COPD Comorbidity Profiles and 2-Year Trajectory of Acute and Postacute Care Use

Background Multiple morbidity is the norm in advanced COPD and contributes to high symptom burden and worse outcomes. Research Question Can distinct comorbidity profiles be identified and validated in a community-based sample of patients with COPD from a large integrated health care system using a standard, commonly used diagnostic code-based comorbidity index and downstream 2-year health care use data? Study Design and Methods In this retrospective cohort study, we used latent class analysis (LCA) to identify comorbidity profiles in a population-based sample of 91,453 patients with a COPD diagnosis between 2011 and 2015. We included specific comorbid conditions from the Charlson Comorbidity Index (CCI) and accounted for variation in underlying prevalence of different comorbidities across the three study sites. Sociodemographic, clinical, and health-care use data were obtained from electronic health records (EHRs). Multivariate logistic regression analysis was used to compare rates of acute and postacute care use by class. Results The mean age was 71 ± 11 years, 55% of patients were women, 23% of patients were people of color, and 80% of patients were former or current smokers. LCA identified four distinct comorbidity profiles with progressively higher CCI scores: low morbidity (61%; 1.9 ± 1.4), metabolic renal (21%; 4.7 ± 1.8), cardiovascular (12%; 4.6 ± 1.9), and multimorbidity (7%; 7.5 ± 1.7). In multivariate models, during 2 years of follow-up, a significant, nonoverlapping increase was found in the odds of having any all-cause acute (hospitalizations, observation stays, and ED visits) and postacute care use across the comorbidity profiles. Interpretation Distinct comorbidity profiles can be identified in patients with COPD using standard EHR-based diagnostic codes, and these profiles are associated with subsequent acute and postacute care use. Population-based risk stratification schemes for end-to-end, comprehensive COPD management should consider integrating comorbidity profiles such as those found in this study. Multiple morbidity is the norm in advanced COPD and contributes to high symptom burden and worse outcomes. Can distinct comorbidity profiles be identified and validated in a community-based sample of patients with COPD from a large integrated health care system using a standard, commonly used diagnostic code-based comorbidity index and downstream 2-year health care use data? In this retrospective cohort study, we used latent class analysis (LCA) to identify comorbidity profiles in a population-based sample of 91,453 patients with a COPD diagnosis between 2011 and 2015. We included specific comorbid conditions from the Charlson Comorbidity Index (CCI) and accounted for variation in underlying prevalence of different comorbidities across the three study sites. Sociodemographic, clinical, and health-care use data were obtained from electronic health records (EHRs). Multivariate logistic regression analysis was used to compare rates of acute and postacute care use by class. The mean age was 71 ± 11 years, 55% of patients were women, 23% of patients were people of color, and 80% of patients were former or current smokers. LCA identified four distinct comorbidity profiles with progressively higher CCI scores: low morbidity (61%; 1.9 ± 1.4), metabolic renal (21%; 4.7 ± 1.8), cardiovascular (12%; 4.6 ± 1.9), and multimorbidity (7%; 7.5 ± 1.7). In multivariate models, during 2 years of follow-up, a significant, nonoverlapping increase was found in the odds of having any all-cause acute (hospitalizations, observation stays, and ED visits) and postacute care use across the comorbidity profiles. Distinct comorbidity profiles can be identified in patients with COPD using standard EHR-based diagnostic codes, and these profiles are associated with subsequent acute and postacute care use. Population-based risk stratification schemes for end-to-end, comprehensive COPD management should consider integrating comorbidity profiles such as those found in this study.