A latent profile analysis of rest–activity behavior patterns among community-dwelling older adults and its relationship with intrinsic capacity

Abstract

Rest–activity behavior clusters within individuals to form patterns are of significant importance to their intrinsic capacity (IC), yet they have rarely been studied. A total of 1253 community-dwelling older adults were recruited between July and December 2021 based on the baseline survey database of the Fujian Prospective Cohort Study on Aging. Latent profile analysis was used to identify profiles of participants based on rest–activity behaviors, whereas logistic regression analysis was carried out to investigate the relationship between profiles and IC. We identified three latent profiles including: (1) Profile 1—labeled “Gorillas”: High physical activity (PA), moderate sedentary behaviors (SB), screen time (ST) and sleep (n = 154, 12%), (2) Profile 2—labeled as “Zebras”: Moderate PA, low SB, ST and high sleep (n = 779, 62%), and (3) Profile 3—labeled as“Koalas”: High SB, ST, low PA and sleep (n = 320, 26%). Logistic regression revealed a negative correlation between low IC and the “Gorillas” profile (β = − 0.945, P < 0.001) as well as the “Zebras” profile (β = − 0.693, P < 0.001) among community-dwelling older adults, with the “Koalas” profile showing the weakest IC compared to the other profiles. The demographic traits i.e., female, older age, living alone, and low educational level also correlated with low IC. Identifying trends of rest–activity behaviors may help in drawing focus on older adults at risk of decreasing IC, and develop personalized improvement plans for IC.

Introduction

Intrinsic capacity (IC) represents an important concept for healthy aging. It is the sum of an individual’s physical and mental (including psychosocial) capacitie^1,2, including five components, i.e., cognition, psychology, sensory, vitality, and locomotion^3–5. The decline of IC is a popular phenomenon in the elderly population. Previous studies have shown that the incidence of IC impairment ranges between approximately 39.9% and 75.3% across one or multiple domains^6–10. Furthermore, IC is a strong predictor of adverse health outcomes in older adults, including falls, IADLs, and an increased risk of hospitalization^11–13. Additionally, a higher IC score correlates with a decreased likelihood of morbidity and mortality among community-dwelling older adults^14,15. In this regard, it is important to identify controllable factors contributing to the maintenance of IC stability in older individuals, hence promoting effective interventions that preserve their functional capacities and forestalling or mitigating adverse health events.

Studies have confirmed the effect of rest–activity behaviors, including physical activity (PA), sedentary behaviors (SB), and sleep, on IC^16,17. A Copenhagen Consensus of 2019 reported that older adults engaging in PA enjoy better benefits in IC than their inactive counterparts¹⁸. However, older adults spend a substantial amount of time in SB, with an average daily sitting time exceeding 10 h^19,20. Prolonged sitting time, particularly screen time (ST) i.e., watching TV and using smartphones, can result in reduced social opportunities and mental health deterioration^21–23. Current research indicates that prolonged sitting can lead to declines in IC. Specifically, it results in a diminished locomotion domain, evidenced by a slower walking pace, and contributes to impairments of the psychology domain, including the onset of depression^20,24. Additionally, prolonged SB has been associated with a decline in the cognition domain²⁵. Moreover, recent research has indicated that optimal sleep duration significantly reduces the incidence of IC in older adults. Specifically, sleeping for 6–8 h per night could lead to a 23% decrease in the risk of poor IC²⁶. Nevertheless, existing studies only focus on the correlation between single rest–activity behavior and IC, ignoring interrelationships and interactions among these behaviors, and less attention paid to screen-based sedentary time. Therefore, identifying the daily rest–activity behavior profiles of older adults as well as the demographic differences between different profiles can provide information for targeting behavioral change interventions and identifying populations that mostly need rest–activity behavior interventions. However, previous analytical methods have failed to address the group heterogeneity generated by multifactorial interactions, nor have they allowed us to understand the mechanisms by which rest–activity behaviors cluster within individuals to build patterns that influence their IC.

This issue can be efficiently addressed using person-centered latent profile analysis (LPA), which clusters heterogeneous populations based on a specific combination of factors²⁷. Unlike conventional statistical methods, LPA can examine statistical co-exposures and reveal combined effects obscured by the collinearity among independent variables²⁸. The optimal categories of physical activity behaviors for older adults based on various activity–rest areas can be ascertained through LPA. With the availability of medical resources, the older adults can be managed based on their rest–activity behavior profiles to maximize the effect of intervention. At present, cluster analysis methods are used to investigate the profile of rest–activity behavior among the elderly^29,30. Park et al. carried out LPA on 85 elderly people from assisted living facilities and assessed aspects including physical health, functionality, light-intensity activities, and moderate to vigorous-intensity physical activity (MVPA). They discovered that mental health status significantly varies among the elderly with different activity patterns³⁰. However, screen time, an essential part of rest–activity behavior has received little attention in the few studies that have used cluster analysis for this demographic. Furthermore, research on the relationship between rest–activity behaviors and IC remains scarce.

Therefore, this study aims to uncover the potential profiles of rest–activity behaviors among community-dwelling older adults through LPA and establish whether these latent profiles are associated with IC. This could eventually provide a reference for community healthcare workers to assist in formulating rest–activity behavior schedules for the elderly, hence helping in preventing the decline of IC.

Methods

Study design and participants

This study is a cross-sectional study carried out in Fuzhou, Fujian Province, China. Study participants were enrolled based on the baseline survey database of Fujian Prospective Cohort Study on Aging, which began in 2020 and involves follow-ups every one to two years to monitor the health status and cardiovascular events among non-hospitalized elderly population in Fujian Province through comprehensive geriatric assessment. Eligible elderly individuals were continuously recruited using methods including telephone calls and posters between July and December 2021.

The inclusion criteria for participants included: (1) individuals aged 60 and above, (2) voluntary participation (3) consenting individuals, and (4) those who could complete questionnaires and undergo clinical assessments. The exclusion criteria included: (1) individuals with a life expectancy of less than six months due to severe illness or advanced cancer, (2) individuals bedridden or completely disabled for extended periods, or (3) those who had severe impairments in vision, hearing, or speech.

Ethical approval for the study was granted by the Ethics Committee of Fujian Provincial Hospital (K2021-05-033). This research have been performed in accordance with the Declaration of Helsinki. The objectives and methodology were thoroughly explained to participants at the beginning, and informed consent was obtained before collecting data. Data collection and scale evaluations were implemented by trained researchers through in-person interviews. A total of 1272 individuals completed the survey, among which 19 were excluded due to incomplete information, leaving a final sample size of 1253 for data analysis, accounting for 98.51% of all the questionnaires distributed.

Measurement

Physical activity

A Chinese version of the International Physical Activity Questionnaire (IPAQ)^31,32, shown to be moderately to substantially reliable in Chinese populations, was used to record the time spent in light-intensity physical activity (L-PA), moderate-intensity physical activity (MPA), and vigorous-intensity physical activity (VPA). Participants were asked about the number of days and the duration they had engaged in the following PA in the past week: L-PA (e.g. walking, housework activities), MPA (e.g. dancing, cycling, but not walking), and VPA (e.g. jogging, swimming, running), with corresponding metabolic equivalents (METs) of 3.3, 4.0, and 8.0, respectively. MPA and VPA were combined to form MVPA. MET-minutes per week were computed for each category of PA, leading to the total computation of MET-minutes per week.

Sedentary behavior

Data on sedentary behavior (SB) were obtained using the Sedentary Behaviour Questionnaire (SBQ)³³. The questionnaire documented various sedentary activities and their respective durations over the past 7 days for older adults. These activities included: (a) watching TV, (b) using a mobile phone, (c) socializing, (d) reading, (e) commuting, and (f) sitting around. Before collecting information, participants were asked about their TV and mobile phone ownership. Daily time spent in each sedentary category was calculated using the formula [(weekday sitting minutes * 5) + (weekend day sitting minutes * 2)]/7. The sum of the average times spent sitting in all categories represented the total daily sitting time³¹. Screen time refers to the duration of engagement with electronic devices while seated.

Sleep

Data on sleep were collected by trained researchers who inquired about the typical sleep and wake times of participants each night. Nightly sleep duration was computed based on the interval between going to bed and waking up. Actual sleep time was estimated to constitute approximately 85% of time spent in bed³⁴ and the self-reported sleep time is about 95% consistent with the sleep time monitored by the wrist-worn accelerometers³⁵. Based on the recommendations of the National Sleep Foundation, sleeping less than 7 h per night was considered indicative of sleep deprivation³⁶.

Intrinsic capacity

The IC screening tool in this study was based on the WHO ICOPE guidelines³⁷: (1) The Cognitive domain was assessed using the Mini-Cog Scale^38,39. This involved participants recalling three words, drawing a clock, and then recalling the initial three words. A score ranging from 0 to 5 was derived, with scores of ≤ 3 indicating cognitive impairment. (2) The locomotion domain was evaluated with a 4-m walk test^9,15,40. A gait speed < 0.8 m/s was considered indicative of locomotion domain impairment⁴¹. (3) The vitality domain was measured using the Short-Form Mini Nutritional Assessment (MNA-SF)^9,42,43. The total score ranged from 0 to 14, with ≤ 11 suggesting a risk of malnutrition or overt malnourishment, signifying impairment in the vitality domain. (4) Psychological status was appraised using the Geriatric Depression Scale (GDS-4)^44,45. A score from 0 to 4 was obtained, with scores of ≥ 2 denoting impairment in the psychology domain. (5) The sensory domain was determined through self-reported visual and hearing impairments and by reviewing the health records of participants^7,11,43,46. Any condition from mild to severe was considered an impaired sensory domain. Each impaired domain scored one point; normal domains scored none, culminating in a total score of 0–5. Higher scores indicated a more significant IC decline. Based on other relevant studies⁴⁷, participants with IC scores ≥ 3 points were classified into the low IC group, whereas those with ≤ 2 points were classified into the high IC group.

Covariate

Drawing from previous literature and clinical experience, the following covariates were identified: Sociodemographic data were collected through face-to-face interviews by trained researchers using a detailed questionnaire, which encompassed: (1) General demographic information, including sex and age, (2) socioeconomic attributes, including marital status (single, married, divorced, and widowed), educational level (primary school or below, middle school, high school, college or above), and economic income, (3) lifestyle factors, including alcohol consumption and/or smoking habits, (4) comorbidities, including the presence of hypertension, diabetes and other diseases. Social support was assessed using the validated Social Support Assessment Scale (SSRS)⁴⁸.

Data analysis

The analysis was structured into three key stages including (1) sample interpretation and the development of LPA models. (2) Assessment of demographic variations across the identified profiles. (3) Investigation of the associations between the profiles, additional covariates, and IC.

The characteristics of L-PA, MVPA, SB, ST, and sleep were used to determine rest–activity behavior patterns among community-residing older adults. Starting with 1 category as the initial model, the number of model categories was gradually increased to 5. The most suitable model was selected based on fit indices, including the Loglikelihood test, Akaike information criterion (AIC), Bayesian information criterion (BIC), adjusted Bayesian information criterion (αBIC), Entropy, Lo-Mendell-Rubin likelihood ratio test (LMR), and Bootstrapped likelihood ratio test (BLRT)²⁸. Lower values of AIC, BIC, and αBIC indicated a superior model fit^28,49. Researchers assessed the quality of the classification using entropy. Scores ≥ 0.8 indicated a clear separation between the different classes. Likelihood Ratio (LMR) and Bartlett’s test (BLRT) were employed to compare neighboring models and evaluate their improvement in fit. A statistically significant P value from these tests suggests that the model with k classes is a significant improvement over the model with k − 1 classes²⁸. Once profiles were identified, they were named to accurately indicate their features and differentiate them from other profiles. Subsequently, demographic differences across profiles were analyzed using Chi-square and ANOVA tests.

Logistic regression analysis was conducted to investigate the relationship between the identified rest–activity behavior profiles and IC. The chi-square test was used to investigate the factors related to IC. Multivariate logistic regression analysis was conducted with IC as the dependent variable (high IC = 0, low IC = 1), including statistically significant rest–activity behavior profiles and sociodemographic characteristics from the Chi-square test as independent variables. A two-tailed P < 0.05 was considered statistically significant. All statistical analyses were conducted using the R version 4.4.0 (2024-04-24 ucrt) and Mplus 8.3 software.

Ethics approval and consent to participate

The research design was approved by the ethics committee of Fujian Provincial Hospital (K2021-03-015). All participants participated in the study voluntarily and filled out the informed consent ethics approval form.

Results

Descriptive characteristics of the participants

The present study classified and analyzed 1253 participants. The average age of the participants was 71.4 years, with a gender distribution of 61.9% females. Among all the participants, 80.8% were married; 52.8% had attained at least a high school education level; and 55.7% reported a monthly economic income of 3000–6000 RMB. Regarding social support, 49.0% of participants indicated a medium level, whereas 41.0% indicated low support levels. Furthermore, non-smokers comprised 91.8% of the cohort, and 88.2% abstained from alcohol consumption.

Profile outcomes of rest–activity behaviors

Table 1 illustrates 1–5 profile models used to determine the optimal number of profiles. In light of profile comparison results, classification accuracy, and substantive interpretability, profile 3 was the optimal model. To validate the reliability of LPA results, we evaluated the average probability of the three profiles belonging to each profile. The results showed that the correct classification probabilities of profiles 1–3 are 0.922, 0.938, and 0.885, respectively, all exceeding the 0.85 thresholds. This indicates that the results of the three LPA models are reliable.

Table 1.

Fit statistics for latent profile analysis of five rest–activity behaviors.

	Loglikelihood	AIC	BIC	αBIC	LMR	BLRT	Entropy	Class probability
class1	− 8884.311	17788.622	17839.955	17808.190				1
class2	− 8515.695	17063.390	17145.523	17094.699	< 0.001	< 0.001	0.804	0.70/0.30
class3	− 8356.387	16756.774	16869.707	16799.825	< 0.001	< 0.001	0.833	0.12/0.62/0.26
class4	− 8229.297	16514.594	16658.327	16569.386	< 0.001	< 0.001	0.869	0.21/0.22/0.48/0.09
class5	− 8145.678	16359.357	16533.889	16425.889	0.34	< 0.001	0.882	0.07/0.19/0.47/0.23/0.04

BIC, Bayesian information criterion; AIC, Akaike information criterion; aBIC, adjusted Bayesian information criterion; LMR, Lo-Mendell-Rubin likelihood ratio; BLRT, bootstrap likelihood ratio test.

Based on the findings of this study, cluster 1 was labeled as “Gorillas (n = 154, 12%)”, profile 2 as “Zebras (n = 779, 62%)”, and profile 3 as “Koalas (n = 320, 26%)”, referring to previous studies⁵⁰. Unlike other profiles of older adults, “Gorillas” were more predisposed to various intensity levels of PA, particularly MVPA, and they appeared to spend less time both sitting and sleeping. The distinguishing features of the “Zebras” included less sedentary time and more time spent sleeping. The “Koalas” were characterized by a lower level of physical activity and excessive sedentariness, particularly in terms of screen time (Fig. 1). Descriptive data on general demographics, socioeconomic status, educational level, lifestyle, comorbidities, and social support were presented based on profile results, revealing significant differences across the three profiles in age, marital status, living conditions, and economic income (P < 0.05) (Table 2).

Figure 1.

Latent profiles of rest–activity behaviors pattern. Note: MVPA, vigorous-intensity Physical Activity; L-PA, light-intensity physical activity; SB, sedentary behavior; ST, screen time.

Table 2.

Characteristics of individuals in baseline and different rest–activity behavior profiles.

	Total (n = 1253 (%))	“Gorillas” (n = 154 (%))	“Zebras” (n = 779 (%))	“Koalas” (n = 320 (%))	P
Gender					0.588
Male	490 (39.1)	63 (5.0)	296 (23.6)	131 (10.5)
Female	763 (60.9)	91 (7.3)	483 (38.5)	189 (15.1)
Age					< 0.001
60–69	568 (45.3)	95 (7.6)	350 (27.9)	123 (9.8)
70–79	523 (41.7)	51 (4.1)	326 (26.0)	146 (11.7)
≥ 80	162 (12.9)	8 (0.6)	103 (8.2)	51 (4.1)
Marital status					0.003
Married	1013 (80.8)	130 (10.4)	645 (51.5)	238 (19.0)
Not Married*	240 (19.2)	24 (1.9)	134 (10.7)	82 (6.5)
Living status					0.005
Live with family	1107 (88.3)	147 (11.7)	687 (54.8)	273 (21.8)
Live alone	146 (11.7)	7 (0.6)	92 (7.3)	47 (3.8)
Education					0.815
Primary school and below	250 (20.0)	29 (2.3)	163 (13.0)	58 (4.6)
Middle school	342 (27.3)	40 (3.2)	213 (17.0)	89 (7.1)
High School and Above	661 (52.8)	85 (6.8)	403 (32.2)	173 (13.8)
Economic income					0.017
< 3000 RMB	445 (35.5)	57 (4.5)	297 (23.7)	91 (7.3)
3000 ~  < 6000 RMB	698 (55.7)	85 (6.8)	409 (32.6)	204 (16.3)
≥ 6000 RMB	110 (8.8)	12 (1.0)	73 (5.8)	25 (2.0)
Social support					0.388
Low-level	514 (41.0)	62 (4.9)	322 (25.7)	130 (10.4)
Medium-level	614 (49.0)	83 (6.6)	377 (30.1)	154 (12.3)
High-level	125 (10.0)	9 (0.7)	80 (6.4)	36 (2.9)
Smoking					0.277
No	1150 (91.8)	145 (11.6)	717 (57.2)	288 (23.0)
Yes	103 (8.2)	9 (0.7)	62 (4.9)	32 (2.6)
Drinking					0.473
No	1105 (88.2)	132 (10.5)	693 (55.3)	280 (22.3)
Yes	148 (11.8)	22 (1.8)	86 (6.9)	40 (3.2)
Presence of chronic diseases					0.369
No	550 (43.9)	73 (5.8)	346 (27.6)	131 (10.5)
Yes	703 (56.1)	81 (6.5)	433 (34.6)	189 (15.1)
MVPA (Met-min/w, mean (95% CI))	959.7 (897.9, 1021.5)	3344.8 (3240.5, 3449.1)	673.7 (626.6, 720.8)	508.0 (436.2, 579.9)
LPA (Met-min/w, mean (95% CI))	1191.3 (1153.7, 1228.8)	1266.9 (1160.1, 1373.7)	1188.3 (1139.7, 1236.9)	1162.0 (1090.5, 1233.5)
SB (min/d, mean (95% CI))	267.2 (259.4, 275.0)	270.1 (250.1, 290.1)	202.5 (195.9, 209.1)	423.4 (410.3, 436.5)
ST (min/d, mean (95% CI))	160.2 (154.6, 165.7)	154.3 (138.9, 169.7)	109.7 (105.5, 113.9)	285.9 (279.1, 292.8)
Sleep (min/d, mean (95% CI))	428.2 (422.6, 433.8)	429.6 (415.0, 444.3)	435.1 (428.0, 442.1)	410.8 (399.1, 422.6)

Not Married*: Widowed, divorced or others.

Association of IC with sociodemographic and rest–activity behaviors profiles

We used logistic regression to examine the relationship between IC and sociodemographic characteristics and rest–activity category. Through the chi-square test of the decline of IC among community-dwelling older adults, we found that the decline of IC was statistically differed significantly in age (χ² = 21.750, P < 0.001), gender (χ² = 11.278, P = 0.001), marriage (χ² = 27.602, P < 0.001), living status (χ² = 48.791, P < 0.001), education (χ² = 16.187, P < 0.001), and social support (χ² = 25.451, P < 0.001). Statistically significant differences were also observed among the three rest–activity profiles (χ² = 34.382, P < 0.001) (Table S1).

The logistic regression analysis showed that unlike the “Koalas” profile, both the “Gorillas” profile (OR: 0.389, 95% CI 0.249–0.607, P < 0.001) and “Zebras” profile (OR: 0.500, 95% CI 0.379–0.660, P < 0.001) had a lower risk of IC decline. Males showed a lower likelihood of IC decline unlike females (OR: 0.717, 95% CI 0.551–0.935, P = 0.014). Individuals aged 70–79 had a decreased risk of IC decline relative to those aged ≥ 80 (OR: 0.675, 95% CI 0.455–1.000, P = 0.005). Furthermore, living alone increased the risk of IC decline compared to residing with family (OR: 0.717, 95% CI 0.551–0.935, P = 0.014). Additional findings from the logistic regression suggested a positive correlation between IC and educational level; the lower the education level, the lower the IC (Fig. 2).

Figure 2.

Forest plot of the association between the decline of intrinsic capacity and social demographic characteristics and three rest–activity behavior profiles.

Discussion

To our knowledge, this is arguably the first study to use the LPA to investigate rest–activity behavior distribution profiles based on varying intensities of PA, SB, ST, and sleep among 1253 community-dwelling older adults, and examine the relationship between different rest–activity behavior profiles and IC decline. We used LPA to categorize the elderly population into three distinct profiles, i.e., “Gorillas”, “Zebras” and “Koalas”. Statistically significant differences were observed across these profiles in terms of age, marital status, living status, and economic income. Logistic regression analysis revealed that the “Koalas” profile was linked to an increased risk of CI decline, unlike other profiles. These findings provide an innovative perspective on the importance of rest–activity behavior allocation in mitigating the deterioration of CI among community-dwelling older adults.

Among the three identified profiles, the “Zebras” constituted the largest segment of the elderly population, accounting for 62%, and characterized by moderate PA of varying intensity, low sedentary time, and long sleep duration. The “Koalas” profile was the second largest, representing 26% of older adults, reporting the highest sedentary time and the least amount of PA and sleep. The third profile, named “Gorillas” (12% of older adults), was distinguished by high levels of PA, specifically in MVPA. These findings suggest that older adults have diverse patterns in how they allocate their time between physical rest and activity, implying individual variations. Most older adults engage in more PA, less SB, and balanced sleep. Furthermore, those who engage in higher levels of daily PA tend to sleep more and spend less time being sedentary⁵¹. This may be attributed to the positive effects of regular exercise on sleep quality by regulating adenosine levels and body temperature⁵². Additionally, increased sleep and PA appear to reduce time spent sitting. This is also consistent with the 24-h movement behavior paradigm's assumption that a 24-h day is fixed and finite, implying that an increase in time spent on one behavior necessarily causes a decrease in time spent on at least one other activity⁵³. Our findings are partially consistent with that of Park et al. on the relationship between physical function, rest–activity behaviors, and mental health in older adults, which identified three unique physical rest–activity behavior profiles³⁰. Specifically, they found that 41.1% had moderate levels, 31.8% had active behavior, and 21.7% had SB. The results of the above studies differ from our findings because of variations in sample characteristics, culture, or geographic region. Unlike previous studies on SB among the elderly, we noted that even the “Koalas” group, which is the most predisposed to sitting, had lower sedentary time⁵⁴. This could be attributed to the generally better health status among the elderly in that community, as well as the availability of more recreational and activity facilities including walking paths, parks, and sports equipment, which is conducive for PA among older adults⁵⁵.

Noteworthy, we investigated screen sitting besides sitting time, lacking in older adults' current rest–activity behavior studies. Our findings showed that a significant part of the sedentary duration of older adults’ is spent engaging with screens; besides, there is a positive relationship between SB time and ST consumption, in line with previous research⁵⁶. This may be due to the effect of COVID-19 containment measures, which required many public places to show an electronic QR code of a negative nucleic acid test result. This change prompted several elderly individuals to start using smartphones. Restrictions on outdoor activities resulted in a rise in screen time, leading to a gradual shift towards electronic devices as the primary source of entertainment⁵⁷. Moreover, these groups indicate that various rest–activity levels are closely linked to age, marital status, and economic income, warranting more research attention.

Logistic regression analysis revealed that both the “Gorillas” and “Zebras” were less likely to belong to the low IC group compared to the “koalas” profile. This implies that the higher frequency of PA at various intensities throughout the week is associated with reduced sedentary time or improved sleep quality, correlating with a lower likelihood of IC decline⁵⁸. That’s because a rest–activity behavior pattern characterized by less SB and more PA could enhance the domain of vitality, which is a potential physiological determinant of IC. This vitality domain reflects the body’s energy and metabolism, neuromuscular function, immune response, and stress reaction capabilities, and it can influence other domains of IC⁵⁹. Prolonged PA has important benefits for alleviating or maintaining the vitality domain, and it can counteract age-related declines in IC caused by physiological changes in systems/organs such as endocrine, neuromuscular, metabolic, and cardiopulmonary functions⁶⁰. PA is considered a vital intervention for improving overall functional health in older adults, particularly the frequency and duration of weekly activities⁶¹. The WHO guidelines recommend that older adults should weekly engage in at least 150–300 min of MPA or 75–150 min of VPA⁶². However, older adults tend to be more sedentary, often spending up to 60% of their waking hours seated, engaged in screen-based sitting time, and only dedicating 2% of their time to MVPA⁵⁶. An interesting observation we found is that the IC levels among these older adults remained higher despite the “Gorillas” profile being more sedentary. This means that increasing the time spent on MVPA could partially offset the negative impacts of sedentary behavior on IC⁶³. Consequently, government agencies should enhance community fitness facilities to encourage daily PA among the elderly. In addition, healthcare professionals should intensify education efforts regarding the dangers of excessive sitting, specifically screen-based SB, and motivate older adults to participate in regular PA, especially MVPA. With additional resources, mobile health technologies can be integrated to design customized exercise programs suiting the individual needs of older adults. Furthermore, logistic regression analysis revealed that women, with low education levels, living alone, and older individuals are more susceptible to IC decline, consistent with previous research findings^46,64,65.

This study has several limitations. Firstly, it is cross-sectional in design, only allowing for the identification of relationships, rather than causal relationships between various rest–activity behavior profiles and IC. To substantiate these associations, more rigorous study designs and longitudinal research are essential to track the dynamic relationship between changes in rest–activity behavior patterns and IC over time. Secondly, the assessment of rest–activity behaviors was based on subjective questionnaires and self-reports, which can be affected by recall bias and lack of objectivity. The use of objective measurement tools, including wearable technology or SOMNO screen PSG, could yield more dependable data. Thirdly, we did not consider other factors that may affect IC in older adults, including environmental elements (e.g., community green space); these should be incorporated into future research to provide a more comprehensive perspective on IC among the elderly. Finally, the data was solely collected from elderly individuals residing in Fuzhou communities, which may not adequately represent the entire population of older adults across China, thereby limiting the generalizability of these findings to other geographical regions to some extent. Despite these constraints, the findings of the present study demonstrate the value of LPA in identifying rest–activity behavior profiles and their effect on IC among community-dwelling older adults, providing novel perspectives for clinicians to precisely identify and address IC decline in this population.

Conclusion

In conclusion, we utilized LPA to identify three distinct profiles of rest–activity behavior in community-dwelling older adults, i.e., the “Gorillas”, “Zebras”, and “Koalas”. Within this demographic, individuals who were more physically active and spent less time sitting appeared to have a greater IC. Our study results can help community healthcare workers identify older adults at risk of a decline in IC through patterns of rest–activity behaviors. They can also provide targeted and actionable interventions to improve the IC of older adults in the community.

Abbreviations

IC

Intrinsic capacity

PA

Physical activity

SB

Sedentary behaviour

ST

Screen time

LPA

Latent profile analysis

MVPA

Moderate to vigorous-intensity physical activity

IPAQ

International Physical Activity Questionnaire

MPA

Moderate-intensity physical activity

VPA

Vigorous-intensity physical activity

L-PA

Light-intensity physical activity

METs

Metabolic equivalents

SBQ

Sedentary Behaviour Questionnaire

MNA-SF

Short-Form Mini Nutritional Assessment

GDS-4

Geriatric Depression Scale

SSRS

Social Support Assessment Scale

AIC

Akaike information criterion

BIC

Bayesian information criterion

αBIC

Adjusted Bayesian information criterion

LMR

Lo-Mendell-Rubin likelihood ratio test

BLRT

Bootstrapped likelihood ratio test

Author contributions

N.L.: Data collection, Data analysis, Writing-original draft. N.W.: Data analysis, Writing-original draft. S.L. and Y.Y.: Data collection. F.H.: Writing—review & editing, Coordinated the study. P.Z.: Methodology, Funding, Writing—review & editing. All authors contributed to reviewing and revising the manuscript, read and approved the final manuscript.

Funding

This work was supported by the Fujian Province Science and Technology Innovation Joint Fund Project (2023Y9341) and the Major Project of Fujian Provincial Health Commission (2022ZD01006). The sponsors had no role in the study design and collection, analysis, and interpretation of data; writing of the report; or decision to submit the article for publication.

Data availability

The datasets analysed in the study are currently not publicly available. We cannot make the data publicly available because the owners of the original data, Fujian Provincial Hospital and the Shengli Clinical Medical College of Fujian Medical University, do not agree to make the data publicly available. However, the data are available upon reasonable request and with the permission of the corresponding author.

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-69114-y.