Skip directly to site content Skip directly to page options Skip directly to A-Z link Skip directly to A-Z link Skip directly to A-Z link
Volume 29, Number 2—February 2023

Longitudinal Analysis of Electronic Health Information to Identify Possible COVID-19 Sequelae

Eleanor S. Click1Comments to Author , Donald Malec1, Jennifer R. Chevinsky, Guoyu Tao, Michael Melgar, Jennifer E. Giovanni, Adi V. Gundlapalli, S. Deblina Datta, and Karen K. Wong
Author affiliations: Centers for Disease Control and Prevention, Atlanta, Georgia, USA (E.S. Click, J.R. Chevinsky, G. Tao, M. Melgar, J.E. Giovanni, A.V. Gundlapalli, S.D. Datta, K.K. Wong); Centers for Disease Control and Prevention, Hyattsville, Maryland, USA (D. Malec)

Cite This Article


Ongoing symptoms might follow acute COVID-19. Using electronic health information, we compared pre‒ and post‒COVID-19 diagnostic codes to identify symptoms that had higher encounter incidence in the post‒COVID-19 period as sequelae. This method can be used for hypothesis generation and ongoing monitoring of sequelae of COVID-19 and future emerging diseases.

SARS-CoV-2 causes acute COVID-19 and may cause post–COVID-19 conditions, which include a range of long-term sequelae (1,2). Review of multiple studies describes ongoing symptoms after acute COVID-19 (3). Post–COVID-19 conditions might include symptoms of nonspecific chest pain, fatigue, and malaise, as well as cardiomyopathy, renal failure, lung disease, and venous thromboembolism. Identifying possible sequelae of an emerging disease has traditionally required aggregating clinical experiences; this approach might miss sequelae that are rare or where the increase above baseline is not obvious (4).

Large electronic health information databases might aid in detecting these early signals, especially when potential sequelae events are temporally and geographically dispersed. The International Classification of Diseases, 10th revision (ICD-10), code for post‒COVID-19 conditions was not available for use in the United States until October 2021; thus, examining other diagnosis codes is needed to infer potential sequelae (5). We evaluated feasibility of a method comparing pre‒ and post‒COVID-19 diagnosis healthcare codes to identify possible sequelae from a large national database of healthcare encounters.

The Study

The Premier Healthcare Database, Special COVID-19 Release (PHD-SR) is a large, hospital-based, service-level, all-payer database with >900 contributing hospitals and healthcare systems. The database includes diagnostic codes from the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM), for inpatient and selected outpatient encounters with representation in all US census regions (68) (Appendix).

Using PHD-SR (release date February 4, 2021), we analyzed variables for encounter type (inpatient, outpatient including emergency), encounter date sequence variables (length of stay, admission and discharge month, and days between encounters), and discharge ICD-10-CM codes. We included patients with a first COVID-19 inpatient or outpatient encounter (i.e., a COVID-19 discharge ICD-10-CM code). COVID-19 index date was the first day of the first COVID-19 encounter. Pre‒COVID-19 encounters were any encounters within 365 days before the patient’s first COVID-19 encounter. Post‒COVID-19 encounters were the first COVID-19 encounter and all subsequent encounters.

We included encounters with discharge dates during January 1, 2019–December 31, 2020. We calculated relative rates (RRs) of post‒ to pre‒COVID-19 diagnoses in the post‒COVID-19 intervals of 60–89, 90–119, and 120–149 days, where day 0 is the COVID-19 index date. Rates were total number of encounters with a particular ICD-10-CM code observed in the specified time interval divided by total number of days in that interval that patients were also in the database (some patients might have died or been no longer observed in the database (i.e., right-censored). Pre-COVID-19 rates were calculated similarly for the whole interval, accounting for when they were first observed in the dataset (i.e., left-censored).

Because the day of reported diagnosis is only known to have occurred sometime between day of admission and day of discharge, we assigned a specific day (assigned randomly over the encounter duration) as the day of diagnosis for analysis. We generated 5 versions of the dataset with imputed diagnosis dates to capture this uncertainty. To compare diagnosis rates for the pre‒ and post‒COVID-19 intervals, we used 1-sided t-tests of the equality of rates performed on a log RR scale (9). We limited analyses to ICD-10-CM codes that occurred in >5 encounters during >1 post‒COVID-19 interval because of difficulty in interpreting RR for rare events (Appendix).

We evaluated whether RR was >1 in the post‒ versus pre‒COVID-19 intervals by using a t-test that includes variability due to multiple imputations (10). We report results significant at p<0.05 after performing the Benjamini-Hochberg adjustment procedure that excludes marginally significant results that could have occurred by chance because of performing a large number of significance tests. We performed analyses in R 3.6.0 (The R Foundation for Statistical Computing, We defined diagnoses with significantly increased encounter rates >60 days after COVID-19 index date relative to pre–COVID-19 as possible postacute sequelae.

We identified 385,067 patients with a COVID-19 discharge date January–December 2020 and >1 visit within the previous 365 days. Median encounters per patient was 4 (interquartile range [IQR] 3–7; pre–COVID-19, 2 [IQR 1–5]; post–COVID-19, 1 [IQR 1–2]); 87% were outpatient encounters. The cohort was 59% female. Median age was 54 (IQR 35–69) years; 5.1% were <18 years of age. Median length of stay for inpatient encounters was 4 (IQR 2–8) days.

Encounters for sequelae of specified infectious and parasitic diseases were increased at least through 149 days after the index date (RR 11.6 at 120–149 days) (Table). Encounters were increased for several months after acute illness for postviral fatigue syndrome, headache, and certain respiratory diseases, including pneumonia and acute respiratory distress syndrome. We identified general sequelae of treatment in intensive care, including polyneuropathy (RR 9.1 at 90–119 days) and myopathy (RR 5.0 at 60–89 days), nonscarring hair loss (RR 2.3–3.5 in multiple intervals beyond 60 days), and pressure ulcers (stage 3 and 4, RR 1.6–1.7 at 60–89 days).

Viral cardiomyopathy (RR 9.8 at 60–89 days) and sepsis codes were only increased in the first 90 days after index date. Rates of nonfollicular diffuse lymphomas were increased in the 60–119 day periods (RR 272.6–411), but most encounters were for 1 patient. Encounters for stage 3 chronic kidney disease (RR 2.5–6.4 beyond 60 days) and for increased liver aminotransferase levels (RR 4.8–6.5 beyond 60 days) were higher for several months after the index date; infective myocarditis (RR 12.6) was increased for 90–119 days.

The possible cardiac, respiratory, kidney, and liver sequelae identified through this method are consistent with those of previous studies (1113). For kidney injury, new diagnoses of stage 3 kidney illness (glomerular filtration rate 30–59 mL/min/1.73 m2) were higher than pre–COVID-19. Stage 3 kidney injury might occur when there is more permanent damage requiring repeated healthcare encounters. This method might generate useful hypotheses about the duration of possible sequelae because we found that encounters for increased aminotransferase levels remain increased at least through the 120–149-day interval after acute illness.

The first limitation of our study is that increased encounter rates might be caused by health-seeking behavior. Encounters for new diagnoses are not equivalent to new disease entities, and rates of encounter diagnosis codes might not represent the rates of disease. For long hospitalizations, diagnosis timing might be mischaracterized because actual diagnosis date is uncertain; however, we used imputation to account for this uncertainty. Counting the initial COVID-19 visit as part of the post–COVID-19 period might identify complications of acute illness as sequelae; however, we focused on sequelae with increased rates >60 days after the COVID-19 index date to mitigate that factor. This analysis does not capture exacerbations of underlying conditions, such as worsening heart failure or reactive airway disease, unless disease exacerbation is captured by a different diagnosis code.

The data in this analysis are more representative of adults than children. We excluded pregnancy-related conditions from the analysis. Our findings might not be generalizable to patients with asymptomatic or mild COVID-19 who might not seek healthcare, and we did not control for factors such as aging and changes in societal behavior, so we cannot attribute increased rates of new diagnoses solely to COVID-19. Advantages of our method include rapid application to large longitudinal healthcare datasets and extension over time to identify possible sequelae occurring long after acute illness.


Our findings are consistent with those of other studies using different methods to identify sequelae, including a matched cohort analysis of PHD-SR during the same period and a direct survey of persons with and without previous SARS-CoV-2 test results (14,15). This hypothesis-generating method can provide early signals of possible sequelae for novel diseases and inform additional studies to identify, characterize, and refine potential sequelae for COVID-19 or other emerging diseases.

Dr. Click is the lead for Extramural Innovation, Office of Advanced Molecular Detection, National Center for Emerging and Zoonotic Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, GA. Her primary research interests are infectious diseases and molecular epidemiology.



We thank Sean Browning for providing indispensable work in extracting, organizing, and maintaining the Premier data needed for this study.



  1. Al-Aly  Z, Xie  Y, Bowe  B. High-dimensional characterization of post-acute sequelae of COVID-19. Nature. 2021;594:25964. DOIPubMedGoogle Scholar
  2. Datta  SD, Talwar  A, Lee  JT. A proposed framework and timeline of the spectrum of disease due to SARS-CoV-2 infection: illness beyond acute infection and public health implications. JAMA. 2020;324:22512. DOIPubMedGoogle Scholar
  3. Nalbandian  A, Sehgal  K, Gupta  A, Madhavan  MV, McGroder  C, Stevens  JS, et al. Post-acute COVID-19 syndrome. Nat Med. 2021;27:60115. DOIPubMedGoogle Scholar
  4. Carfì  A, Bernabei  R, Landi  F, Gemelli Against  C-P-ACSG; Gemelli Against COVID-19 Post-Acute Care Study Group. Persistent symptoms in patients after acute COVID-19. JAMA. 2020;324:6035. DOIPubMedGoogle Scholar
  5. Centers for Disease Control and Prevention. Public health recommendations: evaluating and caring for patients with post-COVID conditions: interim guidance [cited 2022 Aug 22].
  6. PINC. AI™ healthcare data - special release: COVID-19. October 2021. PINC AI™ Applied Sciences, 2021 [cited 2022 Mar 24].
  7. Rosenthal  N, Cao  Z, Gundrum  J, Sianis  J, Safo  S. Risk factors associated with in-hospital mortality in a US national sample of patients with COVID-19. JAMA Netw Open. 2020;3:e2029058. DOIPubMedGoogle Scholar
  8. World Health Organization. ICD-10: international statistical classification of diseases and related health problems, 10th revision. 2nd edition. 2004 [cited 2022 Nov 18].
  9. Lachin  J. Biostatistical methods: the assessment of relative risks. Hoboken (NJ): John Wiley & Sons; 2009.
  10. Rubin  D, Little  RJ. Statistical analysis with missing data. New York: John Wiley & Sons; 1987.
  11. Peleg  Y, Kudose  S, D’Agati  V, Siddall  E, Ahmad  S, Nickolas  T, et al. Acute kidney injury due to collapsing glomerulopathy following COVID-19 infection. Kidney Int Rep. 2020;5:9405. DOIPubMedGoogle Scholar
  12. Puntmann  VO, Carerj  ML, Wieters  I, Fahim  M, Arendt  C, Hoffmann  J, et al. Outcomes of cardiovascular magnetic resonance imaging in patients recently recovered from coronavirus disease 2019 (COVID-19). JAMA Cardiol. 2020;5:126573. DOIPubMedGoogle Scholar
  13. Zhang  C, Shi  L, Wang  FS. Liver injury in COVID-19: management and challenges. Lancet Gastroenterol Hepatol. 2020;5:42830. DOIPubMedGoogle Scholar
  14. Chevinsky  JR, Tao  G, Lavery  AM, Kukielka  EA, Click  ES, Malec  D, et al. Late conditions diagnosed 1-4 months following an initial coronavirus disease 2019 (COVID-19) encounter: a matched-cohort study using inpatient and outpatient administrative data—United States, 1 March–30 June 2020. Clin Infect Dis. 2021;73(Suppl 1):S516. DOIPubMedGoogle Scholar
  15. Wanga  V, Chevinsky  JR, Dimitrov  LV, Gerdes  ME, Whitfield  GP, Bonacci  RA, et al. Long-term symptoms among adults tested for SARS-CoV-2—United States, January 2020‒April 2021. MMWR Morb Mortal Wkly Rep. 2021;70:123541. DOIPubMedGoogle Scholar




Cite This Article

DOI: 10.3201/eid2902.220712

Original Publication Date: December 23, 2022

1These authors contributed equally to this article.

Table of Contents – Volume 29, Number 2—February 2023

EID Search Options
presentation_01 Advanced Article Search – Search articles by author and/or keyword.
presentation_01 Articles by Country Search – Search articles by the topic country.
presentation_01 Article Type Search – Search articles by article type and issue.



Please use the form below to submit correspondence to the authors or contact them at the following address:

Eleanor S. Click, Centers for Disease Control and Prevention, 1600 Clifton Rd NE, Mailstop H16-5, Atlanta, GA 30329-4027, USA

Send To

10000 character(s) remaining.


Page created: November 18, 2022
Page updated: January 23, 2023
Page reviewed: January 23, 2023
The conclusions, findings, and opinions expressed by authors contributing to this journal do not necessarily reflect the official position of the U.S. Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.