This paper is in the following e-collection/theme issue:

Longitudinal and Cohort Studies in Public Health (254) JMIR Theme Issue: COVID-19 Special Issue (2388) Outbreak and Pandemic Preparedness and Management (1673) Theme Issue: Novel Coronavirus (COVID-19) Outbreak Rapid Reports (1481) Infectious Diseases (non-STD/STI) (1478) Public (e)Health, Digital Epidemiology and Public Health Informatics (636) Innovative Methods in Public Health and Surveillance (305) Statistical Methods for Surveillance and Population Health (146) Public Health Informatics (78) Secondary Use of Clinical Data for Research and Surveillance (317) Chronic Conditions (212)

Published on in Vol 10 (2024)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/49841, first published .
Defining the Subtypes of Long COVID and Risk Factors for Prolonged Disease: Population-Based Case-Crossover Study

Defining the Subtypes of Long COVID and Risk Factors for Prolonged Disease: Population-Based Case-Crossover Study

Defining the Subtypes of Long COVID and Risk Factors for Prolonged Disease: Population-Based Case-Crossover Study

Authors of this article:

Skyler Resendez1 Author Orcid Image ;   Steven H Brown2 Author Orcid Image ;   Hugo Sebastian Ruiz Ayala1 Author Orcid Image ;   Prahalad Rangan1 Author Orcid Image ;   Jonathan Nebeker2 Author Orcid Image ;   Diane Montella2 Author Orcid Image ;   Peter L Elkin1, 2 Author Orcid Image
Article Authors Cited by Tweetations (4) Metrics

    Original Paper

    1Department of Biomedical Informatics, University at Buffalo, State University of New York, Buffalo, NY, United States

    2Office of Health Informatics, Department of Veterans Affairs, Washington, DC, United States

    Corresponding Author:

    Peter L Elkin, MD

    Department of Biomedical Informatics

    University at Buffalo, State University of New York

    77 Goodell Street, Suite 540

    Buffalo, NY, 14203

    United States

    Phone: 1 5073581341

    Email: elkinp@buffalo.edu


    Background: There have been over 772 million confirmed cases of COVID-19 worldwide. A significant portion of these infections will lead to long COVID (post–COVID-19 condition) and its attendant morbidities and costs. Numerous life-altering complications have already been associated with the development of long COVID, including chronic fatigue, brain fog, and dangerous heart rhythms.

    Objective: We aim to derive an actionable long COVID case definition consisting of significantly increased signs, symptoms, and diagnoses to support pandemic-related clinical, public health, research, and policy initiatives.

    Methods: This research employs a case-crossover population-based study using International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) data generated at Veterans Affairs medical centers nationwide between January 1, 2020, and August 18, 2022. In total, 367,148 individuals with ICD-10-CM data both before and after a positive COVID-19 test were selected for analysis. We compared ICD-10-CM codes assigned 1 to 7 months following each patient’s positive test with those assigned up to 6 months prior. Further, 350,315 patients had novel codes assigned during this window of time. We defined signs, symptoms, and diagnoses as being associated with long COVID if they had a novel case frequency of ≥1:1000, and they significantly increased in our entire cohort after a positive test. We present odds ratios with CIs for long COVID signs, symptoms, and diagnoses, organized by ICD-10-CM functional groups and medical specialty. We used our definition to assess long COVID risk based on a patient’s demographics, Elixhauser score, vaccination status, and COVID-19 disease severity.

    Results: We developed a long COVID definition consisting of 323 ICD-10-CM diagnosis codes grouped into 143 ICD-10-CM functional groups that were significantly increased in our 367,148 patient post–COVID-19 population. We defined 17 medical-specialty long COVID subtypes such as cardiology long COVID. Patients who were COVID-19–positive developed signs, symptoms, or diagnoses included in our long COVID definition at a proportion of at least 59.7% (268,320/449,450, based on a denominator of all patients who were COVID-19–positive). The long COVID cohort was 8 years older with more comorbidities (2-year Elixhauser score 7.97 in the patients with long COVID vs 4.21 in the patients with non–long COVID). Patients who had a more severe bout of COVID-19, as judged by their minimum oxygen saturation level, were also more likely to develop long COVID.

    Conclusions: An actionable, data-driven definition of long COVID can help clinicians screen for and diagnose long COVID, allowing identified patients to be admitted into appropriate monitoring and treatment programs. This long COVID definition can also support public health, research, and policy initiatives. Patients with COVID-19 who are older or have low oxygen saturation levels during their bout of COVID-19, or those who have multiple comorbidities should be preferentially watched for the development of long COVID.

    JMIR Public Health Surveill 2024;10:e49841

    doi:10.2196/49841

    Keywords

    long COVID; PASC; postacute sequelae of COVID-19; public health; policy initiatives; pandemic; diagnosis; COVID-19 treatment; long COVID cause; health care support; public safety; COVID-19; Veterans Affairs; United States; COVID-19 testing; clinician; mobile phone 


    Numerous symptoms are cited as long-term sequelae of COVID-19. “The symptoms may affect a number of organ systems, occur in diverse patterns, and frequently get worse after physical or mental activity” [1]. Early studies found that the most common long-term symptoms were fatigue, dyspnea, joint pain, and chest pain [2]. Others reported gastrointestinal tract disorders correlated with gut microbiome shifts after COVID-19 infection [3,4]. Cognitive dysfunction, often referred to as brain fog, is another commonly reported long-term symptom [5]. Cognitive dysfunction is particularly concerning given evidence that COVID-19 can alter brain structure [6]. The most common self-reported symptoms documented via a smartphone app were fatigue, headache, dyspnea, and anosmia [7,8].

    More recent studies have added to the knowledge base concerning symptoms of long COVID (post–COVID-19 condition). For example, a study observing cohorts in 4 Chinese cities showed that fatigue, cough, sore throat, difficulty in concentrating, feeling of anxiety, myalgia, and arthralgia were common severe long COVID symptoms. While there is considerable overlap, there is still value in new studies as they help validate previous studies and add new insights, such as identifying a previously underappreciated increase in anxiety among those who had COVID-19 [9].

    Recent reviews have analyzed and integrated long COVID research to date [10]. Reviews like these contribute knowledge (eg, that over 200 symptoms have been identified, affecting multiple organ systems while proposing potential mechanisms). They also point out where our knowledge base is lacking. For example, Davis et al [10] observed a study stating that postural tachycardia syndrome can be a potential complication [11] but research has recently shown that long COVID can also greatly increase the likelihood of complications such as atrial fibrillation [12]. Such contributions are why continued research is critically important to combat the detrimental effects of long COVID.

    Concerningly high long COVID frequencies have been reported since near the start of the pandemic. A cohort study from the Netherlands found that approximately 1 in 8 patients with COVID-19 developed long-term somatic symptoms [13]. Another study showed that approximately 30% of their cohort reported persistent symptoms, with many experiencing worse health-related quality of life compared with baseline and negative impacts on at least one activity of daily living [14]. More recent studies confirm these alarming statistics, indicating that 1 in 7 adults in the United States have reported symptoms of long COVID [15]. Furthermore, long COVID’s impacts extend beyond individual morbidity to include the health care system and economic consequences. Cutler [16] noted long COVID resulted in reduced workforce participation (eg, 44% out of the workforce), direct earning losses, and worker shortages in service jobs. This is likely directly related to the fatigue associated with long COVID, which has now been linked to muscular abnormalities and overall dysfunction of mitochondria within these tissues [17].

    The widespread occurrence of lingering ailments and their impacts on individuals and society make clear the need for a long COVID definition. US public health officials note that we must balance our need for an accurate long COVID definition that includes all afflicted individuals against our need for interim long COVID definitions to expedite immediate action and mobilization [18]. In particular, a working definition of long COVID based on routinely collected coded data could support the identification of at-risk or undiagnosed patients for monitoring, referral, or therapeutic interventions. In this study, we empirically derive an actionable broad-based long COVID definition to support current clinical, public health, research, and policy initiatives related to the pandemic.


    Overview

    We selected veterans who had laboratory-confirmed positive COVID-19 tests. We examined the veterans’ electronic health records for novel International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes between 1 and 7 months after a positive COVID-19 test. We grouped codes with a novel frequency of 1/1000 or greater by diagnosis type creating ICD-10-CM functional groups and performed χ2 testing with Bonferroni correction to compare diagnosis frequencies before and after a positive COVID-19 test. We defined ICD-10-CM functional groups that significantly increased in frequency as “upregulated” (see Figure 1). We then manually aggregated upregulated ICD-10-CM functional groups into medical specialties to organize our empiric definition of long COVID.

    ‎
    Figure 1. Workflow and data curation in the sequence used to generate our long COVID definition. ICD-10-CM: International Classification of Diseases, 10th Revision, Clinical Modification; M: months; T0: date of positive COVID-19 test.

    Population Definition and Data Extraction

    We selected patients with laboratory-confirmed positive COVID-19 studies and followed them for 13 months (6 months before the COVID-19 test result and 7 months after COVID-19) to create a long COVID definition. We used the electronic health records of all the patients who tested positive for COVID-19 at Veterans Affairs (VA) medical facilities nationally between January 1, 2020, and August 18, 2022. In total, 2,377,720 patients were tested for COVID-19 during this time period.

    We applied SQL queries to VA Informatics and Computing Infrastructure, Corporate Data Warehouse data tables [19] to generate 2 diagnosis files for analysis. The first file (“before”) contains a row of retrospectively collected information for each patient and each ICD-10-CM diagnosis assigned to them in the 6-month control window before their COVID-19 test. The row includes the ICD-10-CM code and its description, a unique patient identifier, the COVID-19 test date, and the calculated number of months between ICD-10-CM code entry and COVID-19 testing. This “before” file contained 14,980,288 observations across 426,970 patients.

    We followed the patients for 7 months. The second file (“after”) was created 7 months after the last patient was included. The after file contained ICD-10-CM codes assigned during the 7 months following COVID-19 testing and similar related information as the “before” file. This “after” file contained 15,493,587 observations across 389,677 patients.

    We limited the analysis to the 367,148 patients that appeared in both the “before” and “after” files to ensure that we had a diagnostic history for each patient and eliminated acute findings by removing all ICD-10-CM codes documented less than a month after the positive COVID-19 test (Figure 1). We used the date of the first positive COVID-19 test for patients with multiple positive tests. Multiple repeating ICD-10-CM codes for a single patient were counted once. We wrote R (R Development Core Team) and Python (Python Software Foundation) programs to remove all data concerning patients who tested negative for COVID-19, ICD-10-CM codes that were documented less than a month after the positive COVID-19 test, and patients who were not present in both the “before” and “after” files. The methodology used to generate the patient cohort is depicted in Figure 2.

    We collected additional data to examine the association of demographics, comorbidities, vaccination status, and COVID-19 case severity with the incidence of long COVID. Demographic data collected included age, sex, race, and ethnicity. Comorbidities were evaluated using 2-year Elixhauser Comorbidity Indices Scores. Patients who were vaccinated were defined as having at least one COVID-19 vaccine dose recorded for at least two weeks and no more than 9 months before their positive COVID-19 test. We defined 2 classes of severe COVID-19 based on the minimum recorded oxygen saturation. The first class, severe COVID-19, was defined by a minimum oxygen saturation of <94% [20]. The second class, severe COVID-19 with severe desaturation, was defined by a minimum oxygen saturation of <88% [21].

    ‎
    Figure 2. Patient selection flowchart for this study. The first step excludes all patients that test negative for COVID-19. The second step excludes all patients that are not in both the “before” and “after” files, indicating either a lack of diagnosis history or a lack of follow-up, respectively. The final step excludes any patient that does not have a novel diagnosis 1 to 7 months after their positive test. A diagnosis is novel if it was not observed in the “before” file but was observed during this time. The “before” file consists of data concerning every ICD-10-CM code assigned to each patient who was COVID-19 positive in our cohort for up to 6-months before their test. The “after” file consists of data concerning every ICD-10-CM code assigned to each COVID-19 positive patient in our cohort for up to 7-months after their test. ICD-10-CM: International Classification of Diseases, 10th Revision, Clinical Modification.

    Ethical Considerations

    An institutional review board protocol was developed and approved for this research by the Department of Veterans Affairs (number 1580090). It uses electronic health care records data for patients of VA hospitals across the country. These data are protected health care data and are only used outside of the VA system in a summarized, deidentified format. The original informed consent from patients allows these analyses without additional consent. No additional compensation was given. A HIPAA (Health Insurance Portability and Accountability Act) waiver was approved (00004461).

    Data Analysis

    We chose 6 months “before” and “after” COVID-19 test control and case windows to allow patients to serve as their own controls. The 6 month “after” case window began 1 month after the positive COVID-19 test. We defined “novel” ICD-10-CM codes as those that appeared in a patient’s “after” file but not in the “before” file. We calculated the frequency of each novel ICD-10-CM code as the percentage of this study’s cohort assigned with the code. We excluded novel codes with a frequency of <1:1000 from further analysis. We defined codes as upregulated when the frequency of that code in the after file was significantly increased as compared to the frequency in the before file, if it had a χ2 with a Bonferroni corrected P<.00006. All resulting codes were grouped for additional analysis and organization (see ICD-10-CM Functional and Medical Specialty Groupings subsection of Methods).

    We also defined ICD-10-CM functional groups as “upregulated” if they were statistically more frequent post COVID-19 by χ2 analysis with Bonferroni correction with a P<.00006. We limited the long COVID ICD-10-CM code functional groups to those that were significantly increased in frequency. We used “before” and “after” frequencies to calculate odds ratios and CIs for each novel ICD-10-CM functional group. We calculated odds ratios from frequency data.

    We analyzed potential risk factors including vaccination status and COVID-19 severity for long COVID by creating 2 × 2 tables and applying Pearson χ2 testing. We applied a similar approach to the analysis of demographic factors. We used R (version 4.1.2) and RStudio (Posit Software, PBC) to perform the statistical analysis.

    ICD-10-CM Functional and Medical Specialty Groupings

    We grouped ICD-10-CM codes in 3 steps. We first combined ICD-10-CM codes that had the same initial 3 characters. We then grouped ICD-10-CM codes with different initial characters if the diagnoses were functionally similar to create our ICD-10-CM functional groups. For example, we grouped I47.1 (supraventricular tachycardia), I47.2 (ventricular tachycardia), and R00.0 (tachycardia, unspecified) as tachycardia. Finally, we manually curated each of these ICD-10-CM functional groups into medical specialties for organizational purposes.

    Long COVID Definition

    We included in our long COVID definition each ICD-10-CM code with an incidence over 6 months (T0 + 1M – T0+7M) >1:1000 (M: months; T0: date of positive COVID-19 test) and a significant overall frequency increase. Patients with long COVID were defined as having any of the 323 upregulated ICD-10-CM codes between 2 and 7 months after their positive COVID-19 diagnosis, but not in their pre–COVID-19 diagnoses.

    Risk Factors for Long COVID

    The multivariate regression models were done for each risk factor one including age, gender, race, ethnicity, and 2-year Elixhauser score; a second with age, gender, race, ethnicity, 2-year Elixhauser score, and O2 saturation <94%; and a third with age, gender, race, ethnicity, 2-year Elixhauser score, and COVID-19 vaccination status. We present the univariate rates as well as the results of the regression analysis using R (version 4.1.2) and RStudio.


    Long COVID Definition

    We extracted ICD-10-CM diagnosis codes assigned to 367,148 patients who underwent a positive COVID-19 test at VA. A total of 268,320 patients had one or more novel COVID-19–related diagnoses. The remaining 98,828 patients had no novel long COVID ICD-10-CM diagnoses in their post–COVID-19 period when compared to their pre–COVID-19 period. Table 1 contains the demographic characteristics of this study’s cohort. Men were significantly older than women on average, 60.29 years (95% CI 60.24-60.35) versus 47.85 years (95% CI 47.73-47.97), respectively.

    We developed a definition of long COVID consisting of 323 ICD-10-CM diagnosis codes grouped into 143 ICD-10-CM functional groups that were significantly increased in our 367,148 patient post–COVID-19 population. We define 17 medical specialty long COVID subtypes including cardiology long COVID, neurology long COVID, and pulmonary long COVID. Multimedia Appendix 1 shows the ICD-10-CM functional groups and medical specialties. Within each field, the ICD-10-CM code groups are sorted in descending order by their odds ratios. Combined odds ratios were calculated for each medical specialty category in Table S2 in Multimedia Appendix 2. Additional information about specific codes can be found in Table S3 in Multimedia Appendix 3.

    Figures 3-6 show the signs, symptoms, and diagnoses with significantly increased relative risks in the post–COVID-19 period with their respective 95% CIs sorted by medical specialty. The data used to create these figures can be found in Table S4 in Multimedia Appendix 4.

    Case counts were greatest for the specialties of cardiology (196,632), neurology (159,358), ophthalmology (149,817), and pulmonary (138,470). The lowest case counts were for oncology (7256), rheumatology (10,543), and dermatology (13,233; see Table 2 for more details).

    Patients who were COVID-19 positive were assigned novel signs, symptoms, or diagnoses included in our definition of long COVID at a rate of between 59.7% (268,320/449,450, the percentage is based on patients who were COVID-19 positive and tested at the VA) and 76.6% (268,320/350,315, the percentage is based on all patients who were COVID-19 positive with a diagnostic history and follow-up diagnoses 1 to 7 months after test).

    Most patients with long COVID were documented with at least one ICD-10-CM code found in our long COVID definition within 3 months of their positive COVID-19 test (168,194/268,320, 62.7%). The percentage of patients documented with their first long COVID ICD-10-CM code decreased with each subsequent month.

    Table 1. Demographic data and 2-year Elixhauser scores for patients with long COVID and patients with non–long COVID.
    DemographicPatients with non–long COVID (N=98,828)Patient with long COVID (N=268,320)P value
    Age (years), mean (95% CI)52.14 (52.03-52.24)60.85 (60.79-60.91)—a
    Elixhauser score, mean (95% CI)3.03 (2.99-3.07)7.05 (7.01-7.09)—
    Gender, n (%)

    		Men75,418 (76.31)234,720 (87.48)<.001

    		Women16,854 (17.05)31,651 (11.8)<.001

    		Not listed6556 (6.63)1949 (0.73)<.001
    Ethnicity, n (%)

    		Hispanic or Latino10,147 (10.27)26,171 (9.75)<.001

    		Not Hispanic or Latino71,595 (72.44)227,404 (84.75)<.001

    		Not listed17,086 (17.29)14,745 (5.5)<.001
    Race, n (%)

    		American Indian or Alaska Native743 (0.75)2243 (0.84).011

    		Asian1420 (1.44)2973 (1.11)<.001

    		Black or African American20,779 (21.03)65,218 (24.31)<.001

    		Native Hawaiian or other Pacific Islander905 (0.92)2522 (0.94)=.50

    		White55,359 (56.02)173,169 (64.54)<.001

    		Not listed19,622 (19.85)22,195 (8.27)<.001

    aNot available.

    ‎
    Figure 3. Odds ratios <3 for long COVID ICD-10-CM functional groups by medical specialty subtype: cardiology, dentistry, dermatology, endocrinology, gastroenterology, and otolaryngology. ICD-10-CM: International Classification of Diseases, 10th Revision, Clinical Modification.
    ‎
    Figure 4. Odds ratios <3 for long COVID ICD-10-CM functional groups by medical specialty subtype: general internal medicine, hematology, infectious disease, neurology, and oncology. ICD-10-CM: International Classification of Diseases, 10th Revision, Clinical Modification. MGUS: monoclonal gammopathy of undetermined significance; MRSA: methicillin-resistant Staphylococcus aureus; MSSA: meticillin-sensitive Staphylococcus aureus.
    ‎
    Figure 5. Odds ratios <3 for long COVID ICD-10-CM functional groups by medical specialty subtype: nephrology, ophthalmology, psychiatry or psychology, pulmonary, rheumatology, and urology. ICD-10-CM: International Classification of Diseases, 10th Revision, Clinical Modification.
    ‎
    Figure 6. Odds ratios >3 for long COVID ICD-10-CM functional groups by medical specialty subtype. ICD-10-CM: International Classification of Diseases, 10th Revision, Clinical Modification.
    Table 2. Case counts by medical specialty.
    SubspecialtyDiagnosesCases, n
    Cardiology38196,632
    Neurology38159,358
    Ophthalmology20149,817
    Pulmonary42138,470
    Endocrinology2397,884
    Gastroenterology2787,302
    Nephrology3386,582
    Psychiatry or psychology975,292
    Hematology2046,372
    Urology939,336
    General internal medicine933,130
    Infectious diseases2230,998
    Dentistry1330,202
    Otolaryngology423,583
    Dermatology713,223
    Rheumatology210,543
    Oncology47256
    Totals3201,225,980

    Risk Factors for Long COVID

    We presented in Table 1 a comparison of demographic characteristics and Elixhauser comorbidity scores of patients with long COVID and patients with non–long COVID. The long COVID cohort was older with more comorbidities. The long COVID cohort also had higher percentages of White and Black individuals and non-Hispanic and non-Latino ethnicities. The patients with a 2-year Elixhauser score of greater than 21 had a much higher proportion to develop long COVID (P<.001, Pearson χ2; see Table 3).

    Our data did not indicate that vaccination was protective against the development of long COVID. However, vaccination resulted in significantly lower rates of novel acute respiratory distress syndrome in the post–COVID-19 period (13.2%, 95% CI 10.4%-16.9%) as compared with the unvaccinated population (19.6%, 95% CI 18.1%-21.2%; P<.001).

    Patients with minimum O2 saturations constituting severe COVID-19 and severe COVID-19 with severe desaturation were significantly more likely to develop long COVID (both had P<.001, Pearson χ2; see Table 4).

    The multivariate regression models all confirmed that patients with COVID-19 during the Omicron variant predominant period were at a slightly higher risk of developing long COVID at P<.001.

    Table 3. Proportion of patients that developed long COVID comparing different 2-year Elixhauser score ranges.
    2 year Elixhauser scoreNon–long COVID countLong COVID countPercent long COVIDa (95% CI)
    0-2196,055242,10871.60 (71.44-71.75)
    22-42245422,71690.25 (89.88-90.61)
    43-63308331791.50 (90.55-92.36)
    64-841117994.21 (89.93-96.77)

    aThe percentages come from the numbers to the left of each percentage. It is the long COVID count divided by the sum of both categories.

    Table 4. Low oxygen saturations and the proportion of patients that developed long COVID.
    Severe COVID-19Non–long COVID countLong COVID countPercent long COVIDa (95% CI)
    Low O2 (NIHb definitionc)390329,41188.28 (87.93-88.63)
    No low O2 (NIH definitionc)94,925238,90971.57 (71.41-71.72)
    Low O2 (severe desaturationd)637556689.73 (88.95-90.46)
    No low O2 (severe desaturationd)98,191262,75472.80 (72.65-72.94)

    aThe percentages come from the numbers to the left of each percentage. It is the long COVID count divided by the sum of both categories.

    bNIH: National Institutes of Health.

    cMinimum (O2 saturation) <94%.

    dMinimum (O2 saturation) <88%.


    Conclusions

    Numerous reports document specialty-specific signs, symptoms, and diagnoses correlated with long COVID. We present a novel analysis based on a large national data set and the full multispecialty breadth of ICD-10-CM diagnosis codes to create a holistic long COVID definition that confirms and extends previous reports.

    We allowed patients to be their own controls and used the entire cohort before and after COVID-19 infection to determine the relative risk of signs, symptoms, and disorders. This ensured that the signal was both novel and upregulated. We found patients who were COVID-19 positive developed signs, symptoms, or diagnoses included in our long COVID definition at a proportion of between 59.7% (268,320/449,450, the percentage is based on a denominator of all patients who were COVID-19 positive and tested at the VA) and 76.6% (268,320/350,315, the percentage is based on a denominator of all patients who were COVID-19 positive with a diagnostic history and follow-up diagnoses 1 to 7 months after test). More than three-fourths of patients with long COVID met our long COVID definition within 4 months of their positive COVID-19 test.

    We found long COVID frequency differences based on race and ethnicity. These differences may be related to socioeconomic status, which is directly correlated with the presence of comorbidities [22-24]. The long COVID cohort was 8 years older with more comorbidities (2-year Elixhauser score 7.97 in the patients with long COVID vs 4.21 in the patients with non–long COVID). In our cohort, the men were significantly older than the women on average, 60.29 years (95% CI 60.24-60.35) versus 47.85 years (95% CI 47.73-47.97), respectively. We found that long COVID frequency was increased in patients who were more severely ill before infection and patients who had a more severe bout of COVID-19 as judged by their minimum oxygen saturation.

    We found 143 upregulated diagnostic groups, with odds ratios as high as 23. We also found 17 upregulated medical specialty groupings containing between 3 and 21 signs, symptoms, or diagnoses. This provides strong evidence for a broad definition of long COVID.

    Carfi et al [2] found that the most common long-term symptoms were fatigue, dyspnea, joint pain, and chest pain. Each except joint pain is represented in our long COVID definition. However, joint pain may be related to findings in our definition such as difficulty walking and an overall decrease in mobility. COVID-19 is known to cause lung abnormalities, especially in cases with pneumonia [25]. We found that the likelihood of developing pneumonia after COVID-19 infection is significantly upregulated, potentially interconnected with the numerous findings in our pulmonary long COVID definition. Autopsy evaluation of COVID-19 victims’ lung tissue demonstrated diffuse alveolar damage with perivascular T-cell infiltration and severe endothelial injury [26]. Patients with long COVID have been found to have abnormal 129Xe magnetic resonance imaging gas exchange and computed tomography vascular density measurements, which we postulate could be related to the pulmonary fibrosis (J84.10) or emphysema (J43.9) diagnoses identified in our definition [27].

    Our definition shows that the long-term effects of COVID-19 are associated with damage to numerous body systems including the kidneys, heart, eyes, and nervous system. Our results are corroborated by other studies. Cognitive dysfunction (brain fog) is often associated with long COVID and can be difficult to diagnose and treat [5]. COVID-19 infection is far more likely to cause cardiac complications than vaccination [28]. The gastrointestinal codes we observed reflect previous literature [29] and may relate to reported alterations to the gastrointestinal tract after COVID-19 [3,4]. Finally, previous studies have noted that COVID-19 can alter ocular physiology, supporting our ophthalmology-related findings [30].

    Patients with more severe cases of COVID-19, as manifested by low oxygen saturations, should be watched carefully for the development of long COVID as they are significantly more likely to develop long COVID. Sicker patients with higher 2-year Elixhauser scores were significantly more likely to develop long COVID. Patients with multiple comorbidities should be made aware of this risk and participate in active surveillance for the development of signs and symptoms of long COVID.

    The American Medical Association notes there are 3 categories of patients with long COVID: those who do not recover completely and have ongoing symptoms, those with symptoms related to chronic hospitalization, and those who develop new symptoms after recovery [31]. In our study, we did not differentiate by these subtypes and instead leave that to future research. It is possible that some of these signs and symptoms may have occurred during the first month and may be the persistent subtype. It is possible that some of the upregulated codes may be found with other serious illnesses, though only 9.1% (33,314/367,148) of our cohort had severe COVID-19 based on oxygen saturation <94%. We are not able to distinguish conditions that represent an acceleration of pre-existing disease from those that represent de novo COVID-19–related conditions. For example, is the increased incidence of nonsore throat elevation myocardial infarction (I21.4) related to the general stress of acute illness impacting pre-existing coronary artery disease or to an underlying de novo long COVID–related condition? A better understanding will require additional research. In any event, whether causal or associative, de novo disease or exacerbation of chronic disease and new or persistent clinical problems require assessment, treatment, and monitoring.

    Limitations include that the cohort study population is 84% men, reflective of the overall patient population of VA which is between 87% and 95% men (depending upon data source and whether gender has been self-reported) [32,33]. Additionally, the male veteran population who use the VA health care system is older than the population of female veterans who use the VA. Our study did not include home testing for COVID-19 that went unreported to the VA health care system. Patients who tested positive during the omicron dominant time period were slightly more likely to develop long COVID when compared to the earlier strains (66,643/87,522, 76% vs 201,677/279,626, 72%; P<.001). The reality of emerging viral variants emphasizes the need for a well-defined and well-maintained definition of long COVID over time and with variant-specific derivation. This study was not powered to show the independence of the individual risk factors for long COVID.

    We hope that our empirically defined long COVID definition will lead to more consistent identification of long COVID and its medical specialty subtypes and support of a variety of COVID-19–related initiatives. Our definition is actionable as individuals who have multiple comorbidities and more severe bouts of COVID-19 should be followed more closely for the development of long COVID signs or symptoms. Our definition can also inform screening questions for high-risk patients. For example, helping clinicians identify patients with enhanced long COVID risk who may benefit from monitoring programs or patients with previously undiagnosed long COVID for whom it may be appropriate to create a referral to a long COVID clinic. We also anticipate that our long COVID definition will support the standardization of future subspecialty-specific long COVID research.

    Future research should look at health outcomes for each long COVID-19 medical specialty subtype to identify those at greatest risk of developing severe morbidity. Predictive analytics should be used to help refer these individuals earlier to monitoring and treatment programs.

    As of December 17, 2023, there have been over 772 million confirmed cases of COVID-19 worldwide [34]. Case counts are ever-increasing. As Levine [18] notes, immediately useful long COVID definitions are needed as are ultimately more fully inclusive definitions. We offer our long COVID definition as a public health contribution to our pandemic response.

    Acknowledgments

    This work has been supported in part by grants from the National Institutes of Health (National Library of Medicine T15LM012495 and R25LM014213, National Institute on Alcohol Abuse and Alcoholism R21AA026954 and R33AA0226954, and National Center for Advancing Translational Sciences UL1TR001412). This study was funded in part by the Department of Veterans Affairs.

    Data Availability

    All data generated or analyzed during this study are either protected patient data that require a Veterans Affairs research appointment and institutional review board protocol to access or summarized and included in this published paper (and its supplementary information files).

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    Long COVID definition results by International Classification of Diseases, 10th Revision, Clinical Modification functional group description and medical specialty.

    DOCX File , 24 KB
    Multimedia Appendix 2

    Combined odds ratios calculated for each medical specialty category.

    XLSX File (Microsoft Excel File), 12 KB
    Multimedia Appendix 3

    Additional information about specific codes.

    XLSX File (Microsoft Excel File), 24 KB
    Multimedia Appendix 4

    Data used to create these figures.

    XLSX File (Microsoft Excel File), 27 KB
    1. Phillips S, Williams MA. Confronting our next national health disaster—long-haul covid. N Engl J Med. 2021;385(7):577-579. [CrossRef] [Medline]
    2. Carfì A, Bernabei R, Landi F, Gemelli Against COVID-19 Post-Acute Care Study Group. Persistent symptoms in patients after acute COVID-19. JAMA. 2020;324(6):603-605. [FREE Full text] [CrossRef] [Medline]
    3. Ng SC, Tilg H. COVID-19 and the gastrointestinal tract: more than meets the eye. Gut. 2020;69(6):973-974. [FREE Full text] [CrossRef] [Medline]
    4. Villapol S. Gastrointestinal symptoms associated with COVID-19: impact on the gut microbiome. Transl Res. 2020;226:57-69. [FREE Full text] [CrossRef] [Medline]
    5. Davis HE, Assaf GS, McCorkell L, Wei H, Low RJ, Re'em Y, et al. Characterizing long COVID in an international cohort: 7 months of symptoms and their impact. EClinicalMedicine. 2021;38:101019. [FREE Full text] [CrossRef] [Medline]
    6. Douaud G, Lee S, Alfaro-Almagro F, Arthofer C, Wang C, McCarthy P, et al. SARS-CoV-2 is associated with changes in brain structure in UK Biobank. Nature. 2022;604(7907):697-707. [FREE Full text] [CrossRef] [Medline]
    7. Menni C, Valdes AM, Freidin MB, Sudre CH, Nguyen LH, Drew DA, et al. Real-time tracking of self-reported symptoms to predict potential COVID-19. Nat Med. 2020;26(7):1037-1040. [FREE Full text] [CrossRef] [Medline]
    8. Sudre CH, Murray B, Varsavsky T, Graham MS, Penfold RS, Bowyer RC, et al. Attributes and predictors of long COVID. Nat Med. 2021;27(4):626-631. [FREE Full text] [CrossRef] [Medline]
    9. Wong MCS, Huang J, Wong YY, Wong GLH, Yip TCF, Chan RNY, et al. Epidemiology, symptomatology, and risk factors for long COVID symptoms: population-based, multicenter study. JMIR Public Health Surveill. 2023;9:e42315. [FREE Full text] [CrossRef] [Medline]
    10. Davis HE, McCorkell L, Vogel JM, Topol EJ. Long COVID: major findings, mechanisms and recommendations. Nat Rev Microbiol. 2023;21(3):133-146. [FREE Full text] [CrossRef] [Medline]
    11. Larsen NW, Stiles LE, Shaik R, Schneider L, Muppidi S, Tsui CT, et al. Characterization of autonomic symptom burden in long COVID: a global survey of 2,314 adults. Front Neurol. 2022;13:1012668. [FREE Full text] [CrossRef] [Medline]
    12. Katsoularis I, Jerndal H, Kalucza S, Lindmark K, Fonseca-Rodríguez O, Connolly AMF. Risk of arrhythmias following COVID-19: nationwide self-controlled case series and matched cohort study. Eur Heart J Open. 2023;3(6):oead120. [FREE Full text] [CrossRef] [Medline]
    13. Ballering AV, van Zon SKR, Hartman TCO, Rosmalen JGM, Lifelines Corona Research Initiative. Persistence of somatic symptoms after COVID-19 in the Netherlands: an observational cohort study. Lancet. 2022;400(10350):452-461. [FREE Full text] [CrossRef] [Medline]
    14. Logue JK, Franko NM, McCulloch DJ, McDonald D, Magedson A, Wolf CR, et al. Sequelae in adults at 6 months after COVID-19 infection. JAMA Netw Open. 2021;4(2):e210830. [FREE Full text] [CrossRef] [Medline]
    15. Blanchflower DG, Bryson A. Long COVID in the United States. PLoS One. 2023;18(11):e0292672. [FREE Full text] [CrossRef] [Medline]
    16. Cutler DM. The costs of long COVID. JAMA Health Forum. 2022;3(5):e221809. [FREE Full text] [CrossRef] [Medline]
    17. Appelman B, Charlton BT, Goulding RP, Kerkhoff TJ, Breedveld EA, Noort W, et al. Muscle abnormalities worsen after post-exertional malaise in long COVID. Nat Commun. 2024;15(1):17. [FREE Full text] [CrossRef] [Medline]
    18. Levine RL. Addressing the long-term effects of COVID-19. JAMA. 2022;328(9):823-824. [FREE Full text] [CrossRef] [Medline]
    19. Souden M. Overview of VA data, information systems, National Databases and Research uses. Health Systems Research. URL: https:/​/www.​hsrd.research.va.gov/​for_researchers/​cyber_seminars/​archives/​video_archive.​cfm?SessionID=1203 [accessed 2024-02-29]
    20. COVID-19 Treatment Guidelines Panel. Coronavirus Disease 2019 (COVID-19) Treatment Guidelines. National Institutes of Health. 2022. URL: https:/​/files.​covid19treatmentguidelines.nih.gov/​guidelines/​archive/​covid19treatmentguidelines-12-20 -2023.​pdf [accessed 2024-03-11]
    21. Albert RK, Au DH, Blackford AL, Casaburi R, Cooper JA, Criner GJ, et al. A randomized trial of long-term oxygen for COPD with moderate desaturation. N Engl J Med. 2016;375(17):1617-1627. [FREE Full text] [CrossRef] [Medline]
    22. Sahni S, Talwar A, Khanijo S, Talwar A. Socioeconomic status and its relationship to chronic respiratory disease. Adv Respir Med. 2017;85(2):97-108. [FREE Full text] [CrossRef] [Medline]
    23. Leng B, Jin Y, Li G, Chen L, Jin N. Socioeconomic status and hypertension: a meta-analysis. J Hypertens. 2015;33(2):221-229. [CrossRef] [Medline]
    24. Jaffiol C, Thomas F, Bean K, Jégo B, Danchin N. Impact of socioeconomic status on diabetes and cardiovascular risk factors: results of a large French survey. Diabetes Metab. 2013;39(1):56-62. [CrossRef] [Medline]
    25. Ding X, Xu J, Zhou J, Long Q. Chest CT findings of COVID-19 pneumonia by duration of symptoms. Eur J Radiol. 2020;127:109009. [FREE Full text] [CrossRef] [Medline]
    26. Ackermann M, Verleden SE, Kuehnel M, Haverich A, Welte T, Laenger F, et al. Pulmonary vascular endothelialitis, thrombosis, and angiogenesis in Covid-19. N Engl J Med. 2020;383(2):120-128. [FREE Full text] [CrossRef] [Medline]
    27. Matheson AM, McIntosh MJ, Kooner HK, Lee J, Desaigoudar V, Bier E, et al. Persistent 129Xe MRI pulmonary and CT vascular abnormalities in symptomatic individuals with post-acute COVID-19 syndrome. Radiology. 2022;305(2):466-476. [FREE Full text] [CrossRef] [Medline]
    28. Kuehn BM. Cardiac complications more common after COVID-19 than vaccination. JAMA. 2022;327(20):1951. [CrossRef] [Medline]
    29. Meringer H, Mehandru S. Gastrointestinal post-acute COVID-19 syndrome. Nat Rev Gastroenterol Hepatol. 2022;19(6):345-346. [FREE Full text] [CrossRef] [Medline]
    30. Costa Í, Bonifácio LP, Bellissimo-Rodrigues F, Rocha EM, Jorge R, Bollela VR, et al. Ocular findings among patients surviving COVID-19. Sci Rep. May 26, 2021;11(1):11085. [CrossRef] [Medline]
    31. What is long COVID? American Medical Association. 2022. URL: https://www.ama-assn.org/delivering-care/public-health/what-long-covid [accessed 2023-02-10]
    32. National healthcare quality and disparities report chartbooks. Agency for Healthcare Research and Quality. 2020. URL: https://www.ahrq.gov/research/findings/nhqrdr/chartbooks/index.html [accessed 2024-02-29]
    33. Peltzman T, Rice K, Jones KT, Washington DL, Shiner B. Optimizing data on race and ethnicity for veterans affairs patients. Mil Med. 2022;187(7-8):e955-e962. [FREE Full text] [CrossRef] [Medline]
    34. COVID-19 epidemiological update—22 December 2023. World Health Organization. URL: https://www.who.int/publications/m/item/covid-19-epidemiological-update---22-december-2023 [accessed 2024-01-06]

    ‎
    HIPAA: Health Insurance Portability and Accountability Act
    ICD-10-CM: International Classification of Diseases, 10th Revision, Clinical Modification
    VA: Veterans Affairs

    Edited by A Mavragani; submitted 11.06.23; peer-reviewed by Y Xie, A Balas; comments to author 01.12.23; revised version received 19.01.24; accepted 15.02.24; published 30.04.24.

    Copyright

    ©Skyler Resendez, Steven H Brown, Hugo Sebastian Ruiz Ayala, Prahalad Rangan, Jonathan Nebeker, Diane Montella, Peter L Elkin. Originally published in JMIR Public Health and Surveillance (https://publichealth.jmir.org), 30.04.2024.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on https://publichealth.jmir.org, as well as this copyright and license information must be included.