure, ASA class, emergency or not, functional status, and sepsis or not). Both the training set and the validation set had excellent discrimination, both with a c-statistic of 0.89 (0.5 being no better than chance).
A few examples of their calculator’s predictions:
- An independent person, ASA class 3, going for thoracic surgery (non-esophageal) has a 5% chance of postop respiratory failure.
- A partially functionally dependent person, ASA class 3, going in for aortic surgery: postop respiratory failure risk 13%.
- Someone who’s totally dependent, ASA class 4 with sepsis, going for intestinal surgery has a 55% chance of postop respiratory failure.
Two important predictors of pulmonary complications — obstructive sleep apnea and venous thromboembolism history — were not captured in the NSQIP database, unfortunately, so couldn’t be analyzed as possible predictors.
Interestingly, neither the existence of lung disease or even age make it in as an independent predictor of postop respiratory failure. That makes me wonder how precise the calculator can be — these seem like broad categories it’s capturing — but it’s certainly more informative than what we have been using (nothing).
Best of all, they’ve made the calculator available to all of us, for free. Get your own copy today at:
Gupta H et al. Development and Validation of a Risk Calculator Predicting Postoperative Respiratory Failure. CHEST 2011;140:1207-1215.