Objectives: To revise and update a risk adjustment model for in-hospital mortality following percutaneous coronary intervention (PCI) procedures using all data from the 1.1 version of the American College of Cardiology National Cardiovascular Data Registry (ACC-NCR).
Background: A model based on data received at the ACC-NCDR from 1998-2000 was previously reported. The revision of this mortality model reflects all of the data submitted using 1.1 data specifications and collected through the second quarter of 2001. The model was applied to selected high-risk subgroups from a sample of data collected during the year 2001 from version 2.0 of the NCDR.
Methods: Data on 173,743 PCI procedures collected at the ACC-NCDR between January 1, 1998 and March 31, 2001 were analyzed. A mortality model was generated as well as separate models for presentation with and without acute myocardial infarction within 24 hours. The model was used to generate predicted mortalities that were compared to observed mortalities in more current high-risk patient subgroups in the NCDR.
Results: The same factors that were previously found to be associated with increased risk of PCI mortality were re-verified in the current analysis. Inclusion of the complete 1.1 dataset produced some changes in the regression weights and the constant value. Excellent discrimination was achieved in the revised model (C-Index = 0.89). The model was applied to high-risk patient groups from data collected on 76,249 during the calendar year 2001 using the 2.0 NCDR data elements and definitions. These analyses showed a high level of agreement between observed mortality of each subgroup and the predicted mortality rates generated from the revised 1.1 PCI mortality model.
Conclusions: Risk adjustment models for in-hospital mortality following PCI for all patients and for those with and without recent MI were regenerated using all data collected from the 1.1 data specifications of the ACC-NCDR and validated on high-risk groups from data collected during 2001 under data version 2.0 of the NCDR. These models reflect the most up-to-date analysis of mortality prediction from this large, multi-center national database.