Publication Date
2013
Journal
The Texas Heart Journal
PMID
23678213
Publication Date(s)
2013
Language
English
PMCID
PMC3649797
PubMedCentral® Posted Date
2013
PubMedCentral® Full Text Version
Post-Print
Published Open-Access
yes
Keywords
Cardiac surgical procedures/mortality, evaluation studies as topic, hospital mortality, logistic models, models, statistical, outcome assessment (health care)/methods, predictive value of tests, regression analysis of tests, risk assessment/classification/methods/statistics & numerical data, risk factors, ROC curve, statistics as topic, United States/epidemiology
Copyright
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Abstract
Preoperative risk-prediction models are an important tool in contemporary surgical practice. We developed a risk-scoring technique for predicting in-hospital death for cardiovascular surgery patients. From our institutional database, we obtained data on 21,120 patients admitted from 1995 through 2007. The outcome of interest was early death (in-hospital or within 30 days of surgery). To identify mortality predictors, multivariate logistic regression was performed on data from 14,030 patients from 1995 through 2002 and risk scores were computed to stratify patients (low-, medium-, and high-risk). A recalibrated model was then created from the original risk scores and validated on data from 7,090 patients from 2003 through 2007. Significant predictors of death included urgent surgery within 48 hours of admission, advanced age, renal insufficiency, repeat coronary artery bypass grafting, repeat aortic aneurysm repair, concomitant aortic aneurysm or left ventricular aneurysm repair with coronary bypass or valvular surgery, and preoperative intra-aortic balloon pump support. Because the original model overpredicted death for operations performed from 2003 through 2007, this was adjusted for by applying the recalibrated model. Applying the recalibrated model to the validation set revealed predicted mortality rates of 1.7%, 4.2%, and 13.4% and observed rates of 1.1%, 5.1%, and 13%, respectively. Because our model discriminates risk groups by using preoperative clinical criteria alone, it can be a useful bedside tool for identifying patients at greater risk of early death after cardiovascular surgery, thereby facilitating clinical decision-making. The model can be recalibrated for use in other types of patient populations.