(091) Assessing COVID-19’s impact on condition measurement in claims data

Publication Number: 428

Background: The COVID-19 pandemic reduced healthcare utilization for many non-COVID-19-related conditions, especially in spring 2020. For observational studies using automated healthcare databases, this phenomenon may alter the types of patients that can be enrolled during this period or the types of comorbidities that can be captured as potential confounders.

Objectives: To quantify the impact of COVID-19 on condition measurement defined by the number of clinic visits and diagnoses captured using different look-back periods (LBPs).

Methods: We conducted a secular trend analysis using MarketScan 2019-2020 data. We identified patients aged ≥18 yrs with diagnoses of diabetes, cancer, or urinary tract infections (UTI) on outpatient encounter claims each month (index date). Patients had to have 1-yr continuous enrollment before index date. We calculated monthly prevalence for each condition and conducted pairwise comparisons for each month in 2020 vs 2019. For each patient in each month, we imposed 3, 6, and 12-month LBPs to capture comorbidities grouped by Clinical Classifications Software Refined (CCSR). We then conducted similar pairwise comparisons of the # of CCSRs per patient.

Results: Diabetes
The difference in # of visits were largest in April: 142,099 in 2020 (1.5% of all eligible beneficiaries) vs 239,756 (2.1%) in 2019. Using all 3 LBPs, patients in April had similar comorbidity burden in 2020 vs 2019. The difference in CCSR capture was the largest in June using a 3-month LBP (≤5 CCSR: 58% vs 49%, >15 CCSR: 5% vs 8%). For longer LBPs, the largest difference was: 6-m LBP, September, ≤5 CCSR: 31% vs 27%, >15 CCSR: 16% vs 19%; 12-m LBP, December, ≤5 CCSR: 11% vs 10%, >15 CCSR: 39% vs 42%).
Cancer
The # of visits in April was 92,286 (0.9%) in 2020 versus 237,565 (2.1%) in 2019. In April, patients had more comorbidities in 2020 than 2019 (e.g., using 12-month LPB, >15 CCSR 56% vs 43%). The difference in CCSR capture was the largest in June for 3-m LBP (≤5 CCSR: 51% vs 43%; >15 CCSR: 7% vs 10%); September for 6-m LBP (≤5 CCSR: 29% vs 25%; >15 CCSR: 20% vs 23%); and December for 12-m LBP (≤5 CCSR: 10% vs 10%, >15 CCSR: 45% vs 47%).
UTI
The # of visits in April was 35,840 (0.4%) in 2020 vs 54,684 (0.5%) in 2019. Using all 3 LBPs, patients in April had similar comorbidity burden. The difference in CCSR capture was largest in June for 3-m LBP (≤5 CCSR: 51% vs 43%; >15 CCSR: 7% vs 10%); August for 6-m LBP (≤5 CCSR: 36% vs 32%; >15 CCSR: 17% vs 20%), and November for 12-m LBP (≤5 CCSR: 17% vs 16%, >15 CCSR: 38% vs 40%).

Conclusions: The lockdown in spring 2020 caused a significant reduction in clinical visits as well as variation in patients’ characteristics. Under-detection of previous dx due to reduced visits during the lockdown was attenuated by long LBPs.