Abstract
BACKGROUND: The Charlson Comorbidity Index (CCI) is a frequently used mortality predictor based on a scoring system for the number and type of patient comorbidities health researchers have used since the late 1980s. The initial purpose of the CCI was to classify comorbid conditions, which could alter the risk of patient mortality within a 1-year time frame. However, the CCI may not accurately reflect risk among American Indians because they are a small proportion of the US population and possibly lack representation in the original patient cohort. A motivating factor in calibrating a CCI for American Indians is that this population, as a whole, experiences a greater burden of comorbidities, including diabetes mellitus, obesity, cancer, cardiovascular disease, and other chronic health conditions, than the rest of the US population. METHODS: This study attempted to modify the CCI to be specific to the American Indian population utilizing the data from the still ongoing The Strong Heart Study (SHS) - a multi-center population-based longitudinal study of cardiovascular disease among American Indians. A 1-year survival analysis with mortality as the outcome was performed using the SHS morbidity and mortality surveillance data and assessing the impact of comorbidities in terms of hazard ratios with the training cohort. A Kaplan-Meier plot for a subset of the testing cohort was used to compare groups with selected mCCI-AI scores. RESULTS: A total of 3038 Phase VI participants from the SHS comprised the study population for whom mortality and morbidity surveillance data were available through December 2019. The weights generated by the SHS participants for myocardial infarction, congestive heart failure, and high blood pressure were greater than Charlson's original weights. In addition, the weights for liver illness were equivalent to Charlson's severe form of the disease. Lung cancer had the greatest overall weight derived from a hazard ratio of 8.31. CONCLUSIONS: The mCCI-AI was a statistically significant predictor of 1-year mortality, classifying patients into different risk strata χ(2) (8, N = 1,245) = 30.56 (p = 0.0002). The mCCI-AI was able to discriminate between participants who died and those who survived 73% of the time.