Associations of sociodemographic and behavioral factors with frailty transition patterns: a multi-state Markov analysis of the China Health and Retirement Longitudinal Study (CHARLS)

doi:10.21203/rs.3.rs-8149868/v1

Download PDF

Research Article

Associations of sociodemographic and behavioral factors with frailty transition patterns: a multi-state Markov analysis of the China Health and Retirement Longitudinal Study (CHARLS)

https://doi.org/10.21203/rs.3.rs-8149868/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Frailty state is dynamic and reversible, but there is a lack of clarity about the patterns of frailty transition and its influencing factors. This study aimed to investigate frailty transitions and the impact of sociodemographic and behavioral factors on these transitions in Chinese middle-aged and older adults.

Methods

We used five-wave data from the China Health and Retirement Longitudinal Study (CHARLS), and 20140 Chinese adults aged ≥ 45 years were included. Frailty was assessed using a 35-item frailty index. A multi-state Markov model was used to systematically analyze the transition patterns of frailty states (robust, prefrail, frail) and death, and to explore the associations of sociodemographic and behavioral factors with frailty transitions.

Results

Among the 20140 participants at baseline, the proportions of robust, prefrailty, and frailty were 30.4%, 51.8%, and 17.8%, respectively. In the short term, participants tended to remain in their original state, while over time, the probability of deterioration and recovery increased significantly. Increasing age [HR (hazard ratio): 1.020; 95% CI (confidence interval): 1.016–1.023], being female (1.394, 1.268–1.533), living in a rural area (1.336, 1.255–1.423), being illiterate (1.207, 1.124–1.296), current smoking (1.108, 1.013–1.212), and sleep < 6 or > 8 hours (1.228, 1.157–1.304) increased the risk of frailty deterioration. While current drinking (0.924, 0.859–0.994) reduced the risk of frailty deterioration. Stratified analyses by age and gender showed consistent results with the main analysis.

Conclusions

Targeted interventions should be developed for at-risk populations and intervenable behavioral factors should be taken to slow or reverse the progression of frailty.

Frailty

Frailty transition

Multi-state Markov model

Sociodemographic and behavioral factors

Middle-aged and older adults

A multi-state Markov model was used to evaluate 10-year frailty transition patterns and their influencing factors.
Participants largely stayed the initial state in short-term, but deterioration and recovery probabilities increased over time.
Age, gender, marital status, education level, smoking, drinking, and sleep duration were associated with frailty transitions.

Frailty is a syndrome manifested as increased vulnerability to stressors due to declines in multi-system reserve, leading to elevated risk of adverse outcomes. ^[1] Actually, frailty is not an inevitable part of ageing, nor is it an irreversible one-way process of disability or mortality, but rather a dynamic process. ^[2] Previous studies have reported that transitions between the three frailty states (robust, prefrailty, and frailty) were reasonably common, and individuals can either recovery or deteriorate over time. ^[3–7] For frailty state transitions, the multi-state Markov (MSM) model was deemed as an ideal model to fully explore the transition patterns between different states and death. ^{[8, 9]} However, existing studies usually only report transition frequency or probability in a limited manner, failing to conduct systematic analysis of key parameters of frailty state transition. ^{[9, 10]} Furthermore, previous studies usually used data from two assessments and lacked long-term follow-up, and some studies have not considered the absorption state of death. ^{[11, 12]}

Given that frailty state is reversible and improvable, identifying the key factors that influence frailty transition is of major significance. While prior studies have explored the effects of age, gender, education, smoking, and drinking on the frailty transition, findings regarding these associations remain inconsistent. ^{[3, 5, 6, 11, 13, 14]} In addition, previous studies usually considered data of two time points and used logistic regression models, ignoring dynamic transitions of frailty over time. Notably, the MSM model outperformed traditional logistic regression and unidirectional transition analyses in modeling disease progression, which can simultaneously characterize multidirectional transitions (including progression to death) while systematically quantifying the differential effects of sociodemographic and behavioral factors across distinct frailty transition pathways. ^{[11, 12]} Nevertheless, no studies have yet focused on the effects of exposure factors on the multiple pathways of transition between robust, prefrailty, frailty, and death.

Therefore, based on five-wave panel data from the China Health and Retirement Longitudinal Study (CHARLS) with a total of ten years of follow-up from 2011 to 2020, this study used a MSM model and aimed to systematically analyze the transition pattern of frailty state and the associations of sociodemographic and behavioral factors with frailty transition.

Study design and population

The CHARLS study collects detailed micro-level data from Chinese residents aged 45 years and above, covering demographic characteristics, health status, cognitive function, and biological markers. Initiated in 2011 with a national baseline survey, CHARLS employed a multistage cluster sampling approach across 28 provinces, with a sample size of approximately 17000 individuals from 10000 households. The study has conducted subsequent waves of data collection in 2013, 2015, 2018, and 2020, serving as a valuable resource for cross-disciplinary aging research in China. More detailed information about the CHARLS has been previously published. ^[15] Ethical approval for all the CHARLS waves was granted from the Institutional Review Board at Peking University (IRB approval number: IRB00001052-11015) and each respondent provided informed consent form.

A total of 25586 participants in 2011, 2013, 2015, 2018, and 2020 survey were selected for this study. We excluded participants younger than 45 years at baseline and whose age and gender variables were missing at all five waves (N = 2304). The remaining 23282 participants were remained to construct the frailty index (FI). The state of death was determined by answering the questions “Whether Individual Died?” and “What was the date on which R died?”. We excluded participants who had less than two transition state data (N = 3108) and those with missing values on sociodemographic and behavioral factors (N = 34). Finally, 20140 participants were remained for the primary analysis (Fig. 1).

Assessment of frailty

By referring to previous research, ^{[16, 17]} we employed a previously validated FI by incorporating 35 variables related to diseases, symptoms, disabilities, depression, and cognition (Supplementary Table S1). ^[18] The FI value was computed as the number of items present in an individual divided by the total items considered (i.e., 35). Thus, FI was a continuous variable ranging from 0 to 1, with higher FI indicating higher level of frailty in the participants. According to previous study, ^[19] FI was further divided into three categories, robust (FI ≤ 0.10), prefrailty (0.10 < FI < 0.25), and frailty (FI ≥ 0.25).

Assessment of sociodemographic and behavioral factors

Information on sociodemographic and behavioral factors were obtained by questionnaire survey. The factors we considered were as follows: age (continuous), gender (male, female), residence areas (urban, rural), marital status (married, others), highest educational levels (literate, illiterate), smoking status (not current smokers, current smokers), drinking status (not current drinkers, current drinkers), sleep times (6–8 hours, < 6 or > 8 hours). For participants with missing values, we imputed them through forward/backward filling using adjacent wave data, ^{[20, 21]} ensuring complete data across all five waves for each participant.

Statistical analysis

Baseline characteristics of participants according to each frailty states were described, expressed as mean (standard deviation) for continuous variables and proportions for categorical variables; one-way ANOVA and χ² test was used to test possible differences across groups.

In this study, we constructed MSM models using “msm” package in R software to analyze the panel data. The MSM model is a statistical model for describing the transitions of an individual between states in a continuous time process. ^[8] The model is particularly suitable for dealing with observational data obtained in follow-up studies, which typically consist of observations of an individual's state at different waves in time, whereas the exact time of state transitions is often not straightforward to observe. Further details are provided in the Supplementary appendix.

In order to investigate the transition patterns among frailty states, four states were defined in this study: state 1 (robust, S1), state 2 (prefrailty, S2), state 3 (frailty, S3), and state 4 (death, S4). Among them, robust, prefrailty and frailty were defined as transient states, while death was an absorbing state (Fig. 2). Specifically, participants could not only experience state deterioration or improvement between adjacent states, but could also transition directly from any state to the death state. In this study, state transitions encompass three primary pathways: recovery transitions (i.e., S2→S1 or S3→S2), deterioration transitions (i.e., S1→S2 or S2→S3), and transitions to death (i.e., S1→S4, S2→S4 or S3→S4).

In this study, the MSM model constructed was used to derive the whole population transitions, as well as the transition intensity matrix, the mean sojourn time, and the probability of transitions at 1-, 3-, 5-, and 10-year for the frailty states. Hazard radios (HRs) and 95% Confidence intervals (CIs) were calculated for the associations between sociodemographic and behavioral factors and frailty state transitions. Finally, the fitting effect of the MSM model was evaluated through the plot function in the R software by the degree of matching of the observed and expected frequency and percentage curves for each state. Considering age and gender differences, we further conducted stratified analyses by age (45–60 vs. ≥ 60 years) and gender (men, women).

All statistical tests were two-side, and P-value < 0.05 was considered statistically significant. In this study, Stata 15.0 and R 4.0.0 software were used for data analysis.

A total of 20140 participants were included in this study. At baseline, the mean age of all participants were 58.2 years, 49.4% were male (Table 1). And 6123 (30.4%) participants were robust, 10436 (51.8%) prefrailty, and 3581 (17.8%) frailty. Compared to the robust group, participants in the frailty group were on average older, tended to be females, illiterate, not current smokers, not current drinkers, and sleep < 6 or > 8 hours.

Table 1

Baseline characteristics of participants according to frailty states groups
Characteristics	All (n = 20140)	Robust (n = 6123)	Prefrailty (n = 10436)	Frailty (n = 3581)	P Value
Age, mean (SD)	58.2 (9.5)	56.2 (9.0)	58.1 (9.3)	61.8 (10.1)	0.038
Gender, n (%)					< 0.001
Male	9942 (49.4)	3547 (57.9)	5040 (48.3)	1355 (37.8)
Female	10198 (50.6)	2576 (42.1)	5396 (51.7)	2226 (62.2)
Residence areas, n (%)					< 0.001
Urban	8184 (40.6)	2650 (43.3)	4356 (41.7)	1178 (32.9)
Rural	11956 (59.4)	3473 (56.7)	6080 (58.3)	2403 (67.1)
Marital status, n (%)					< 0.001
Married	16722 (83.0)	5309 (86.7)	8666 (83.0)	2747 (76.7)
Others*	3418 (17.0)	814 (13.3)	1770 (17.0)	834 (23.3)
Highest educational levels, n (%)					< 0.001
Literate	14985 (74.4)	4915 (80.3)	7839 (75.1)	2231 (62.3)
Illiterate	5155 (25.6)	1208 (19.7)	2597 (24.9)	1350 (37.7)
Smoking status, n (%)					< 0.001
Not current smokers	11907 (59.1)	3380 (55.2)	6170 (59.1)	2357 (65.8)
Current smokers	8233 (40.9)	2743 (44.8)	4266 (40.9)	1224 (34.2)
Drinking status, n (%)					< 0.001
Not current drinkers	13269 (65.9)	3559 (58.1)	6965 (66.7)	2745 (76.7)
Current drinkers	6871 (34.1)	2564 (41.9)	3471 (33.3)	836 (23.4)
Sleep times, n (%)					< 0.001
6-8h	12526 (62.2)	4384 (71.6)	6502 (62.3)	1640 (45.8)
< 6h,>8h	7614 (37.8)	1739 (28.4)	3934 (37.7)	1941 (54.2)
Note: Data were expressed as mean ± standard deviation or frequency (proportion);
One-way ANOVA was used to compare the continuous variables of each frailty states group.
Pearson’s χ² test was used to compare categorical variables of each frailty states group.
*Others include married but not living with spouse temporarily for reasons such as work, separated, divorced, widowed, and never married.

In this study, participants were followed for a ten-year period, and dynamic transitions in their frailty states were recorded (Table S2). For participants who were currently robust, 55.2% remained robust at the next follow-up, 38.4% progressed to a prefrailty state, 4.2% progressed to a frailty state, and 2.2% transitioned to death at the next follow-up. The frailty state transition intensities and mean sojourn times are shown in Table S3. The transition intensity for participants in the robust state to the prefrailty (0.301, 0.293–0.309) was 2.2 times greater than the transition intensity for participants in the prefrailty state to the frailty state (0.136, 0.132–0.139). The intensity of the transition from a frailty state to a prefrailty state (0.217, 0.210–0.224) was 1.5 times greater than the intensity of the transition from a participant in a prefrailty state to a robust state (0.146, 0.142–0.151). For mean sojourn time, the longest stay was in the frailty state (3.880 years), followed by 3.425 years in the prefrailty state, and 3.247 years in the robust state.

The transition probabilities of frailty states within 1-, 3-, 5-, and 10-year observation intervals were estimated (Table 2). For participants in the robust state, the 1-year transition probabilities from robust to prefrailty, frailty, and death were 0.226, 0.015, and 0.008, whereas the 10-year probabilities increased to 0.429, 0.208, and 0.122, respectively. The probability of frailty deterioration and transition to death was significantly higher in individuals aged ≥ 60 years compared to those aged 45–60. Additionally, females exhibited a higher likelihood of frailty progression but a lower probability of transition to death than males (Supplementary Tables S4-S5).

Table 2

Transition probabilities of frailty states within 1, 3, 5, and 10-year observation intervals
	1-year				3-year				5-year
	Robust	Prefrailty	Frailty	Death	Robust	Prefrailty	Frailty	Death	Robust	Prefrailty	Frailty	Death	Robust	Prefrailty	Frailty	Death
Robust	0.751	0.226	0.015	0.008	0.481	0.409	0.082	0.027	0.356	0.451	0.142	0.051	0.241	0.429	0.208	0.122
Prefrailty	0.110	0.774	0.104	0.012	0.199	0.562	0.198	0.041	0.219	0.482	0.227	0.072	0.209	0.414	0.228	0.150
Frailty	0.012	0.167	0.784	0.037	0.064	0.317	0.524	0.095	0.110	0.364	0.386	0.140	0.162	0.366	0.247	0.225

Table 3 presents the HRs and 95% CIs for the associations between sociodemographic and behavioral factors and frailty state transitions. Increasing age was associated with higher risk of deteriorate transition (robust to prefrailty, HR = 1.010; prefrailty to frailty, HR = 1.020) and death transition (robust to death, HR = 1.097; prefrailty to death, HR = 1.084; frailty to death, HR = 1.100) and lower likelihood of recovery transition (prefrailty to robust, HR = 0.987; frailty to prefrailty, HR = 0.986). Compared to males, females were at higher risk of deteriorate transition (robust to prefrailty, HR = 1.252; prefrailty to frailty, HR = 1.394) and lower likelihood of recovery transition (prefrailty to robust, HR = 0.779; frailty to prefrailty, HR = 0.835); nevertheless, females were at lower risk of progressing to death (robust to death, HR = 0.454; prefrailty to death, HR = 0.574; frailty to death, HR = 0.482). Compared to urban residence, rural residence had higher risk of deteriorate transition (robust to prefrailty, HR = 1.172; prefrailty to frailty, HR = 1.336). Compared to married, participants in other marital status were at increased risk of transition to death, i.e., prefrailty to death (HR = 1.390), frailty to death (HR = 1.295); they were also at lower probability of recovery transition, i.e., frailty to prefrailty (HR = 0.914). Compared to being literate, illiterate participants were at increased risk for deterioration and death transition (prefrailty to frailty, HR = 1.207; frailty to death, HR = 1.386).

Table 3

Hazard radios and 95% confidence intervals of covariates on transitions among frailty states transitions
Covariates	Recovery transition		Deteriorate transition		Death transition
Covariates	Prefrailty→Robust	Frailty→Prefrailty	Robust→Prefrailty	Prefrailty→Frailty	Robust→Death	Prefrailty→Death	Frailty→Death
Age, continuous	0.987 (0.984,0.991)	0.986 (0.982,0.990)	1.010 (1.006,1.013)	1.020 (1.016,1.023)	1.097 (1.072,1.124)	1.084 (1.067,1.101)	1.100 (1.090,1.110)
Gender (ref = Male)
Female	0.779 (0.708,0.857)	0.835 (0.752,0.926)	1.252 (1.147,1.367)	1.394 (1.268,1.533)	0.454 (0.242,0.850)	0.574 (0.373,0.885)	0.482 (0.390,0.595)
Residence areas (ref = Urban)
Rural	1.228 (1.151,1.310)	1.074 (0.996,1.157)	1.172 (1.107,1.242)	1.336 (1.255,1.423)	0.969 (0.631,1.488)	1.038 (0.809,1.332)	0.984 (0.844,1.148)
Marital status (ref = Married)
Others*	1.023 (0.939,1.114)	0.914 (0.838,0.996)	1.038 (0.959,1.123)	0.992 (0.920,1.070)	1.188 (0.720,1.960)	1.390 (1.054,1.834)	1.295 (1.110,1.512)
Highest educational levels (ref = Literate)
Illiterate	1.397 (1.289,1.514)	1.019 (0.940,1.104)	1.068 (0.991,1.152)	1.207 (1.124,1.296)	1.273 (0.781,2.075)	1.140 (0.828,1.569)	1.386 (1.179,1.630)
Smoking status (ref = Not current smokers)
Current smokers	0.899 (0.822,0.983)	0.865 (0.782,0.956)	1.068 (0.985,1.159)	1.108 (1.013,1.212)	0.860 (0.525,1.409)	2.183 (1.501,3.175)	1.107 (0.913,1.342)
Drinking status (ref = Not current drinkers)
Current drinkers	1.068 (0.995,1.147)	1.191 (1.096,1.294)	0.978 (0.917,1.044)	0.924 (0.859,0.994)	0.902 (0.562,1.450)	0.683 (0.519,0.899)	0.678 (0.558,0.824)
Sleep times (ref = 6-8h)
< 6h,>8h	0.929 (0.871,0.991)	0.856 (0.799,0.918)	1.166 (1.098,1.237)	1.228 (1.157,1.304)	1.545 (0.962,2.482)	1.078 (0.827,1.405)	0.911 (0.785,1.058)
Note: Bold fonts data represent statistical significance (P ≤ 0.05).
In marital status: *Others include married but not living with spouse temporarily for reasons such as work, separated, divorced, widowed, and never married.

Compared to not current smokers, current smokers increased the risk of deterioration transition and transition to death, i.e., prefrailty to frailty (HR = 1.108), prefrailty to death (HR = 2.183), they were also at lower probability of recovery transition (prefrailty to robust, HR = 0.899; frailty to prefrailty, HR = 0.865). Compared to not current drinkers, current drinkers were at higher likelihood of recovery transition (frailty to prefrailty, HR = 1.191), they were also at lower risk of deterioration and death transition, i.e., prefrailty to frailty (HR = 0.924), prefrailty to death (HR = 0.683), frailty to death (HR = 0.678). As compared to sleep times of 6–8 hours, sleep times of < 6 or > 8 hours were at higher risk of deterioration transition (robust to prefrailty, HR = 1.166; prefrailty to frailty, HR = 1.228), they were also at lower probability of recovery transition, i.e., prefrailty to robust (HR = 0.929), frailty to prefrailty (HR = 0.856).

The results of the age- and sex-stratified analysis were largely consistent with the primary analysis (Supplementary Table S6-S9). The fitting of the MSM model for the frailty state is shown in Figure S1. The two curves, the expected frequency dashed line and the observed frequency solid line, tended to coincide, thus the model fit well.

In this longitudinal study with a ten-year follow-up, we incorporated 20140 Chinese adults aged 45 years and above as participants. By constructing the MSM model, we found that the transition intensity (i.e., the instantaneous transition risk) from robust to prefrailty states was significantly higher (2.2 times) than from prefrailty to frailty states. Similarly, the recovery intensity from frailty to prefrailty states exceeded (1.5 times) that from prefrailty to robust states. The transition probabilities suggested that most of the study subjects were tended to keep the original state unchanged at 1-year, but the probabilities of both deterioration and improvement of each frailty state were significantly increased over time. Overall, increasing age, female, illiterate, current smokers, not current drinkers, and sleeping < 6 or > 8h increased the risk of frailty deterioration and transition to death.

The present study showed that, for the actual transition frequency, the vast majority of participants remained unchanged in their original state, transitions between adjacent states were more common than transitions across states (i.e., robust ⇌ frailty), which was in line with previous evidence. ^{[4, 5]} Also, prefrail participants were more likely to have both state improvement (prefrailty → robust: 18.3% vs frailty → robust: 3.1%) and deterioration (prefrailty → frailty: 18.3% vs robust → frailty: 4.2%) than frail and robust state participants, indicating that preventive interventions in the prefrailty period would be more effective. ^{[5, 22]} The results of our study were similar to the results of a meta-analysis among older people, ^[23] and our study extended these findings to a broad age range of 45 years and above.

There is limited evidence on the intensity of frailty state transitions and mean sojourn time. In our study, by using MSM models, we found that the transition intensity from frailty state to death was significantly higher than that from robust/prefrailty to death. The mean sojourn time in the frailty state was the longest, indicating that the probability of reversion after entering frailty state was lower than entering robust and prefrailty states. Over the ten-year follow-up, the results of transition probabilities showed that in the short term, the tendency was to keep the original state unchanged. However, the probability of deterioration and improvement of each state increased significantly over time. The results of the transfer probability in this study were consistent with the conclusions of the Irish Longitudinal Study on Ageing. ^{[9, 10]} Thus, a deeper understanding of the pattern of frailty state transition can help to clarify the optimal target population for frailty interventions. Some research suggest early intervention for prefrailty populations. ^[23]

Frailty is reversible, we then comprehensively examined the associations between multiple influencing factors and frailty transitions. We found that advancing age significantly increased the risk of both frailty deterioration and transition to death while reducing the likelihood of frailty improvement, which is consistent with longitudinal studies conducted in Europe and Southeast Asia. ^{[6, 12]} Regarding gender, our results showed that compared with males, females were more likely to undergo deteriorating transitions, but less likely to undergo improving transitions or transitioning to death. Existing evidence also indicated that although women exhibited a higher risk of frailty onset, they demonstrated a relatively lower probability of transition to death, ^{[22, 24]} which is consistent with our study. Existing evidence on the association of residence and marital status with frailty transition is quite limited. In our study, we found that the effects of rural residence on frailty were bidirectional: it may either facilitated the improvement of frailty, possibly through increasing physical activity opportunities and strengthening social support networks, ^{[25, 26]} or accelerated the deterioration of frailty due to lower availability of healthcare resources and higher chance of occupational hazardous exposures. ^[27–30] Regarding marital status, we found separated or unmarried status significantly increased the risk of transition to death and decreased the likelihood of improvement in frailty status, which was in line with established studies. ^{[31, 32]}

For education, Emiel O. Hoogendijk et al ^[33] found that shorter years of education was associated with the progression of frailty. Also, previous study have found that higher level of education was a protective factor for frailty. ^[32] The results of our study showed that being illiterate increased the risk of deterioration of the frailty state and the transition to the death state, which was consistent with the results of previous studies. We also found that current smokers and sleep < 6 hours or > 8 hours increased the risk of frailty state deterioration, while current drinkers decreased the risk of worsening frailty and transition to death. These findings are largely consistent with those of previous studies. ^{[14, 22, 34, 35]} Noteworthily, when analyzing the influencing factors, we employed the MSM model to explore the association between influencing factors and various transition paths of frailty. In contrast, prior studies predominately relied on two-wave changes of frailty status, typically using logistic regression model. Unlike logistic regression, which only captures transitions between two discrete time points, the MSM model can fully accounts for the dynamic transition process between multiple states, thus providing a more comprehensive understanding of frailty transition over time.

The main strength of this study was that, we used five-wave longitudinal data from the CHARLS study, which was representative of the Chinese mainland population. By using MSM models, we not only revealed the dynamic evolution of frailty status, but also comprehensively investigated the impact of multiple sociodemographic and lifestyle factors on diverse transition paths of frailty. These findings offer novel evidence to inform targeted interventions for frailty prevention and management. Several limitations also should be noted. First, CHARLS data were obtained mainly through self-reported questionnaires, which may lead to recall bias. Second, because some of the variables of the frailty phenotype could not cover the whole follow-up period, only FI was used to assess frailty in this study. Third, for death outcomes, we did not examine specific causes of death, and addressing this question in future studies could reveal specific biological risks associated with the frailty state. Last, residual confounding may exist in our study.

Based on ten-year dynamic data from the CHARLS with five follow-ups, this study revealed that the frailty state has significant time-varying characteristics. Individuals' frailty status tends to stabilize in the short term, but long-term follow-up showed that the probability of both deterioration and recovery of frailty status increased significantly. Further analysis showed that age, gender, residence, marital status, education level, smoking, drinking, and sleep duration all had a significant effect on the transition (recovery, deterioration, and death) of the frailty state. These findings highlight the imperative for early interventions focusing on modifiable behavioral factors to delay frailty progression among middle-aged and older adults.

Ethics approval and consent to participate

Ethical approval for this study was conducted by the Institutional Review Board at Peking University (IRB approval number: IRB00001052-11015) and each respondent provided informed consent form. The study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. All participants were informed about the purpose of the study, assured of confidentiality, and provided written consent prior to participation. Participation was voluntary, and respondents could withdraw at any time without consequence.

Consent for publication

Not applicable.

Availability of data and materials

All CHARLS information is available online on the webpage http://charls.pku.edu.cn/en. Data access is available through applications.

Competing interests

No potential conflicts of interest relevant to this article were reported.

Funding

This work was supported by National Natural Science Foundation of China (82304226), Natural Science Foundation of Shandong Province (ZR2023QH188), China Postdoctoral Science Foundation (2023M731839), Mount Taishan Scholar Youth Program (No.tsqn202306179), Qingdao Postdoctoral Innovation Project (QDBSH20230102012), Qingdao University Scientific Research Startup Fund (DC2200002531). The funders had no role in the study design, data collection, data analysis and interpretation, writing of the report, or the decision to submit the article for publication.

Authors' contributions

JF conceived and designed the study. KZ and JF analyzed the data. KZ drafted the manuscript. YZ, WK, XH and LC helped to organize the information and review the data. JF and KZ helped the interpretation of the results. JF contributed to the critical revision of the manuscript for important intellectual content and approved the final version of the manuscript. All authors reviewed and approved the final manuscript. JF and DZ is the guarantor.

Acknowledgements

We are grateful to the China Health and Retirement Longitudinal Study (CHARLS) participants. This research has been conducted using the CHARLS Resource.

Ensrud KE, Ewing SK, Taylor BC, et al. Frailty and risk of falls, fracture, and mortality in older women: the study of osteoporotic fractures[J]. J Gerontol A. 2007;62(7):744–51.
Clegg A, Young J, Iliffe S, et al. Frailty in elderly people[J]. Lancet. 2013;381(9868):752–62.
Ottenbacher KJ, Graham JE, Al Snih S, et al. Mexican Americans and frailty: findings from the Hispanic established populations epidemiologic studies of the elderly[J]. Am J Public Health. 2009;99(4):673–9.
Gill TM, Gahbauer EA, Allore HG, et al. Transitions between frailty states among community-living older persons[J]. Arch Intern Med. 2006;166(4):418–23.
Espinoza SE, Jung I, Hazuda H. Frailty Transitions in the San Antonio Longitudinal Study of Aging[J]. J Am Geriatr Soc. 2012;60(4):652–60.
Lee JSW, Auyeung TW, Leung J, et al. Transitions in frailty states among community-living older adults and their associated factors[J]. J Am Med Dir Assoc. 2014;15(4):281–6.
Sd NFAM. S, Transitions in frailty status in older adults in relation to mobility: a multistate modeling approach employing a deficit count[J]. J Am Geriatr Soc, 2011, 59(3).
Jackson C. Multi-State Models for Panel Data: The msm Package for R[J]. J Stat Softw. 2011;38:1–28.
Romero-Ortuno R, Hartley P, Davis J, et al. Transitions in frailty phenotype states and components over 8 years: Evidence from The Irish Longitudinal Study on Ageing[J]. Arch Gerontol Geriatr. 2021;95:104401.
Romero-Ortuno R, Hartley P, Knight SP, et al. Frailty index transitions over eight years were frequent in The Irish Longitudinal Study on Ageing[J]. HRB open Res. 2021;4:63.
Lorenzo-López L, López-López R, Maseda A, et al. Changes in frailty status in a community-dwelling cohort of older adults: The VERISAÚDE study[J]. Maturitas. 2019;119:54–60.
Setiati S, Laksmi PW, Aryana I G P, S, et al. Frailty state among Indonesian elderly: prevalence, associated factors, and frailty state transition[J]. BMC Geriatr. 2019;19(1):182.
Lee YS, Nishita Y, Tange C, et al. Association between objective physical activity and frailty transition in community-dwelling prefrail Japanese older adults[J]. J Nutr health aging. 2025;29(4):100519.
To TL, Kuo CP, Yeh CJ, et al. Transitions of self-management behaviors and frailty status among community-dwelling older adults: a national longitudinal population-based cohort study[J]. BMC Geriatr. 2022;22(1):874.
Zhao Y, Hu Y, Smith JP, et al. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS)[J]. Int J Epidemiol. 2014;43(1):61–8.
Li C, Ma Y, Yang C, et al. Association of Cystatin C Kidney Function Measures With Long-term Deficit-Accumulation Frailty Trajectories and Physical Function Decline[J]. JAMA Netw Open. 2022;5(9):e2234208.
Searle SD, Mitnitski A, Gahbauer EA, et al. A standard procedure for creating a frailty index[J]. BMC Geriatr. 2008;8:24.
Zhang K, Chai L, Zhang Y, et al. Association of childhood and adulthood socioeconomic status with frailty index trajectories: Using five-wave panel data from the China Health and Retirement Longitudinal Study (CHARLS)[J]. Arch Gerontol Geriatr. 2025;131:105780.
Fan J, Yu C, Guo Y, et al. Frailty index and all-cause and cause-specific mortality in Chinese adults: a prospective cohort study[J]. Lancet Public Health. 2020;5(12):e650–60.
Luo Y, Nur J, Jin Y. Adjust for non-ignorable panel attrition in the analysis of leaving the parental home[J]. Adv Life Course Res. 2024;60:100605.
Davis MP. Missing Data and the Last Observation Carried Forward[J]. J Pain Symptom Manag. 2024;67(6):e921–2.
Trevisan C, Veronese N, Maggi S, et al. Factors Influencing Transitions Between Frailty States in Elderly Adults: The Progetto Veneto Anziani Longitudinal Study[J]. J Am Geriatr Soc. 2017;65(1):179–84.
Kojima G, Taniguchi Y, Iliffe S, et al. Transitions between frailty states among community-dwelling older people: A systematic review and meta-analysis[J]. Ageing Res Rev. 2019;50:81–8.
Gordon EH, Hubbard RE. Differences in frailty in older men and women[J]. Med J Aust. 2020;212(4):183–8.
Angulo J, El Assar M, Álvarez-Bustos A, et al. Physical activity and exercise: Strategies to manage frailty[J]. Redox Biol. 2020;35:101513.
Nascimento CM, Ingles M, Salvador-Pascual A, et al. Sarcopenia, frailty and their prevention by exercise[J]. Volume 132. Free Radical Biology & Medicine; 2019. pp. 42–9.
Ying M, Wang S, Bai C, et al. Rural-urban differences in health outcomes, healthcare use, and expenditures among older adults under universal health insurance in China[J]. PLoS ONE. 2020;15(10):e0240194.
Thorpe JM, Van Houtven CH, Sleath BL, et al. Rural-urban differences in preventable hospitalizations among community-dwelling veterans with dementia[J]. JRural Health. 2010;26(2):146–55.
Alavanja MCR, Hoppin JA, Kamel F. Health effects of chronic pesticide exposure: cancer and neurotoxicity[J]. Annu Rev Public Health. 2004;25:155–97.
Curl CL, Spivak M, Phinney R, et al. Synthetic Pesticides and Health in Vulnerable Populations: Agricultural Workers[J]. Curr Environ Health Rep. 2020;7(1):13–29.
Kojima G, Walters K, Iliffe S, et al. Marital Status and Risk of Physical Frailty: A Systematic Review and Meta-analysis[J]. J Am Med Dir Assoc. 2020;21(3):322–30.
An S, Ouyang W, Wang S, et al. Marital transitions and frailty among middle-aged and older adults in China: The roles of social support[J]. SSM - Popul health. 2023;24:101497.
Hoogendijk EO, Dent E. Trajectories, Transitions, and Trends in Frailty among Older Adults: A Review[J]. Annals Geriatric Med Res. 2022;26(4):289–95.
Pérez-Tasigchana RF, Sandoval-Insausti H, Donat-Vargas C et al. Combined Impact of Traditional and Nontraditional Healthy Behaviors on Frailty and Disability: A Prospective Cohort Study of Older Adults[J]. Journal of the American Medical Directors Association, 2020, 21(5): 710.e1-710.e9.
Zhu Y, Fan J, Lv J, et al. Maintaining healthy sleep patterns and frailty transitions: a prospective Chinese study[J]. BMC Med. 2022;20(1):354.

No competing interests reported.

Supplementarytransition45ZK1124.docx

Download PDF

Reviewers agreed at journal
08 Jan, 2026
Reviewers agreed at journal
26 Dec, 2025
Reviewers invited by journal
12 Dec, 2025
Editor assigned by journal
25 Nov, 2025
Submission checks completed at journal
25 Nov, 2025
First submitted to journal
18 Nov, 2025

You are reading this latest preprint version

Associations of sociodemographic and behavioral factors with frailty transition patterns: a multi-state Markov analysis of the China Health and Retirement Longitudinal Study (CHARLS)

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Key Points

Introduction

Methods

Study design and population

Assessment of frailty

Assessment of sociodemographic and behavioral factors

Statistical analysis

Results

Discussion

Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1