Given that the adjusted model now predicts hospitalization for COVID-19 as a risk factor for subsequent diagnosis of PCC, we conducted multivariable subgroup analyses on both non-hospitalized and hospitalized veterans (see Fig. 4). In terms of comorbidities, while liver disease, hypertension, and diabetes no longer exhibited significance for increased risk of PCC in the non-hospitalized group, CKD and COPD maintained significant associations with PCC. Likewise, within the hospitalized cohort, only COPD showed a significant 21% increased relative odds for PCC diagnosis. Notably, veterans who were unvaccinated during the COVID-19 illness stage, older adults, those residing in the Southwest U.S., and those with higher BMI had increased relative odds within the hospitalized cohort when compared to those who did not. Black veterans' adjusted odd ratios revealed that they sustained their relatively protective risk association with PCC in both subgroups.
As stated in the introduction, this study tested the hypothesis that hospitalization for the COVID-19 and pre-existing comorbidities would increase the odds of PCC. The data supported the hypothesis, showing hospitalization and pre-existing comorbidities increased the relative odds of PCC diagnosis. Interestingly, the unvaccinated veterans showed significantly increased relative risk for PCC diagnosis. Also, multiple COVID-19 episodes showed a reduced risk for PCC diagnosis in both non-hospitalized and hospitalized veterans, and it is in agreement with literature. Among the veteran population receiving care through the nationwide VA hospital systems, subsequent to the introduction of the ICD-10 U09.9 diagnosis code, the monthly prevalence of PCC diagnosis exhibited a range from 2.0 to 4.5 % (at the commencement and culmination of the study period, respectively). The incidence proportion varied from 2.41 to 0.05%. These proportions met the rare disease assumption; thus, the reported ORs were interpreted as relative risks. On average, PCC diagnosis was documented 73 days, 34 (IQR, 14-77) median days, after the SARS-CoV-2 infection index day, supporting published VA data.
It is worth noting that many symptoms, such as fatigue, deteriorated quality of life, dyspnea, joint pain, chest discomfort, cough, skin rashes, palpitations, headaches, diarrhea, and sensations like 'pins and needles,' along with anxiety, depression, and post-traumatic stress disorder, are reported with PCC. Although the prevalence and incidence of each individual symptom among veterans diagnosed with PCC may vary, we did not attempt to calculate them individually. The veterans' data from EHR indicated that, overall, a lower prevalence of PCC diagnosis (5.3%) compared to what has been reported by the CDC (6.0-11%) and other studies. These differences were not to be expected within the veteran population, given their free access to the healthcare system. These rate differences could also be attributed to the study design and reporting bias. Study designs that prospectively follow patients or use surveys can be expected to measure higher incidence and prevalence than patient medical records that rely on the ICD-10 code U09.9, as with this study design.
Veterans, relative to non-veterans, tend to have multiple chronic conditions. Nittas and colleagues reported prevalence estimates ranging from 7.5 to 41.0% based on an umbrella review and a targeted evidence synthesis of 102 studies. Another meta-analysis reviewing 63 studies noted high heterogeneity in prevalence reporting across different follow-up periods. Raveendran et al. reported a higher prevalence of approximately 87% among hospitalized patients for residual symptoms compared to those treated for COVID-19 on an outpatient basis. Although we have yet to attempt to differentiate between prevalence among hospitalized and non-hospitalized cohorts, these differences may be attributed to the complexity and lack of clarity in diagnosing and reporting the PCC diagnosis. The variability in prevalence data can be attributed to differences in study periods (pre- or post- ICD-10 code availability), symptom definitions (CDC vs. WHO), study designs, and reporting practices. Since the introduction of the ICD-10 code for PCC diagnosis, our data (Fig. 2) clearly indicated that from April 2022, the prevalence has stabilized at around 4.5% with new incidence rates of 0.5%, suggesting increasing clarity with reporting of the ICD-10 code and/or veterans did not seek help unless it was physically debilitating.
Most importantly, our data also aligned with prior research indicating that veterans who were hospitalized for acute COVID-19 exhibit a relative 41% increased risk of having a PCC diagnosis on their health record compared to non-hospitalized veterans. The PCC may have been driven by extended tissue damage during the severe infection period, affecting organs such as the pulmonary, muscular, neural, and cardiac, as well as pathological systemic inflammation. Recent articles suggested that there were differences in exaggerated humoral responses relative to matched controls in a cross-sectional study. Again, our data suggest that COVID-19 severity may be related to the subsequent presentation of PCC.
The presented data demonstrated significantly increased relative risk (p < 0.05) associations with CKD and COPD, and PCC diagnosis in non-hospitalized veterans. However, within the hospitalized cohort, only COPD was associated with increased risk (p < 0.001). As found in the literature, the presence of diabetes is expected to influence the exacerbation of persisting residual symptoms of COVID-19 via various pathophysiological mechanisms, but our analysis did not find an association with PCC diagnosis. Moreover, it has been reported in a prospective study that included 108 patients with Type 2 diabetes who had COVID-19, as compared to those without, had significantly more fatigue after the acute illness, which may be inadequately powered compared to the data presented here. It is also worth mentioning that the above study used subjective scores such as patient-reported assessment of their fatigue, and their statistical analysis was also limited to reporting group differences.
A few studies have consistently suggested that PCC diagnosis is more prevalent in females than their male counterparts. Our data corroborates this trend, indicating that non-hospitalized female veterans had a relative 24% higher odds of having a PCC diagnosis compared to male veterans, which was slightly lower (12%) in the hospitalized group; though it was insignificant (p = 0.05). While many studies have reported a two- to four-fold increased risk for female patients, our findings suggest a more modest increased risk. This discrepancy could potentially be attributed to the sample bias inherent to the veteran dataset, which predominantly consists of males (approximately 85%; Table 1). Although gender disparities have been evident in COVID-19 data, with males exhibiting higher mortality rates during COVID-19, the existing literature has not provided a definitive explanation for why PCC diagnosis disproportionately affects females. This disparity may be linked to the inherent immunological differences between sexes or hormonal factors. It is established that females generally mount a stronger immune response to infections, which could result in a prolonged immune reaction in some cases, potentially elevating the risk of receiving PCC diagnosis. Additionally, hormonal variations between males and females, including the presence of estrogen and progesterone, can influence immune responses and inflammatory processes. These hormonal distinctions may contribute to the development of PCC diagnosis, necessitating further research to elucidate their precise role in this phenomenon.
While our adjusted odds ratio indicates a relative 35-85% increased risk among individuals aged 41 and above (Fig. 3, Right), the broader literature presents conflicting findings regarding the association between age and PCC diagnosis. For instance, Notarte and colleagues analyzed 38 studies and concluded that age is not linked to the presentation of PCC, while another systematic review identified older age as a risk factor. Interestingly, within the subset of literature that employed U09.9 chart records to identify PCC cases, adjusted odds ratios did demonstrate an association between age and PCC diagnosis, supporting our finding. Farmer et al. also found that, in general, veterans who seek health care within the VHA system are generally older and sicker than other veterans who received healthcare elsewhere, possibly contributing to the disparities in risk reporting.
Our adjusted overall odds ratios (Fig. 3) revealed a relative 8 and 24% increased risks for PCC diagnosis among obese and morbidly obese veterans, respectively. This is in agreement with existing research findings from several sources. It is important to underscore that both obesity and veterans' overall health have been established as substantial risk factors for the severity of COVID-19 and the subsequent diagnosis of PCC, thus validating our data. This association could be attributed to the persistent inflammation linked to obesity and the modified immune responses during COVID-19, which contribute to severe illness and serve as risk factors for the emergence of long-term COVID-19 complications (PCC). Consistent with prior research, our veteran-focused data revealed a lower risk of having a PCC diagnosis in the Black race (OR 0.73 for non-hospitalized and OR 0.77 for hospitalized veterans) compared to White counterparts, indicating a degree of protection. This may also be indicative of the underrepresentation of minority races within the study cohort, suggesting underdiagnosis of this group of individuals. When comparing non-Hispanics, there was an elevated relative risk of having a PCC diagnosis in the Hispanic/Latino ethnicity (refer to Fig. 3) in both non-hospitalized and hospitalized cohort. Findings from another veterans' study, which employed an ICD-10 code to identify PCC cases, corroborated our results. Of particular significance, there was an over 3-fold increase in the odds of PCC diagnoses among veterans in the Southwestern U.S. region, where most of the population identified as of Hispanic origin. This observation highlights a disproportionate prevalence of PCC diagnosis within the Hispanic community, further supporting CDC data.
This study comes with certain limitations. First, our data only captures diagnoses documented in the electronic medical records. Consequently, PCC may be underdiagnosed among individuals who were less likely to seek medical care for COVID-19 or PCC either due to mild disease presentation or non-clinical factors such as distance to a healthcare facility. Thus, prevalence and incidence rates may have been underreported. Additionally, the methods used in this manuscript to calculate the incidence and prevalence use the assumption that patients do not recover from PCC. Given that this study is retrospective in nature, it is advisable to conduct a prospective study to comprehensively understand the associations between hospitalization and treatment with the diagnosis of PCC. Furthermore, to investigate the potential association between vaccination and PCC diagnosis, it is essential to confirm these findings through a prospective controlled or propensity-matched study, especially since vaccination provides protection for a limited duration. Additionally, considering that long-term COVID symptoms are diverse and have not been explored, it may be valuable to survey individual symptoms as well, which was not investigated in this study.