OBJECTIVES:

To evaluate the performance characteristics of the ICD-10-CM diagnosis code, U07.1, in identifying COVID-19 cases in administrative claims data. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) nucleic acid amplification test (NAAT) results in linked electronic health records (EHR) were used as the reference method. This study was conducted in the United States Food and Drug Administration (FDA) Biologics Effectiveness and Safety (BEST) network.

METHODS:

Three study populations comprising different care episodes were identified from two linked claims-EHR databases, IBM® MarketScan® Explorys® commercial claims-EHR (CED) and OneFlorida Data Trust linked Medicaid-EHR, using: 1) COVID-19-related diagnoses, 2) SARS-CoV-2 NAAT procedures, and 3) all-cause hospitalizations. The presence or absence of U07.1 code in the claims portion of the database was verified by SARS-CoV-2 NAAT results in the EHR portion of the database. The positive predictive value (PPV) and negative predictive value (NPV) of U07.1 were estimated for each study population.

RESULTS:

Study populations 1, 2 and 3 from CED consisted of 26,686, 26,095 and 2,564 episodes, respectively, and the corresponding number of episodes from OneFlorida was 29,117, 23,412, and 9,629. The PPV ranged from 79.6% to 95.1%, and the NPV ranged from 95.3% to 99.8% across populations and databases, with the highest PPV in population 3 (CED: 91.9%; OneFlorida: 93.1%). PPVs and NPVs did not vary substantially by age. PPVs in populations 1 and 2 fluctuated from April to June 2020 and stabilized in CED but remained unstable in the OneFlorida Medicaid population. NPVs were consistently >90% over time in each study population within both databases.

CONCLUSIONS:

ICD-10-CM code U07.1 performs well in identifying COVID-19 cases and non-cases among care-seeking populations in claims databases, suggesting U07.1 may be used in observational studies, including COVID-19 vaccine safety and effectiveness evaluations. Portability of these findings depends on study period, disease prevalence, and database characteristics.