Abstract 19783: A Prediction Model for Long Term Mortality after PCI: Results from the NCDR
Background: A prediction model for long term mortality of after PCI has not yet been derived from NCDR data.
Methods: Predictors of long term survival after PCI were determined from the NCDR-CathPCI Registry® for the years 2004 through 2007 with long term survival from the Centers for Medicare and Medicaid Services (CMS) 100% denominator file. Of 518,195 patients in the NCDR database undergoing PCI, a CMS probabilistic match was found in 369,668 (71%). Prior studies have shown the probability of a false match to be <1%. Restricting the population to first admission data yielded 343,466 patients for analysis, randomly divided into derivation (n=206,081) and validation (n=137,385) cohorts.
Results: The derivation set included 25,653 with acute STEMI and 180,428 without STEMI at time of admission. Baseline characteristics from the derivation group: age 75 ± 7, 42% women, 10% diabetes, 81% hypertension, 27% prior MI and 44% multivessel disease. Survival at 1 and 2 years with STEMI was 84% and 80%, and without STEMI was 93% and 88%. Preliminary major independent predictors of all-cause mortality by Cox model are shown in the Table.
Conclusions: Many baseline characteristics, both cardiac and non-cardiac, affect all-cause mortality after PCI. Following validation and calibration (results to be presented), the model can be used to predict survival in individual patients for medical decision making, as well as for benchmarking outcomes and evaluating health care policy.
- © 2010 by American Heart Association, Inc.