Associations between Physical Activity Level and Mental Health in the Spanish Population: A Cross-Sectional Study

Physical inactivity and sedentary lifestyles appear to be critical factors in developing mental health problems, including depression, anxiety, and other diseases in developed societies. This study analysed the associations between physical activity level (PAL) and mental health using the Goldberg General Health Questionnaire (GHQ12) in the Spanish population before the COVID-19 pandemic. A cross-sectional design, based on data from the Spanish National Health Survey (ENSE 2017), the last health survey before the pandemic, was carried out with 17,641 participants. Data did not follow a normal distribution, so non-parametric tests were used to analyse intergroup differences, differences at baseline and post hoc, and correlations between variables. Associations were found between the PAL, mental health and all its dimensions. The groups that performed moderate and intense PAL showed lower values in the GHQ12 questionnaire than those who walked or were inactive. Thus, higher PAL was associated with better mental health indicators, including successful coping, self-esteem and stress. This study provides a framework to compare outcomes between the pre- and post-pandemic periods, as the ENSE is performed every five years.


Introduction
Mental disorders are the second leading cause of illness globally [1]. Anxiety and depression are the most prevalent mental disorders in the general population [2]. Mental health problems are associated with a higher prevalence of chronic diseases [3], poor adherence to medical treatment [4], increased morbidity [5][6][7] and premature mortality [8]. As an example, the 12-month prevalence of anxiety disorders was 13.4% in Europe (69.1 million people), costing more than 74 billion euros [9], and 22% in the United States [10]. Moreover, 103 years old [42], and whose interviews were held between October 2016 and 2017 by experienced surveyors.
Regulation 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of individuals concerning the processing of personal data and on the free movement of personal data and derogating from Directive 95/46/EC [43] states that files for public use are not confidential; therefore, neither the application of data protection principles to anonymised information nor the approval of accredited ethics committees is required.

Participants
A stratified three-phase random sampling was carried out in the Spanish population, considering people aged between 15 and 103 years, resulting in a 23,089 sample. A total of 10,595 men and 12,494 women were interviewed. In this research, 5312 individuals were excluded as the ENSE 2017 [41] did not ask about PA in the 69+ age group, and 136 individuals were excluded because they did not present complete data on the variables of interest for this study. Finally, the sample for our study was composed of 17,641 participants, including 8469 men and 9172 women ( Figure 1). 103 years old [42], and whose interviews were held between October 2016 and 2017 by experienced surveyors.
Regulation 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of individuals concerning the processing of personal data and on the free movement of personal data and derogating from Directive 95/46/EC [43] states that files for public use are not confidential; therefore, neither the application of data protection principles to anonymised information nor the approval of accredited ethics committees is required.

Participants
A stratified three-phase random sampling was carried out in the Spanish population, considering people aged between 15 and 103 years, resulting in a 23,089 sample. A total of 10,595 men and 12,494 women were interviewed. In this research, 5312 individuals were excluded as the ENSE 2017 [41] did not ask about PA in the 69+ age group, and 136 individuals were excluded because they did not present complete data on the variables of interest for this study. Finally, the sample for our study was composed of 17,641 participants, including 8469 men and 9172 women ( Figure 1).

Measures and Variables
The considered and created variables for this research were: Age: taken from the AGEa variable of the ENSE 2017 (years). Sex: taken from the SEXOa variable from the ENSE 2017 (male or female). Mental health: the Spanish version of the Goldberg General Health Questionnaire (GHQ-12) was used. This questionnaire evaluates psychological health based on the answers to 12 items graded from 0 to 3, forming an overall index with the sum of all the answers. The total score ranges from 0 (the best condition) to 36 (the worst). The GHQ-12 presents high internal consistency (α = 0.86) [16]. The GHQ-12 is a self-administered screening test for non-psychotic psychiatric disorders, widely used in clinical settings and the general population both for its brevity and its psychometric characteristics. Although its factor structure has been a matter of debate, discussing whether it is composed of one

Measures and Variables
The considered and created variables for this research were: Age: taken from the AGEa variable of the ENSE 2017 (years). Sex: taken from the SEXOa variable from the ENSE 2017 (male or female). Mental health: the Spanish version of the Goldberg General Health Questionnaire (GHQ-12) was used. This questionnaire evaluates psychological health based on the answers to 12 items graded from 0 to 3, forming an overall index with the sum of all the answers. The total score ranges from 0 (the best condition) to 36 (the worst). The GHQ-12 presents high internal consistency (α = 0.86) [16]. The GHQ-12 is a self-administered screening test for non-psychotic psychiatric disorders, widely used in clinical settings and the general population both for its brevity and its psychometric characteristics. Although its factor structure has been a matter of debate, discussing whether it is composed of one factor, two (depression/anxiety and social dysfunction) or three factors, in this study we agree with the three-factor option [44] based on the factor analysis results: successful coping (FI), self-esteem (FII) and stress (FIII) [17,45]: • Successful coping (FI): obtained by summing 6 items (1, 3, 4, 7, 8 and 12); scores ranged from 0 to 18 (0, the best; 18, the worst) and external validity of 0.82 with a p-value of 0.001. • Self-esteem (FII): obtained by summing 4 items (6, 9, 10 and 11), with scores between 0 and 12 (0, the best; 12, the worst) and external validity of 0.70 with a p-value of 0.001. • Stress (FIII): obtained by summing 3 items (2, 5 and 9), with scores between 0 and 9 (0, the best, 9, the worst) and external validity of 0.75, with a p-value of 0.001.
The Physical Activity Index (PAI) [46] was created by combining several PA factors with the answers obtained in the ENSE 2017. The factors were: • Intensity: intense activity (10), moderate activity (5) and mild activity (0). • Frequency: on the question "how many days did you practise intense and moderate PA?" the following values to the possible answers: "0" for zero days, "1" for one day per week, "2" for two or three days per week and "3" for more than three days per week. • Duration: on the questions "how much time did you spend in total on intense PA? and, how much time did you spend in total on moderate PA?" a value of "1" was given for less than 30 min and "1.5" for 30 min or more.
So, the formula to find the PAI was = (intensity factor for intense activity * frequency factor for intense activity × duration factor for intense activity) + (intensity factor for moderate activity × frequency factor for moderate activity × duration factor for moderate activity). The factors were applied to intensity, frequency, and duration questions. PAI values range from 0 to 67.5 (a maximum of 45 for intense and 22.5 for moderate activities). Mild activities did not add value to the PAI. Thus, six PAL were established: • "Inactive": participants with PAI = 0 who answered the question "now think about how much time you spent walking in the last 7 days", with "no day more than 10 min at a time". • "Insufficient": participants with PAI = 0 who answered the question "now think about how much time you spent walking in the last 7 days" or stated, "at least one day more than 10 min consecutively". • "Low": participants with a Physical Activity Index (PAI) score between 1 and 15 (75th percentile). • "Medium": individuals with a PAI score between 16 and 30 (90th percentile). • "High": participants with a PAI score between 31 and 45 (95th percentile). • "Very high": individuals with a PAI over 45 (values above the 95th percentile).

Statistical Analysis
Statistical analysis was conducted using the Statistical Package for the Social Sciences (SPSS, Version 25, IBM SPSS, Armonk, NY, USA) software. Data distribution was analysed using the Kolmogorov-Smirnov test, and deciding to use non-parametric tests based on the results. Then, a descriptive statistical analysis was carried out to characterise the sample, by presenting age, mental health, successful coping, self-esteem, stress, and PAI variables using medians and interquartile ranges, complemented by means and standard deviations. The PAL was characterized using the absolute and relative frequencies of the population in its different levels, in total population and by sex. The Mann-Whitney U and the Chi-square tests for continuous and ordinal variables, respectively, were used to check potential differences between sexes and groups. The Kruskal-Wallis test was carried out to find differences at baseline between PAL and continuous variables from the GHQ-12, in addition to the post hoc Mann-Whitney U test to identify differences between the various PAL groups. Additionally, the effect size was calculated by using the z value (r = Z/ √ N), interpreted as 0.1 = small effect, 0.3 = medium and 0.5 = large effect [47]. Finally, a Spearman correlation study with the Bonferroni adjustment was carried out to analyse the associations between PAL and mental health dimensions. For all analyses, two-sided p-values ≤ 0.05 were considered statistically significant. Table 1 shows sociodemographic sample information (n = 17,641). Significant differences were found between general mental health status, self-esteem, successful coping, stress, PAI and PAL and sex of participants. Specifically, male participants presented higher scores in mental health and lower scores in PAI. IQR: interquartile range; SD: standard deviation; GHQ-12: Goldberg's General Health Questionnaire, scores between 0 and 36; FI Successful Coping: scores between 0 and 18. 0, the best coping and 18, the worst; FII Self-esteem: scores 0-9, 0 the best self-esteem and 9, the worst); FIII Stress: scores between 0 and 9. 0, no stress and 9, very stressed; PAI: Physical Activity Index, considering only intense and moderate physical activity, scores 0-67.5; Inactive: PAI = 0, reporting not going out for more than 10 min at a time; Insufficient: PAI = 0, reporting to walk more than 10 min; a: p-value from Mann-Whitney U test; b: p-value from Chi-square test.

Results
Significant correlations were found between PAL and mental health in the total population (Table 2). Significant differences were found between mental health and all its dimensions (GQH-12) in the "Inactive" and "Insufficient" PAL groups and between these two and the rest of the groups (p < 0.001), finding better mental health at higher PAL. Again, in the total population, a 3 points difference was found between the "Inactive" and "Medium", "High", and "Very high" groups' median scores, representing a reduction of 27.3%. A 3.53 point difference was also found in the GHQ-12 mean scores between the "Inactive" and "Very high" groups, which means a reduction of 31.4%. Thus, a higher PAL was associated with better mental health, according to the GHQ-12 results. In the men's subgroup, differences were found between the "Inactive" and "Inadequate" PAL groups and between these and the other levels ( Table 3). The median decreased by 11 points in the "Inactive" group and to 8 points in the "Medium", "High", and "Very high" groups, as in the total population. The groups' mean scores on the GHQ-12 decreased as the level of PA increased, from a value of 12.04 in the "Inactive" group to 8.51 in the "Very high" group, with a difference of 3.53 points, which means a reduction of 29.3%.
In the women's subgroup, significant differences were also found between the "Inactive" and "Low" PAL groups with the rest of the levels ( Table 4). The median of the different groups decreased as the level of PA increased. The median decreased from 10 to 9 in the "Inactive" and "Low" groups, reaching 8 in the rest of the groups, representing a 20% decrease. Between the "Inactive" and "Poor" groups, the mean difference was 1.77 points on the GHQ-12, with the difference between the "Inactive" and "Very High" groups being 3.43 points less, representing a 27.6% reduction in score. The Successful Coping factor (mental health factor 1) scored better with higher levels of PA in the total population. Significant differences were found between the "Inactive" and "Poor" groups and between these and the other groups ( Table 5). The medians were the same for all groups. However, the groups' median decreased as the PAL increased from 6.80 in the "Inactive" group to 5.76 in the "Very high" group.
The self-esteem factor (mental health factor 2) scored better at a higher PAL, with significant differences found between the "Inactive" and "Insufficient" groups and between these and the rest of the groups ( Table 6). The median was 2 in the "Inactive" and "Insufficient" groups and 1 in the other groups. The group mean decreased as the PAL increased, with a reduction of 1.69 points between the "Inactive" and the "Very high" groups.     IQR: interquartile range; SD: standard deviation; FII, Self-esteem, from the Goldberg's General Health Questionnaire, scores between 0 and 9, 0 being for the best self-esteem and 9, the worst; Medians Diff: mental health medians differences for every physical activity level; Means diff: between mental health means differences for every physical activity level; p * Kruskal-Wallis value: mental health measured by GHQ-12 as response and physical activity level as a factor; PAI: Physical Activity Index, considering only intense and moderate physical activity, scores 0-67.5; Inactive: PAI = 0, reporting not going out for more than 10 min at a time; Insufficient: PAI = 0, reporting to walk more than 10 min; Low: PAI between 1 and 15; Medium: PAI between 16 and 30; High: PAI between 31 and 45; Very high: PAI > 45; ** p Mann-Whitney U test: resulting from the mental health median comparison for every physical activity level.
The Stress factor (in mental health factor 3) scored less as the PAL increased. No significant differences were found between the "Medium" and "High" levels, but significant differences were found between the other groups (Table 7).  IQR: interquartile range; SD: standard deviation; FIII, Stress, from the Goldberg's General Health Questionnaire, scores between 0 and 9, 0 being for the best self-esteem and 9, the worst; Medians Diff: mental health medians differences for every physical activity level; Means diff: between mental health means differences for every physical activity level; p * Kruskal-Wallis value: mental health measured by GHQ-12 as response and physical activity level as a factor; PAI: Physical Activity Index, considering only intense and moderate physical activity, scores 0-67.5; Inactive: PAI = 0, reporting not going out for more than 10 min at a time; Insufficient: PAI = 0, reporting to walk more than 10 min; Low: PAI between 1 and 15; Medium: PAI between 16 and 30; High: PAI between 31 and 45; Very high: PAI > 45; ** p Mann-Whitney U test: resulting from the mental health median comparison for every physical activity level.
As shown in Table 8, weak correlations between mental health and PAL were found in the total population as well as in the men and women subgroups. These correlations were inverse, with the GHQ-12 score decreasing as the PAL increased.

Main Findings and Theoretical Applications
The main findings of this research are the associations between mental health and PAL in the Spanish population during the last pre-pandemic period analysed by the ENSE17 [41,42]. Thus, PAL seems to be linked with better mental health, coping, self-esteem, and stress levels. In addition, moderate and intense PAL showed stronger correlations with higher GHQ-12 scores. Although it is not the best option, in case of not being able to perform intense or moderate PA, walking seems to be a better alternative to physical inactivity for mental health care.
According to the data extracted from the ENSE 2017 and subsequent analysis, the Spanish population's mental health appeared to be at a relatively good level. A median score of 9 was found, with significant differences between men (9) and women (10). Values in the Spanish population were below 12, a threshold that may imply emotional disorders [48,49]. However, significant differences were found in the GHQ-12 median values according to the PAL. Inactive people showed 11 points median on the GHQ-12. The median for people who at least walked was 10, 9 for people with a "Low" PAL, and 8 for the other levels, with medians decreasing by as much as 3.53 points when comparing inactive people and those with a "Very high" PAL. In this case, significant differences were found between the GHQ-12 medians of the groups with a different PAL. Our results recommend at least a "Medium" PAL to protect mental health. Similar findings were found in other studies, in which PA was associated with less psychological distress and improved mental health [50], and sedentary behaviours were associated with poorer mental health [51]. In the same sense, people who performed low PA, such as walking, also improved successful coping, stress, or self-esteem, although lower than those who performed moderate and intense PA. The groups with higher PAL presented better values in the three mental health dimensions and significant differences between sedentary people and those who only walked. Therefore, moderate and intense PA is recommended for mental health care according to the GHQ-12 results. Other studies have indicated that higher PAL protects people from depression [36], anxiety and other disorders [22] compared with those with lower PAL.
These associations were also analysed during the COVID-19 pandemic, showing that people who exercised daily had fewer somatisation symptoms, lower stress, and better sleep levels than those who did not [52,53]. In addition, appropriate PA helped people to release psychological tension during confinement [54]. In this regard, symptoms related to anxiety, depression and stress were found in people with a lack of PA and a sedentary lifestyle [55][56][57][58]. Furthermore, the pandemic has negatively affected PAL, particularly in outdoor activities, which have been shown to have protective effects on well-being [53].
In the analysis of the dimensions, Self-esteem was the dimension that benefited the most from higher PAL, both in the total population and in the subgroups divided by sex. The "Inactive" PAL group had a 3.01 mean, compared to the 1.32 in the "Very high" group, a 56.1% reduction in the FII score. Increased self-confidence could be one of the main ways PA helped with mental health care in the Spanish population. However, the Stress dimension mean decreased from 3.13 in the "Inactive" to 1.82, representing a 41.8% decrease. The Successful Coping dimension was the least benefited by higher PAL: the mean decreased from 6.80 in the "Inactive" to 5.76 in the "Very high" group, a 15.3% decrease. Thus, better mental health due to increased PA could be due to increased self-esteem and reduced stress, with minor improvement in successful coping. In this line, several studies suggest that PA improves self-efficacy and coping with new challenges that build confidence and selfesteem [59,60]. Walking ≥105 min/week compared to <105 min/week was significantly and inversely associated with stress and anxiety [61]. In addition, self-reported PA was associated with lower subjective stress levels; even low/moderate daily PA was associated with significantly lower stress levels [62]. In this sense, PA could have a stress-reducing effect [63].
During the COVID-19 pandemic, stress levels and sleep quality improved in people who exercised regularly, showing decreased risk of depressive and anxiety symptoms in participants who reported ≥30 min of moderate-vigorous PA/day [52]. In contrast, participants who spent ≥10 h per day in sedentary activities were more likely to develop depressive symptoms [23]. Therefore, PA and emotional well-being associations keep before and during the COVID-19 pandemic. Concerning the sex subgroups, inactive women obtained a GHQ-12 median score of 12, the threshold for emotional distress, while this value decreased to 8 in women who performed higher PA. Also, in women, the PAL and self-esteem were related: higher self-esteem was reported at moderate and high PAL [64]. Inactive men had a self-esteem score of 10, which decreased to 8 in those who performed greater PA. According to our data, PA duration and intensity had a more significant influence on mental health in women than in men. Although there were differences between inactive women and those who performed light-intensity PA, those who performed <1 h, 1-2 h, or ≥3 h/week were more likely to develop anxiety. No inverse associations were observed between men and women [65].

Practical Applications and Future Lines
This study provides a baseline to analyse potential changes in the PA-mental health associations in the Spanish population once the future ENSE is published. As these surveys are conducted every 5 years, the next is expected to be published in 2023, with data from 2022, favouring future research on the impact of the pandemic on these associations.
Although the study design does not allow for cause-effect relations, longitudinal studies will provide the necessary information for the development of PA interventions and guidelines as a tool for mental health disease prevention and treatment.

Strengths and Limitations
The use of the ENSE 2017 is an excellent example of a nationally representative survey, and the sample size is the study's main strength.
However, some limitations must be mentioned: (1) as this is a cross-sectional study, it is not possible to establish cause-effect associations; (2) the GHQ-12 questionnaire is a screening instrument that may lead to an overestimation of mental health problems; moreover, some of the limitations of self-report questionnaires include social desirability and response bias, or item clarity, which may affect the validity and reliability of the tool [66]; (3) a 24 h compositional analysis, including objective PA parameters, was not included, so the PAI was determined using the parameters indicated in the ENSE 2017; (4) the survey did not collect PA data from people over 69 years of age, which could have affected the analysis: (5) changes in the future ENSE methodology could prevent comparisons with data obtained in this study.

Conclusions
This research found that PAL was positively related to mental health in the Spanish population before the COVID-19 pandemic. In addition, moderate and intense PAL showed stronger correlations with higher GHQ-12 scores.
These results need to be confirmed with longitudinal studies to recommend PA programs as a valid alternative to promote mental health.