Using a data-driven approach to define post-COVID conditions in US electronic health record data.

Objective: To create a data-driven definition of post-COVID conditions (PCC) by directly measure changes in symptomatology before and after a first COVID episode.
Materials and Methods: Retrospective cohort study using Optum® de-identified Electronic Health Record (EHR) dataset from the United States of persons of any age April 2020-September 2021. For each person with COVID (ICD-10-CM U07.1 "COVID-19" or positive test result), we selected up to 3 comparators. The final COVID symptom score was computed as the sum of new diagnoses weighted by each diagnosis' ratio of incidence in COVID group relative to comparator group. For the subset of COVID cases diagnosed in September 2021, we compared the incidence of PCC using our data-driven definition with ICD-10-CM code U09.9 "Post-COVID Conditions", first available in the US October 2021.
Results: The final cohort contained 588,611 people with COVID, with mean age of 48 years and 38% male. Our definition identified 20% of persons developed PCC in follow-up. PCC incidence increased with age: (7.8% of persons aged 0-17, 17.3% aged 18-64, and 33.3% aged 65+) and did not change over time (20.0% among persons diagnosed with COVID in 2020 versus 20.3% in 2021). For cases diagnosed in September 2021, our definition identified 19.0% with PCC in follow-up as compared to 2.9% with U09.9 code in follow-up.
Conclusion: Symptom and U09.9 code-based definitions alone captured different populations. Maximal capture may consider a combined approach, particularly before the availability and routine utilization of specific ICD-10 codes and with the lack consensus-based definitions on the syndrome.
Competing Interests: Dr Andersen, Mr Khan, Dr Park, Dr Wiemken, Dr Emir, Ms Malhotra, and Dr McGrath are employees of Pfizer Inc and may hold stock or stock options. Dr Alhanai and Dr Ghassemi are co-founders of Ghamut Corporation which has received consulting fees from Pfizer Inc. This does not alter our adherence to PLOS ONE policies on sharing data and materials
(Copyright: © 2024 Andersen et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)