(A-148) Differences in the prevalence of diagnosis code groups for comorbidities using Medicare Advantage encounter data and Fee-For-Service claims

Background: Medicare enrollees can enroll in either fee-for-service (FFS) plans, administered federally, or Medicare Advantage (MA) plans, administered through private insurers. FFS claims are a well-established resource for pharmacoepidemiologic research in the US. MA encounter data became available for research in 2019; though data quality concerns remain.

Objectives: We evaluated whether the prevalences of diagnosis code-based comorbidities were consistent across FFS claims and MA encounter data.

Methods: We used data from Round 7 (2017) of the National Health and Aging Trends Study (NHATS), a nationally representative cohort of Medicare enrollees (65+ years), linked to Medicare enrollment, claims, and encounter data (2016-2017). We included NHATS participants with continuous enrollment in FFS or MA during the 365 days prior to the Round 7 interview. Comorbidities were identified in the FFS and MA files using ICD-10-CM diagnosis code groupings from the Gagne combined comorbidity index, which includes Elixhauser and Charlson conditions. To account for differences between FFS and MA populations, we standardized NHATS covariates to the MA population: demographics (age, gender, race, ethnicity), socioeconomic status (rurality, education, wealth), and clinical variables (frailty, dementia, self-reported physician-diagnosed comorbidities, smoking history, sensory impairment, falls, hospitalizations). Missing NHATS variables were multiply imputed. We compared the standardized prevalence of the diagnosis code-based comorbidities between MA and FFS enrollees.

Results: We identified 5158 enrollees (median age 74 years; 56% female), 40% in MA and 60% in FFS. MA enrollees were more likely to be Black, Hispanic, have less than a high school degree, and reside in a metropolitan area than FFS enrollees. After standardization, the prevalences of diagnosis code-based comorbidities were similar between groups. The largest standardized prevalence differences (PDs) were observed for peripheral vascular disease (PVD; MA: 26.0%, FFS: 18.7%, PD=7.3%, 95% CI 3.9 to 10.7%), renal failure (MA: 20.0%, FFS: 16.3%, PD=3.7%, 95% CI 0.9 to 6.5%), fluid and electrolyte disorders (MA: 11.0%, FFS: 14.2%, PD=-3.2%, 95% CI -5.4 to -0.9%), and deficiency anemias (MA: 16.9%, FFS: 21.6%, PD=-4.7%, 95% CI -7.3 to -2.1%). All other PDs were less than three percentage points.

Conclusions: Comorbidity prevalence was largely consistent between MA encounter and FFS claims. Remaining differences may reflect true differences in disease prevalence (e.g., research suggests MA enrollees have higher prevalence of PVD), and differences in coding practices by plan type. Our findings support the use of MA encounter data for pharmacoepidemiologic research in older adults with comorbidities.