Descriptive, reliability, and validity analysis

Table 1 shows the mean and standard deviations (SD) for men, women, and the complete sample in the two measurement points. Several t-tests for independent samples by sex were performed, showing that at both T1 and T2 women scored significantly higher than men on all items that measure depressive symptoms, except for the item “tired,” and in six items of ten of the negative affect questionnaire. In contrast, men scored significantly higher than women on positive affect in some items (see Table 1). Regarding PA levels, women showed significantly higher values on days of LPA and men showed higher values on days of VPA levels. Repeated-measures t-tests were also performed between T1 and T2 for the whole sample. The results showed a significant increase in four of the six depressive symptoms and the items pa4 “enthusiastic” and na5 “hostile.” An increase was also observed between T1 and T2 in days of VPA and days of MPA.

Table 1. Descriptive analyses of physical activity levels and mental health variables

Note: * p < .05, ** p < .01, and *** p < .001. DVPA, DMPA, and DLPA represent days per week of vigorous, moderate, and light physical activity, respectively.

Cronbach’s alpha, composite reliability rh, and average variance extracted were used as measures of reliability and validity. In depressive symptoms, Cronbach’s alpha was .84; composite reliability (rh_a) was 0.87; composite reliability (rh_c) was 0.89; and average variance extracted was 0.57. In positive affect, Cronbach’s alpha was 0.89; composite reliability (rh_a) was 0.90; composite reliability (rh_c) was 0.91; and average variance extracted was 0.51. In negative affect, Cronbach’s alpha was 0.85; composite reliability (rh_a) was 0.88; composite reliability (rh_c) was 0.88; and average variance extracted was 0.51.

Inter-subject network analysis

The network shown in Figure 1 (top) indicates that the three dimensions of PA grouped together to form a first cluster. A second cluster was formed by the variables making up positive affect. A third cluster was formed, due to proximity, by depressive symptoms, and a fourth, relatively separate grouping, by the dimensions of negative affect. The four clusters were not completely separate, as they were connected with weaker links, except for the strong links between “nervous” (nervous) (an8m) and “hopeless” (D4m) and between “distressed” (distressed) (an1m) and “unhappy” (D3). The connections between the PA dimensions and the other clusters were weak. The three PA dimensions were positively connected with the variable “attentive” (ap10m). The frequency of light PA (DLPAm) was furthermore negatively connected with “embarrassed” (ashamed) (an7m) and “guilty” (guilty) (an3m) and positively connected with “alert” (alert) (ap6m). DMPAm was negatively related to distressed (an1m), and DVPAm was negatively related to irritable (an2m) and positively related to enthusiastic (ap4m).

Figure 1 (middle and bottom) shows the results in the inter-subject network by participant gender. The four previously noted clusters that emerged in the full sample were present for both men and women. The inter-cluster links were weaker, but the links between “nervous” (an8m) and “hopeless” (D4m) were stronger for women than men and between “anguish” (an1m) and “unhappiness” (D3) were the same for both genders. The three dimensions of PA frequency were positively connected with the symptom “attentive” (ap10m), except for DLPA in women. The frequency of light PA (DLPA) was still negatively connected with “ashamed” (an7m) and “guilty” (an3m) for women, but only with “guilty” (an3m) for men. For women, there was a new negative connection between DLPA and “tired” (D1m). The negative relationship between DMPAm and “anxious” (an1m) remained only in women, and there was a new negative connection between DMPAm and “nervous” (an8m) for women only. Interestingly, for the men there was a new positive connection between DMPAm and “fearful” (an10). The positive relationship between DVPAm and “enthusiastic” (ap4m) only remained for women, and there was a new positive connection between DVPAm and “interested” (ap1) for women. For men, there was a negative relationship between DVPAm and “hopeless” (D4).

Figure 1. Inter-subject network analysis for the full sample (top), men (middle), and women (bottom). Note. Men = above, women = below.

Intra-subject network analysis

Figure 2 (i.e., the intra-subject network (top)) shows that the four previously noted clusters in the inter-subject network analysis appeared clearer. The links between the clusters were weaker except for the strongest link between “nervousness” (an8c) and “hopelessness” (D4c). There were fewer connections between the PA frequency dimensions and the other clusters than in the inter-subject network. Only two (DMPAc and DVPAc) were still positively connected with the “attentive” variable (ap10c). The other link worth noting is the negative relationship of DVPAc with “distressed” (an1c).

Figure 2. Intra-subject network analysis for the full sample (top), men (middle), and women (bottom). Note. Men = above, women = below.

Figure 2 (middle and button) shows the results in the within-subjects network by participant gender. The four previously noted clusters that emerged in the full sample were present for both men and women. The links between clusters were also weaker, except between “nervous” (an8c) and “hopeless” (D4c) for women and between “irritable” (an2c) and “hopeless” (D4c), between “distressed” (distressed) (an1c) and “unhappy” (D3c), and between “excited” (excited) (ap2c) and “distressed” (distressed) (an1c) for men. For women, DLPA was not connected with any variable, while in men it was negatively related to “irritable” (an2c) and “guilty” (an3) and positively related to “sleep” (D2c) and “unhappiness” (D3). In the DMPAc, the only connection worth mentioning was the positive connection seen in women with “attentive” (ap10c), while in the DVPAc, the most important difference was the positive connection with “inspired” (ap2c), which was not apparent for men.

Network indexes

The CS coefficients obtained in both inter- and intra-subject network analyses were above 0.74, indicating a proper value. Figure 3 shows the centrality indices of all estimated nodes in the inter-subject and intra-subject networks for the whole sample and men and women separately. In the inter-subject network, the “unhappiness” (D3m) node showed the highest centrality of betweenness, closeness, and strength for both the total sample and men. There was a similar pattern for women except for the centrality of intermediation. In this case, the highest variable was “tiredness” (D1m). More differences were seen in the intra-subject network. In terms of closeness centrality, “embarrassed” (an7c) was the strongest variable connected to other nodes in the shortest path, both for the total sample and for women, while in men the strongest was “unhappiness” (D3m). The strength centrality data indicate that “inspired” (ap7c) was the most central node for both the total sample and for women, while in men the strongest was “determined” (ap8c). Analyses of betweenness centrality identified the symptom of “unhappiness” (D3m) as the node most frequently found on the shortest path between other nodes for both the full sample and for men, while in women it was “embarrassed” (an7c).

Figure 3. Network indexes for intra-subject network analysis and inter-subject network analysis. Note. Above = inter-subjects; below = intra-subjects.

Discussion

To our knowledge, this is the first study that has adopted a network approach to examine associations between PA, depression, positive affect, and negative affect at two time points during the confinement of the population to their homes during the COVID-19 pandemic. The results showed four clusters (PA, depression, positive affect, and negative affect), which were consistent in the within-subject network and the between-subjects network for both men and women. The four clusters were not completely separated, as they were connected with weaker links, except for the strong links between nervousness and hopelessness and between distress and unhappiness. The isolation of people in their homes by COVID-19 seems to have increased the state of nervousness, which was related to a feeling of hopelessness for the future. In addition, the anguish of feeling confined led to a feeling of sadness and unhappiness, and vice versa.

Depression nodes play a key role in the inter-subject network, in line with previous studies (Fitzpatrick et al., Reference Fitzpatrick, Harris and Drawve2020; Zhang et al., Reference Zhang, Lu, Zeng, Zhang, Du, Jiang and Du2020). The sadness and unhappiness symptom node showed the highest centrality of mediating closeness and strength for both the full sample and for men. For women, we found the same pattern, except for the centrality of betweenness, a parameter space that was occupied by another depression symptom (i.e., feeling too tired to get things done). These results are consistent with findings from Di Blasi et al. (Reference Di Blasi, Gullo, Mancinelli, Freda, Esposito, Gelo and Coco2021), which put depression in a central place, meaning it could have a considerable impact on both positive and negative affect symptoms.

In the within-subjects network, centrality was shared by symptoms of depression and affection. However, the centrality of closeness and intermediation in men was occupied by the same depression symptom as in the inter-subject network – “unhappiness.” In women, the centrality of closeness and mediation was occupied by the negative affect node “ashamed.” The strength centrality data were occupied by nodes of positive affect, “inspired” (for the total sample and women), and “determined” (men).

In the between-subjects network for the full sample, the connections of the PA dimensions were with the two affect clusters, which were the closest. In particular, there were positive connections with positive affect symptoms and negative connections with negative affect symptoms. In other words, PA during isolation had a positive effect on confined people’s emotional states. All three PA dimensions were positively connected with the attentive positive affect variable (attentive). This may be because PA is related to neuroelectrical activity, suggesting better overall sustained attention, which demonstrates a better ability to allocate attentional resources over time (Luque-Casado et al., Reference Luque-Casado, Perakakis, Hillman, Kao, Llorens, Guerra and Sanabria2016). Feelings of guilt and shame seemed to be able to be mitigated by the frequency of light PA, which also seemed to positively influence alertness. The frequency of moderate PA seemed to reduce distress, while the frequency of vigorous PA decreased irritability and increased enthusiasm. Ultimately, the frequency of all three PA dimensions exhibited beneficial effects on confined people’s emotions.

In the within-subjects network, there were fewer connections of the PA frequency dimensions with the other clusters than in the between-subjects network. Only DMPAc and DVPAc were still positively connected with the “attentive” symptom. The other link worth noting was the negative relationship of DVPAc with “distressed.”

In the inter-subject network, direct relationships between PA and depressive symptoms only appeared in women, with a negative connection between PLDm and feeling too tired to do things. In men, there was a negative connection between PWDm and feeling hopeless about the future. In the within-subjects network, curiously, there was a positive relationship in men between PWDm and having problems falling asleep and feeling unhappy.

Stress, such as social isolation, is an environmental factor that precipitates a potent physiological response involving the autonomic nervous system and the hypothalamic–pituitary–adrenal axis (Lee, Reference Lee2022). The brain is especially susceptible to the catabolic effects of stress and its glucocorticoid hormone cortisol (GC), as prolonged exposure to stress or cortisol leads to the development of psychiatric disorders such as anxiety and depression. Social isolation further affects synaptic plasticity and increases basal synaptic transmission in hippocampal CA1 pyramidal neurons, and it appears that physical exercise prevents stress-induced synaptic effects (Aberg et al., Reference Aberg, Mattsson, Graae, Galan, Aski, Canto and Brene2012).

A recent systematic review examining the association between physical activity and mental health during the first year of the COVID-19 pandemic (Marconcin et al., Reference Marconcin, Werneck, Peralta, Ihle, Gouveia, Ferrari, Sarmento and Marques2022) reported that greater PA is associated with greater well-being, quality of life, and lower depressive symptoms, anxiety, and stress, regardless of age. Sang et al. (Reference Sang, Menhas, Saqib, Mahmood, Weng, Khurshid, Iqbal and Shahzad2021) tested a structural equation model to examine the psychological impacts of COVID-19 home confinement and PA, indicating that PA is the best strategy to manage psychological issues such as depression and anxiety. In the present study, PA was found to be associated with better-sustained attention, increased alertness, and enthusiasm. In addition, it mitigates feelings of guilt and shame and decreases confinement distress and irritability. In women, it also reduced fatigue, and in men, it reduced feelings of unhappiness. To our knowledge, this is the first study to address the relationship between PA and depressive and affective symptoms, so it is not possible to compare results with previous research. Nevertheless, the results support the need to increase PA during confinement. However, studies that collected measures before and after the stay-at-home order found a significant reduction in PA (Savage et al., Reference Savage, James, Magistro, Donaldson, Healy, Nevill and Hennis2020; Suzuki et al., Reference Suzuki, Maeda, Hirado, Shirakawa and Urabe2020), and depressive symptoms increased as weeks of isolation passed (Cecchini et al., Reference Cecchini, Carriedo, Fernández-Río, Méndez-Giménez, González, Sánchez-Martínez and Rodríguez-González2021). In addition, increased levels of physical activity have been associated with stronger effects on well-being (O’Brien and Forster, Reference O’Brien and Forster2021), with a decrease in depressive symptoms (Marconcin et al., Reference Marconcin, Werneck, Peralta, Ihle, Gouveia, Ferrari, Sarmento and Marques2022) and an increase in positive affect (Carriedo et al., Reference Carriedo, Cecchini, Fernández-Río and Méndez-Giménez2020b).

There are several research studies that have analyzed the relationships between PA and different psychological factors during home confinement (e.g., López-Bueno et al., Reference López-Bueno, Calatayud, Andersen, Balsalobre-Fernández, Casaña, Casajús, Smith and López-Sánchez2020; Carriedo et al., Reference Carriedo, Cecchini, Fernández-Río and Méndez-Giménez2020a, Reference Carriedo, Cecchini, Fernández-Río and Méndez-Giménez2020b; Cecchini et al., Reference Cecchini, Carriedo, Fernández-Río, Méndez-Giménez, González, Sánchez-Martínez and Rodríguez-González2021; Dai et al., Reference Dai, Sang, Menhas, Xu, Khurshid, Mahmood and Alam2021; Méndez-Giménez et al., Reference Méndez-Giménez, Cecchini and Fernández-Río2021; Sang et al., Reference Sang, Menhas, Saqib, Mahmood, Weng, Khurshid, Iqbal and Shahzad2021 and Fernandez-Rio et al., Reference Fernandez-Rio, Cecchini, Mendez-Gimenez and Carriedo2022). However, the results of this study are novel because it uses a new methodology, which is complementary to other types of study, to understand the relationship between depression, positive and negative affects, and physical activity, where symptoms are secondary to a common underlying cause (Borsboom, Reference Borsboom2008; Schmittmann et al., Reference Schmittmann, Cramer, Waldorp, Epskamp, Kievit and Borsboom2013).

PA was an effective option for mitigating the negative effects of the COVID-19 pandemic on mental health during its first year. Public health policymakers should be alert to possibilities of increasing physical activity during stay-at-home orders in many countries around the world. Moreover, programs should be developed to promote physical activity to be able to cope with confinement or similar scenarios in the future. For instance, programs that describe recommendations and guidelines for staying active at home, with aerobic exercise training on a bike or rowing ergometer, bodyweight training, dance, and videos with full-body active routines, can be useful to achieve the recommended physical activity levels that might mitigate the negative effects of the COVID-19 pandemic. Also, fitness apps, live-streaming workout classes, and virtual reality fitness for PA could be useful (Liu et al., Reference Liu, Menhas, Dai, Saqib and Peng2022).

This study has some limitations. For instance, this methodological approach might lead to the overfitting of the model. For this reason, although network approaches are a useful exploratory tool, the presented results need to be confirmed in future research (cross-validation or experimentation; Epskamp et al., Reference Epskamp, Waldorp, Mõttus and Borsboom2018). Nevertheless, this network approach can be a very useful tool for modeling different dynamic relationships between psychological constructs addressing emotional, psychological, and social well-being and PA, for example, affective-type disorders, or personality and behavioral disorders, psychological developmental disturbances, and their relationship with PA. Another limitation is related to the sample criteria. Depressive symptoms and affective symptoms could be affected by one’s medical conditions. In this study, this information had not been collected; therefore, this should be also considered in future studies.