Assistant Director, Epidemiology Regeneron Pharmaceuticals Tarrytown, United States
Background: While administrative claims are often used to estimate disease prevalence by selecting enrollees who fulfill a minimum enrollment period (MEP) in a healthcare plan, the impact of the MEP on the composition of the eligible population or prevalence estimates is rarely assessed.
Objectives: To examine the impact of varying the MEP on the characteristics of the eligible population as well as on the overall and the age-specific prevalence estimates of three diseases with varying prevalence, age of onset, and mortality rate in a US claims database.
Methods: Using 2019 Optum Clinformatics data, we estimated the prevalence of the following conditions (CDC 2019-estimated prevalence/age of onset/annual disease-specific mortality rate) asthma ( < 10%/adolescent/low), Type 2 diabetes (T2D) ( < 5% in 18-44 years (y) and >25% in 65+ y/midlife/low), and acute myeloid leukemia (AML) ( < 1%/older adults/high) using 4 different MEPs. The numerator consisted of enrollees with ≥1 ICD-CM diagnosis code for the selected condition in 2019 who met the requirements for the denominator (D). For the D, we identified enrollees in healthcare plans for following MEPs in 2019: D1) ≥1 day; D2) on January 1; D3) on July 1; and D4) the entire year. Prevalence estimates were age- and sex-standardized to the 2019 US Census population.
Results: The number of eligible people varied depending on the MEP (D1=20,418,682; D2=16,905,152; D3=16,867,722; D4=13,654,956). While the sex distribution was similar in D1-D4, the population in D1 was the youngest and was the oldest in D4 (21.1% vs 16.3% aged 18-34 y and 27.1% vs 35.6% aged 65+ y). Age distribution was similar in D2-D3. The overall prevalence of each condition was similar across D1-D4 with ranges from 4.7%-5.5%, 8.9%-9.3%, and 0.021%-0.024% for asthma, T2D, and AML respectively. The prevalence in the least restrictive MEP (D1) was 26%, 20%, and 22%, higher in patients aged 18-34 y compared to the most restrictive MEP (D4) for asthma, T2D, and AML, respectively. Among those aged 65+ y, use of D1-D4 resulted in similar prevalence estimates for asthma and T2DM, but prevalence estimates for AML derived using D3 and D4 were 10% and 24.7% lower than when estimated by using D1, respectively.
Conclusions: The choice of MEP impacted the age distribution of the eligible population. The effect of MEP selection on prevalence estimates varied depending on underlying disease characteristics and was most evident in the age-specific prevalence estimates with a notable inflection point at 65+ y. As this study was limited to examining age and sex characteristics in commercial claims data, further investigation of impact of MEP on these variables as well as race/ethnicity in other databases in warranted.
