Quality of Life and Daily Function in People with Long COVID Compared by Whether They Received Rehabilitation

Despite the rapid growth of the PCC literature, several gaps remain clinically important.

These gaps have consequences. Large population cohorts show substantial decrements in HRQoL on generic instruments (EQ-5D-5L, SF-36), with quantifiable utility losses now documented in big-data resources (e.g., Open-PROMPT/OpenSAFELY), underscoring the need for scalable, real-world solutions [27].

In the context of post-COVID-19 condition, rehabilitation is conceptualised as a set of structured interventions delivered or supervised by health professionals (e.g., physiotherapists, physicians, occupational therapists, psychologists) to address limitations in physical, cognitive and psychosocial functioning. These may include supervised exercise or activity management, respiratory and musculoskeletal physiotherapy, education about pacing and energy conservation, cognitive strategies to address attention and memory problems, and psychological or behavioural interventions targeting anxiety, low mood and adjustment difficulties. In routine clinical practice, such programmes are heterogeneous in content, format (inpatient, outpatient or community-based) and intensity, and may be complemented by elements of home-based self-management.

In light of these gaps between protocolised clinical trials and heterogeneous real-world rehabilitation, the present study aimed to address three questions. First, we sought to describe health-related quality of life and functional status, as measured by SF-36 domain and summary scores and the Post-COVID Functional Status (PCFS) scale, in a cross-sectional sample of adults with post-COVID-19 condition recruited from two national online support communities. Second, we examined whether self-reported completion of rehabilitation was associated with differences in these outcomes. Third, we explored which patient factors (including age, sex, symptom duration, comorbidities and other covariates) were associated with health-related quality of life and functional status in multivariable models. Given the cross-sectional design and the non-standardised nature of rehabilitation programmes, these analyses were intended to explore associations and generate hypotheses for future prospective evaluations, rather than to infer causal treatment effects.

Recruitment was conducted between 15 April and 15 May 2024, via two private, moderator-supervised Facebook support communities for individuals with ongoing post-COVID-19 symptoms—"long COVID a postcovidový syndrom—ČESKO A SLOVENSKO" (≈4500 members) and "COVID následné potíže" (≈2700 members)—where access requires prospective members to confirm a history of SARS-CoV-2 infection and persistent health complaints. An anonymous online survey was posted as a pinned announcement in each group. From a combined membership of approximately 7200, a total of 406 individuals (308 women, 98 men) completed the survey (≈5.6% response). Participants were recruited exclusively from these two Facebook support communities for individuals with ongoing post-COVID-19 complications. No additional recruitment channels (e.g., hospital clinics, primary care practices) were used, and the resulting sample is therefore best described as a non-probability convenience sample of members of these online support communities.

In addition to general data (gender, age, weight, and height), patients completed the Patient Self-Report Methods for the Post-COVID-19 Functional Status Scale (PCFS) and a structured Quality of Life questionnaire (SF-36). The respondents also answered the question of whether they had completed any form of rehabilitation, which was the key information used to divide the obtained questionnaires into two groups: those that had completed rehabilitation (n = 182) and those that had not (n = 224).

The inclusion criteria were men and women aged ≥18 years with persistent symptoms following COVID-19, lasting at least four weeks after recovery from the last episode of COVID-19. This long-lasting threshold captures individuals with post-acute and longer-term post-COVID-19 symptoms. As shown in Table 1, the majority of respondents reported symptom duration of ≥4 months, but approximately one-fifth had symptoms lasting 1–3 months at the time of the survey. The persistence of symptoms was assessed using specific questions in the questionnaire in the appendix. The full versions of all study-specific questionnaire items used in this study are provided in Supplementary File S1. Standardised instruments (SF-36 and the Post-COVID Functional Status Scale) are not reproduced in full due to copyright restrictions.

As shown in Table 1, over 80% of respondents reported persistent symptoms for at least four months (4–6, 7–12 or >12 months), whereas around 20% were in the 1–3-month range at the time of the survey. The sample therefore represents a symptomatic post-COVID-19 group with predominantly long-lasting symptoms rather than an exclusively ≥12-week 'long-COVID' population.

The exclusion criteria included recovery from COVID-19 without any persistent or long-term respiratory, cardiovascular, neurological, haematological, psychological, or infectious symptoms associated with the disease; presence of difficulties listed in the questionnaire that were already experienced before contracting COVID-19; and lack of willingness to cooperate or refusal to consent to participate in the research study. The study protocol was approved by the Ethics Commission of the University of Prešov in Prešov under the number 04/2023.

Participants were asked whether they had completed any rehabilitation programme specifically intended to address their post-COVID-19 symptoms, broadly defined to include physiotherapy-based programmes, multidisciplinary rehabilitation, or other structured interventions prescribed or supervised by health professionals. For the purposes of the present analysis, rehabilitation was coded as a binary variable (completion of any such programme: yes/no), and we did not standardise or stratify by specific modality, setting or intensity.

Continuous outcomes (SF-36 domain scores and PCS/MCS) were summarised as mean ± SD or median (IQR), and categorical variables as counts (%). Age was compared between groups using Welch's t-test. Other continuous or ordinal outcomes were compared using the Mann–Whitney U test. Where applicable, categorical comparisons used the χ² test (or Fisher's exact test when expected counts were small). To control family-wise error, Bonferroni correction was applied in two contexts. For between-group comparisons of SF-36 domains and component summaries, raw p-values were adjusted within the family of eight SF-36 domains, with PCS and MCS considered separately. For the multivariable regression analyses, we controlled the family-wise error rate using a global Bonferroni correction across 37 prespecified predictor–outcome tests. For each predictor–outcome pair we calculated a Bonferroni-adjusted p-value (p_adj = min [p × 37, 1.00]; two-sided) and reported it alongside the corresponding standardised β coefficient in a separate table. Effect sizes are presented as rank-biserial correlation for Mann–Whitney tests and standardised β coefficients for multivariable models. For Mann–Whitney tests, effect size was expressed as the rank-biserial correlation r, oriented in a benefit-directed manner such that r > 0 indicates better outcomes in the rehabilitation group (including lower PCFS grades).

Multiple linear regression was used to examine associations between covariates (age, sex, symptom duration, comorbidities, prior COVID-19 episodes, history of hospitalisation, BMI, physical workload, physical activity, rehabilitation status) and SF-36 outcomes (domains and PCS/MCS). Two-sided tests with α = 0.05 were applied. Available-case analyses were performed (sample sizes per analysis reported in the Results). Analyses were conducted in JASP, version 0.19.3.

Given the absence of robust prior data on the prevalence of persistent post-COVID-19 symptoms in our target communities, we used a standard single-proportion formula with finite-population correction as a rough benchmark for the number of respondents required to estimate proportions with reasonable precision. Assuming a 95% confidence level (Z = 1.96), a 5% margin of error (e = 0.05), an expected proportion p = 0.50 (conservative), and using the Slovak resident population (n = 5,424,687) as a reference, the uncorrected size was n₀ = 384.16 and the corrected size n ≈ 384. We therefore aimed to recruit ≥385 respondents. However, recruitment was conducted via two online support groups and thus represents a non-probability convenience sample, including both Slovak and Czech residents. The sample size calculation should therefore be interpreted as illustrative rather than as defining a formal sampling frame or inferential precision for the general population. Calculations were verified with DataStatPro, version 1.1.7 Beta (DataStatPro Team, 2024; accessed May 2024).

The most commonly reported symptoms included fatigue (77.1%), muscle pain (68.6%), headache (47.9%), sleep disturbances (42.9%) and joint pain (42.9%). Respondents could indicate more than one symptom; less frequent complaints included hypersensitivity to sound, visual disturbances and so-called "COVID fingers" (Figure 1).

In the next step (Table 3), we compared quality-of-life and functional outcomes between respondents who reported rehabilitation and those who did not.

Effect size is the rank-biserial correlation (r) reported in a benefit-oriented direction: r > 0 indicates better outcomes in the rehabilitation group. For SF-36 domains (higher = better), positive r corresponds to higher scores in the rehabilitation group. For PCFS (higher = worse), we reversed the direction so that positive r corresponds to lower PCFS (less limitation) in the rehabilitation group. Magnitude: ~0.10 small, ~0.30 medium, ≥0.50 large. All p-values shown are unadjusted (raw, two-tailed); family-wise error for SF-36 domains and component summaries was controlled using Bonferroni correction as described in Section 2.3.

All indicators except Physical Functioning differed significantly between groups (p < 0.001). Effects favoured the rehabilitation group, with the largest differences for Vitality (r = 0.743) and Mental Health (r = 0.572), whereas Physical Functioning showed only a borderline difference (p = 0.052). Positive r denotes better outcomes in the rehabilitation group for all measures, including lower functional limitation on PCFS.

These between-group differences are illustrated in Figure 2, which shows consistently higher mean SF-36 domain and summary scores, as well as better perceived health, in respondents who completed rehabilitation compared with those who did not. The largest gaps are visible for Vitality, Mental Health and the Mental Component Summary, indicating clinically meaningful advantages in energy, mood and psychosocial functioning among participants who reported rehabilitation.

Participants who reported completing rehabilitation differed from those who did not across multiple health-related parameters. In multivariable linear regression, completion of rehabilitation was positively associated with Vitality (β = 0.645, p < 0.001) and the Mental Component Summary (β = 0.475, p < 0.001). Additional positive associations were observed for General Health (β = 0.285, p < 0.001), Mental Health (β = 0.486, p < 0.001) and Perceived Health Status (β = 0.406, p < 0.001). Completion of rehabilitation was also positively associated with the Physical Component Summary (β = 0.285, p < 0.001) and with Bodily Pain (β = 0.274, p < 0.001). Significant associations were likewise noted for Role Limitations due to Physical Health (β = 0.202, p < 0.001), Social Functioning (β = 0.302, p < 0.001) and Role Limitations due to Emotional Problems (β = 0.261, p < 0.001). Finally, rehabilitation status was inversely associated with Post-COVID Functional Status (PCFS) scores (β = −0.246, p < 0.001), consistent with lower functional limitation among those who completed rehabilitation. Because multiple linear regression analyses were conducted, a Bonferroni correction was applied to account for multiple comparisons and to minimise the risk of Type I error, yielding a more conservative interpretation of statistical significance.

Bonferroni-adjusted p-values were computed using a global correction across 37 prespecified predictor–outcome tests (p_adj = min [p × 37, 1.00]; two-sided tests, coefficient degrees of freedom ≈ 396). For each predictor–outcome pair, Table 5 reports the corresponding standardised regression coefficient, raw p-value and Bonferroni-adjusted p-value. Where raw p-values were extremely small, we either report the numerical value (to four decimal places) or indicate p < 0.001; adjusted p-values were obtained by multiplying the raw p by 37 and truncating at 1.00. Associations with p_adj < 0.05 were considered statistically significant after correction for multiple testing.

After adjustment for multiple comparisons, completion of rehabilitation remained significantly associated with better scores in several SF-36 domains, including Role Limitations due to Physical Health, Bodily Pain, General Health, Vitality, Social Functioning, Role Limitations due to Emotional Problems and the Mental Health domain, as well as with higher Physical and Mental Component Summary scores, more favourable Perceived Health Status and lower PCFS grades (all p_adj < 0.05). In contrast, the association between rehabilitation and Physical Functioning did not reach the adjusted significance threshold (p_adj = 0.074). Age and longer symptom duration also remained significant negative predictors for multiple physical and mental health outcomes (for example, both were associated with lower General Health, Social Functioning, PCS, MCS and Perceived Health), whereas sex remained significantly associated only with Physical Functioning and the Physical Component Summary, and the number of past COVID-19 infections did not retain statistical significance for any outcome after Bonferroni correction.

After Bonferroni correction for multiple testing, the association between rehabilitation and Physical Functioning did not meet the adjusted significance threshold, whereas the associations with several other domains, both component summaries, perceived health and PCFS remained statistically significant.

Figure 3 illustrates the distribution of PCFS grades by rehabilitation status. Participants who completed rehabilitation were more likely to report mild or no functional limitations (PCFS 0–1), whereas moderate to severe limitations (PCFS 3–4) were more frequent among those who did not complete rehabilitation, underscoring the clinical relevance of the observed differences in functional status.

Effect sizes were directionally consistent and clinically coherent: negative associations for age and symptom duration were modest (β in the approximate range −0.16 to −0.22), indicating somewhat poorer health status among older participants and those with more prolonged symptoms. In contrast, rehabilitation showed some of the largest positive associations (for example, β ≈ 0.65 for Vitality and β ≈ 0.30 for Social Functioning), indicating better patient-reported health among respondents who reported having completed rehabilitation. These findings should, however, be interpreted within the constraints of the cross-sectional design and the potential for residual confounding, and are best viewed as hypothesis-generating rather than as definitive evidence of causal treatment effects.

Respondents who had completed rehabilitation reported higher scores in most SF-36 domains and on both component summaries, with the largest differences between groups observed for Vitality and Mental Health, whereas Physical Functioning showed only a borderline difference. Functional status on the PCFS scale was also more favourable in the rehabilitation group. These patterns are broadly consistent with earlier work demonstrating that long COVID is associated with substantial impairments across physical, psychological and social dimensions of health [9,10,11,32]. In a regional study, Líška et al. reported that patients with long COVID had significantly lower SF-36 scores than healthy controls across all domains, with particularly marked impairments in physical functioning and role limitations due to physical health [32]. Our between-group comparisons by rehabilitation status extend this picture by suggesting that, within a symptomatic long-COVID population, there is measurable heterogeneity in HRQoL and functional limitation that co-varies with reported rehabilitation exposure.

The magnitude and pattern of the differences observed in our study are compatible with findings from structured rehabilitation programmes. Multidisciplinary programmes combining supervised exercise, education and psychological support have been associated with improvements in physical performance, dyspnoea, fatigue and mental health in cohorts of patients post-COVID-19 (with long COVID) [15,16,17,18,33]. Recent trials of brief, protocolised interventions, including a behaviourally informed outpatient programme and an online multimodal programme, have also reported clinically relevant gains in SF-36 Physical Functioning and overall HRQoL [22]. Unlike these interventional studies, however, our analysis is observational and relies on self-reported rehabilitation of heterogeneous content and intensity. The differences between groups may therefore reflect a combination of rehabilitation effects, selection into rehabilitation, baseline severity, health-seeking behaviour and other unmeasured contextual factors.

The particularly pronounced differences in Vitality and Mental Health are noteworthy given the central role of fatigue and psychological distress in long COVID. Fatigue is one of the most frequent and persistent symptoms reported after COVID-19, with high prevalence in both hospitalised and non-hospitalised cohorts [34,35]. Longitudinal research has identified fatigue as a key predictor of overall HRQoL, persistent symptoms and perceived stress one year after severe COVID-19 [36]. In this context, our observation that vitality and mental health scores were higher in the rehabilitation group is consistent with the broader literature and with interventional studies suggesting that multidisciplinary or psychologically informed rehabilitation can alleviate fatigue, mood symptoms and cognitive complaints [12,33,37].

The absence of a clearly significant difference in Physical Functioning warrants cautious interpretation. It may indicate that perceived improvements in vitality and mental health emerge even when limitations in physical activities persist, that changes in physical capacity lag behind psychological adaptation, or that the study was underpowered to detect more modest between-group differences in this domain. It is also possible that the SF-36 Physical Functioning scale is less sensitive than targeted performance-based measures for capturing changes relevant to long-COVID rehabilitation at the time point studied [7,32].

In multivariable models, older age and longer symptom duration were consistently associated with lower scores across several SF-36 domains and component summaries, and with less favourable perceived health. These associations align with large cohort and registry studies showing that older adults and those with more prolonged symptom trajectories tend to report lower HRQoL and greater functional impact after COVID-19 [7,9,27,35]. They are also in line with meta-analytic evidence that a proportion of patients experience symptoms persisting beyond two years, with fatigue, pain and cognitive problems contributing to long-term burden [8].

Rehabilitation status emerged as a robust correlate of better outcomes in several domains, even after adjustment for age, sex and symptom duration. Positive associations with Vitality, Social Functioning, Bodily Pain, General Health, Perceived Health Status and lower PCFS scores remained statistically significant after stringent Bonferroni correction, whereas the association between rehabilitation and the Physical Component Summary did not meet the adjusted significance threshold. These patterns are directionally consistent with the benefits reported in structured rehabilitation cohorts and trials [15,16,18,33,38].

From a mental health perspective, participants who completed rehabilitation reported better scores in several relevant SF-36 domains, including Vitality, Social Functioning and the Mental Health domain, and these domain-level differences remained statistically significant after Bonferroni correction. In line with this pattern, the Mental Component Summary (MCS) also showed a substantial association with rehabilitation (β ≈ 0.48), and this association persisted after applying the conservative Bonferroni adjustment (p_adj < 0.001). Taken together, these findings suggest that rehabilitation may be linked not only to improved physical functioning and symptom burden but also to better perceived mental health and psychosocial adjustment. However, because of the cross-sectional design and the possibility of residual confounding, these results should be interpreted as hypothesis-generating rather than as definitive evidence of causal treatment effects.

Several variables that might be expected to influence outcomes—such as BMI and history of hospitalisation—were not consistently associated with SF-36 domains or PCFS in our models. This may be due to limited variability in these measures within our sample, the relatively small number of previously hospitalised respondents, or the limitations of self-reported anthropometry in an online survey. Prior COVID-19 infections and physically demanding work showed only selective associations, underscoring the multifactorial nature of long-COVID sequelae and the need for more detailed characterisation of comorbidities, occupational demands and psychosocial context [4,19,37].

In our analysis, rehabilitation was treated as a binary, self-reported exposure [12,13,18]. Incomplete specification of "real-world" rehabilitation content, dose and fidelity across studies remains a major challenge for evidence synthesis and guideline development [12,37].

Our findings also speak to concerns about access and system-level constraints. Only a minority of long-COVID patients in many settings appear to receive structured rehabilitation, despite growing evidence that such programmes can improve symptoms and functional outcomes [15,16,21,24]. Workforce shortages, long waiting times and fragmented service configurations have been reported in several countries, including within Europe [20,26]. A recent WHO multi-country survey among rehabilitation workers in five European countries documented widespread disruption and suspension of services and increased patient waiting lists, indicating ongoing pressure on rehabilitation capacity beyond the acute phase of the pandemic [25]. Against this backdrop, our observation that respondents who reported rehabilitation tended to have better HRQoL and functional scores—while not implying causality—supports calls to systematically integrate multidisciplinary rehabilitation into long-COVID care pathways and to address inequities in access [19,31].

Several plausible mechanisms may underpin the observed associations between rehabilitation and better outcomes, although our cross-sectional design does not permit causal inference. First, structured rehabilitation programmes typically provide education about pacing, energy conservation and symptom-contingent activity planning, which may help patients avoid cycles of overexertion and post-exertional symptom exacerbation. Second, supervised physical and respiratory training can address deconditioning, dyspnoea and musculoskeletal pain, thereby improving physical functioning and fatigue perception. Third, psychological and behavioural components, including cognitive-behavioural or supportive interventions, may alleviate anxiety, low mood and health-related worry, which are known to interact bidirectionally with physical symptoms. Finally, regular contact with a multidisciplinary team may enhance self-efficacy, provide social validation and improve navigation of health and social care systems. These mechanisms may act in combination and could partly explain why participants who reported completion of rehabilitation also reported better Vitality, social functioning and perceived general health.

Our findings are broadly consistent with emerging interventional and systematic evidence on rehabilitation in adults with post-COVID-19 condition. A large systematic review and meta-analysis of rehabilitation interventions for PCC reported that structured programmes lead to improvements in physical capacity (e.g., 6 min walk distance), dyspnoea and health-related quality of life compared with usual care [39]. Randomised controlled trials of supervised, multicomponent exercise-based rehabilitation (RECOVE trial) and cardiopulmonary rehabilitation in long-COVID cohorts likewise demonstrated clinically meaningful gains in exercise tolerance, symptom burden and patient-reported health status [40]. Furthermore, a recent living systematic review of long-COVID interventions concluded that combined physical and mental-health rehabilitation programmes probably improve overall health status, functional outcomes and quality of life when compared with usual care [12]. In contrast to these protocolised trials with clearly defined interventions and eligibility criteria, our study reflects a heterogeneous, real-world mix of rehabilitation modalities accessed by a self-selected group of patients with varying symptom duration and severity. The observation that participants who reported completing rehabilitation had higher Vitality and social functioning scores, better general and mental health and lower functional limitation is therefore in line with the direction of effects seen in controlled settings and suggests that some of the benefits of rehabilitation may extend into routine practice. These observations may help to inform the design and targeting of future prospective rehabilitation programmes for post-COVID-19 condition, by highlighting which patient-reported domains (such as vitality, social functioning and perceived health) are most likely to improve and which subgroups may merit particular attention in controlled studies.