Frailty progression before and after diabetes onset among older adults in china: a nationwide cohort study

Abstract

Background

Although epidemiological studies have established a link between diabetes and frailty, the dynamic changes in frailty before and after diabetes diagnosis remained inadequately characterized. This study aimed to examine the trajectories of frailty preceding and following diabetes diagnosis in older adults and to identify potential modifiers of the association between diabetes and trajectories of frailty.

Methods

We included adults aged≥65 years without baseline diabetes from the Chinese Longitudinal Healthy Longevity Survey (2002–2018). Incident diabetes was defined by self-reported physician diagnosis or diabetes-related hospitalization. The date of diagnosis was estimated as the midpoint between the last non-diabetic interview and the first report of diabetes. Frailty was assessed repeatedly using a 43-item frailty index (FI). Linear mixed-effects models were used to compare frailty trajectories before and after diabetes onset.

Results

Among 9,992 participants (mean age 82.12 years; 51.31% women), 420 developed diabetes over a median follow-up of 8.67 years. After diagnosis, the rate of change in FI increased significantly, with an annual additional increase of 0.52 (95% CI: 0.33, 0.71). The most pronounced acceleration was observed in instrumental activities of daily living (β = 0.80/year, 95% CI: 0.27, 1.32). Subgroup analyses revealed stronger associations in participants aged ≥80 years (β = 1.01/year, 95% CI: 0.42, 1.59), those who were single (β = 0.60/year, 95% CI: 0.24, 0.95), and those with BMI <22 kg/m² (β = 0.72/year, 95% CI: 0.41, 1.03).

Conclusions

Incident diabetes was associated with accelerated progression of frailty in older adults, particularly in the domain of instrumental activities of daily living. Regular functional assessments and targeted interventions following diabetes diagnosis was important, especially for high-risk subgroups including the oldest-old, unmarried individuals, and those with lower BMI.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-025-06681-3.

Background

Population aging represents a global demographic trend characterized by an increasing proportion of older adults and a higher prevalence of age-related chronic diseases and geriatric conditions [1]. Among these conditions, diabetes is highly prevalent in this population, affecting more than one in five individuals aged 65 years or older [2]. Although the elevated risks of macrovascular and microvascular complications were well documented [3, 4], recent evidence indicated that frailty is emerging as an additional complication among older patients with diabetes [5, 6]. Epidemiological studies show that frailty affects approximately 50% of individuals with diabetes, with these individuals experiencing a three- to five fold higher likelihood of developing frailty compared to those without diabetes [7]. Moreover, frailty increases the risk of adverse outcomes such as fractures, falls, disability, and hospitalization, and contributes to higher healthcare expenditures in people with diabetes [8]. As frailty is considered a reversible condition, it is essential to understand its dynamic trends within this high-risk group to implement early interventions that may delay its onset and progression.

Diabetes and frailty are both age-related conditions sharing common pathophysiological mechanisms and risk factors, suggesting that individuals with diabetes may experience frailty progression even prior to its clinical onset [9]. However, most previous research has focused on individuals with established diabetes, thereby overlooking frailty in the prediabetic phase [10, 11]. Currently, limited studies have simultaneously examined frailty changes occurring both before and after a diabetes diagnosis, leaving the overall frailty trajectory largely undetermined. Moreover, because frailty tends to progress with age even among healthy older adults [12], it is crucial to assess prediabetes-related frailty changes to accurately determine the impact of diabetes on subsequent frailty trajectories.

To address these gaps, we utilized data from a nationally representative cohort, the Chinese Longitudinal Healthy Longevity Survey (CLHLS), to map the frailty trajectory before and after a diabetes diagnosis and identify factors that modify the risk associated with changes in the frailty trajectory after diabetes diagnosis.

Methods

Study design and participants

The CLHLS is an ongoing nationwide, open cohort study in China. A detailed description of the study design has been published elsewhere [13]. The study began in 1998, with follow-up surveys conducted in 2000, 2002, 2005, 2008, 2011, 2014, and 2018. To maintain sample size and representativeness, new participants were enrolled at each follow-up wave. The study has accumulated over 100 thousand participant-records from more than 57,000 unique older adults (aged ≥ 65 years) across 23 provinces in China (Figure S1). Participants completed face-to-face, interviewer-administered questionnaires, anthropometric evaluations, and laboratory measurements of biological samples.

For the present analysis, we obtained data from all waves of the CLHLS (1998–2018). The flowchart for participant selection is shown in Figure S2. Of the 57,071 participants enrolled across all baseline survey waves, we applied the following exclusion criteria: 9,146 individuals participated only in the 1998–2000 survey and lacked outcome variable information; 12,443 were first enrolled in the 2018 baseline survey and had no follow-up data; 18,412 were lost to follow-up; 3,045 had missing diabetes status or prevalent diabetes at baseline; 1,770 had missing covariate data; and 2,263 had missing information on more than 10% of the frailty index items. Ultimately, 9,992 participants were included in our analysis, each having complete baseline data and at least one reassessment of diabetes status and frailty index during follow-up.

Assessment of incident diabetes

Information on diabetes was collected at each survey wave through two questions: (1) “What is the name of the disease suffered at the 1st/2nd/last hospital visit?”, and (2) “Have you been suffering from diabetes?”. Diabetes was defined as a history of hospitalization due to diabetes or self-reported physician-diagnosed diabetes. The date of diabetes diagnosis was recorded as the midpoint between the last interview at which diabetes was not reported and the interview at which incident diabetes was first reported. Notably, we did not distinguish between type 1 and type 2 diabetes.

Assessment of frailty

Frailty was evaluated at each wave using the frailty index (FI), a measure that captures the accumulation of multiple age-related health deficits [14]. In this study, we followed the 44-item FI methodology proposed by Liu and colleagues [15]. Of the original 44 items, we selected 43 (excluding diabetes) for FI construction, comprising 5 items related to psychological well-being, 6 to basic activities of daily living, 8 to instrumental activities of daily living, 7 to functional limitations, 4 to health status, and 13 to chronic diseases (Table S1). Each item was dichotomized or polychotomized and mapped to the range of 0 to 1, representing the severity of the deficit. For each participant, the FI was calculated by summing the presence of these deficits and dividing by the total number of deficits (n = 43). As a result, the FI ranged from 0 to 1, with higher scores indicating a greater level of frailty. For clarity in presenting our results, we multiplied the original FI by 100.

Covariates

Covariates were collected via face-to-face questionnaires at each wave. These covariates included age (years), sex (men, women), years of education, residential area (urban, rural), marital status (married and living with spouse, single), tobacco smoking (current or former, never), alcohol drinking (current or former, never), regular physical activity (yes, no), and body mass index (BMI, kg/m²). Individuals were classified as smokers if they had smoked at any time in the past or were currently smoking at the time of the interview, and as drinkers if they had consumed alcohol previously or were currently drinking at the time of the survey. Height and weight were measured while participants wore lightweight clothing and no shoes. BMI was calculated as weight (kg) divided by the square of height (m²) and further categorized as underweight (< 22.0 kg/m²), normal (22.0–<26.9 kg/m²), and overweight (≥ 26.9 kg/m²) in accordance with recent guidelines [16].

Statistical analysis

Participants were divided into two groups: those who remained diabetes-free and those who incident diabetes during follow-up. Baseline characteristics were reported as mean (standard deviation, SD) for continuous variables and frequency (percentage) for categorical variables.

To analyze the longitudinal data with repeated measurements, we employed linear mixed models with years since baseline as the time scale and included random intercepts at the individual level. The following variables were incorporated into the model:

1. Calendar time since baseline (T);

2. Time after diabetes onset (T_after): coded 0 for the diabetes-free group and for the incident-diabetes group during the prediabetes period.

In our primary analysis, we anticipated an accelerated progression of the FI rather than an abrupt change immediately following a diabetes diagnosis. Accordingly, we used the following regression model:

FI = β₀ + β₁ × T + β₂ × T_after.

Where β₀ represents the baseline FI level, β₁ is the rate of change (slope) in FI for all individuals over the period without diabetes, β₂ is the difference in slope post-diabetes relative to pre-diabetes and can be interpreted as the long-term impact of diabetes on the rate of change in FI.

All models were adjusted for time-varying confounders associated with diabetes and FI progression, including age, sex, years of education, residential area, marital status, tobacco smoking, alcohol drinking, physical activity, and BMI.

For illustrative purposes, we calculated predicted FI values for an 85-year-old, unmarried, divorced, or widowed woman living in a rural area with no formal education. She did not smoke or consume alcohol, did not engage in regular physical activity, and had a BMI of 21 kg/m². We assumed that incident diabetes occurred (or did not occur) midway through the follow-up (i.e., at year 5).

Differences in FI progression after diabetes onset across various subgroups were examined by including a subgroup variable and its interaction with the time variable denoting the post-diabetes onset period. We also performed two sensitivity analyses to evaluate the robustness of our primary results. First, we repeated the main analysis after including participants with missing covariates, using median imputation for continuous variables and mode imputation for categorical variables. Second, we restricted the analysis to those participants who contributed data from at least four survey waves (n = 2,407, including 215 incident diabetes and 2,192 remained diabetes-free).

Results

Characteristics of participants

A total of 9,992 participants with 28,545 visit records were included in the final analysis. The mean (SD) baseline age was 82.12 (11.17) years, 5,127 (51.31%) were women, and 420 (4.20%) developed diabetes during follow-up. The median (interquartile range [IQR]) follow-up duration was 8.67 (3.58, 11.93) years for the incident diabetes group and 3.45 (3.00, 6.20) years for the diabetes-free group. From 2002 to 2018, the number of available records was 222, 297, 357, 294, 228, 127 in the incident-diabetes group, respectively; the corresponding numbers were 4,961, 6,848, 6,240, 4,363, 3,026, and 1,582 in the diabetes-free group.

Table 1 presents the baseline characteristics of participants stratified by incident diabetes status. Compared with diabetes-free participants, those who developed diabetes during follow-up were younger, more likely to be women, more educated, more likely to live in urban areas and be married and living with spouse, engaged in physical activity more frequently, and had higher BMIs, but showed a similar prevalence of tobacco smoking and alcohol drinking. Baseline characteristics of participants who were enrolled in the CLHLS but were excluded from the current analysis are provided in Table S2.

Table 1.

Baseline characteristics of participants by diabetes status during follow-up

	Diabetes status
	Diabetes-free	Incident diabetes
Number of participants	9,572	420
Age (mean ± SD, years)	82.09 ± 11.16	75.18 ± 9.38
Age groups (n, %)
< 80	4,092 (42.7)	306 (72.9)
≥ 80	5,480 (57.3)	114 (27.1)
Sex (n, %)
Men	4,682 (48.9)	183 (43.6)
Women	4,890 (51.1)	237 (56.4)
Education (mean ± SD, years)	2.50 ± 3.68	3.33 ± 4.20
Residential area (n, %)
Urban	3,906 (40.8)	222 (52.9)
Rural	5,666 (59.2)	198 (47.1)
Marital status
Married and living with spouse	4,018 (42.0)	260 (61.9)
Single	5,554 (58.0)	160 (38.1)
Tobacco Smoking (n, %)
Current or former	3,550 (37.1)	151 (36.0)
Never	6,022 (62.9)	269 (64.0)
Alcohol drinking (n, %)
Current or former	3,221 (33.7)	133 (31.7)
Never	6,351 (66.3)	287 (68.3)
Regular physical activity (n, %)
Yes	3,962 (41.4)	215 (51.2)
No	5,610 (58.6)	205 (48.8)
BMI (mean ± SD, kg/m2)	21.00 ± 4.82	23.12 ± 5.29
BMI categories (n, %)	6,338 (66.2)	196 (46.7)
Underweight (< 22.0 kg/m2)	2,383 (24.9)	152 (36.2)
Normal (22.0-<26.9 kg/m2)	851 (8.9)	72 (17.1)
Overweight (≥ 26.9 kg/m2)	21.00 (4.82)	23.12 (5.29)

Rate of change in FI before and after incident diabetes

The rate of change in FI among all individuals without prior diabetes was 0.31/year (95% CI: 0.28, 0.35). In the adjusted model including all participants, the difference in the rate of change in FI was 0.52/year (95% CI: 0.33, 0.71) after diabetes onset (Table 2; Fig. 1A). Prior to diabetes onset, the rate of change in FI in the incident diabetes group was 0.33/year (95% CI: 0.15, 0.50), which was similar to that in the diabetes-free group (0.31/year, 95% CI: 0.28, 0.35) (Table 2; Fig. 1B). When analysis was restricted to participants with incident diabetes, the estimated difference in the rate of change after versus before diabetes diagnosis was 0.29/year, (95% CI: −0.02, 0.59). This estimate was attenuated compared to that from the overall population and was no longer statistically significant (p = 0.065).

Table 2.

Difference in the rate of change in frailty index and its components before and after incident diabetes ^a

	All participants		Incident diabetes only		Diabetes-free only
	β (95% CI)	p	β (95% CI)	p	β (95% CI)	p
Number of participants	9,992		420		9,572
Number of observations	28,545		1,525		27,020
Frailty index
Slope without incident diabetes (TIS, point/y)	0.31 (0.28, 0.35)	< 0.001	0.33 (0.15, 0.50)	< 0.001	0.31 (0.28, 0.35)	< 0.001
Difference in postdiabetes relative to TIS (TSD; point/y)	0.52 (0.33, 0.71)	< 0.001	0.29 (−0.02, 0.59)	0.065	NA	NA
Psychological well-being
Slope without incident diabetes (TIS, point/y)	0.15 (0.10, 0.20)	< 0.001	0.22 (−0.03, 0.46)	0.088	0.15 (0.10, 0.21)	< 0.001
Difference in postdiabetes relative to TIS (TSD; point/y)	0.10 (−0.20, 0.39)	0.519	−0.16 (−0.6, 0.28)	0.478	NA	NA
Basic activities of daily living
Slope without incident diabetes (TIS, point/y)	0.16 (0.12, 0.21)	< 0.001	0 (−0.19, 0.20)	0.962	0.17 (0.13, 0.22)	< 0.001
Difference in postdiabetes relative to TIS (TSD; point/y)	0.31 (0.07, 0.56)	0.013	0.39 (0.04, 0.74)	0.031	NA	NA
Instrumental activities of daily living
Slope without incident diabetes (TIS, point/y)	0.46 (0.37, 0.55)	< 0.001	0.21 (−0.23, 0.64)	0.346	0.47 (0.38, 0.56)	< 0.001
Difference in postdiabetes relative to TIS (TSD; point/y)	0.80 (0.27, 1.32)	0.003	1.16 (0.40, 1.93)	0.003	NA	NA
Functional limitations
Slope without incident diabetes (TIS, point/y)	0.18 (0.13, 0.23)	< 0.001	0.04 (−0.19, 0.26)	0.737	0.19 (0.14, 0.24)	< 0.001
Difference in postdiabetes relative to TIS (TSD; point/y)	0.38 (0.10, 0.66)	0.008	0.45 (0.05, 0.85)	0.026	NA	NA
Health status
Slope without incident diabetes (TIS, point/y)	0.59 (0.54, 0.64)	< 0.001	0.82 (0.52, 1.11)	< 0.001	0.58 (0.53, 0.63)	< 0.001
Difference in postdiabetes relative to TIS (TSD; point/y)	0.79 (0.48, 1.09)	< 0.001	−0.09 (−0.61, 0.43)	0.729	NA	NA
Chronic diseases
Slope without incident diabetes (TIS, point/y)	0.26 (0.23, 0.28)	< 0.001	0.49 (0.31, 0.67)	< 0.001	0.25 (0.22, 0.28)	< 0.001
Difference in postdiabetes relative to TIS (TSD; point/y)	0.77 (0.60, 0.93)	< 0.001	−0.05 (−0.37, 0.27)	0.766	NA	NA

Abbreviations: CI  Confidence interval, TIS  Time in the study, TSD  Time since diabetes

^a Adjusted for calendar time, and time varying information in age, sex, education years, urban/rural residence, marital status, tobacco smoking, alcohol drinking, physical activity and body mass index

Fig. 1.

Predicted mean change in frailty index before and after incident diabetes at year 5 (Panel A: All participants analysis; Panel B: Subgroup analyses) . Predicted values of the frailty index were calculated for an 85-year-old unmarried, divorced, or widowed woman residing in a rural area. She had no formal education, did not smoke or consume alcohol, did not engage in regular physical activity, and had a body mass index of 21 kg/m². The Solid lines represent predicted values and the dashed lines represent the 95% confidence intervals

Rate of change in FI components before and after incident diabetes

As shown in Table 2; Fig. 2, all FI components increased with age prior to incident diabetes. The annual rates of change were 0.15/year (95% CI: 0.10, 0.20) for psychological well-being, 0.16/year (95% CI: 0.12, 0.21) for basic activities of daily living, 0.46/year (95% CI: 0.37, 0.55) for instrumental activities of daily living, 0.18/year (95% CI: 0.13, 0.23) for functional limitations, 0.59/year (95% CI: 0.54, 0.64) for health status, and 0.26/year (95% CI: 0.23, 0.28) for chronic diseases. Long-term outcomes of diabetes significantly influenced the rate of change in all FI components except psychological well-being (β = 0.10/year, 95% CI: −0.20, 0.39). The difference in the rate of change after diabetes onset was greatest for instrumental activities of daily living (β = 0.80/year, 95% CI: 0.27, 1.32) and smallest for basic activities of daily living (β = 0.31/year, 95% CI: 0.07, 0.56).

Fig. 2.

Predicted mean change in 6 components of frailty index before and after incident diabetes at year 5 in all participants (Panel A: Psychological well-being; Panel B: Basic activities of daily living; Panel C: Instrumental activities of daily living; Panel D: Functional limitations; Panel E: Health status; Panel F: Chronic diseases). Predicted values of the frailty index were calculated for an 85-year-old unmarried, divorced, or widowed woman residing in a rural area. She had no formal education, did not smoke or consume alcohol, did not engage in regular physical activity, and had a body mass index of 21 kg/m². The Solid lines represent predicted values and the dashed lines represent the 95% confidence interval

Subgroup analysis

Most subgroup interactions were not statistically significant, except for age (p for interaction < 0.001), marital status (p for interaction = 0.042), and BMI (p for interaction = 0.035) (Fig. 3). Compared with their respective reference groups, participants aged ≥ 80 years (β = 1.01/year, 95% CI: 0.42, 1.59), those who were single (β = 0.60/year, 95% CI: 0.24, 0.95), and those with BMI < 22 kg/m² (β = 0.72/year, 95% CI: 0.41, 1.03) experienced a prominent increase in the rate of change in FI than younger individuals (β = 0.62/year, 95% CI: 0.45, 0.79), those married and living with a spouse (β = 0.54/year, 95% CI: 0.33, 0.75), and those with higher BMI (β = 0.45/year, 95% CI: 0.21, 0.69), respectively.

Fig. 3.

Post-diabetes increase in the rate of change in FI, according to subgroups defined by covariates. β and 95% CI were estimated after adjusting for calendar time, and time varying information in age, sex, education years, urban/rural residence, marital status, tobacco smoking, alcohol drinking, physical activity and body mass index. To calculate p for interaction, likelihood ratio tests were conducted by comparing the maximum likelihood χ² value from a nested model with and without the multiplicative interaction term of years after diabetes variable and each subgroup variable

Sensitivity analyses

In sensitivity analyses, when missing covariates were included using imputation, the effect estimates were slightly attenuated relative to those in the primary analysis (Table S3). Furthermore, when the analysis was restricted to participants with complete data across all four waves, the results remained smilar with the main findings (Table S4).

Discussion

In this national cohort of older adults initially free of diabetes, the rate of change in frailty increased significantly following a new diabetes diagnosis, with an additional increase of 0.52/year compared to the pre-diabetes period. This acceleration was observed in the domains of basic activities of daily living, instrumental activities of daily living, functional limitations, health status, and chronic diseases. Subgroup analyses further revealed that the association between diabetes onset and rate of change in frailty was stronger among older adults (≥ 80 years), those who were single, and individuals with lower BMI.

The relationship between diabetes and an increased risk of incident frailty has been well-documented in previous epidemiological studies [9, 17, 18]. However, relatively few investigations have examined the frailty trajectory before and after new-onset diabetes. To our knowledge, only one study has compared the trajectory of daily functioning before and after diabetes onset [19]. Using data from the English Longitudinal Study of Ageing (ELSA), Ji and colleagues reported that post-diabetes changes in daily functioning (0.093 points/year; 95% CI: 0.056–0.131) accelerated compared with the pre-diagnosis period. Our findings align with this study and extend the evidence to overall frailty. Furthermore, a Mendelian randomization study of more than 0.45 million individuals found a bidirectional causal relationship between frailty and type 2 diabetes, indicating a complex, mutually reinforcing interaction [20]. Nevertheless, our study makes a unique contribution by tracking the dynamic frailty trajectory both before and after diabetes diagnosis within the same cohort.

Our subgroup analyses revealed several notable effect modifiers. The age-dependent effect is particularly noteworthy. Participants aged 80 years and older experienced a markedly higher rate of change in frailty, withapproximately 1.6 times higher than that of their younger counterparts (β = 1.01/year vs. β = 0.62/year). This marked disparity may reflect the diminished physiological reserve associated with advanced age, rendering older adults more susceptible to metabolic stresses arising from diabetes [5, 21]. Notably, the current study is among the few investigations to focus on such an elderly population (mean age: 82.12), offering a unique perspective on the detrimental impact of diabetes on frailty in the oldest-old. The stronger association observed in unmarried (including divorced or widowed) individuals may reflect the important role of social support and marital status in managing chronic conditions [22]. Lack of spousal support could exacerbate the metabolic and functional challenges posed by diabetes, thereby accelerating frailty progression. Furthermore, unmarried older adults may experience poorer access to coordinated care or lower adherence to complex diabetes management regimens, compounding the detrimental effects on physical function [23].​ Finally, our finding that a BMI < 22 kg/m² was associated with increased rate of change in frailty aligns with emerging evidence highlighting the adverse outcomes linked to underweight status among older adults with diabetes [24].

Several mechanisms may account for the accelerated frailty progression observed following a diabetes diagnosis [9, 25, 26]. First, diabetes appears to exacerbate the age-related loss of muscle mass and function through direct effects on skeletal muscle metabolism and structure [27]. Second, insulin resistance may play a critical role; indeed, muscle biopsies of older individuals with metabolic syndrome have revealed alterations in muscle fiber type content and increased intramyocellular lipid accumulation [28]. In this context, non-diabetic insulin resistance worsens the anabolic response in older adults, and this situation is further exacerbated when glucose metabolism is impaired in diabetes. Our finding of a significantly steeper increase in instrumental activities of daily livingafter diabetes onset aligns with these observations. Beyond muscular effects, diabetes negatively impacts vascular function, aggravating the age-related decline in vascular health. Such dysfunction compromises tissue perfusion, particularly in skeletal muscles, thereby contributing to poorer physical performance [29]. Finally, the psychological impact of a diabetes diagnosis, including potential changes in self-care behaviors, medication burden, and healthcare utilization, could indirectly influence frailty progression [30]. However, our finding that the psychological well-being component did not differ significantly between groups suggested that the accelerated frailty trajectory among Chinese older adults may be primarily driven by physiological rather than psychological factors.

This study has several strengths. We utilized a large, national cohort with longitudinal follow-up and examined frailty trajectories over time in relation to diabetes onset, which was limited reported before. Nonetheless, several limitations should be noted. First, diabetes was assessed via self-report, which may have led to misclassification. Given the relatively low diabetes awareness rate in China and low diabetes incident rate in the current study, the association between incident diabetes and frailty progression could be underestimated. Second, although the CLHLS is a nationally representative cohort at inception, the substantial exclusions applied in our analysis(approximately 83% of the baseline sample)have resulted in a study population that systematically differs from the original cohort. This raises the possibility of selection bias, which could limit the generalizability of our findings, particularly if the factors leading to exclusion are associated with diabetes incidence or frailty progression. Third, although we adjusted for multiple potential confounders, residual confounding is possible, particularly from unmeasured factors such as diet quality, C-reactive protein, cholesterol levels, healthcare access, or genetic predispositions. Fourth, we were unable to differentiate between type 1 and type 2 diabetes; however, most incident cases in this older population likely represent type 2 diabetes. Lastly, because the CLHLS primarily involved Chinese older adults, the generalizability of our findings to other racial or ethnic groups is limited.

Conclusions

In conclusion, this study demonstrated that incident diabetes was associated with accelerated progression of frailty, particularly in the domains of basic activities of daily living, instrumental activities of daily living, functional limitations, health status, and chronic diseases. Future research should investigate how modifying risk factors in later life may influence frailty trajectories in both diabetic and non-diabetic older adults.

Abbreviations

 CLHLS

Chinese Longitudinal Healthy Longevity Survey

 FI

Frailty Index

 BMI

Body Mass Index

 SD

Standard Deviation

 IQR

Interquartile Range

 ELSA

English Longitudinal Study of Ageing

Authors’ contributions

Ruomei Yang, Fang Sun, Hongbo He and Zhiming Zhu designed the study. Ruomei Yang, Fang Sun, Zongming Yang, Zhencheng Yan and Zhigang Zhao were responsible for the analysis and interpretation of data. All the authors were involved in the writing, revision and final approval of the manuscript, and agree to be accountable for all aspects of the work.

Funding

This work was supported by the National Natural Science Foundation of China (grant number: 82495173) and Noncommunicable Chronic Diseases-National Science and Technology Major Project (No. 2023ZD0508300).

Data availability

The data that support the findings of this study are available from the authors but restrictions apply to the availability of these data. Data are, however, available from the authors upon reasonable request and with permission from the PKU Center for Healthy Aging and Development at https://chads.nsd.pku.edu.cn/index.htm.

Declarations

Ethics approval and consent to participate

The CLHLS study was approved by the Research Ethics Committee of Peking University (IRB00001052–13074). All participants or their guardians provided written informed consent. The study adhered to the principles of the Declaration of Helsinki. All research procedures were conducted following the established guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.