Introduction
Numerous studies contend that regular exercise or health promoting or leisure time physical activity (PA) has a variety of health benefits. Some researchers assert that regular exercise can effectively lower the prevalence of many diseases [
1‐
3], and postpone the onset of disabilities [
4]. In addition, health promoting PA can also improve physical health [
2], and mental health [
5‐
11].
Health-related quality of life (HRQOL) has emerged as an important multidimensional concept in clinical and public health research in past decades [
9,
12,
13], and been recognized as a desired health outcome for evaluating preventive or therapeutic health plans [
14,
15]. The relationship between PA (including exercise) and HRQOL has been widely explored, and consistently demonstrated to be a positive association across age groups and countries [
8,
16,
17]. Higher levels of PA were related to better HRQOL [
18]. Higher dose of exercise exerted larger improvements in physical and mental quality-of-life [
10].
Risk and protective factors have been found to be related to HRQOL. Older age, female gender, living alone, lower education, and poorer physical health (i.e., more diseases or medications) would be associated with worse HRQOL [
19‐
21]. Moreover, unhealthy behaviors (i.e., smoking, alcohol drinking, betel-nut chewing) and mental/psychiatric disorders were related to poorer HRQOL [
22‐
26].
Although the impacts of PA on HRQOL have received great attentions, there are research gaps in the literature. First, the vast majority of studies on the relations of PA to HRQOL have been conducted in Western countries, whereas fewer studies were carried out in Asian nations. Because of the differences in life style, healthcare, and social determinants between Western nations and non-Western nations, further replication of prior observations are warranted. Second, although past studies explored the effects of different intensity or time of PA on physical or mental health, most of the findings of positive association of PA with HRQOL were usually found in cross-sectional studies or in the only two-wave longitudinal studies [
8,
11,
27‐
31]. The time effect of physical activities on HRQOL was little explored, and the time-varying covariates were not considered.
This study aims to explore the impacts of different degree of regular exercise on the trajectories of physical and mental dimensions of HRQOL for adults who participated in a multiple-wave community-based health screening program in Taiwan during 2006–2014.
Results
Table
2 exhibits the characteristics of participants at first four waves of LIONS. In contrast to stable PCS scores ranging from 53.0 to 53.6, MCS scores steadily increased from 49.7 to 52.2 during 2006–2014. Of all 6182 enrollees, 54.7% were women, the average age was 53.5 years. Although the proportion of participants without doing exercise remained around 35.5 to 38.4%, the percentage of participants spending more than 150 min weekly on doing exercise increased obviously from 12.7 to 18.4%.
Table 2
Descriptive statistics of HRQOL and covariates at 4 waves
PCS scorea | 53.0 ± 7.3 | 53.6 ± 6.8 | 53.4 ± 7.1 | 53.2 ± 7.1 |
MCS scorea | 49.7 ± 8.8 | 50.6 ± 8.3 | 50.9 ± 8.3 | 52.2 ± 8.1 |
Time-constant covariates |
Gender (female)e | 54.7% | | | |
Age (years)a | 53.5 ± 13.1 | | | |
30–64 years old | 80.2% | | | |
≧65 years old | 19.8% | | | |
Education (years)a | 10.1 ± 4.7 | | | |
Time-varying covariates |
Time (years)a,b | 0 ± 0 | 2.1 ± 0.3 | 1.9 ± 0.9 | 2.9 ± 0.5 |
Marital status |
Living with the spouse | 89.0% | 89.4% | 87.6% | 87.6% |
Living alone | 11.0% | 10.6% | 12.4% | 12.4% |
Tobacco smoking (yes) | 18.0% | 14.0% | 11.4% | 11.0% |
Betel-nut chewing (yes) | 3.2% | 2.0% | 1.4% | 1.0% |
Psychiatric disorder (CHQ-12≧4)e | 26.3% | 21.6% | 18.9% | 17.2% |
# of chronic diseasesa,c | 1.6 ± 1.4 | 1.7 ± 1.5 | 2.0 ± 1.5 | 2.0 ± 1.5 |
# of Medicationsa,d | 0.1 ± 0.4 | 0.1 ± 0.3 | 0.1 ± 0.4 | 0.1 ± 0.4 |
Regular exercise status |
No exercise | 36.9% | 35.5% | 38.4% | 36.7% |
Irregular exercise | 40.0% | 41.7% | 37.8% | 35.3% |
<150 min | 10.4% | 10.0% | 9.6% | 9.5% |
150–299 min | 5.7% | 5.9% | 6.8% | 8.1% |
≧300 min | 7.0% | 6.9% | 7.3% | 10.3% |
The characteristics of the participants according to their exercise level at the baseline were examined (see Table
3). There were differences in PCS and MCS across exercise levels, indicating a higher score accompanied with higher exercise level. The “no exercise” and “irregular exercise” groups were more likely to be female, older, lower educated, and having more chronic diseases.
Table 3
Similarity of HRQOL scores and covariates by regular exercise status at wave 1
PCS scorea,b | 52.3 ± 8.1 | 52.8 ± 7.3 | 54.7 ± 5.9 | 54.8 ± 5.1 | 54.5 ± 6.0 | < 0.001 |
PF | 86.8 ± 21.2 | 87.2 ± 19.1 | 93.1 ± 12.3 | 93.4 ± 11.9 | 92.8 ± 13.4 | < 0.001 |
RP | 79.7 ± 37.1 | 82.1 ± 35.0 | 86.6 ± 29.1 | 87.2 ± 28.4 | 88.2 ± 28.4 | < 0.001 |
BP | 81.8 ± 21.0 | 83.9 ± 18.9 | 83.7 ± 18.0 | 85.7 ± 16.8 | 85.3 ± 18.8 | < 0.001 |
GH | 62.0 ± 20.6 | 65.6 ± 19.8 | 68.1 ± 18.9 | 69.2 ± 17.6 | 70.3 ± 18.7 | < 0.001 |
MCS scorea,b | 48.5 ± 9.3 | 50.2 ± 8.7 | 49.1 ± 8.2 | 50.9 ± 7.2 | 51.9 ± 8.2 | < 0.001 |
VT | 63.1 ± 18.8 | 67.7 ± 18.1 | 67.2 ± 16.2 | 70.7 ± 17.0 | 73.7 ± 17.7 | < 0.001 |
SF | 84.8 ± 18.4 | 86.7 ± 15.8 | 87.2 ± 14.1 | 89.2 ± 13.0 | 89.5 ± 14.5 | < 0.001 |
RE | 82.9 ± 34.9 | 85.6 ± 32.5 | 85.1 ± 31.3 | 89.1 ± 28.5 | 88.6 ± 28.0 | 0.001 |
MH | 69.9 ± 17.5 | 73.1 ± 16.6 | 72.1 ± 16.0 | 74.4 ± 15.0 | 77.3 ± 16.9 | < 0.001 |
Gender (female)c | 56.1% | 58.9% | 42.9% | 48.8% | 45.9% | < 0.001 |
Age (years)a,b,c | 51.1 ± 12.4 | 57.3 ± 13.6 | 46.2 ± 9.4 | 51.1 ± 10.0 | 55.8 ± 11.0 | < 0.001 |
30–64 years old | 85.3% | 70.9% | 96.7% | 90.7% | 80.2% | < 0.001 |
≧65 years old | 14.7% | 29.1% | 3.3% | 9.3% | 19.8% | |
Education (years)a,b | 9.7 ± 4.6 | 9.4 ± 4.8 | 12.9 ± 3.3 | 11.8 ± 4.0 | 10.5 ± 4.1 | < 0.001 |
Marital statusd |
Living with the spouse | 87.4% | 90.0% | 87.6% | 91.8% | 90.1% | 0.016 |
Living alone | 12.6% | 10.0% | 12.4% | 8.2% | 9.9% | |
Tobacco smoking (yes)c | 24.0% | 12.3% | 21.1% | 15.4% | 15.6% | < 0.001 |
Alcohol drinking (yes)c | 13.0% | 9.5% | 19.5% | 14.2% | 13.5% | < 0.001 |
Betel-nut chewing (yes)c | 5.1% | 1.2% | 5.3% | 2.4% | 2.2% | < 0.001 |
Psychiatric disorder (CHQ-12≧4)c | 29.8% | 25.3% | 24.4% | 22.5% | 19.7% | < 0.001 |
# of chronic diseasesa,b,d | 1.64 ± 1.44 | 1.77 ± 1.51 | 1.32 ± 1.26 | 1.39 ± 1.28 | 1.58 ± 1.30 | < 0.001 |
# of Medicationsa,b,e | 0.11 ± 0.39 | 0.10 ± 0.36 | 0.08 ± 0.33 | 0.07 ± 0.29 | 0.10 ± 0.35 | 0.425 |
Table
4 shows the trajectories of PCS & MCS with contemporary regular exercise and other covariates by LMM. For the trajectory of PCS, Model 0 indicated an increasing growth curve (β = 0.72,
p < 0.001), but with decelerated growth rate (β = − 0.24,
p < 0.001), of PCS score over time. After controlling for regular exercise and other covariates, the deceleration growth curve of PCS score across time was confirmed in Models 1 and 2, because of significantly positive linear time slope (β = 0.81,
p < 0.001; β = 1.01,
p < 0.05) but negative quadratic time slope (β = − 0.25 & -0.22,
p < 0.001). Consequently, the trajectory of PCS score would increase at the beginning but slow down later on (i.e., a nonlinear, reversed U-shaped curve).
Table 4
Trajectories of HRQOL by regular exercise and covariates (n = 6182 persons, 16,281 observations)
Fixed effects |
For intercept |
Intercept | 53.02*** (52.83~53.20) | 52.25*** (51.96~52.54) | 60.67*** (59.26~62.07) | 49.74*** (49.51~49.96) | 48.52*** (48.17~48.87) | 44.51*** (42.80~46.21) |
Gender (female)d | | | −0.59** (− 0.99~ − 0.19) | | | − 0.26 (− 0.74~0.22) |
Education (years) | | | 0.03 (−0.01~0.08) | | | −0.002 (− 0.06~0.05) |
AGE (years) | | | −0.13*** (− 0.15~ − 0.11) | | | 0.09*** (0.07~0.11) |
Time-varying covariates |
Living with the spoused | | | 0.62* (0.09~1.14) | | | 1.77*** (1.14~2.40) |
Tobacco smoking (yes) | | | 0.14 (−0.36~0.64) | | | 0.06 (− 0.54~0.67) |
Betel-nut chewing (yes) | | | −0.18 (−1.20~0.83) | | | −0.32 (− 1.55~0.91) |
Psychiatric disorderd | | | −2.93*** (−3.31~ − 2.54) | | | −6.17*** (−6.64~ − 5.71) |
# of chronic diseasesb | | | −0.71*** (− 0.83~ − 0.59) | | | 0.09 (− 0.05~0.24) |
# of Medicationsc | | | −2.22*** (− 2.68~ − 1.76) | | | − 2.42*** (− 2.97~ − 1.86) |
Regular exercise statusd |
≧300 min | | 2.27*** (1.57~2.97) | 2.54*** (1.86~3.22) | | 3.39*** (2.55~4.24) | 2.03*** (1.21~2.85) |
150–299 min | | 2.68*** (1.91~3.45) | 1.99*** (1.25~2.72) | | 2.52*** (1.59~3.44) | 1.55** (0.65~2.44) |
<150 min | | 2.39*** (1.78~3.00) | 1.24*** (0.66~1.83) | | 0.71 (− 0.02~1.45) | 0.65 (− 0.06~1.36) |
Irregular exercise | | 0.55** (0.15~0.94) | 1.27*** (0.89~1.66) | | 1.89*** (1.41~2.36) | 0.87*** (0.40~1.33) |
For time slope |
Time (years)a | 0.72*** (0.42~1.03) | 0.81*** (0.48~1.15) | 1.01* (0.06~1.97) | 0.79*** (0.42~1.17) | 0.95*** (0.54~1.35) | 1.49** (0.37~2.62) |
Time2 (years) | −0.24*** (−0.35~ − 0.14) | − 0.25*** (− 0.36~ − 0.15) | −0.22*** (− 0.33~ − 0.10) | −0.06 (− 0.19~0.07) | −0.06 (− 0.19~0.07) | −0.05 (− 0.19~0.08) |
Gender (female)d | | | 0.02 (−0.23~0.27) | | | −0.14 (− 0.43~0.15) |
Education (years) | | | −0.003 (− 0.03~0.03) | | | −0.01 (− 0.05~0.02) |
AGE (years) | | | −0.01 (− 0.02~0.003) | | | −0.005 (− 0.02~0.01) |
Time-varying covariates |
Living with the spoused | | | −0.09 (−0.40~0.23) | | | −0.40* (− 0.77~ − 0.03) |
Tobacco smoking (yes) | | | 0.01 (−0.32~0.34) | | | −0.15 (− 0.54~0.24) |
Betel-nut chewing (yes) | | | 0.50 (−0.25~1.25) | | | 0.69 (−0.21~1.59) |
Psychiatric disorderd | | | 0.11 (−0.13~0.35) | | | 0.19 (−0.09~0.48) |
# of 15 diseasesb | | | 0.02 (−0.05~0.10) | | | −0.08 (− 0.17~0.005) |
# of Medicationsc | | | 0.05 (−0.23~0.32) | | | 0.10 (−0.22~0.43) |
Regular exercise statusd |
≧300 min | | −0.02 (− 0.41~0.37) | 0.03 (− 0.36~0.41) | | −0.43 (− 0.89~0.04) | 0.001 (− 0.46~0.47) |
150–299 min | | −0.24 (− 0.67~0.18) | 0.08 (− 0.34~0.50) | | −0.50 (−1.01~ − 0.001) | −0.11 (− 0.62~0.39) |
<150 min | | − 0.43* (− 0.78~ − 0.08) | 0.03 (−0.32~0.38) | | −0.10 (− 0.53~0.32) | −0.12 (− 0.54~0.30) |
Irregular exercise | | − 0.03 (− 0.26~0.20) | 0.04 (− 0.19~0.28) | | −0.21 (− 0.48~0.07) | 0.07 (− 0.20~0.35) |
Random effects |
Variance (Wave 1) | 53.76*** | 53.01*** | 44.84*** | 78.02*** | 77.07*** | 65.81*** |
Variance (Wave 2) | 29.31*** | 29.39*** | 27.81*** | 50.18*** | 50.46*** | 48.30*** |
Variance (Wave 3) | 35.20*** | 35.03*** | 32.82*** | 51.28*** | 50.76*** | 47.63*** |
Variance (Wave 4) | 23.73*** | 23.85*** | 26.37*** | 34.90*** | 35.25*** | 37.01*** |
Variance (Time) | 4.25*** | 4.06*** | 2.59*** | 4.60*** | 4.44*** | 2.80*** |
-2LL | 103,650.8 | 102,351.7 | 94,864.7 | 109,273.9 | 107,958.8 | 100,337.1 |
AIC | 103,660.8 | 102,361.7 | 94,874.7 | 109,283.9 | 107,968.8 | 100,347.1 |
BIC | 103,699.0 | 102,399.9 | 94,912.6 | 109,322.1 | 108,007.0 | 100,385.0 |
Model 1 exhibited that doing exercise regularly had significantly positive effects on the intercept of PCS score, including those irregular PA. That means those participants who were doing exercise at the beginning had better physical at the beginning. For the effects of regular exercise on the time slope, only doing exercise 150 min weekly would reduce the PCS score over time (β = − 0.43, p < 0.05). With the inclusion of other covariates in Model 2, regular exercise still had significantly positive effects on the initial status of PCS score (at the intercepts), which seemed to be a dose-response relationship based on the levels of effective PA (β = 1.20–2.55, p < 0.001). However, regular exercise became insignificant for the changes of PCS score over time (p > 0.05), indicating constant effects of regular exercise on PCS score across time.
Model 2 also showed that some covariates had substantial and varying effects on the intercept, but not on the time slope, of PCS score. For time-constant variables, female (β = − 0.59, p < 0.01) or older β = − 0.13, p < 0.001) participants had a stably lower initial PCS score over time. Similarly, there were five time-varying covariates with significant effects only on the intercept of PCS score. Both living with a spouse (β = 0.62, p < 0.05) was positively related to the initial status of PCS score. Psychiatric disorder (β = − 2.93, p < 0.001), the number of chronic diseases (β = − 0.71, p < 0.001) and taken medications (β = − 2.22, p < 0.001) had negative effects on the intercept of PCS score. However, these covariates were in-significant across time.
For the trajectory of MCS score, Model 0 showed an increasing growth curve (β = 0.79, p < 0.001) across time. Even if regular exercise and other covariates were controlled in Models 1 and 2, the increasing tendency of MCS score over time still existed, due to significantly positive linear time slope (β = 0.95, p < 0.001; β = 1.49, p < 0.01). That means the growth curve of MCS score would be an increasing linear trajectory over time.
Model 1 indicated that regular exercise had significantly positive and varying impacts on the intercept of MCS score (p < 0.001), except for doing low level exercise (< 150 min weekly). However, doing exercise regularly had no time effect on MCS score in Model 1.
After controlling for other covariates in Model 2, most levels of regular exercise, except for the low level (<150 min per week), were significantly and positively associated with MCS score at the intercept. Additionally, for those significant regular exercise, higher level of effective PA (i.e., ≧300 min weekly) was related to better initial status of MCS score (β = 2.03, p < 0.001). The effect of irregular physical activities was still significant. However, regular exercise had no significant effect on changing the MCS score over time.
Model 2 exhibited that several covariates had significant and varying impacts on the intercept and linear time slope of MCS score. Age (β = 0.09, p < 0.001) was positively related to the initial status of MCS score, indicating that older participants would perceive better mental QOL constantly. Living with a spouse had a positive effect on the intercept (β = 1.77, p < 0.001) but a negative impact on the time slope (β = − 0.40, p < 0.05) of MCS score. These results revealed that participants living with a spouse had higher MCS score but it would decrease over time. Both psychiatric disorder (β = − 6.17, p < 0.001) and the number of taken medications (β = − 2.42, p < 0.001) only had negative effects on the initial status of MCS score. Namely, participants with the psychiatric disorder or taking more drugs would have lower MCS score at the very beginning. However, morbidities and medications were not significantly on the time slope.
Discussion
This study used a multiple-wave longitudinal panel data of community-dwelling adults in Northern Taiwan to examine the effects of regular exercise on health-related quality of life over time among community adults. Those with regular exercise showed a significant improving effect on physical health than those without doing exercise at the intercepts, especially if the regular exercise time was more than 150 min a week. Regular exercise also showed positive effects (at the intercepts) on mental health if the exercise time was 150 min or more in a week. However, the regular exercise effects were not significant over time (at time slope) on either physical or mental health scores. In addition, physical health and mental health improved over time, but the time effect would offset a little bit on physical health. Moreover, being female, increasing age, living alone, or poorer health status were related to lower physical health, but drinking alcohol was related to better physical health. The participants who were younger, lived alone or had poorer health status would have worse mental health.
The present research demonstrates that compared to non-exercising almost all levels of PA had positive effects on PCS and MCS. These results were consistent with the literature that usually speculated the varying positive associations of exercise/PA with overall quality-of-life [
16,
18,
45,
46], PCS [
10,
30,
46‐
50], MCS [
10,
11,
30,
47], or subjective well-being [
27,
28,
51,
52].
In our study, 76.9% of our participants did not have exercise habits or did not meet the recommendation criteria. Previous research indicated that 86% of Taiwanese did not meet the national recommendation [
28]. However, the present research showed that even irregular levels of PA lower than currently recommended for beneficial health effects [
38] were also related to the improvement in PCS. Previous research suggested the positive association of light-intensity PA to different aspects of well-being [
28], and the low PA level (40–179 min per week) [
47] or even irregular excise [
46] still had significantly positive effects on PCS. Therefore, doing exercise regularly would be beneficial for improving physical health, no matter the amount and intensity of PA. Further, this study found the insignificant impact of PA on the linear time slope, indicating that the effects of PA on HRQOL would be constant over time. A curvilinear trend between PA levels and HRQOL, particularly physical functioning and vitality in previous study [
47], was not discovered in this study. It is possibly attributed to using different measures of PA, or missing some confounders to physical health.
Regarding to the PA effect on mental health, although the level of <150 min was insignificant, further analyses of MCS subscales (see Additional file
2) revealed that it had significantly positive effects on three domains, including vitality, social functioning and mental health. Wicker and Frick [
27] suggested that light-intensity PA would be beneficial for improving individuals’ mental health too. However, some past research found the insignificant relationship between PA and role limitations due to emotional problems [
11,
50]. Therefore, all levels of PA would have varying influences on MCS and its domains, including irregular exercise. Besides, our results support the conclusion that changes in HRQOL were dose dependent on the amount of PA [
10,
18,
46‐
49].
The trajectory of physical health would increase at first and decrease after a period of time, consistent with the literature. Previous studies have found the trajectories of HRQOL for chronic diseases may be improved by medical interventions [
53‐
55]. Moreover, this study found the aging effect would offset the physical health in quality-of-life, consistent with previous research [
56]. The trajectory of mental health was also found to be positively related to time linear effect as previous research [
57,
58]. The time effect to improve mental health can be explained by the socioemotional selectivity theory [
59]. Increasing age will limit the future prospections of people and then make them focusing on positive information and seeking for the emotional satisfaction [
60,
61], and thus people would employ various coping strategies to focus on potential positive outcomes of unfavorable events such as chronic diseases or disabilities.
Previous studies usually contended that women had worse HRQOL than men in general or in specific domains, even after controlling for sociodemographic characteristics, economic status and chronic diseases [
20,
62‐
64]. However, our results indicated that female adults only had significantly lower PCS score. The insignificant gender difference in MCS score could be attributed to the inclusion of psychiatric/mental disorder and PA/exercise in the models, which would reduce the direct effect of gender on HRQOL, especially MCS. Previous studies concluded that psychiatric disorder (e.g., anxiety, depression) was more prevalent in women and most important predictor of MCS score [
25], and PA had larger positive effects on MCS for women [
30,
65]. This study discovered that chronic diseases had negative impacts on PCS score, consistent with previous literature. However, the relationship between chronic diseases and MCS score was found to be nonlinear, which could be partially explained by psychological adjustment to chronic diseases [
59,
60]. Namely, the positivity effect of psychological adjustment could be offset by the cumulative burden of chronic diseases over time, resulting in the curvilinear association of MCS score with chronic diseases. Additionally, only one unhealthy behavior (i.e., alcohol drinking) was positively related to PCS score, which was supported by some studies [
24,
66] but contrary to other research [
23]; nevertheless, other unhealthy behaviors had insignificant and inconsistent impacts on HRQOL. These results could be attributed to binary non-quantity measures [
67] and small cases with risky behaviors (e.g., betel-nut chewing).
Study limitations
This study could be improved in several ways. First, although this study was based on a sizable sample with four repeated observations during 2006–2014, the sample could not be generalized to the district or other populations. Further replications with data collected in Taiwan and other Asian countries would be required to extend the generalizability of our findings. Second, self-reported exercise data could be refined by the adoption of objective instruments, which may reduce measurement errors (i.e., potential overlapping between self-reported data and subjective HRQOL) and avoid incorrect statistical inference [
18,
28,
52]. Third, the intensity of regular exercise and all the forms of physical activity can be measured by metabolic equivalents (METS) for a better measure of physical activity. Fourth, a refined model with fewer significant covariates but more interaction terms of PA and time could generate more accurate effects of PA on the trajectory of HRQOL over time. Fourth, alcohol drinking behavior was not included in the model, because the amount of alcohol drinking was not available. Alcohol drinking may be a confounder to HRQOL. In fact, we tried to put alcohol drinking (yes/no) in the models, and it showed a positive effect on PCS at the intercept but non-significant on the time slope or on MCS. A more detailed alcohol intake is suggested to examine the longitudinal effect of alcohol drinking on HRQOL. Fifth, there were attritions in the follow-ups. Only 1.5%~ 4.2% of the non-participants because of death or loss in the follow-ups. Most of the non-participants (more than 95%) did receive the invitation in the next following waves and considered or agreed to participate in the survey, but eventually they did not show up. We further compared the baseline characteristics between the participants and non-participants, and the participants were more likely to be healthier than the non-participants (please see Additional file
3). Therefore, the sample may be over-estimated the health and healthy lifestyle from the population.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.