Sujita W Narayan1, Prasad S Nishtala2. 1. School of Pharmacy, University of Otago, P O Box 56, Dunedin, 9054, New Zealand. sujita.narayan@otago.ac.nz. 2. School of Pharmacy, University of Otago, P O Box 56, Dunedin, 9054, New Zealand.
Abstract
PURPOSE: An index for estimating multimorbidity based on prescription claims data is important for predicting health outcomes for older people in pharmacoepidemiological studies. We aimed to develop a Medicines Comorbidity Index (MCI) based on nationwide prescription claims data and evaluate its performance in predicting adverse outcomes in older individuals. METHODS: The index was developed on a retrospective cohort comprising of all individuals aged ≥ 65 years old, captured in the claims dataset from 1st January to 31st December 2012. The cohort was followed for 1 year to identify an event of hospitalisation or mortality. A list of medications for 20 comorbidities based on the Chronic Disease Score framework was collated. Predictive performance of the MCI was evaluated against the Charlson Comorbidity Index (CCI) using measures of discrimination (Receiver Operating Characteristic curves), sensitivity and specificity (c-statistic) and calibration (Brier scores) for regression models. RESULTS: The MCI was validated for an outcome of mortality (n = 161,461) and hospitalisation (n = 149,729). For mortality, MCI had a marginally lower c-statistic in comparison to CCI (0.70, 95% CI 0.70-0.71 vs 0.72, 95% CI 0.71-0.72 at p < 0.05) with Brier scores of 0.07 at p < 0.05. For hospitalisation, the Hazard Ratio was higher with MCI (1.08, 95% CI 1.08-1.08, p < 0.001) compared to CCI (0.92, 95% CI 0.91-0.92, p < 0.001). CONCLUSION: Initial testing indicates that the MCI is a valid and appropriate tool for measuring multimorbidity and predicting health outcomes for older individuals, and can be an important index for adjusting comorbidity in pharmacoepidemiological studies.
PURPOSE: An index for estimating multimorbidity based on prescription claims data is important for predicting health outcomes for older people in pharmacoepidemiological studies. We aimed to develop a Medicines Comorbidity Index (MCI) based on nationwide prescription claims data and evaluate its performance in predicting adverse outcomes in older individuals. METHODS: The index was developed on a retrospective cohort comprising of all individuals aged ≥ 65 years old, captured in the claims dataset from 1st January to 31st December 2012. The cohort was followed for 1 year to identify an event of hospitalisation or mortality. A list of medications for 20 comorbidities based on the Chronic Disease Score framework was collated. Predictive performance of the MCI was evaluated against the Charlson Comorbidity Index (CCI) using measures of discrimination (Receiver Operating Characteristic curves), sensitivity and specificity (c-statistic) and calibration (Brier scores) for regression models. RESULTS: The MCI was validated for an outcome of mortality (n = 161,461) and hospitalisation (n = 149,729). For mortality, MCI had a marginally lower c-statistic in comparison to CCI (0.70, 95% CI 0.70-0.71 vs 0.72, 95% CI 0.71-0.72 at p < 0.05) with Brier scores of 0.07 at p < 0.05. For hospitalisation, the Hazard Ratio was higher with MCI (1.08, 95% CI 1.08-1.08, p < 0.001) compared to CCI (0.92, 95% CI 0.91-0.92, p < 0.001). CONCLUSION: Initial testing indicates that the MCI is a valid and appropriate tool for measuring multimorbidity and predicting health outcomes for older individuals, and can be an important index for adjusting comorbidity in pharmacoepidemiological studies.
Authors: Aurélie Bannay; Christophe Chaignot; Pierre-Olivier Blotière; Mickaël Basson; Alain Weill; Philippe Ricordeau; François Alla Journal: Med Care Date: 2016-02 Impact factor: 2.983
Authors: Jeremy Walker; Nynke Halbesma; Nazir Lone; David McAllister; Christopher J Weir; Sarah H Wild Journal: J Epidemiol Community Health Date: 2015-12-17 Impact factor: 3.710