Use of Long COVID Clinics in the Veterans Health Administration: Implications for the path forward

Abstract

Long COVID is a serious chronic illness that can present in many forms and impact daily functioning and quality of life. Without curative treatments, management of long COVID requires coordination and ongoing access to multidisciplinary care. Starting in 2020, the Veterans Health Administration (VHA), established a national network of Long COVID Clinics (LCCs). In this retrospective cohort study of 494 547 veterans with documented SARS-CoV-2 infection in the VHA from March 2020 to April 2022 (n = 494 547), we examined trends in ICD-10 U09.9 diagnosis code use for long COVID and LCC use in the VHA up to May 2024. Overall, 5.9% (n = 29 195) of patients in our cohort had a documented U09.9 code and 2% had at least 1 LCC visit. Among veterans with a U09.9 code, 17.4% (n = 5089) used LCCs. LCC use rates were low across all patient subgroups. LCCs were more available to veterans residing in the South census region (28% vs <7% use rate) than veterans in other regions. Developing evidence about LCC effectiveness and ensuring equitable access to LCCs within and beyond the VHA will be critical in meeting the evolving needs of people with long COVID.

Introduction

Long COVID is a serious chronic illness caused by SARS-CoV-2 infection that affects multiple body systems, and is characterized by persistent symptoms and conditions that may range from mild to severe.^1-3 Given the range of definitions and methodological limitations of published studies, there is much uncertainty about the incidence and prevalence of long COVID.⁴ Large and representative national surveys estimate that approximately 7% of US adults or approximately 18 million adults have ever had long COVID,^5,6 with significant estimated economic costs from reduced quality of life, lost earnings, and medical spending.⁷ With regard to incidence, it is estimated that 10% to 35% of people infected with SARS-CoV-2 go on to develop long COVID.⁸ Current scientific understanding implicates several pathophysiologic abnormalities in patients with long COVID that may be driving the illness, including immune dysregulation, endothelial dysfunction, dysfunctional neurologic signaling, and microbiota dysbiosis.⁹ The impacts of long COVID on daily functioning and quality of life due to chronic fatigue, post-exertional malaise, cognitive impairments, and autonomic dysfunction are the key reasons patients seek care and motivated the emergence of Long COVID Clinics (LCCs).^1,10

Given the complex nature of long COVID illness and lack of established treatments, multidisciplinary care offered by LCCs to manage symptoms and improve functional ability may be the most promising approach to management of long COVID. Starting in 2020, LCCs were established by local clinical champions to provide specialty care to patients with long COVID across the Veterans Health Administration (VHA), the largest integrated health care network in the United States.¹¹ These LCCs vary widely across the VHA in both size and structure, with clinics ranging from 1 provider to large interdisciplinary teams. Clinics may provide in-person care, virtual care, or a hybrid of both. Clinics also vary in reach, providing local, regional, or even national care through established agreements with other VHA medical centers. Clinics may be administratively located within various departments such as Physical Medicine and Rehabilitation, Primary Care, Pulmonary, Infectious Disease, or another specialty. Models of care vary across the VHA, and include focus on care coordination and referrals to specialists, management of symptoms within clinics with limited consultation, and “tumor board”–style interdisciplinary discussions; however, there is generally an interdisciplinary focus regardless of the clinical model or primary specialty.

The goal of this analysis was to examine trends in LCC use in the VHA up to May 2024, and examine the extent of variation in LCC use among patient subgroups and those with and without International Classification of Diseases, Tenth Revision (ICD-10), U09.9 diagnosis code documentation for long COVID. We also discuss implications for the future of LCCs in the VHA and considerations for other health care systems. These results provide a benchmark for future work evaluating the effectiveness of LCCs and emerging care models.

Data and methods

Study design and population

We assembled a national retrospective cohort of all VHA enrollees with a documented positive SARS-CoV-2 test between March 1, 2020, and April 30, 2022, in the 50 states and District of Columbia; SARS-CoV-2 infection was indicated by either a positive SARS-CoV-2 polymerase chain reaction test recorded in patients’ electronic health records (EHRs) or documented in clinical notes (as described in reference 12). To minimize missing data among veterans who do not regularly use VHA services, we excluded veterans not assigned to a primary care team or without at least 1 primary care visit in the 2 years before infection. Veterans not residing in the 50 states or District of Columbia were also excluded.

Data sources and variables

Data on patient demographics, pre-existing health conditions, and LCC use were gathered from the VHA Corporate Data Warehouse, a central repository of patient EHR data. Patient demographics and health conditions were assessed as of the date of their first positive SARS-CoV-2 test, and we used a 2-year lookback to identify pre-existing health conditions. To ensure comprehensive assessment of long COVID, we included linked Centers for Medicare and Medicaid Services (CMS) data on inpatient and outpatient diagnoses (from Medicare Inpatient and Outpatient Fee-for-Service tables) as well as diagnoses that veterans received in visits to community clinics that were reimbursed by the VHA (from the Integrated Veteran Care Consolidated Data Sets). Veterans were identified as having long COVID if they had a documented outpatient or inpatient ICD-10 U09.9 diagnosis code for “Post COVID-19 condition, unspecified” in their EHR anytime between the date of initial infection and May 2024. Patients without an ICD-10 U09.9 code were considered as not having long COVID. We used Managerial Cost Accounting 4-Character National Codes (CHAR4), which are administrative codes that indicate different outpatient clinics in each VHA medical center, to identify LCC visits from September 2020 to May 2024 and their use. The CHAR4 codes have some national standards, but others (such as LCCs) are set locally, and may be delayed in being implemented following national standards.¹³ For this analysis, we used the “CGLC” and “DEBC” codes that are reserved exclusively for LCCs. Long COVID Clinic use could occur before or after U09.9 documentation.

We collected data on patient demographics, pre-existing conditions, and geography (state of residence, US Census region) to determine whether U09.9 diagnoses and LCC use varied by patient subgroup. Patients’ Care Assessment Needs (CAN) scores, an automatically generated and validated comorbidity score that accurately predicts 1-year hospitalization and mortality risk (higher scores indicating poorer health), were collected to assess overall health and collected at the time of infection.¹⁴ Diagnosis codes for pre-existing health conditions considered high risk for severe COVID-19 were also collected,¹⁵ such as heart failure and pulmonary disease.

Statistical methods

In descriptive analyses, we compared characteristics of cohort members with and without U09.9 documentation and cohort members who did and did not have 1 or more LCC visit during the observation period via standardized mean differences (SMDs) both in the overall cohort and in the subgroup with diagnosed long COVID. We considered variables with SMDs greater than 0.10 to indicate important differences between comparison groups.¹⁶ We calculated rates (per 100 persons) of U09.9 documentation and LCC use by patient demographics and characteristics to identify patient subgroups with high rates of U09.9 coding and potential patient gaps in LCC use. We also examined geographic variation in LCC use among veterans with U09.9 documentation. In post hoc analyses, we conducted a sensitivity analysis excluding patients residing in Texas, given that they were early LCC adopters (clinic established in March 2020) and conducted active long COVID outreach and screening for all patients with newly diagnosed COVID-19.

Results

Our cohort included 494 547 veterans with a documented first case of SARS-CoV-2 in the VHA between March 1, 2020, and April 30, 2022, who met our inclusion/exclusion criteria (Figure S1). The cohort was predominantly male (88%), with a median age of 61 years (IQR: 47, 73 years); the majority lived in urban areas (71%) and the South census region (45%) (Table 1). Approximately one quarter (26%) of patients resided in Texas (8.9%), Florida (8.8%), and California (8.0%) (Table S1). Most veterans were non-Hispanic (78%) and White race (69%). The most prevalent pre-existing health conditions in the cohort were hypertension (58%), diabetes (32%), and coronary heart disease (28%).

U09.9 Documentation in the Cohort

Approximately 6% (n = 29,195 of 454 547) of veterans had U09.9 codes documented between October 2021 and May 2024. Compared with veterans without a U09.9 diagnosis code, those with U09.9 codes were slightly older (median age: 63 vs 61 years; SMD = 0.124), and more likely to be Hispanic (15% vs 8%; SMD = 0.275) and reside in the South census region (55% vs 45%; SMD = 0.215) (Table 1). Veterans who received a U09.9 code had slightly higher CAN scores compared with those who did not (median CAN score: 65 vs 60; SMD = 0.113). Pulmonary conditions (28% vs 22%; SMD = 0.137), coronary heart disease (33% vs 27%; SMD = 0.111), and diabetes (36% vs 31%; SMD = 0.106) were also more common among those with a U09.9 code (Table 1). The distribution of sex, urbanicity, COVID-19 vaccination history, smoking status, and number of prior health visits was similar (all SMD < 0.10) between those with and without U09.9 documentation.

When assessing U09.9 documentation rates by subgroups we found high rates of U09.9 coding in several subgroups. Among the pre-existing conditions examined, we found the U09.9 documentation rate was highest among veterans with a history of organ transplantation (9%) (Figure S2). U09.9 documentation rates were higher among Hispanic than non-Hispanic veterans (11% vs 5%) and highest in the South (7%) compared with other census regions (Figure S3). Poorer health appeared to be associated with U09.9 documentation as U09.9 rates increased as CAN scores and body mass index (BMI) increased (Figure S3).

Long COVID clinic use

Overall, 2% (n = 7764 of 454,547) of the cohort had 1 or more LCC visits (see Table S2 for comparison of all LCC users with nonusers). The LCC visits occurred in 16 of the 18 Veteran Integrated Service Networks (VISN); no LCC visits were identified in VISN 9 (MidSouth Network) or VISN 15 (Heartland Network).¹⁷ Among the subset of cohort members with a documented U09.9 diagnosis code, 17% (n = 5089 of 29 159) had at least 1 recorded LCC visit (Table 2). Among veterans with a U09.9 code, those who used LCCs were younger than nonusers (mean age = 56 vs 62 years; SMD = 0.409), and more likely to be female (18% vs 12%; SMD = 0.168) and Hispanic (29% vs 12%; SMD = 0.435). Compared with nonusers, LCC users were more likely to live in urban areas (78% vs 69%; SMD = 0.220) and predominantly in the South census region (88% vs 48%; SMD = 0.944). The LCC users tended to be healthier with lower CAN scores than nonusers (median CAN score: 55 vs 65; SMD = 0.264), and with lower rates of pre-existing conditions like chronic kidney disease (19% vs 28%; SMD = 0.194) and coronary heart disease (23% vs 34%; SMD = 0.245). The LCC users had a higher prevalence of mental health conditions like major depression (41% vs 36%; SMD = 0.103) and post-traumatic stress disorder (35% vs 27%; SMD = 0.159), and were more likely to not have received a COVID-19 vaccination (66% vs 60%; SMD = 0.128). BMI and rarer pre-existing conditions did not appear to differ between LCC users and nonusers (SMD < 0.10).

Within demographic subgroups of those with a U09.9 diagnosis code, Hispanic veterans had the highest LCC use rate, with over one third having used LCCs (34%) (Figure 1). The LCC use rate was highest among veterans younger than 45 years old (26%) and decreased with age and higher CAN scores (see Figure S4 for LCC use rates by pre-existing conditions). Among the 4 regions, the South census region had the highest LCC use rate (28%). The LCC use rates were slightly higher in veterans residing in urban areas (19%) than in nonurban areas (13%). Notably, Texas had the highest cumulative LCC use rate among patients with long COVID (44%), followed by Washington, DC (30%), Florida (21%), and Maryland (18%) (Figure 2). Use rates accelerated through 2023 and have continued at a slower rate in 2024.

Sensitivity analysis

After excluding 43 947 veterans residing in Texas, rates of U09.9 documentation and LCC use in the overall cohort were similar and differences in patient characteristics by U09.9 documentation (Table S3, Figure S5) and LCC use were largely the same (Table S4, Figure S6).

Discussion

As it became clear that many veterans infected with COVID-19 had ongoing health care needs beyond the period of acute infection, clinical champions throughout the VHA began to establish LCCs. In this analysis of veterans with documented SARS-CoV-2 infection in the VHA between March 2020 and April 2022, only 6% had U09.9 documentation of long COVID between October 2021 and May 2024. The LCC use rates were low overall, even among those with U09.9 documentation, and we found large regional disparities in LCC use rates.

While older veterans and those with a higher comorbidity burden tended to have higher rates of U09.9 documentation, LCC use rates declined with age and poorer health. It is possible that medically complex patients with long COVID have care needs met in ongoing primary care teams, specialty care, or in home-based primary care.^18,19 Due to the broad definition of long COVID, the U09.9 diagnosis code may also be used in different ways for younger and older patients; among younger patients, U09.9 may be more frequently given for an incident syndrome that providers recognize as requiring specialized care, but for the elderly it may signify worsening pre-existing conditions that their existing care teams address.^20,21 Given the high rate of U09.9 documentation in older veterans, however, their primary care clinicians may need additional support and training for managing long COVID, and guidance for when referral to LCCs may be necessary.^21,22

Whether long COVID care needs of older veterans are fully met in non-LCCs or LCCs is unknown and warrants further inquiry to inform the development of standards of care and care practices that can be more widely implemented in other settings. Given the potential serious long-term impacts of long COVID on daily functioning and quality of life,^1-3 if LCC care is found to be more effective in managing long COVID, efforts to increase regional access to LCCs in the VHA may be needed. While outside the scope of the current study, future work to understand the effect of LCCs on patient outcomes is warranted to inform policies for expanding LCC access. A clear understanding of potential benefits of LCC care is complicated by the heterogeneity of the condition; however, recent work supports the use of individualized care with idiographic reasoning in an interdisciplinary setting and such resources may be more readily available in specialized LCCs.²³

While racial, ethnic, and gender disparities in COVID-19 outcomes and access to treatment have been documented,^24,25 we did not observe these disparities in LCC access, possibly because LCC use rates were low. Nevertheless, the vast majority of veterans with long COVID in all demographics examined did not use LCCs. In the VHA and other health systems,²⁶ LCCs are still emerging and therefore not equally available in all areas. Further, they may not be systematically identified in VHA administrative data based on the CHAR4 codes that we used. Long COVID Clinics use multidisciplinary teams of specialists often located in urban areas, which can leave rural areas under-resourced. Given the proliferation of telehealth and clinical hub-and-spoke models that bring together resources across VHA facilities, further consideration of regional resource sharing might also be advantageous to meet demand for long COVID care in rural areas.^22,27,28 The extent to which these approaches and care models may be effective in reaching distal and rural areas has yet to be reported and requires further evaluation. We have much to learn about the effectiveness of various LCC care models that have emerged.^29,30 Long COVID Clinic care models range from initial assessments with multidisciplinary providers to hub-and-spoke models with an anchor clinician providing referrals to various specialists.^31-35 In the United Kingdom, implementation of a multidisciplinary consulting team has been used with the LCC model, especially for complex cases.²³ Gathering such multidisciplinary perspectives is possible in an integrated system like the VHA and other US health systems, yet has not been reported so far.

Currently, Texas and Florida have 2 of the largest VHA LCCs.³⁶ The VHA LCC in south Texas, the largest and oldest in the VHA, operates as a virtual-care model offering active outreach and comprehensive interdisciplinary care to veterans in the convalescence phase of COVID-19. Based within Geriatrics and Extended Care, this clinic serves as a nationally designated telehealth hub, expanding its reach across the VHA. By utilizing telehealth visits, the VHA LCC in south Texas improves access to long COVID care for veterans, particularly those living in rural areas of Texas. The VHA LCC in Miami is a generalist-led model with internal medicine clinicians providing and coordinating comprehensive long COVID–specific care in a multidisciplinary setting.³⁷ This care may be tailored to patient needs and may look like chronic fatigue syndrome care in the appropriate patient or complex care management in a patient with pulmonary fibrosis.³⁸ The VHA LCC in Miami provides a hybrid of both in-person and virtual (telehealth) care locally and has also set up telehealth service agreements with other facilities in Florida to provide remote care regionally following a hub-and-spoke model similar to that of south Texas.³⁷ While pathways to access LCCs vary across the VHA, referrals are typically required and may be placed by any provider in the health care system. Informal survey of clinics participating in the VHA Long COVID Community of Practice confirms that most referrals come from primary care, followed by medical specialties such as pulmonary, cardiology, and rheumatology. These referrals often occur when specialists recognize that symptoms for which patients sought care are occurring in the context of systemic symptoms outside of their area of specialty. Additionally, 4 clinics (south Texas, Miami, VISN 21 Clinical Resource Hub, and the Bronx Veterans Affairs [VA] Medical Center) participate in an active digital outreach screening,³⁶ and at least 1 LCC, in Tampa, has a direct scheduling option through a clinical call center that does not require a provider consultation.

There remains uncertainty about the mechanisms underlying SARS-CoV-2 infection and long COVID symptomatology, and consequently, effective therapeutics for long COVID are lacking. Long COVID Clinics might serve as an important nexus for coalescing both the multispecialty clinical expertise and the willing patient populations to evaluate the effectiveness of therapies to treat long COVID. The LCCs could serve as an important platform on which to deliver and evaluate new and existing medications and other therapies with robust study designs in real-world settings.^39,40 The extent to which standardized care pathways can be developed and evaluated for patients with long COVID can be examined through such networks. Similar research networks under the National Institutes of Health, such as Researching COVID to Enhance Recovery (RECOVER),⁴¹ could also be leveraged to obtain real-world evidence on effective therapies and practices. The evolution of long COVID and its effects over time and in different cohorts would benefit from rigorous pragmatic research to understand disease trajectory, risks, and effectiveness of treatments. Opportunities to have communities of practice that include patients and clinicians in co-designing studies and from a wide range of populations and settings may prove useful to prioritize addressing the most debilitating long COVID symptoms, as is currently underway in the Researching COVID to Enhance Recover-Treating Long COVID (RECOVER-TLC) program.⁴² Further, developing a registry of available trials and studies would foster access to a more diverse population of patients and increase the generalizability of results.

Limitations

This analysis had several important limitations. Long COVID is difficult to diagnose and the ICD-10 U09.9 code may be inconsistently used, resulting in an undercount of long COVID cases.^43,44 Moreover, the U09.9 diagnostic code is an imperfect proxy for long COVID diagnosis; a validation study across 3 large medical systems found a positive-predictive value of only 23%–62% based on the World Health Organization definition and 45%–80% for the Centers for Disease Control and Prevention case definition.⁴⁵ In the VHA, CHAR4 codes are updated regularly and mapped to individual clinics and there may be administrative errors in coding clinics¹³; a number of LCCs may not use the 2 CHAR4 codes we analyzed as these are restricted to clinics that provide only long COVID care, while, in some instances, clinicians may provide a facility's long COVID specialty care within their own general clinic schedule (eg, physiatry) instead of within a labeled LCC. The count of LCCs and number of veterans in LCCs may be under-counted. Future work to better categorize and/or more reliably ascertain LCC care will be important for future studies. Further, our analysis was limited to LCCs in the VHA and it is unclear to what extent veterans’ long COVID care needs were met in community (non-VHA) clinics; this is another important area for future research. Our study was limited in scope to veterans infected early in the pandemic and is unlikely to be representative of all veterans with COVID-19 in the VHA presently. Similarly, as SARS-CoV-2 infection was measured over 2 years, U09.9 documentation and LCC use may be more common among veterans infected earlier in the pandemic.

Conclusions

In this analysis of LCC use among veterans in the VHA, most veterans with a long COVID diagnostic code did not use LCCs and there were large regional variations in the use of these clinics. Given ongoing transmission of COVID-19, the lack of curative treatments for long COVID, and its long-term effects on quality of life, further resources, and interventions, to ensure needed access to LCCs may be necessary. Greater understanding about the accessibility and effectiveness of LCCs within and beyond the VHA and the opportunities to enhance resources to address the needs of those with long COVID are critically needed.

Supplementary material

Supplementary material is available at Health Affairs Scholar online.

Funding

The study was supported by the grants C19 21-278, C19 21-279, and RCS 10-391 from the VA Health Services Research and Development. The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. All statements in this article, including its findings and conclusions, are solely those of the authors and do not necessarily represent the views of the US Department of Veterans Affairs or the US government.

Data availability

The United States Department of Veterans Affairs (VA) places legal restrictions on access to veteran’s health care data, including identifying and sensitive patient data. The analytic data used in this study are not permitted to leave the VA firewall without a Data Use Agreement. This limitation is consistent with other studies based on VA data. VA data are made freely available to researchers behind the VA firewall with an approved VA study protocol.

Notes

Author notes