812

Abstract

Introduction: The Charlson Comorbidity Index (CCI) Deyo method is widely employed in acute care.This scoring system uses abstraction from administrative databases, with 17 comorbidities assigned a score (1-6 based on risk of dying associated with the condition) and summed for a mortality risk score. Since CCI items were empirically derived, this study sought to provide psychometric information about the CCI Deyo through exploratory factor analysis (EFA). Hypothesis: EFA of CCI items will expose factor structure of items and construct validity. Methods: CCI items and scores were electronically abstracted from 634 neurotrauma step-down unit patient records. Univariate analyses of individual items were explored. Inter-item correlations, multivariate outliers with Mahalanobis distance, and KMO and Bartlett’s tests were verified for EFA application. Principal Component Analysis (PCA) using promax rotation, assuming associations among the factors, produced form factors and subscales.Cronbach’s alpha for factors reliability was computed. Pearson correlation of each subscale score was compared to overall CCI scores testing construct validity. Results: The sample mean age was 57 yrs,and 49% had a CCI score of 0 (24%=CCI 1, 15%=CCI 2, 6%=CCI 3, 6%? 4). Diabetes (DM) was the most prevalent comorbidity (21%) followed by COPD (18%), them MI, PVD and CHF (10% each) and cerebral vascular disease (6%), and the other 11 comorbidities prevalence? 4%. A KMO value of 0.546 and Bartlett’s p<0.001 denoted adequate sample size for EFA, and no multicollinearity (no inter-item correlation > 0.68). PCA wih eigenvalues > 1 extracted 7 factors which explained 59% of the variance. Pearson correlation of Factor 2 loaded items (CHF, MI, COPD, DM and PVD) with overall CCI score had the highest r value of 0.707, indicating strong association. However, all 7 subscale scores obtained p< 0.01 compared to overall CCI score. Conclusions: EFA of the CCI Deyo revealed co-occurrence of certain disease items, indicating common underlying pathological mechanisms, as well as propensity of multiple comorbidity groupings. However, as a scoring mechanism the composite CCI Deyo may be only moderately helpful as construct validity was poor.