COVID-19 pandemic impact on clinical condition capture in real-world data: an assessment framework

The COVID-19 pandemic has impacted healthcare utilization and, consequently, real-world data. In this study, we used analytical and data visualization approaches to untangle effects on condition measurement and true shifts in the patient population seeking healthcare. We used Merative MarketScan 2018-2020 commercial claims data to develop 24 monthly cohorts of patients aged ≥18 years with 12 months baseline enrollment and an encounter for diabetes, cancer, hypertension, depression, myocardial infarction, atrial fibrillation, or urinary tract infections as the index condition in a given month. We compared monthly prevalence of each condition in 2020 versus 2019. We then imposed 3-, 6-, and 12-month look-back periods (LBP) to capture comorbidities grouped by Clinical Classifications Software Refined (CCSR) or summarized in the Charleson Comorbidity Index (CCI) and conducted similar 2020 versus 2019 prevalence comparisons. Changes in condition prevalence varied across conditions with strongest declines for cancer in April 2020 (-57.4%) and strongest increases for depression in December 2020 (+11.8%). The mean CCI was higher for most conditions during the spring of 2020, and this difference was accentuated by applying a longer LBP. Similar trends were found regarding the number of CCSR categories. Pandemic-related changes in condition capture were complex, involving both increases and decreases in encounters for specific conditions and in comorbidities, along with variations in comorbidity capture dependent on LBP. We provided a practical approach to untangle these phenomena along with open-source algorithms and visualization tools to assess these changes and inform study design and analysis.