Joint associations of physical activity and sleep duration with falls: evidence from the China health and retirement longitudinal study

Abstract

Objective

Population aging is accelerating in China, making falls a critical public health challenge. This study aimed to examine the independent and joint associations of physical activity (PA) and sleep duration with the risk of falls among Chinese older adults.

Method

We analyzed 10,232 participants aged 60 years and older from the 2020 China Health and Retirement Longitudinal Study (CHARLS) wave. The occurrence of falls was identified based on self-reported incidents since the previous interview wave (2018) for returning participants or at the time of entry for new participants. Sleep duration was classified into three categories: short (< 6 h/night), optimal (6–8 h/night), and long (> 8 h/night). Physical activity was assessed using the International Physical Activity Questionnaire (IPAQ) short form and dichotomized into low physical activity (LPA) and moderate-to-vigorous physical activity (MVPA). Chi-square tests and multivariate logistic regression models were employed to estimate the independent and joint associations of sleep duration and PA with falls. Subgroup and sensitivity analyses were further conducted to evaluate the robustness of the findings.

Results

The prevalence of falls among the participants was 20.1%. Compared with LPA, MVPA was associated with a lower odds of falls (OR = 0.842, 95% CI: 0.745–0.952). Compared with optimal sleep duration, those with short sleep duration (OR = 1.582, 95% CI: 1.424–1.756) and long sleep duration (OR = 1.200, 95% CI: 1.003–1.436) had a higher likelihood of falls. The risk of falls was approximately 46.4% lower for individuals with LPA and optimal sleep duration compared to those with short sleep and LPA (Model 3, OR = 0.536, 95% CI: 0.425 = 0.677). In MVPA group, the mitigation effect of exercise on falls was observed regardless of sleep duration (Model 3, short: OR = 0.797, 95% CI: 0.669–0.949; optimal: OR = 0.524, 95% CI: 0.440–0.626; long: OR = 0.564, 95% CI: 0.440–0.722). Subgroup analyses confirmed that the protective association of MVPA and optimal sleep duration was particularly pronounced among males and those aged 60–74 years. Sensitivity analyses demonstrated the consistent robustness of these findings.

Conclusion

Both physical activity and sleep duration are significantly associated with the risk of falls in Chinese older adults. Our findings suggest that the combination of MVPA and optimal sleep duration (6–8 h) is associated with the lowest risk of falling. While the cross-sectional nature of this study limits causal inference, integrated interventions targeting both sleep hygiene and physical activity may be beneficial for fall prevention strategies in the older adults. Future studies should employ prospective designs and objective measures to investigate the causal relationships and underlying mechanisms in greater depth.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28089-3.

Introduction

Falls in older adults have become a major public health concern worldwide. According to statistics from the World Health Organization, approximately 684,000 fatal falls occur globally each year, with individuals over 60 accounting for more than half of these cases [1]. Between 1990 and 2019, China also experienced a marked increase in the burden of falls among older adults [2]. As the aging population increases, more individuals will be at risk of falling [3]. The consequences of falls among older adults are multifaceted, spanning physical, psychological, and economic dimensions. Physically, injurious falls often result in severe outcomes such as hip fractures and traumatic brain injuries, which are leading causes of disability and mortality in this age group [4, 5]. Beyond immediate physical harm, even non-injurious falls can trigger a “fear of falling,” leading to anxiety, depression, and a self-imposed restriction of movement [6]. Moreover, the economic burden associated with fall-related care is substantial and continues to escalate globally [7]. Therefore, early screening that incorporates modifiable risk factors is essential to preserve the health and social participation of the aging population. It is widely recognized that falls in older adults are multifactorial events, influenced by a complex interplay of behavioral, environmental, nutritional, and pharmacological factors [8]. While environmental hazards, nutritional status, and medication use are critical determinants, behavioral lifestyles such as physical activity (PA) and sleep patterns can provide modifiable avenues for fall intervention.

PA is any bodily movement that results in increased energy expenditure, and can be achieved by a variety of leisure-time, work or transportation-related activities [9, 10]. Ample evidence now exists that regular physical activity is key to preventing falls and managing balance impairments common to older adults [11–13]. A recent ten-year population-based longitudinal study in Australia found that older adults who increased their own moderate-to-vigorous physical activity (MVPA) further reduced the risk of falls [14], while some studies found that light physical activity (LPA) were associated with falls in older adults [15, 16]. Furthermore, PA is vital for preserving physical function, mental health, and cardiovascular function; these collective benefits enhance postural control and gait stability, thereby promoting mobility and independence while significantly reducing the risk of falls [17]. Even simple activities like walking contribute significantly to cardiometabolic health, with research suggesting that maintaining a brisk pace can further enhance these protective benefits compared to slower speeds [18]. In fact, older adults deeply value elements such as healthcare access, social relationships, functional autonomy, and staying active as core components of their well-being; thus, engaging in PA constitutes a fundamental element of their overall quality of life (QOL) [19].

In addition to PA, sleep duration has emerged as a critical determinant of health outcomes in the ageing population [20], and it is important for maintaining physiological homeostasis, cognitive integrity, and physical recovery [21]. Growing evidence indicates that both short and long sleep duration are significantly associated with an increased risk of falls among older adults [22–24]. Specifically, short sleep duration may lead to daytime drowsiness [25], impaired executive function [26], and slowed reaction times [27]; these cognitive deficits diminish old adults’ ability to identify and avoid environmental hazards, thereby elevating fall risk. Additionally, prolonged sleep duration is often linked to underlying frailty [28] and reduced muscle strength [29], both of which make old adults more susceptible to losing their balance and falling.

Despite the established importance of both PA and sleep in fall prevention, most existing literature has examined these two behaviors in isolation. However, PA and sleep are intrinsically linked; for instance, regular PA is known to enhance sleep quality [30], while sufficient sleep provides the necessary energy and cognitive alertness required to maintain an active lifestyle [31, 32]. Therefore, exploring how PA levels combines with sleep duration to influence fall risk is essential for developing comprehensive public health interventions.

To address this research gap, this study utilized the data from the fifth Wave of the China Health and Retirement Longitudinal Study (CHARLS) to: (1) investigate the independent associations of PA and sleep duration with the risk of falls; (2) examine the joint associations of these two behavioral factors on fall risk; and (3) explore whether these joint associations vary across different gender and age subgroups. We hypothesized that the combination of meeting recommended PA levels and optimal sleep duration would be associated with the lowest risk of falls, although the magnitude of this association may differ by sex and age.

Method

Study design and population

This study used a cross-sectional design and was based on a secondary analysis of the 2020 China Health and Retirement Longitudinal Study (CHARLS). CHARLS is a nationally representative, large-scale longitudinal survey covering 28 provinces (including autonomous regions and municipalities) across China. It aims to collect comprehensive information on the health, economic status, and social condition of Chinese adults aged 45 years and older. The study protocol was approved by the Institutional Review Board of Peking University (No. IRB00001052-110155), and all the participants or their proxies provided written informed consent [33]. Further details are available in the cohort profile [34].

The 2020 CHARLS database initially comprised a total of 19,395 participants. To establish the final analytical sample, the following inclusion criteria were applied: (1) participants aged 60 years or older, (2) availability of complete data from the sleep duration and physical activity, and (3) available self-reported data on fall incidents. We excluded participants with missing data on the physical activity, sleep duration or self-reported fall status. Further exclusions were made for missing values in key covariates, including demographic characteristics, chronic disease history and lifestyle habits. Ultimately, 10,232 individuals were included in the final analysis for the year 2020 (Fig. 1).

Fig. 1.

Flowchart of subject selection process

Assessments

Fall

The primary outcome of this study was the occurrence of falls. In the 2020 CHARLS questionnaire, fall incidents were assessed based on participants’ self-reported history. Specifically, participants were asked, “Have you fallen since the last interview wave (2018) ?” (for returning participants) or “Have you ever fallen?” (at the time of entry for new participants). A “Yes” response was defined as a fall occurrence. This approach ensured a continuous and comprehensive record of fall events across the study population since their previous assessment or baseline enrollment.

Physical Activity (PA)

PA was assessed using questionnaire items from CHARLS, which adopted a format consistent with the short form of the International Physical Activity Questionnaire (IPAQ) [35]. CHARLS categorized physical activity intensity into three levels: (1) light physical activity (LPA), such as walking for transportation, leisure, or light household duties; (2) moderate physical activity (MPA), including tasks like carrying light loads, Tai Chi, or brisk walking; and (3) vigorous physical activity (VPA) including activities such as heavy lifting, digging, aerobic exercise, or fast cycling. Researchers asked each Participant, “Do you usually participate in at least 10 minutes of LPA/MPA/VPA per week?” If the participant answered “no”, it was concluded that they do not regularly engage in that type of physical activity. If the participants answered “yes”, researchers further inquired about the weekly frequency (1–7 days) and duration of each type of LPA/MPA/VPA (≥ 10 and < 30 min, ≥ 30 min and < 2 h, ≥ 2 h and < 4 h, and ≥ 4 h). Since no specific duration was mentioned in the questionnaire, drawing in the treatment of other scholars [36], we transformed the time range. That is, “≥ 10 and < 30 min” was recorded as 30 min, “≥ 30 min and < 2 h” was recorded as 60 min, “≥ 2 h and < 4 h” was recorded as 180 min, and “≥ 4 h” was recorded as 240 min. And in order to ensure the accuracy of grouping, this study processed physical activity data based on the data truncation principle. For each intensity of physical activity, this principle allows for a maximum of 21 h (1260 min) to be reported per week, which means a maximum of 180 min to be reported per day. Therefore, for individuals who engage in physical activity of a certain intensity for more than 3 h per day, their activity time is uniformly encoded as 180 min to avoid extreme values affecting the analysis results [37].

Based on IPAQ, metabolic equivalent (MET) were assigned to each type of physical activity [35]. The MET values assigned as follows: 3.3 for LPA, 4.0 for MPA, and 8.0 for VPA. The formula for calculating the total physical activity (TPA) is as follows: TPA = (duration of VPA per day × number of days of per week × 8.0 MET) + (duration of MPA per day × number of days of per week × 4.0 MET) + (duration of LPA per day × number of days of per week × 3.3 MET) [35]. TPA was categorized into LPA (< 600 MET-min/w) and MVPA (≥ 600 MET-min/w) based on the standard scoring protocol of the IPAQ [35] and the World Health Organization (WHO) guidelines on physical activity, which define 600 MET-min/week as the minimum threshold for achieving health benefits [38]. To further justify the rationality of this categorization, a sensitivity analysis was performed by treating TPA as a continuous variable in the regression model, suggesting that no statistically significant between TPA and falls (Waldχ² = 0.039, P > 0.05) (Supplementary Table S1).

Sleep duration

Sleep duration was available from the lifestyle and health behaviors section of the CHARLS questionnaire. More specifically, the item was “during the past month, how many hours of actual sleep did you get on average per night?” The participants provided the answers in hours and minutes. Self-reported nighttime sleep duration has been widely used in previous studies and is considered to have acceptable reliability [39]. Based on the results of previous studies, we divided sleep duration into three groups: short sleep (< 6 h), optimal sleep (6–8 h), and long sleep (≥ 8 h) [40].

Covariates

Covariates were selected based on the prior literature and the availability of relevant data in CHARLS [41, 42], with consideration given to their potential role as confounders of the primary association. The following variables were included: age (60–74 years = 0, ≥ 75 years = 1), gender (male = 0, and female = 1), marital status (married = 0, unmarried = 1), educational attainment (illiteracy = 0, elementary school and below = 1, middle school and above = 2), residence (urban = 0, rural = 1), chronic disease history (no = 0, yes = 1), drinking status (no = 0, yes = 1), smoking status (no = 0, yes = 1), daytime napping (no = 0, yes = 1).

Statistical analysis

Categorical data were presented as frequency (n) and percentage (%). Group comparisons were performed using the chi-square test. Multivariate logistic regression analysis was used to analyse the joint associations of PA and sleep duration with falls among older adults. Subgroup analysis was used to evaluate differences across different age groups and genders. Sensitivity analysis was used to ensure the robustness of our primary findings. The results were expressed by OR, 95% CIs and corresponding P-value. The confounding factors of age, gender, marital status, educational attainment, residence, chronic disease history, drinking status, smoking status and daytime napping were adjusted. P-value < 0.05 was considered as significant difference. SPSS v27.0 and Stata v18.0 software were used for data statistical analysis. GraphPad Prism v10.0 was used to draw forest plots of subgroup analysis and sensitivity analysis.

Results

Characteristics of participants

The participants in this study exhibited the following characteristics (Table 1). First, 20.1% of the participants had experienced falls based on their self-reported history. In terms of age, the majority (80.2%) were between 60 and 74 years old. The sex distribution were relatively balanced, with 48.9% males and 51.1% females. Marital status showed that the majority of participants (77.1%) were married, and 22.9% were unmarried. Regarding educational level, nearly half of the participants (44.2%) had an education level below elementary school, and only 26.0% had received a middle school education or above. In terms of residence, 65.5% of participants lived in rural areas, whereas 34.4% lived in urban areas.

Table 1.

Characteristics of participants and comparison of characteristics between non-fall and fall groups

Variable	Total (n = 10,232)	Non-Fall Count (n = 8,178)	Fall Count (n = 2,054)	χ2/t
Age				31.124***
60–74	8,205 (80.2%)	6,648 (81.3%)	1,557 (75.8%)
≥ 75	2,027 (19.8%)	1,530 (18.7%)	497 (24.2%)
Gender				95.681***
Male	5,002 (48.9%)	4,196 (51.3%)	806 (39.2%)
Female	5,230 (51.1%)	3,982 (48.7%)	1,248 (60.8%)
Marital status				39.348***
Married	7,885 (77.1%)	6,409 (78.4%)	1,476 (71.9%)
Unmarried	2,347 (22.9%)	1,769 (21.6%)	578 (28.1%)
Education				57.009***
Illiteracy	3,052 (29.8%)	2,308 (28.2%)	744 (36.2%)
Elementary school and below	4,523 (44.2%)	3,655 (44.7%)	868 (42.3%)
Middle school and above	2,657 (26.0%)	2,215 (27.1%)	442 (21.5%)
Residence				7.412**
Urban	3,524 (34.4%)	2,869 (35.1%)	655 (31.9%)
Rural	6,708 (65.6%)	5,309 (64.9%)	1,399 (68.1%)
Chronic disease				79.196***
No	1,387 (13.6%)	1,232 (15.1%)	155 (7.5%)
Yes	8,845 (86.4%)	6,946 (84.9%)	1,899 (92.5%)
Drinking status				8.955**
No	6,885 (67.3%)	5,446 (66.6%)	1,439 (70.1%)
Yes	3,347 (32.7%)	2,732 (33.4%)	615 (29.9%)
Smoking status				39.205***
No	5,623 (55.0%)	4,368 (53.4%)	1,255 (61.1%)
Yes	4,609 (45.0%)	3,810 (46.6%)	799 (38.9%)
Napping				0.783
No	3,497 (34.2%)	2,778 (34.0%)	719 (35.0%)
Yes	6,735 (65.8%)	5,400 (66.0%)	1,335 (65.0%)
PA level				21.758***
LPA	1,919 (18.8%)	1,460 (17.9%)	459 (22.3%)
MVPA	8,313 (81.2%)	6,718 (82.1%)	1,595 (77.7%)
Sleep duration				121.224***
Short sleep (< 6 h)	4,063 (39.7%)	3,037 (37.1%)	1,026 (50.0%)
Optimal (6–8 h)	5,228 (51.1%)	4,390 (53.7%)	838 (40.8%)
Long sleep (> 8 h)	941 (9.2%)	751 (9.2%)	190 (9.3%)

Abbreviations: PA Physical activity, LPA Light physical activity, MVPA Moderate-to-vigorous physical activity

^**P-value < 0.01, ^*** P-value < 0.001

Chronic disease history shows that majority of participants (86.4%) had chronic diseases, whereas 13.6% of participants do not have. Regarding lifestyle habits, 45% of participants have smoked, 67.3% did not drink alcohol, and 65.8% of participants took a nap after lunch. In terms of PA level, 81.2% engaged in moderate-to-vigorous activity, and 18.8% engaged in light activity. Regarding sleep duration, 39.7% of participants slept less than 6 h per night, 51.1% of participants slept between 6 and 8 h, and 9.2% slept more than 8 h.

Comparison of characteristic between non-fall and fall group

Table 1 provides a detailed comparison of the distribution between non-fall and fall individuals across three key domains: demographic characteristics, chronic disease history and lifestyle habits. The Chi-square test confirmed significant differences across most categories (P < 0.01), with the exception of napping behaviour. Regarding demographic characteristics, the non-fall group predominantly consisted of individuals aged 60–74 years (81.3%), whereas the fall group had a higher proportion of participants age 75 years and above (24.2% vs. 18.7%). Female were significantly prevalent in the fall group (60.8%) than in the non-fall group (48.7%). Education attainment also showed disparities: 42.3% of the fall group had an educational level below elementary school, and only 21.5% of the fall group had received middle school education or above, compared with 27.1% in the non-fall group. In terms of residence, 68.1% of fall individuals lived in rural areas, a higher percentage than 64.9% observed in the non-fall group. By contrast, 35.1% of the non-fall group resided in urban areas compared to 31.9% in the fall group.

In terms of chronic disease history, a higher proportion of fall individuals (92.5%) had at least one chronic disease compared with the non-fall group (84.9%). Regarding lifestyle habits, the fall group had a lower proportion of drinkers (29.9%) than did the non-fall group (33.4%). Similarly, the proportion of smokers was lower in the fall group (38.9%) than that in the non-fall group (46.6%). Regarding napping behavior, both groups showed nearly identical coverage, with 65.0% in the fall group and 66.0% in the non-fall group, indicating no significant difference.

Finally, the analysis of core exposure variables revealed significant differences in both PA level and sleep duration. Regarding PA level, 22.3% of the fall group engaged in LPA, which was significantly higher than the 17.9% of the non-fall group. Conversely, a small proportion of the fall group (77.7%) engaged in MVPA compared with the non-fall group (82.1%). Sleep duration analyses showed that 50.0% of fall individuals slept less than 6 h per night, a substantially higher percentage than 37.1% observed in the non-fall group.

Univariate logistic regression analyses of factors affecting falls in old adults

To identify potential risk factors, univariable logistic regression analysis was performed for each candidate (Table 2). Several factors were found to be significantly associated with the odds of falling among old adults (P < 0.01). The analysis revealed that the ≥ 75 years group had significantly higher risks compared with the 60–75 years group (OR = 1.387; 95% CI: 1.236–1.556). Females (OR = 1.632; 95%CI: 1.478–1.801) and unmarried individuals (OR = 1.419; 95% CI: 1.271–1.583) also exhibited a significantly increased risk. Having chronic diseases was the strongest individual risk factor, associated with more than double the odds of falling (OR = 2.173; 95% CI: 1.825–2.588). Conversely, higher education attainment, living in urban areas, and smoking or drinking status were associated with lower odds of falls in this sample (all P < 0.01).

Table 2.

Univariate logistic regression analysis evaluated the effect of various factors on the incidence of fall

Variable	B	SE	Waldχ2	OR (95% CI)
Age group (Ref = 60–74)
≥ 75	0.327	0.059	30.942	1.387 (1.236–1.556)***
Gender (Ref = Male)
Female	0.490	0.050	94.678	1.632 (1.478–1.801)***
Marital status (Ref = Married)
Unmarried	0.350	0.056	39.100	1.419 (1.271–1.583)***
Education (Ref = Illiteracy)
Elementary school and below	-0.306	0.057	29.150	0.737 (0.659–0.823)***
Middle school and above	0.480	0.067	51.215	0.619 (0.543–0.706)***
Residence (Ref = rural)
Urban	-0.143	0.053	7.405	0.866 (0.781–0.961)**
Chronic disease (Ref = No)
Yes	0.776	0.089	75.925	2.173 (1.825–2.588)***
Drinking status (Ref = No)
Yes	-0.160	0.054	8.944	0.852 (0.767–0.946)**
Smoking status (Ref = No)
Yes	-0.315	0.050	39.032	0.730 (0.661–0.806)***
Napping (Ref = No )
Yes	-0.046	0.052	0.783	0.955 (0.863–1.057)
PA level (Ref = LPA)
MVPA	-0.281	0.060	21.662	0.755 (0.671–0.850)***
Sleep duration (Ref = Optimal, 6–8 h)
Short (< 6 h)	0.571	0.052	119.589	1.770 (1.598–1.960)***
Long (> 8 h)	0.282	0.090	9.899	1.325 (1.112–1.580)**

Abbreviations: SE Standard error, Ref Reference group, PA Physical activity, LPA Light physical activity, MVPA Moderate-to-vigorous physical activity, OR Odds ratio, CI Confidence interval

^**P-value < 0.01, ^***P-value < 0.001

Regarding our core exposure variables, both PA level and sleep duration showed significant associations with fall risk. Compared with individuals engaging in LPA, those who participated in MVPA had a 24.5% lower risk of falling (OR = 0.755; 95% CI: 0.671–0.850). In terms of sleep duration, compared with optimal sleep (6–8 h), both short sleep (< 6 h; OR = 1.770; 95% CI: 1.598–1.960) and long sleep (> 8 h; OR = 1.325; 95% CI: 1.112–1.580) were significantly associated with increased fall risk.

Notably, napping behaviour was the only variable that did not show a statistically significant association with the incidence of falls in the univariate model (OR = 0.955; 95% CI: 0.863–1.057; P > 0.05). All variables with significant associations in the univariate analysis, along with napping behavior as a theoretically relevant confounder, were included in the subsequent multivariate analysis.

Independent associations of physical activity and sleep duration with falls

The results of the multivariate logistic regression analysis were presented in Table 3. After adjusting for all potential confounders, including demographic characteristics (age, gender, marital status, education, and residence), chronic disease history, and lifestyle habits (smoking, drinking, and napping behavior), both PA level and sleep duration remained significantly and independently associated with the risk of falls.

Table 3.

Multivariate logistic regression analysis of factors influencing falls in older adults

Variable	B	SE	Waldχ2	OR (95% CI)
PA level (Ref = LPA)
MVPA	-0.172	0.063	7.516	0.842 (0.745–0.952)**
Sleep duration (Ref = Optimal, 6–8 h)
Short (< 6 h)	0.458	0.054	73.402	1.582 (1.424–1.756)***
Long (> 8 h)	0.182	0.092	3.963	1.200 (1.003–1.436)*

Analyses were adjusted for PA level when nighttime sleep duration was the independent variable, and for nighttime sleep duration when PA level was the independent variable.Adjusted for age, gender, marital status, education level, residence, chronic disease history, drinking status, smoking status, and napping

Abbreviations: SE Standard error, Ref Reference group, PA Physical activity, LPA Light physical activity, MVPA Moderate-to-vigorous physical activity, OR Odds ratio, CI Confidence interval

^*P-value < 0.05, ^**P-value < 0.01, ^***P-value < 0.001

Specifically, compared with individuals engaging in LPA, those participating in MVPA showed a significantly lower risk of falls (OR = 0.842; 95% CI: 0.745–0.952; P < 0.01). Regarding sleep duration, the protective association remained robust in the fully adjusted model. Compared with optimal sleep (6–8 h), short sleep (< 6 h) increased the odds of falling by 58.2% (OR = 1.582; 95% CI: 1.424–1.756; P < 0.001). Similarly, long sleep (> 8 h) was also independently associated with a increased risk of falls (OR = 1.200; 95% CI: 1.003–1.436; P < 0.05). These results suggest that sufficient sleep and higher intensity of physical activity are both independent protective factors against falls in older adults.

Joint associations of physical activity and sleep duration with falls

To further elucidate the combined impact of PA and sleep duration on falls, participants were categorized into six groups based on their PA levels and sleep duration. The joint association were evaluated using three hierarchical logistic regression models (Table 4). In Model 1, using the “LPA and Short sleep” group as the reference, all other combination exhibited varying degrees of protective associations against falls. Notably, the MVPA and Optimal sleep group showed the strongest protective effect (OR = 0.432; 95% CI: 0.364–0.513; P < 0.001). This association remained highly significant after adjusting for demographic factors in Model 2 (OR = 0.502; 95% CI: 0.421–0.598; P < 0.001). In fully adjusted model (Model 3), which further controlled for chronic disease history, smoking, drinking, and napping behavior, the MVPA and Optimal sleep group still demonstrated the lowest risk of falls (OR = 0.524; 95% CI: 0.440–0.626; P < 0.001). Similarly, the LPA and Optimal sleep (OR = 0.536; 95% CI: 0.425–0.677; P < 0.001) and the MVPA and Long sleep (OR = 0.564; 95% CI: 0.440–0.722; P < 0.001) groups also maintained significant inverse associations with fall risk compared to the reference group.

Table 4.

Multivariate logistic regression analysis of the combined effect of PA and sleep dration on falls in older adults

Group	Model 1	Model 2	Model 3
	OR (95% CI)	OR (95% CI)	OR (95% CI)
LPA and Short sleep	1.000 (Ref)	1.000 (Ref)	1.000 (Ref)
LPA and Optimal sleep	0.483 (0.384–0.608)***	0.521 (0.413–0.657)***	0.536 (0.425–0.677)***
LPA and Long sleep	0.850 (0.614–1.177)	0.853 (0.614–1.186)	0.889 (0.639–1.237)
MVPA and Short sleep	0.729 (0.615–0.865)***	0.791 (0.665–0.940)**	0.797 (0.669–0.949)*
MVPA and Optimal sleep	0.432 (0.364–0.513)***	0.502 (0.421–0.598)***	0.524 (0.440–0.626)***
MVPA and Long sleep	0.505 (0.396–0.645)***	0.543 (0.424–0.694)***	0.564 (0.440–0.722)***

Model 1 was unadjusted for any confounding factors; Model 2 was adjusted for age, gender, marital status, education level, and residence; Model 3 was further adjusted for chronic disease history, drinking status, smoking status, and napping on the basis of Model 2

Abbreviations: Ref Reference group, PA Physical activity, LPA Light physical activity, MVPA Moderate-to-vigorous physical activity, OR Odds ratio, CI Confidence interval

^*P-value < 0.05, ^**P-value < 0.01, ^***P-value < 0.001

As shown in Supplementary Table S2, we formally tested the interaction between PA levels and sleep duration. The results indicated that while the joint variables were significant associated with falls, the multiplicative interaction term between PA and sleep duration did not reach statistical significance (Waldχ² = 5.164; P = 0.076). This suggests that the observed reduction in fall risk in the combined groups likely follows an additive pattern rather than a synergistic multiplicative interaction. In other words, while the combination of MVPA and optimal sleep provide the greatest risk reduction, the two factors act as independent contributions to fall prevention without a significant extra synergistic effect in this population.

Subgroup analysis

To evaluate the consistency of the joint associations between PA, sleep duration, and falls, subgroup analyses were performed stratified by age and gender. Among participants aged 60–74 years, the combined effect of MVPA and Optimal sleep showed a robust protective association (OR = 0.500; 95% CI: 0.404–0.620; P < 0.001) compared with the LPA and Short sleep group. Notably, for participants aged 75 years and above, while the protective trend remained, the effect size was slightly attenuated (OR = 0.607 for MVPA and Optimal sleep; OR = 0.592 for MVPA and Long sleep; both P < 0.05). Additionally, the interaction test for age was not statistically significant (P > 0.05). When stratified by gender, the joint associations remained significant for both males and females. In the male subgroup, all combinations involving MVPA or optimal sleep duration were significantly associated with reduced odds of falls (all P < 0.01). Specifically, the combination of MVPA and Long sleep showed the lowest risk (OR = 0.416; 95% CI: 0.281–0.618). In the female subgroup, the protective effects of MVPA and Optimal sleep (OR = 0.570; P < 0.001) and MVPA and Long sleep (OR = 0.680; P < 0.05) were also confirmed. However, the protective association of MVPA and Short sleep seen in males (OR = 0.659; P < 0.01) was not statistically significant in females (OR = 0.878; P > 0.05). And the interaction between gender and these joint categories was not significant (P > 0.05). These findings, detailed in Supplementary Fig S1, suggest population-specific differences in the joint associations of PA and sleep duration with falls.

Sensitivity analysis

Beyond the primary and subgroup analysis, sensitivity analysis were performed to further verify the stability of our findings. Specifically, we excluded participants with a history of major chronic conditions (including hypertension, dyslipidemia, diabetes, chronic lung disease, heart disease, stroke, psychiatric problems, memory-related diseases, Parkinson’s disease, and arthritis) to minimize the potential confounding effect of severe physical or cognitive impairment. After these exclusions, the final analysis included 2,017 individuals. The results of this restricted sample remained consistent with our main findings; notably, the joint association between MVPA and optimal sleep continued to show a significant protective effect against falls (OR = 0.392; 95% CI: 0.222–0.694; P < 0.01). These findings, detailed in Supplementary Fig S2, confirm that the observed associations are robust and not primarily driven by severe underlying health conditions.

Discussion

To the best of our knowledge, this is the first study to examine the independent and joint associations of PA and sleep duration with falls among older adults in China. This study revealed two key findings. First, both MVPA and optimal sleep (6–8 h) were independently associated with reduced odds of falling. Second, the findings indicated a joint association between PA and sleep duration with falls, suggesting that combinations of PA and sleep duration interact to influence falls. Specifically, the MVPA group generally had the lower odds of falling regardless of sleep duration; however, those with LPA and optimal sleep duration also had reduced odds of falling. Overall, these findings indicated the important combined role of PA and sleep duration in the prevention of falls.

Physical activity is a cornerstone of active ageing and overall well-being. In our study, older adults engaging in MVPA had significantly lower odds of falls compared with those engaging in LPA. This aligns with several studies which have investigated the relationship between PA and falls. For example, a recent ten-year population-based longitudinal study in Australia found that older adults who increased their own MVPA further reduced the risk of falls [14]. And a cross-sectional study based on Shandong Rural Elderly Health Cohort (SREHC) showed that older adults engaging in MVPA had a 42% lower odds of falling [43]. Additionally, a systematic review and meta-analysis which involved ten prospective cohort studies with a total of 58,241 older adult participants further confirmed that MVPA was associated with lower risk of falls. Furthermore, our findings are consistent with the exercise recommendations from the global guideline for the prevention and management of falls in older adults, which advocates for 150–300 min of moderate-intensity PA or 75–150 min of vigorous-intensity PA per week to effectively reduce fall risk [44]. Regarding the underlying mechanisms, a 12-month pragmatic intervention study among Chinese older adults demonstrated that MVPA promotion leads to significant improvements in muscle strength and neuromuscular coordination [45], both of which are vital for maintaining postural balance to prevent falls. Furthermore, regular MVPA can promote endogenous vitamin D synthesis. A previous research has indicated that each 10-minute increment in self-reported or objectively measured moderate- or vigorous-intensity PA is associated with a 0.18 ng/ml and 0.32 ng/ml increase in circulating vitamin D concentrations, respectively, alongside an elevation in cardiac output [46]. Elevated circulating vitamin D levels are associated with increased lower limb muscle strength, better gait control, and shortened reaction times, all of which act as critical physiological barriers against falls [46–49]. Additionally, higher PA levels can significantly improve the clinical management of chronic conditions such as stroke and arthritis, thereby reducing the risk of falls secondary to acute exacerbations of these diseases [50].

Sleep duration is a another key changeable aspect of lifestyle that profoundly influences biological homeostasis and energy balance [51, 52]. Our findings indicated that older adults with short sleep (< 6 h) had higher odds of falls compared with those with optimal sleep duration (6–8 h). This is consistent with a previous CHARLS-based study, which used < 6 h of sleep as the reference group and reported that 6–8 h of sleep was associated with significantly reduced risks of hip fracture and falls in older adults [53]. And there is evidence suggesting that insufficient nighttime sleep contributes to falls. A population-based nationwide study showed that sleep duration of 5 to 6 h/night (OR = 1.49; 95% CI: 1.03–2.15) and of ≤ 5 h/night (OR = 1.63; 95% CI: 1.18–2.25) were associated with a higher fall prevalence compared with sleep duration of 7 to 8 h/night [54]. Mechanistically, short sleep duration can also lead to a decline in cognitive function [55], including reductions in attention, executive function, memory, and reaction speed [56–59]. And many studies have revealed that cognitive impairment is a risk factor for falls among older adults [60, 61]. For example, a study based on CHARLS database showed that there was a significantly association between cognitive impairment and subsequent falls (OR = 0.97; 95% CI: 0.95–0.99). Such cognitive impairment caused by short sleep directly compromises older adults’ ability to identify environmental hazards, evaluate gait stability, and take timely protective measures during a fall event [62, 63], thereby increasing their risk of falls. Meanwhile, our findings indicated that long sleep (> 8 h) was also significantly associated with a increased risk of falls compared to optimal sleep, which is consistent with previous studies suggesting that long sleep duration is associated with a greater risk of falls [64–67]. However, there are some studies suggesting that no significant association is observed at long sleep duration and fall risk [68–70]. This discrepancy may be attribute to differences in the study populations, participant eligibility criteria, and the assessment method of sleep duration. The explanation of the adverse effect of long sleep observed in our study may be that excessive sleep could serve as a marker of frailty and muscle mass loss [29, 66]. A ten-year follow-up prospective cohort study revealed that the risk of frailty was increased among older adults with long sleep duration (HR: 1.26; 95% CI: 1.14–1.38) compared with those with middle sleep duration [71], while a meta-analysis also proved that frailty was significantly associated with a higher risk for falls [72].

A major contribution of this study is identifying the joint additive effect of PA and sleep duration on falls in older adults. To further explore these combined benefits, we conducted multivariable logistic regression across six stratified groups, using LPA and short sleep group as the reference. After full adjustment for confounding factors, the MVPA and optimal sleep group presented the lowest odds of falls. This additive protective pattern may operate via multiple underlying biological and behavioral mechanisms. The first potential mechanism involves the reciprocal relationship between PA and sleep duration. Regular MVPA has been consistently shown to enhance sleep quality and efficiency, while partially alleviating the adverse health impacts of sleep disorders [30]. For example, epidemiological studies have found that higher levels of MVPA are associated with a lower likelihood of sleep disturbance and sleep-disordered breathing (SDB) [73, 74]. Meanwhile, optimal sleep duration facilitates post-exercise bodily recovery and maintains sufficient energy reserves [31, 32], which reduces daytime fatigue, improves long-term PA adherence, and enables older adults to engage in PA at higher intensity levels [75]. This positive cycle independently and concurrently lowers the risk of falls. Second, Skeleton muscle health is vital for preventing falls among older adults. So another potential mechanism underlying this additive protective effect is the joint regulation of skeleton muscle health by greater engagement in PA and maintain sufficient sleep, which is mediated by the modulation of inflammatory cytokine levels. Some studies has been confirmed that poor sleep quality can induce a sustained chronic inflammatory state, elevating pro-inflammatory markers such as interleukin-6 (IL-6) and C-reactive protein (CRP) [76–79]. And high levels of IL-6 can promote muscle protein breakdown and insulin resistance via the nuclear factor-κB (NF-κB) signaling pathway, while elevated CRP can also trigger insulin resistance and impair myocyte function. These adverse effects may lead to diminished muscle strength and higher sarcopenia risk, which increases the susceptibility to falls in older adults [80, 81]. On the other hand, greater engagement in PA can exert robust anti-inflammatory effects by inhibiting the expression of tumor necrosis factor-α (TNF-α) and stimulating synthesis of anti-inflammatory mediators [82], which is beneficial to skeleton muscle regeneration [83]. Furthermore, regular PA modulates the circadian rhythm molecular clock to improve the sleep-wake cycle, and stimulates the secretion of myokines like irisin. The latter not only promotes skeletal muscle repair and hypertrophy [84, 85], but has also been well documented to increase slow-wave sleep activity, thereby preserving skeletal muscle function while improving sleep quality [86]. In summary, optimal sleep duration creates a low-inflammatory milieu that provides favorable conditions for exercise-induced muscle protein synthesis [87]. In turn, regular PA can directly counteracts inflammation and enhances muscle function, while positively modulates sleep architecture [88, 89]. These complementary pathways form a positive feedback loop that optimizes neuromuscular function and maximizes the reduction risk of falls in older adults.

Furthermore, in order to explore whether these joint associations vary across different gender and age subgroups, we conducted a subgroup analysis revealing that compared with the LPA and short sleep group, the reduction in odds of falls in the MVPA and optimal sleep group was more pronounced in male older adults and those aged 60–74 years. Previous studies have demonstrated that the oldest-old adults aged ≥ 75 years face a higher risk of frailty compared with young-old adults [90]. Meanwhile, frailty characterized by the decline in physiological reserve and reduced resistance has been shown to be closely associated with an increased risk of falls [91]. Additionally, multimorbidity is highly prevalent in the oldest-old adults [92], which act as strong risk factors for falls [93], and thus may attenuate the joint protective effect of MVPA and optimal sleep against falls. Regarding sex-related discrepancy, it may be attributed to the sharp decline in estrogen levels among postmenopausal women, which more easily leads to impaired anti-inflammatory capacity and diminished muscle protein synthesis capacity [94, 95]. And women exhibit a higher annual rate of lower limb muscle strength loss compared with men, which renders older women more prone to developing fear of falling [96]. In addition, due to the constraints of traditional social gender roles, there are notable differences in the types of PA engaged in by older men and women. Older men are more likely to participate in structured targeted exercise training, whereas older women predominantly engage in household chores [97]. The latter has a relatively limited beneficial effect on improving muscle strength and balance capacity [98].

Despite these adjusting efforts and subgroup insights, a critical methodological consideration is the potential for residual confounding from several unmeasured factors that are intricately linked to the multifactorial nature of falls. As noted in existing literature, environment hazards [99], nutrition status [100], and specific medication regimens [101] are potent determinants of fall risk that were not fully captured in our statistical model. These unmeasured confounders may have influenced our observed independent and joint associations in distinct directions. For instance, the use of sedative or polypharmacy is highly prevalent among older adults with sleep disorder or multiple chronic conditions [102]. And due to these medication can induce drowsiness or orthostatic hypotension, omitting them from regression might have overestimated the direct adverse effects of short/long sleep duration or LPA on fall risk. Conversely, regarding environmental and nutritional factors, older adults living in hazardous home environments [99] or suffering from severe vitamin D deficiency [100] might instinctively restrict their PA due to physical frailty or a fear of falling. Since these factors can drive fall incidents, their absence in our models could introduce residual confounding that potentially magnified or distorted the observed joint protective effects of MVPA and optimal sleep. Therefore, future studies employing comprehensive geriatric assessments are required to futher disentangle these multifactorial etiologies.

Clinical implication

From a clinical and preventive standpoint, our findings underscore the necessity of developing integrated lifestyle intervention strategies for older adults, aiming not only to prescribe tailored physical activity intensities but also to safeguard optimal sleep hygiene. Given that MVPA presents significantly lower odds of falls compared to LPA, clinicians should move beyond generalized recommendations to “move more” and instead encourage robust older individuals to safely achieve the MVPA threshold to maximize neuromuscular and balance benefits. Concurrently, the observed U-shaped association between sleep duration and falls indicates that extreme sleep duration (both < 6 h and > 8 h) serve as vital behavioral indicators of fall vulnerability, suggesting that routine sleep tracking should be integrated into standard geriatric risk assessments. Crucially, the primary evidence from our joint analysis demonstrates that the coexistence of insufficient or excessive sleep alongside low-intensity activity compounding fall risks, whereas the combination of MVPA and optimal sleep (6–8 h) yields the lowest odds of falls. Therefore, early screening for sleep disturbances, precision exercise counseling, and multidisciplinary wellness programs targeting the intersection of daytime activity and nighttime sleep are essential to comprehensive fall prevention in aging populations.

Strengths and limitations

This study has several notable strengths. The large sample size of 10,232 participants from the CHARLS ensures robust statistical power and allows for accurate estimation of the relationships between PA, sleep duration and falls. Additionally, the study focused on PA but also distinguished between different duration of sleep (such as < 6 h, 6–8 h, > 8 h), providing more detailed and specific health recommendations. However, our study had some limitation. First, due to the cross-sectional design of this study, causal inferences cannot be established. And the potential for bidirectional relationships between PA, sleep duration and falls, where a history of previous falls or the psychological fear of falling may subsequently induce older adults to restrict their daytime physical activities and experience fragmented sleep architecture, cannot be ruled out. Consequently, our interpretations of the potential biological mechanisms underlying these associations are largely inferential based on literature; because we could not measure real-time neuromuscular or physiological mediators. Second, both PA and sleep duration were primarily obtained through self-reported questionnaires, which may have introduced recall bias or social desirability bias and lead to potential misclassification. Similarly, falls was assessed via a single self-reported question, which may affect measurement validity. Third, as with all observation studies, residual confounding from unmeasured or uncontrolled confounders (e.g., environmental hazards, nutritional status, and medication use) may persist despite adjustment for known potential confounders. Additionally, although a wide range of health-related covariates were adjusted for, detailed parameters regarding overall sleep quality, specific sleep disturbances, and precise circadian rhythm indicators were not fully assessed due to data availability constraints in the primary database. Fourth, our findings were exclusively on a older Chinese population, which may limit the generalizability of this results to other cultural, geographical, or healthcare contexts. Future research should incorporate more robust causal inference methods, and objective assessments of PA and sleep duration, while also considering the practical challenges and policy implications for promoting PA and sleep duration among diverse ageing population.

Conclusions

Our findings suggest that MVPA and sleep duration of 6–8 h are independently associated with lower odds of falls, whereas short sleep duration (< 6 h) and long sleep duration (> 8 h) are associated with higher odds of falls. In addition, this study highlighted the joint association between PA and sleep duration in relation to falls. Specifically, individuals engaging in MVPA combined with an optimal sleep duration (6–8 h) exhibited the lowest odds of falls, while those engaging in LPA coupled with an optimal sleep duration also showed a reduced likelihood of falling. While the cross-sectional nature of this study limits causal inference, integrated interventions targeting both sleep hygiene and physical activity may be beneficial for fall prevention strategies in the older adults.

Abbreviations

CHARLS

China Health and Retirement Longitudinal Study

PA

Physical activity

LPA

Light physical activity

MVPA

Moderate-to-vigorous physical activity

VPA

Vigorous physical activity

SE

Standard error

CI

Confidence interval

OR

Odds ration

Authors’ contributions

W-JN: Conceptualization, Methodology, Data analysis, Resources, Writing - original draft, Writing - review & editing. FY: Data curation, Resources, Writing - review & editing. Z-LQ: Methodology, Resources, Writing - review & editing. All authors have reviewed and approved the final version of the manuscript.

Funding

No.

Data availability

The original data sets are publicly available on the website of the China Health and Retirement Longitudinal Study (CHARLS) (https://charls.pku.edu.cn/). The data sets generated and analyzed during this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The studies involving humans were approved by Biomedical Ethics Committee of Peking University (Approval No: IRB00001052-110155). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.