Long COVID is likely much more prevalent than indicated in electronic health record (EHR) diagnostic or referral codes, London School of Hygiene and Tropical Medicine researchers report in eClinicalMedicine.
The investigators analyzed National Health Service clinical data from more than 19 million adults in England on COVID-19 test results, hospitalizations, and vaccinations from November 2020 to January 2023. The aim was to detail trends in the documentation of 16 clinical codes tied to long COVID. Median follow-up was 2.2 years.
Codes didn't reflect all pandemic trends
A total of 55,465 patients were flagged for long COVID, with 20,025 diagnostic codes and 35,440 referral codes. The incidence of new long COVID rose steadily in the records during 2021, peaking in January 2022 and then declining.
"The pattern of long COVID recording over time did peak at the same time as SARS-COV-2 infections at a national level, but did not reflect the decline in infections in early 2021 or the waves of infections in 2022," the study authors wrote.
In comparison, 2.1 million people self-reported having long COVID in the proactively sampled Office for National Statistics community infection survey in January 2023
"If we assume a crude 10% of SARS-CoV-2 infections result in long COVID, as elsewhere, and with approximately 20 million recorded infections in England the number of recorded long COVID cases in primary care is an order of magnitude below the estimated incidence of long COVID in England given the number of SARS-CoV-2 infections," the researchers wrote, meaning the true prevalence is 10 times higher.
The rate of long COVID per 100,000 person-years was 177.5 in women and 100.5 in men. Most patients with a long-COVID code (59%) didn't have documentation of a positive SARS-CoV-2 test in the previous 12 weeks or more, and 6.5% were hospitalized.
"There were systematic differences between those with and without a positive test amongst all participants with a long COVID record: those with a previous positive test result were more likely to be female, older, from a more deprived IMD [index of multiple deprivation] quintile, vaccinated, and to have not been hospitalised with COVID-19," the team wrote.
Validation of outcome measures
Crude rates of long-COVID codes were highest for women, those aged 40 to 60 years, White patients, those with at least one underlying medical condition, and people who continued to take infection-prevention precautions because they were at high-risk for severe COVID-19. Crude rates of long-COVID records were lowest in recipients of at least three vaccine doses and lower in those who received an mRNA vaccine as their first dose.
Our findings agree with previous work, that there are serious limitations with simply using EHR records as a measure of long COVID, and alternative approaches may be preferable.
But the raw rate of long-COVID codes was higher in those who received one or two vaccine doses. Also, crude rates of long COVID in the EHR depended on whether referral codes were included in the condition's definition.
"Our findings agree with previous work, that there are serious limitations with simply using EHR records as a measure of long COVID, and alternative approaches may be preferable," the investigators concluded. "However, our analysis highlights that these other methods may be limited as well, especially if they depend on a recorded positive SARS-CoV-2 test result, since we found systematic differences between those with long COVID recorded, with and without a positive test result."
The authors said that validation of outcome measures is needed to better capture long-COVID cases. "National survey data suggests that many people in the UK suffer with long COVID, but relatively few cases are recorded in primary care," they concluded. "We have shown that using EHR diagnostic or referral codes unfortunately has major limitations in identifying and ascertaining true cases and timing that severely limit its utility in shedding light on causal pathways to prevent or treat Long COVID."
