OBJECTIVES: Real-world data (RWD) are being employed to understand post-acute sequelae of COVID-19, or long-COVID. A recent study employed a machine learning approach to EHR data to identify four clinical phenotypes: cardiac and renal (CR), respiratory sleep and anxiety (RSA), musculoskeletal and nervous (MN), and digestive and respiratory (DR). The current study maps these clinical phenotypes to administrative claims data and aims to estimate the incidence rate of symptoms and conditions following a diagnosis of long-COVID within each.

METHODS: Adult patients with long-COVID were identified within HealthVerity claims data between 01JAN2021 and 31AUG2022. Index was defined as the date of long-COVID diagnosis via ICD-10 code U09.9 or B94.8 preceded by U07.1 within 365 days. Clinical phenotypes were identified as subgroups via a relevant diagnosis code within 90 days prior through 90 days following the index long-COVID date. Overlap of the phenotypes was described in addition to incidence rates of long-COVID symptoms following the index diagnosis date.

RESULTS: Among 188,232 patients identified with long-COVID, 84.0% were in the RSA subgroup, 83.7% in the DR subgroup, 77.6% in the MN subgroup, and 59.4% in the CR . Approximately 4% of patients were not in any subgroup and 42% were in all subgroups. Incidence rates of long-COVID symptoms were highest among patients in the CR subgroup. Rates of dyspnea and fatigue/malaise ranked among the highest at 70,753.84 (69,192.73, 72,314.94) and 42,256.94 (41,101.84, 43,412.05) per 100,000 person-years, respectively. Rates of long-COVID symptoms were similar among other subgroups. Observed trends in incidence rates were generally consistent within gender subgroups across phenotypes.

CONCLUSIONS: Differences in incident long-COVID symptoms were observed between clinical phenotypes, particularly among patients with cardiovascular or renal conditions. Aligned with findings reported from EHR data, patients with cardiovascular diagnoses concurrent with their COVID-19 diagnosis more frequently had documented symptoms around the time of long-COVID diagnosis