How common and what features long-term symptoms after COVID-19 in recovered patients have

CASPER, a cross-sectional two-stage cluster sampling methodology, was utilized to collect the data that underwent secondary analysis. CASPER is a validated method developed for rapid needs assessment by the Centers for Disease Control and Prevention (CDC) to rapidly obtain information about the needs of a community (20). While it was originally designed for emergency management disaster settings, there are opportunities for CASPER to influence public health in non-disaster settings as CASPER has been previously used to estimate community needs and assess public health perceptions. In using data from the CASPER methodology in this study, household-based population estimates are determined to ensure the sample is generalizable to the larger population. By design, CASPER instruments collect cross-sectional data about the entire household with data obtained from an individual household member. All questions analyzed were asked at the household level.

The CASPER was conducted in the Third Ward in Houston, Texas. The Third Ward has a resident population of 38, 920 and has seen a population increase of 1.49 percent since 2020. The population by race is Black/African American (38.43%), White (30.75%), American Indian (0.56%), Asian (14.01%), Native Hawaiian/Pacific Islander (0.05%), Other (7.46%) and/or Mixed (8.74%). The median household income is $59,026 with 18.72% of families living below the poverty line ($23,400) for a family of four. One hundred sixty-eight unique households completed the survey instrument, representing a weighted population size of 14, 769 total households. Exploring the baseline health in Third Ward was a cornerstone of this study and was needed to explore associations between comorbid conditions and long COVID. The prevalence of disease in Third Ward prior to COVID-19 was higher than national rates for many of the health indicators selected in this study.

The data used for secondary analysis was collected by an instrument that included information regarding household demographics (age, income, education, race/ethnicity, employment status), prevalence of chronic conditions (heart disease. Stroke, asthma, cancer, cholesterol, obesity, diabetes, depression, anxiety, stroke, COPD, and chronic kidney disease) and post-COVID impact (history of hospitalization, length of time since diagnosis, symptom severity, presence of persistent symptoms, outpatient visits, impact on activities of daily living).

One hundred sixty-eight unique households completed the interviewer-administered survey. Each household for which an interview was completed was assigned a weight based on the household's probability of selection. This weighting ensured that the resulting estimates were generalizable to all household in the sampling frame, with each cluster weighted equally. SAS 9.4 was utilized for the analysis allowing for appropriate multistage sample design weighting. The following weight formula was used: Weighted frequencies, percentages, and 95% confidence intervals were then calculated for each of the interview questions for cells with five or more observations. A total of 14, 769 households were included in the 77,004-sampling frame and used in the weight calculation. Once weight was assigned, frequencies and corresponding percentages were calculated for each question.

Descriptive assessments were done among the full cohort, but research questions (inferential statistics) were only conducted and among weighted households (N = 7,088) with members who tested positive for COVID or had symptoms. Participants who reported not having any COVID symptoms were then further excluded from the inferential analyses, bringing the study cohort for the associations between risk factors and long COVID to 6,418 weighted households.

Bivariate (crude) and multivariable logistic regression models were constructed to determine associations between risk factors and long COVID. Five separate logistic regression models were constructed with unadjusted and adjusted odds ratios (OR) and 95% confidence intervals (CI) reported for each risk factor, where applicable. The study outcome was a binary measurement of the development of post-acute sequalae of COVID-19. The variable long COVID was created together by a combination of those experiencing symptoms at least 30 days beyond infection and having been diagnosed at least 12 weeks (3 months) prior to the survey. Odds ratios were obtained, and confidence intervals were used to determine whether statistical significance had been reached.

Due to limited data, some variables (household income and educational level) required recoding. In the original survey instrument, respondents were asked to "check all that apply" to indicate the presence of COVID-19 symptoms. For analysis, each symptom variable was separated, and a binary yes/no response was created.

Potential risk factors for COVID-19 were identified using existing literature, clinical expertise and the Charlson Comorbidity Index, supplemented by additional comorbidities documented in prior studies (21). Research questions, variable measurement levels, and corresponding statistical analyses are summarized in Table 3.

The final descriptive analysis includes 168 households in Third Ward representing 14,769 weighted households. Weighted frequencies, percentages, and 95% confidence intervals were then calculated for each of the interview questions for cells with five or more observations. Of the participants, 57.78% (n = 8,432) of households identified as Non-Hispanic Black/African American and 16. 78% (n = 2,449) identified as Hispanic. The majority of households identified as employed (42.45%, n = 6,076), had a household income of $75,000 or more (29.70%, n = 3,999), and had at least one individual in the home who had attended college (21.94%, n = 9,938). Most respondents identified the members of their household as representing the 18–64 age group (73.35%, n = 10,833). Complete descriptive characteristics of the surveyed population are shown in Table 4.

Approximately 70 % of households (78.01%, n = 11,233) of households identified as not having health insurance coverage and cited a doctor's office as the place visited most often to see a doctor (52.81%, n = 7,707). 83.86%, n = 12,011 of households indicated no difficulty in getting medical services in the prior 12 months. Of those who cited difficulty (16.14%, n = 2,312), lack of insurance represented the greatest barrier to care (39.33%, n = 950). The clinical characteristics of the study population are shown in Figure 2 and Table 5.

Approximately 81.5% of households rated their health as excellent (24.74%, N = 3,082), very good (25.54%, N = 3,181), or good (31.24%, N = 3,892). Figure 3 presents the weighted distribution of self-rated personal and household health status among surveyed households. High blood pressure (36.79%, n = 5,321), high cholesterol (25.92%, n = 3,701), poor mental health (19.18%, n = 2,740), obesity (18.79%, n = 2,699) and diabetes/high blood sugar (17.53%, n = 2,511) were cited as the more prevalent health conditions. Weighted and unweighted frequencies of health status/well-being for households can be found in Table 6.

Approximately half of households (48.87%, n = 7,088) reported having tested positive for COVID-19 with the majority of households (76.46%, n = 5,346) reporting it had been 12 months or more since their diagnosis (Figure 4; Table 7). From Figure 4, most respondents had been diagnosed over 12 months ago, suggesting long-term persistence of symptoms and relevance to long COVID. 18.06%, n = 1,134 reported long COVID (still experiencing symptoms beyond 30 days after their initial COVID-19 infection). The majority of households reported moderate symptoms (50.77%, n = 3,505) with 12.05% (n = 832) of households indicating symptoms to be severe. Systemic symptoms including tiredness or fatigue, joint or muscle pain, taste or smell changes, headache, brain fog, and shortness of breath were the most frequently experienced household symptoms anytime during infection.

The most persistent symptoms lasting 30 days or more included anxiety, brain fog, depression, joint or muscle pain, shortness of breath, exercise inability, tiredness or fatigue, and weight loss. To better illustrate the comparative burden of symptoms during acute infection and their persistence into the post-acute period, Figure 5 presents the weighted percentage of households reporting each symptom both during infection and 30 days or more afterward. This visualization reveals that while symptoms like fatigue (37.4%), joint or muscle pain (27.4%), and changes in taste or smell (20.5%) were highly prevalent during the acute phase, their persistence into the long COVID period declined substantially. In contrast, symptoms such as anxiety, brain fog, and depression, though less prevalent during acute illness, demonstrated notable persistence over time.

Eight hundred and forty (12.23%) households indicated at least one household member was hospitalized for COVID with 20.27% (n = 1,399) reporting at least one outpatient clinic visit after COVID recovery. 1,134 (18.06%) households reported long COVID (Table 7). We also explored activities of daily living impacted by COVID-19 symptoms, but given these characteristics fall outside of the scope of work comprehensive data was not reported. However, shopping, meal preparation, and household chores were cited to be among the activities of daily living most impacted.

This study sought to explore associations between select demographic, clinical, epidemiological, and non-clinical factors and long COVID. Both bivariate (crude) and multivariable logistic regression models were employed to determine these associations. The study outcome was a binary measurement of the development of post-acute sequalae of COVID-19. The variable long COVID was created together by a combination of those experiencing symptoms at least 30 days beyond infection and having been diagnosed at least 12 weeks (3 months) prior to the survey in accordance with the CDC definition for long COVID. Odds ratios (ORs) with corresponding confidence intervals were calculated to assess statistical significance. The complete results of the unadjusted and adjusted logistic regression models predicting long COVID are presented in Tables 8–10.

After adjusting for age; age and race; and age, race, and income, individuals who experienced severe acute COVID-19 symptoms had nearly 30 times higher odds of long COVID (OR = 29.58; 95% CI: 1.38–632.53) compared to those with mild symptoms. Even moderate symptoms were associated with elevated odds (OR = 2.64) (Figure 6). This strong association suggests that severe acute illness is a notable predictor of long COVID. In contrast, no statistically significant association was observed between mild symptoms and the development of long COVID.

After adjustment for covariates, individuals who reported heart disease had significantly higher odds of developing long COVID compared to those without heart disease (OR = 6.00, 95% CI [1.15, 31.28]). Although asthma was significantly associated with long COVID in the unadjusted (crude) model, the association was attenuated and no longer statistically significant after adjustment (OR = 3.49, 95% CI [1.05, 11.59]). Poor physical health remained a significant predictor of long COVID after adjusting for age and race, with affected individuals exhibiting 4.20 times higher odds of developing the condition compared to those reporting good physical health (OR = 4.20, 95% CI [1.12, 15.75]). These findings underscore the importance of preexisting cardiovascular and general health conditions in the risk profile for long COVID.

The association between symptoms experienced at any time during infection and symptoms experiences thirty days or more after infection and long COVID was also explored. This provided additional valuable insight into the conditions that may exacerbate the development of long COVID. Those who experienced anxiety at any time during their infection were 22.94 times higher odds developing long COVID after adjusting for age and race (OR = 22.94, 95% CI [2.01, 262.31]) and those who experienced chest pain at any time during their infection were at increased odds of developing long COVID after adjusting for age, race, income, and health insurance coverage (OR = 7.15, 95% CI [1.13, 45.23]) in comparison to those who did not experience those symptoms during infection. Those who experienced constipation at any time during their infection were 16.81 higher odds of developing long COVID in comparison to those who did not experience constipation (OR = 16.81, 95% CI [1.33, 213.23]). Those who experienced heart palpitations (OR = 6.59, 95% CI [1.08, 40.18]) and shortness of breath (OR = 4.97, 95% CI [1.16, 21.36]) were 6.59 times higher odds of developing long COVID after adjustments for age, race, income, and access to healthcare.

By contrast, no statistically significant associations were observed between long COVID and sociodemographic factors such as race, education, income, or access to healthcare in either the crude or adjusted models.

The relationship between severe acute COVID-19 and subsequent long COVID is affirmed by the findings. Individuals who experienced severe symptoms had nearly 30 times higher odds of developing long COVID compared to those with mild symptoms, consistent with prior research (1, 22). These results align with findings from the NIH RECOVER Initiative, which reported an increased risk of PASC among individuals with more severe initial illness, especially those requiring hospitalization or mechanical ventilation (22). Severe acute illness may trigger prolonged inflammatory responses or tissue damage, contributing to persistent symptoms.

The significant association of heart disease and asthma with long COVID adds to existing literature recognizing chronic comorbidities as risk enhancers (23, 24). In this study, participants with heart disease had six times the odds of developing long COVID compared to those without. Although asthma lost statistical significance after full adjustment, it was notable in crude models. This aligns with evidence from Hill et al. (22), who found chronic lung disease to be one of the strongest predictors of PASC. Moreover, the study found that poor physical health significantly increased the odds of long COVID, consistent with research by Mandal et al. (13) and Houben-Wilke et al. (25), who identified fatigue, breathlessness, and poor functional recovery as common post-COVID conditions.

The presence of specific acute symptoms, particularly anxiety, chest pain, heart palpitations, constipation, hypersomnia, and shortness of breath, demonstrated strong associations with long COVID. This mirrors the symptom clusters identified by the FAIR Health (10) report, which found that fatigue, breathing abnormalities, and persistent cough were among the most commonly co-occurring symptoms across age and gender groups. Our findings also corroborate those of McCorkell et al. (26), whose comprehensive patient-led study emphasized fatigue, cognitive dysfunction, and post-exertional malaise as dominant features of long COVID persisting for months. The high odds associated with hypersomnia and anxiety in this study further support calls for integrated physical and mental healthcare in managing long COVID.

Interestingly, while this study included a racially diverse sample with a majority of African American respondents, race and ethnicity were not significantly associated with long COVID. This finding diverges from some national-level studies, such as those by Jacobs et al. (27) and Hill et al. (22), which documented underdiagnosis of PASC in minority populations, particularly Black Americans. However, it supports Louie and Wu (28), who found that socioeconomic factors modulated the relationship between race and long COVID, with effects appearing buffered in lower-income communities. In our sample, the homogeneity of healthcare access limitations and high comorbidity prevalence in the Third Ward may have attenuated detectable racial differences. Further research is warranted to explore how structural inequities and community-level factors influence long COVID trajectories.

Another important finding is the absence of significant associations between long COVID and educational attainment, income, and healthcare access. While surprising, this may reflect limitations in the survey design or the complexity of long COVID, where individual biological and clinical experiences outweigh traditional socio-demographic predictors. Notably, over 78% of households in the study lacked health insurance, and 16% reported difficulty accessing care, often due to financial barriers. These structural constraints may have affected both the reporting and experience of long COVID, contributing to potential underdiagnosis or delayed treatment.

The role of vaccination was not directly addressed in this study, which is a limitation. Our study occurred at a time when testing and vaccination coverage varied widely, thus the instrument utilized did not capture vaccine status. Research from Antonelli et al. (29) demonstrated that vaccinated individuals had reduced odds of developing long COVID, especially after the second dose, and experienced fewer symptoms if infected. Given that most infections reported by participants occurred 12 months or more before the survey, many may have been exposed before widespread vaccine availability. Additionally, inability to clinically verify vaccination and immunization data contributed to its exclusion from the survey instrument. This limits the ability to contextualize the findings within the shifting epidemiological landscape, including the rise of variants like Delta and Omicron, which have shown different patterns of transmissibility and severity.

Vaccination against SARS-CoV-2 has been associated with a reduction in the risk of developing long COVID, although the magnitude of this effect varies depending on the study design and operational definitions used. Several studies suggest a protective role, but results are nuanced. Antonelli et al. (29) based on app-reported symptoms, observed that individuals who received two vaccine doses had approximately half the risk of experiencing symptoms persisting beyond 4 weeks following infection. However, this study's reliance on self-reported data and a relatively short follow-up period may limit generalizability to definitions of long COVID that require longer symptom duration.

Other studies using broader or more clinically validated definitions have reported stronger effects. For instance, a population-based cohort analysis by Krishna et al. (30) found a 79% reduction in hospital admission for post-acute sequelae 6 months after infection among vaccinated individuals in the United Kingdom. Similarly, Ayoubkhani et al. (31) analyzed data from a UK community-based cohort and found that vaccination after SARS-CoV-2 infection was associated with a 12.8% reduction after the first dose and an 8.8% reduction after the second dose. Population-level registry studies have also supported a protective association. For example, Ayoubkhani et al. (32) found that individuals vaccinated with two doses had signifigantly lower odds of reporting long COVID. Conversely, other large-scale electronic health record (EHR) studies, such as Al-Aly et al. (33) found only modest reductions in long COVID risk among vaccinated individuals. This discrepancy may be due in part to the older population studied, the inclusion of pre-existing comorbidities such as cardiovascular disease, and broader outcome definitions encompassing systemic post-viral sequelae.

Given the heterogeneity in study findings, our study acknowledges the importance of incorporating vaccination status into long COVID assessments. Although our data did not initially include this variable, its significance as a potential confounder or effect modifier is noted, and we recommend that future studies incorporate vaccination history. Despite the lack of vaccination data, the timing of infections in the study provides valuable insights. Most participants experienced symptoms more than a year before the survey, indicating that long COVID persisted over time. The transition from early variants to Delta and Omicron introduced changes in symptom profiles, with Omicron associated with higher transmissibility but lower severity (29). However, this study could not stratify findings by variant or time since infection, underscoring the need for more granular data on timing and viral strain in future studies to enhance data accuracy and interpretability of findings.

The multisystemic nature of symptoms reported—ranging from neurological and gastrointestinal to cardiovascular—confirms the wide-ranging impact of long COVID observed in other studies (1, 34). For instance, this study's findings of persistent constipation and chest pain reflect gastrointestinal and cardiac involvement, respectively, while symptoms such as brain fog, hypersomnia, and anxiety indicate neurological and psychological dimensions. Studies by Blackett et al. (35) and Xu et al. (36) reinforce this, highlighting gastrointestinal and neurological sequelae as part of the long COVID syndrome. These varied symptom clusters necessitate multidisciplinary management approaches that include mental health, cardiology, pulmonology, and rehabilitation services.

Psychological symptoms, particularly anxiety and depression, also emerged as prevalent among respondents. This aligns with findings from Houben-Wilke et al. (25), who reported high levels of anxiety, depression, and PTSD symptoms in patients with persistent complaints at 3 and 6 months. In the current study, anxiety showed one of the strongest associations with long COVID, indicating a need to integrate behavioral health services into post-COVID care models. Similarly, the findings resonate with Ladds et al. (12), whose qualitative research underscored the emotional toll and difficulties patients faced navigating healthcare services during recovery.

Notably, gastrointestinal symptoms, such as constipation, were significantly associated with long COVID, confirming emerging literature that documents persistent GI symptoms even months after infection (35). While often overlooked in clinical assessments, GI manifestations can severely affect quality of life and may be linked to viral persistence or gut microbiome disruption (34). These findings advocate for including gastrointestinal evaluation in long COVID protocols.

The study highlights a complex interplay of acute symptom severity, comorbidities, and individual symptom experiences in predicting long COVID. The absence of racial, economic, and educational associations does not negate the role of structural inequities but rather points to the dominant influence of biological and clinical risk factors in this specific population. However, systemic inequities may still modulate who gets diagnosed, treated, and included in research. The lack of significant findings for diabetes contrasts with studies such as Assad et al. (23) and Fernández-de-las-Peñas et al. (37), which found associations between diabetes and long COVID. The absence of association in our study may be due to limitations in sample size, lack of detailed glycemic control data, or variability in self-reporting. Furthermore, the comorbidity interplay, such as between hypertension and diabetes, may obscure individual associations without detailed stratification or interaction modeling.

These findings have several public health implications. First, they underscore the need for symptom-based screening tools that prioritize clinical severity and comorbid history over demographic profiling. Second, they call for community-based education and support services, especially in historically underserved neighborhoods like the Third Ward, where distrust of the healthcare system may delay care-seeking (38). Third, they emphasize the urgency of integrated care pathways that can address the multisystem nature of long COVID.

Despite its strengths, this study has limitations that warrant consideration. Data were based on self-reported surveys, introducing recall bias and potential misclassification of respondent. The reliance on one household member to report for the entire household may lead to inaccuracies in symptom recall or comorbidity reporting. Additionally, the cross-sectional design precludes causality assessments and limits understanding of symptom progression. With respect to recovery data, our survey did not include longitudinal follow-up or specific questions regarding symptom resolution or duration. Participants were asked to report their current and past symptoms at the time of the survey, but without a time-course element, we were unable to systematically assess recovery trajectories. We plan to incorporate these elements into future studies. The absence of vaccine status, testing date, and viral strain data further restricts interpretations. Moreover, missing data and limited sample size may have underpowered the detection of some associations or contributed to unstable odds ratios. While the use of weighted data and cluster-based sampling enhances generalizability, the relatively small number of directly surveyed households (N = 168) may limit statistical precision, as reflected in some wide confidence intervals. Larger studies are needed to validate and strengthen these estimates. We also acknowledge that the lack of a comparison group limits the interpretability of symptom frequency and severity. Future studies should consider including a comparison group to enhance contextual understanding of the findings. Lastly, although the survey instrument included open-ended items and general questions about the impact of COVID-19 on daily life, a validated scale for functional assessment was not utilized. Future work should incorporate standardized instruments to more accurately assess functional outcomes. Stratified analyses by sex, age, and time since infection are also recommended to improve comparability and analytic precision.

Nevertheless, this study contributes valuable insights to the evolving understanding of long COVID, particularly within a racially diverse and underrepresented urban population. Future research should adopt longitudinal designs, integrate clinical validation of symptoms, and examine variant-specific effects, vaccine influence, and the impact of evolving social determinants. Such work is essential for informing equitable, effective, and patient-centered responses to the long-term aftermath of COVID-19.