(097) COVID-19 Related Changes in Healthcare Utilization Among Insured Persons in the US and its Impact on Pharmacoepidemiologic Research

Background: COVID-19 pandemic containment measures in the US resulted in changes in the provision and utilization of health services. Validity of pharmacoepidemiologic studies using claims data from this time may be impacted.

Objectives: To investigate the impact of changing healthcare utilization on ascertainment of study variables by replicating a known positive association using 2018-2022 data.

Methods: Angiotensin converting enzyme inhibitors (ACEi) are associated with a 3- to 5-fold increase in risk of angioedema (AE) relative to beta blockers (BB) in adults. We identified a cohort of adults newly initiating treatment with ACEi or BB between January 2018 and June 2022, at 5 sites active in the Sentinel System. We required new users to have continuous medical and drug coverage; no evidence of ACEi, BB, or other antihypertensive use; and no history of AE in the 6 months prior to cohort entry. New users were followed from treatment initiation until the earliest occurrence of AE, treatment discontinuation or crossover, disenrollment, recorded death, or 90 days since treatment initiation. Propensity score (PS) methods (1:1 matching and quintile stratification) were used to control for confounding. Hazard ratios (HR) and 95% confidence intervals (CI) for AE risk were estimated overall and stratified by quarter-year of cohort entry to investigate changes in cohort selection, confounder ascertainment, and risk estimates over time. Secondary analyses were conducted restricting the study to exclusive periods before and during the pandemic, modifying the lengths of lookback windows, and including calendar time in the PS model.

Results: We identified 2,240,239 ACEi/2,479,076 BB new users in the overall study period. Prior to adjustment by PS, ACEi new users were more likely to be male (51.5% vs. 42.4% in BB). New users of BB had a greater burden of comorbidities, allergic reaction, heart failure, and ischemic heart disease. After adjustment by PS, cohorts were balanced on known measured confounders. The overall adjusted HR was 4.44 (95% CI 4.06-4.85) by matching and 4.54 (95% CI 4.21-4.90) by stratification. In analyses stratified by quarter-year, there were fewer new users identified in April-June 2020 relative to other quarters but with similar distributions of demographics, known measured confounders, and common chronic conditions.

Conclusions: We replicated a known positive association using pandemic era data and saw no meaningful differences in cohort composition or risk estimates. Stratifying analyses by calendar time can be helpful in the identification of threats to validity of pharmacoepidemiologic research using data from the pandemic era.