Furthermore, although a statistically significant reduction was observed in anxiety scores (p = 0.047), the absolute mean difference was relatively small (mean = 0.94), which raises the issue of clinical relevance. While statistically meaningful, such a minimal change may not represent a substantial improvement in daily psychological functioning. Future studies should consider reporting minimal clinically important differences (MCIDs) to assess real-world effects of PA on mental health.
4.1. Impact of Physical Activity Levels and Domains on Anxiety, Depression, and Stress
A comparison of the results revealed that at the final measurement, certain levels and domains of PA, as measured by the IPAQ questionnaire, had a statistically significant impact on mental health, unlike at the initial measurement. Low-intensity PA, such as walking, and moderate-intensity activity were significantly associated with lower levels of anxiety and stress (Model 4), while vigorous and total PA had no statistically significant impact on anxiety. This is consistent with the findings of a meta-analysis by Xu and associates [
29], who showed that walking positively affects mental health regardless of format (group or individual), duration, or location. Regarding moderate-intensity PA, our findings align with previous studies highlighting its positive impact on mental health. Research has shown that moderate PA can reduce symptoms of anxiety regardless of gender, age, or health status [
30,
31,
32].
Moderate activities, such as brisk walking or recreational exercise, activate psychophysiological mechanisms involving neurotransmitter regulation and reduced physiological arousal, contributing to better emotional well-being [
33]. Our results support the hypothesis of a protective effect of moderate PA on anxiety symptoms, especially in the recovery period following COVID-19.
In terms of depression, the final measurement results indicated that total and vigorous PA were statistically significant predictors (Model 5). However, the positive beta coefficient for vigorous PA (β = 0.182) suggests a possible paradoxical effect, i.e., higher levels of vigorous activity may be associated with increased depressive symptoms. While most studies highlight the protective effect of vigorous activity [
14,
34,
35], some research warns about the negative effects of excessive PA, particularly in vulnerable populations [
36] such as those recovering from COVID-19. Shimura and associates [
36] concluded that optimal PA in terms of duration, intensity, and frequency contributes to better psychological outcomes. Conversely, inadequately adjusted or prolonged PA may negatively impact mental health and overall well-being.
Our findings suggest that while total PA may exert a protective influence, the relationship between vigorous PA and depression might not be linear. It is possible that overly intense or prolonged physical activity in certain individuals leads to overexertion, emotional fatigue, or physiological stress, thereby increasing the risk of depressive symptoms. This observation aligns with the hypothesis of a U-shaped association between PA intensity and mental health, where both inactivity and excessive intensity could be detrimental. However, this hypothesis was not statistically tested in the current models. Future studies should incorporate interaction or quadratic terms to examine nonlinear associations and better capture the complexity of the PA–mental health relationship in post-COVID-19 populations.
Overall, the data imply that individuals may respond differently to PA after COVID-19, depending on personal, psychosocial, or physiological factors, highlighting the need for more tailored PA recommendations in post-COVID-19 recovery strategies. These findings align with a growing body of evidence highlighting the complexity of post-COVID-19 recovery and the necessity of individualized approaches. For instance, Mitroi et al. [
37] demonstrated that the combination of prior COVID-19 infection and sociodemographic characteristics significantly influenced quality of life among tuberculosis patients, a population already burdened by chronic illness and social vulnerability. Their study emphasizes the relevance of contextual factors, such as disease history, economic status, and social support, when assessing recovery outcomes. Similarly, Cioboata et al. [
38] found that both the clinical form of initial COVID-19 illness and the presence of comorbidities were strongly associated with the persistence and severity of post-COVID syndrome symptoms. Although the comorbidities were not statistically analyzed as independent predictors, the authors acknowledged their potential relevance. These findings suggest that the effectiveness of physical activity may be mediated or even constrained by pre-existing health conditions and the overall clinical profile of the individual. Accordingly, our results, although obtained from a community-based sample, should be interpreted with caution and viewed as part of a broader clinical landscape in which PA interventions must be tailored to the individual’s physiological, psychological, and social circumstances.
Regarding stress, our results indicate that low-intensity PA in the form of walking and moderate activity were statistically significantly associated with lower stress levels (Model 6), consistent with previous research [
11,
39]. Although several PA domains and intensity levels emerged as statistically significant predictors of mental health outcomes, the explained variance in all regression models was relatively low (R
2 = 1.8–5.7%), indicating that while PA contributes to psychological well-being, many other unmeasured factors, such as personality traits, social support, and sleep quality, likely play a larger role. Future research should include a wider range of variables to enhance explanatory power and model robustness.
The transport domain of PA showed a statistically significant impact on all three mental health components (anxiety, depression, and stress) at the initial measurement (
Table 6). At the final measurement (
Table 7), home- and transport-related PA predicted anxiety (Model 10), home activity alone predicted depression (Model 11), and both categories predicted stress (Model 12). These results suggest that transport-related PA, as a daily and accessible form of movement, may positively affect mental health, which has been confirmed by earlier studies [
40].
However, in our study, the impact of transport-related PA was not uniform for all psychological outcomes, as its effect on stress was not statistically significant. This suggests that different forms of transport-related activity (e.g., walking, cycling, using public transportation with walking) may not equally affect all aspects of mental health. Further research is needed to explore the specific mechanisms through which transport-related PA influences anxiety, depression, and stress symptoms in individuals recovering from COVID-19.
Regarding household PA, research suggests that tasks like cleaning, cooking, and home maintenance can positively impact mental health. For example, a study by Koblinsky and associates [
41] found that higher household activity levels were associated with greater gray matter volume in the brain among older adults, potentially indicating better cognitive function. Additional studies have confirmed the link between household PA and reduced stress levels [
13].
The explanation for why household activity affects all three mental health parameters may lie in its variety, frequency, and integration into daily life. Hammar and associates (13) support the idea that diverse activities, such as household chores and gardening, are associated with better mental health, provided they last at least 20 min. Additionally, the positive influence of household- and transport-related PA observed in our models may be explained by factors such as daily structure, perceived autonomy, and a sense of purpose. Unlike formal exercise, these domains often reflect meaningful and routine-based engagement, which may support emotional regulation and psychological recovery.
In contrast, the absence of an impact from workplace PA on mental health parameters may be due to the fact that workplace activity is often not a matter of personal initiative, but rather an obligation or burden, diminishing its psychological benefit [
40].
In terms of leisure-time PA, the same authors [
40] confirmed its positive impact on mental health, though our study did not support this, likely due to irregular or limited participation, which may have reduced its effect.
In general, our study’s findings confirm that levels and domains of PA do not always show consistent and stable impacts on mental health and well-being, especially in individuals who have recovered from COVID-19. Variations observed between the two measurement points suggest that each domain and level of PA may differently influence symptoms of anxiety, depression, and stress depending on the period and the participants’ life context. This underscores the importance of studying PA in a dynamic timeframe for a more accurate understanding of its effects.
When interpreting the results, it is essential to consider the potential influence of personal perception and motivation on different levels and domains of PA. Individual perception of activity, as well as intrinsic motivation, can play a crucial role in how PA affects psychological well-being. For example, research among older adults found that internal motivation for PA was significantly associated with life satisfaction and regular exercise [
42]. Satisfying basic psychological needs such as autonomy, competence, and connectedness contributes to better mental health and a higher likelihood of continued PA. Additionally, a study among healthy middle-aged adults found that perceptions of one’s own PA are often inaccurate [
43].
Many participants overestimated or underestimated their activity levels, which affected their mental health. Accurate perception of PA was linked to better psychological outcomes and greater life satisfaction.
Taking all of this into account, it is important to emphasize the need for a personalized approach to promoting PA for mental health improvement. Recommendations that consider individual motivational factors and perceptions of activity may be more effective in encouraging long-term engagement in PA and enhancing psychological well-being.
4.3. Limitations of the Study
Despite its strengths, this study has several important limitations that should be considered when interpreting the results. Using a self-report PA questionnaire (IPAQ) may introduce response bias, and the absence of a control group prevents direct comparisons with individuals who did not experience COVID-19. This limits our ability to isolate the effects of COVID-19 recovery on mental health outcomes. Including appropriate comparison groups in future studies, such as individuals with no history of COVID-19 or those recovering from other illnesses, would allow for more robust causal inference. Such designs would help determine whether observed psychological changes are specifically attributable to COVID-19 recovery or reflect broader population-level trends. Longitudinal studies with matched controls could further strengthen the validity and generalizability of future findings.
Furthermore, other psychosocial factors that could affect mental health, such as social support and living conditions, were not considered in this study. The sample included only participants from the Jablanica District, limiting the generalizability of the results to other regions. Moreover, individuals who were hospitalized with severe COVID-19 were excluded from the study, further narrowing the generalizability to those with milder or moderate disease courses. Additionally, the sample was predominantly female (193 women vs. 95 men), which may introduce gender bias and limit the applicability of the findings to broader populations.
In addition, although the DASS-42 is a validated instrument, it measures emotional states over the previous seven days. Therefore, administering it at only two time points may not fully capture the psychological fluctuations that occur throughout the recovery process. Finally, while self-report instruments were practical and widely used during the pandemic, the inclusion of objective data (e.g., wearable devices or physiological markers) would further strengthen the methodological rigor and accuracy of the findings.
Furthermore, although we used change scores to assess improvements in mental health outcomes, we did not adjust for baseline values as covariates in the regression models, which may introduce statistical artifacts such as regression to the mean. Future studies could improve the methodological robustness by including baseline scores as control variables and exploring nonlinear effects through quadratic modeling techniques.
Finally, several methodological issues warrant further attention. First, multiple statistical tests were conducted across various PA domains and intensity levels without applying corrections for multiple comparisons (e.g., Bonferroni adjustment), which increases the risk of Type I error.
Second, although stepwise linear regression was used to explore predictors of mental health outcomes, this approach is known to carry risks of overfitting and model instability. No cross-validation or robustness checks were performed to confirm the reliability of the derived models. These factors may limit the generalizability and reproducibility of our findings. Future research should adopt more conservative and theory-driven statistical methods and aim to replicate findings using alternative modeling strategies and validation techniques.