Abstract

Background:

The global agricultural labor force has declined from 41% in 1995 to 26% in 2023, while farming populations in Asian newly industrialized countries are aging rapidly. China, as a representative case, faces challenges in agricultural labor force renewal, with older farmers serving as “last guardians” through high-intensity labor. The prevalence of frailty among rural Chinese adults aged 60 years and older is 23.31%, significantly higher than in urban areas. However, previous studies have treated physical activity as homogeneous, failing to distinguish occupational agricultural labor from leisure-time activity. This study examined the association between agricultural labor intensity and frailty among older farmers, identifies risk thresholds, and provides evidence for promoting healthy aging during agricultural transformation.

Methods:

We conducted a multiregion cross-sectional survey in Guizhou Province, China. A total of 1,443 older farmers aged 60–79 years were selected using multifactor stratified sampling. Frailty was assessed using the Fried Frailty Phenotype. Labor intensity was quantified as total daily energy expenditure (kcal/day), calculated as body weight (kg) × daily working hours (h) × metabolic equivalent of task (MET). Multivariable logistic regression analyses were performed with stepwise adjustment for demographic, lifestyle, and health-related confounders.

Results:

The prevalence of frailty among older farmers was 19.7%, with significant regional variation. Advanced age, low educational attainment, malnutrition, polypharmacy, widowhood, and multiple pain sites were independently associated with frailty. A non-linear association was observed between agricultural labor intensity and frailty. Daily labor energy expenditure exceeding 1,752 kcal/day was associated with significantly increased frailty risk (OR = 3.596; 95% CI: 2.386–5.420; P < 0.001). Unlike leisure-time physical activity, high-intensity agricultural labor was independently associated with frailty.

Conclusions:

High labor intensity represents a strong, independent risk factor for frailty among older farmers. These findings underscore the urgent need for targeted interventions and occupational health guidelines to protect this vulnerable population from excessive physical workloads and promote healthy aging in rural communities.

1 Background

Data from the International Labour Organization (ILO) demonstrate that the proportion of the global agricultural labor force declined substantially from 41% in 1995 to 26% in 2023 (1). This rapid transition from agriculture is reshaping population age structures and altering the relationship between demographic change and agricultural production. Newly industrialized countries (NICs) in Asia, experiencing rapid urbanization and expansion of nonagricultural employment, are witnessing pronounced aging of their farming populations (2–4).

As a representative NIC (5), China faces significant challenges in agricultural labor force generational renewal (6). Official data indicate that by 2020, the proportion of farmers younger than 40 years had decreased to 19% and continues to decline (7). Model projections suggest that by 2030, the number of farmers aged 60 years and older will reach 123 million, including approximately 58.57 million aged 65 years or older (8). Notably, despite institutional shifts toward larger-scale farming, Chinese agriculture remains fundamentally characterized by smallholder production. Older farmers continue to identify strongly with farming and serve as the “last guardians” of the agricultural labor force through a strategy characterized by high labor intensity, low willingness to exit farming, and engagement in multiple occupations (8). Under population aging, labor-intensive agriculture faces a fundamental dilemma: a mismatch between labor supply and industry demands, further compounded by mechanization bottlenecks (9).

This phenomenon has triggered cascading health challenges. The prevalence of frailty among rural Chinese adults aged 60 years and older is 23.31% (10), significantly higher than rates reported in major urban areas such as Beijing and Shanghai (11, 12). Because older farmers are exposed to long-term heavy physical workloads and often lack occupational health protection (13), agriculture faces a dual burden: increasing average farmer age and progressive decline in the health of the aging agricultural workforce (14).

However, previous studies have largely treated physical activity as a homogeneous exposure, without distinguishing leisure-time physical activity from occupational agricultural labor. Additionally, few studies have quantified a specific safety threshold for labor intensity in older farmers or explored potential non-linear associations between labor intensity and frailty. Clarifying the effect of labor intensity on frailty and the social mechanisms that may modify this relationship is essential for promoting healthy aging and addressing the key scientific challenge of balancing productivity and sustainability during agricultural transformation. To address these gaps, we proposed three hypotheses:

Hypothesis 1: frailty prevalence differs across regions and socioeconomic levels among older farmers. Hypothesis 2: multiple individual characteristics, including advanced age, low education, malnutrition, polypharmacy, widowhood, and multiple pain sites, are independently associated with frailty. Hypothesis 3: a non-linear association exists between agricultural labor intensity and frailty risk, with higher labor intensity associated with increased frailty risk.

To test these hypotheses, we conducted a cross-sectional study among older farmers. By distinguishing occupational agricultural labor from general physical activity and providing the first quantitative safety threshold for labor intensity, this study seeks to fill these gaps and offer actionable evidence for targeted interventions.

2 Methods2.1 Study design

This cross-sectional study was conducted in Guizhou Province, southwestern China, where more than 92% of the land is mountainous and the rural agricultural labor force constitutes a substantial population proportion. A cross-sectional design was selected to efficiently examine current frailty prevalence and its association with agricultural labor intensity among older farmers.

2.2 Study subjects and sampling methods

The sample size was calculated using the formula: n = ,

According to previous literature, the prevalence of frailty among rural older farmers was reported to be 36.76%. The minimum required sample size was 993 participants. After accounting for an anticipated 20% invalid response and loss to follow-up rate, the target sample size was set at 1,192. A total of 1,443 eligible participants were finally enrolled.

This study used a multifactor stratified sampling design in Guizhou Province. Counties were stratified by geographic region (eastern, southern, western, northern) and relative economic level with reference to the World Bank's regional economic classification framework. Based on the 2023 provincial average GDP (54,200 yuan) as the benchmark, counties were classified into four tiers: developed (≥150%), relatively developed (100%−150%), moderately developed (50%−100%), and less developed (< 50%). Four representative cities were selected: Guiyang (south, developed), Zunyi (north, relatively developed), Tongren (east, moderately developed), and Bijie (west, less developed). This strategy accounted for geographic and socioeconomic heterogeneity and reduced sampling bias.

In China, the concept of “retirement” is ambiguous, and many older farmers remain engaged in agricultural work beyond the official retirement age. According to gerontological standards, individuals aged ≥60 years are defined as older adults. However, those aged ≥80 years (the oldest-old) typically experience severe physical decline and high care dependency, which may confound the association between agricultural labor and frailty (15, 16). We therefore restricted participants to 60–79 years to ensure they were physically capable of sustained farming while minimizing heterogeneity in physical function.

Inclusion criteria were: (1) aged 60–79 years; (2) clear consciousness, good comprehension and communication, and independent mobility to complete the investigation; (3) long-term engagement in agricultural work with no history of other occupations; (4) acknowledged and agreed to participate in this survey.

Exclusion criteria were: (1) severe cardiac, cerebrovascular, renal, pulmonary or other major organ diseases; (2) severe mental disorders; (3) inability to complete the questionnaire or dropout during investigation.

Data were collected from March 1 to May 31, 2025, with a total of 1,443 valid participants included. The participant flowchart is shown in Figure 1.

Flowchart of participant enrollment and selection.

2.3 Questionnaires2.3.1 Assessment of frailty

Frailty was assessed using the Fried Frailty Phenotype (FP) (17). The phenotype includes five criteria: slow walking speed, reduced grip strength, low physical activity, fatigue, and unintentional weight loss. Participants meeting three or more criteria were classified as frail, those meeting one or two criteria as prefrail, and those meeting none as non-frail. The FP evaluates physical frailty using objective and quantitative indicators and has strong predictive validity. The Cronbach's α of the scale in this study was 0.705.

2.3.2 Labor intensity assessment and calculation

Labor intensity reflects actual physical workload and is commonly quantified by energy expenditure (kcal) or physiological load. In this study, daily labor intensity was quantified using the following energy expenditure formula derived from the Compendium of Physical Activities: Energy expenditure (kcal/day) = body weight (kg) × daily working hours (h) × metabolic equivalent of task (MET) (18). MET values were obtained from the Older Adult Compendium of Physical Activities, which provides energy costs specifically for adults aged 60 years and older (19). Based on this database, fixed MET values were assigned to common local agricultural activities: light activities (e.g., planting, potting, transplanting, and watering) were assigned 3.3 METs; moderate activities (e.g., digging, tilling, weeding, and plowing) were assigned 4.8 METs; and vigorous activities (e.g., heavy digging, composting, and carrying heavy loads) were assigned 7.3 METs (19).

During the field survey, a structured questionnaire collected detailed information on agricultural work types performed during the previous 6 months (each activity session lasting at least 30 min) and average daily working hours. For participants engaged in multiple activities of different intensities, the dominant intensity type was defined as the activity accounting for the largest proportion of working time. The corresponding MET value for that activity was then entered into the formula to calculate a continuous variable representing daily labor energy expenditure for subsequent analyses.

For example, consider a participant weighing 60 kg who worked an average of 5 h per day, including 1 h of planting/transplanting (3.3 METs), 3 h of weeding/tilling (4.8 METs), and 1 h of carrying heavy loads or vigorous digging (7.3 METs). According to our method, weeding/tilling (4.8 METs) was identified as the dominant activity because it accounted for the largest share of working time. The daily labor energy expenditure was calculated as: 60 kg × 5 h × 4.8 MET = 1,440 kcal/day. Thus, the estimated daily labor energy expenditure for this participant was 1,440 kcal/day.

2.3.3 Geriatric depression scale (GDS-5)

The GDS-5, derived from the GDS-15, consists of the five items with the strongest predictive value for clinical depression (20). Responses are recorded as binary options (“yes” or “no”). Item 1 is reverse scored (yes = 0, no = 1), whereas items 2–5 are scored in the standard direction (yes = 1, no = 0). Total scores range from 0 to 5, with scores ≥2 indicating depressive symptoms; higher scores indicate greater symptom severity. The GDS-5 demonstrated a sensitivity of 0.94 and a specificity of 0.81 upon validation.

2.3.4 Mini nutritional assessment-short form (MNA-SF)

The MNA-SF is widely used for early malnutrition screening due to its high sensitivity, specificity, and ease of administration (21). It consists of six items, including weight loss, acute stress, mobility, neuropsychological problems, appetite/digestion, and BMI, with a total score of 14. Scores of 0–7 indicate malnutrition, 8–11 indicate risk of malnutrition, and 12–14 indicate normal nutritional status. The instrument has demonstrated a sensitivity of 96%, specificity of 98%, and predictive value of 97%. The Cronbach's α of the scale in this study was 0.711.

2.3.5 Demographic characteristics

Demographic and health-related characteristics included sex, age, marital status, living arrangement, educational level, body mass index (BMI), sleep duration, smoking status, alcohol consumption, number of chronic diseases, number of medications, number of pain sites, nutritional status, and depressive status.

2.4 Data collection

Data collection was conducted from March 1 to May 31, 2025. Investigators conducted one-on-one face-to-face interviews and entered responses directly into the Wenjuanxing online platform. All scales were used in their original validated form. Before formal investigation, all investigators received standardized training, and a pilot study was performed to ensure applicability. During interviews, standardized procedures were followed, with family members providing assistance only when necessary. Each participant received a small gift in appreciation of their participation. A total of 1,472 questionnaires were collected, of which 29 were excluded due to incomplete or inconsistent information, resulting in 1,443 valid responses and an effective response rate of 98%.

2.5 Quality control and survey implementation

Data collection was conducted across four cities in Guizhou Province—Guiyang, Zunyi, Tongren, and Bijie—with the involvement of 13 well-trained investigators, including eight postgraduate nursing students, one chief nurse specialist, two associate chief nurses, and two nursing supervisors. To ensure data quality, rigorous quality control procedures were implemented throughout the study. Validated instruments with established reliability and validity were selected, all investigators received standardized training, and the protocol was refined through a pilot survey. During fieldwork, one-on-one interviews and on-site verification were used. Potential participants were recruited through village committee announcements and household visits, and their status as farmers was confirmed through self-reported long-term agricultural engagement and verification by local village cadres. Two researchers independently entered the data, blinded to each other. Discrepancies were resolved by checking the original questionnaires. Logical verification was also performed. The participant flowchart (Figure 1) summarizes enrollment, screening, and final inclusion.

2.6 Statistical analysis

Participant characteristics are presented as frequencies and percentages for categorical variables, mean ± standard deviation (SD) for normally distributed continuous variables, and median [interquartile range (IQR)] for non-normally distributed continuous variables. Group differences in categorical variables were assessed using the chi-square test. Independent-samples t tests were used for normally distributed continuous variables, and the Mann–Whitney U test was used for non-normally distributed continuous variables.

Univariate analyses and multivariable binary logistic regression were performed to identify factors associated with frailty and evaluate the association between labor intensity and frailty risk. Three hierarchical models were constructed: Model 1 was unadjusted; Model 2 was adjusted for sociodemographic characteristics; and Model 3 was further adjusted for sociodemographic factors, lifestyle variables, and comorbidities. Restricted cubic spline (RCS) analysis was performed to explore potential non-linear relationships between continuous labor intensity and frailty risk. All statistical analyses were conducted using SPSS version 29.0 (IBM Corp., Armonk, NY, USA). A two-sided P value < 0.05 was considered statistically significant.

3 Results3.1 Socio-demographic characteristics and univariate analysis of frailty prevalence

Among 1,443 older farmers, 19.7% were identified as frail, with notable regional variation across the four study sites. As shown in Table 1, frail participants differed significantly from non-frail participants across multiple sociodemographic and health-related characteristics (P < 0.05).

VariableCategoryNon-frailty (n = 351)Pre-frailty (n = 808)Frailty (n = 284)t/χ2/ZPSexMale170 (48.4)373 (46.2)128 (45.1)0.7970.671Female181 (51.6)435 (53.8)156 (54.9)Age(years)–64 (62, 70)65 (62, 71)69 (62, 74)31.218< 0.001Educational levelIlliterate164 (46.7)455 (56.3)193 (68.0)43.183< 0.001Primary school graduate148 (42.2)298 (36.9)89 (31.3)Junior high school and above39 (11.1)55 (6.8)2 (0.7)Marital statusMarried314 (89.5)711 (88.0)224 (78.9)31.824< 0.001Divorced27 (7.7)82 (10.1)37 (13.0)Widowed10 (2.8)15 (1.9)23 (8.1)Living arrangementLiving with spouse and children109 (31.1)245 (30.3)93 (32.7)7.2680.297Living with spouse only149 (42.5)360 (44.6)119 (41.9)Living with children only32 (9.1)69 (8.5)36 (12.7)Living alone61 (17.4)134 (16.6)36 (12.7)Monthly household income (RMB)< 1,00075 (21.4)123 (15.2)68 (23.9)31.468< 0.0011,000–1,999221 (63.0)603 (74.6)190 (66.9)2,000–2,99928 (8.0)27 (3.3)7 (2.5)≥3,00027 (7.7)55 (6.8)19 (6.7)BMI–24.5 (22.2, 26.5)24.2 (21.9, 26.6)23.8 (21.2, 26.2)6.6520.032Smoking statusCurrent smoker296 (84.3)639 (79.1)234 (82.4)7.1480.128Current quitter23 (6.6)61 (8.3)26 (9.2)Never smoker32 (9.1)102 (12.6)24 (8.5)Alcohol consumption statusCurrent drinker234 (66.7)459 (56.8)186 (65.5)21.090< 0.001Current abstainer10 (2.8)64 (6.9)23 (8.1)Lifetime abstainer107 (30.5)285 (35.3)75 (26.4)Number of comorbidities–1 (0, 1)1 (0, 1)1 (0, 1)10.8260.004Number of regular medicationsNone207 (59.0)413 (51.1)101 (35.6)38.139< 0.001Less than 3 types123 (35.0)316 (39.1)143 (50.4)3 or more types21 (6.0)79 (9.8)40 (14.1)Number of body pain sites–0 (0, 1)1 (0, 1)1 (0, 2)22.153< 0.001Depressive symptomsNo318 (90.6)713 (88.2)230 (81.0)14.378< 0.001Yes33 (9.4)95 (11.8)54 (19.0)Nutritional statusMalnutrition87 (24.8)231 (28.6)135 (47.5)44.422< 0.001healthy264 (75.2)577 (71.4)149 (52.5)Sleep duration< 7 h203 (57.8)450 (55.7)196 (69.0)22.416< 0.0017–9 h122 (34.8)324 (40.1)78 (27.5)≥9h26 (7.4)34 (4.2)10 (3.5)Labor intensity (Kcal)1,267 (613, 1964)1,198 (594, 2112)1,387 (789, 2,302)12.2030.002

Socio-demographic characteristics and univariate analysis associated with frailty prevalence.

Due to low cell counts (n < 10), the original “malnourished” category was combined with the “at risk of malnutrition” category into a single group labeled “Malnourished/at risk.”

3.2 Robustness analysis of the association between labor intensity and frailty

In fully adjusted multivariable logistic regression, labor intensity remained significantly associated with frailty both as a continuous variable (OR = 1.000, 95% CI: 1.000–1.001, P < 0.001) and as a categorical variable. Using the lowest intensity tertile (Q1) as reference, the highest tertile (Q3) was associated with substantially greater odds of frailty (OR = 3.596, 95% CI: 2.386–5.420, P < 0.001), supporting a clear dose–response relationship. Results were consistent across sequentially adjusted models (Table 2).

VariablesModel 1Model 2Model 3OR (95% CI)POR (95% CI)POR (95% CI)PLabor intensity1.188 (1.049–1.345)0.0061.440 (1.247–1.663)< 0.0011.000 (1.000–1.001)< 0.001Labor intensity grouping (with Q1 as the reference group)Q21.508 (1.083–2.100)0.0151.834 (1.286–2.615)< 0.0012.122 (1.456–3.093)< 0.001Q31.646(1.186–2.285)0.0032.473 (1.707–3.583)< 0.0013.596 (2.386–5.420)< 0.001

Association between labor intensity and frailty across different models.

Model 1: Crude. Model 2: Adjust: Sex, age, Education status, Marital status, Living arrangement, Monthly household income (RMB). Model 3: Adjust: Gender, age, Education status, Marital status, Living arrangement, Monthly household income (RMB), BMI, Smoking status, Alcohol status, Number of regular medications, Depressive symptoms, Nutritional status, Number of comorbidities, Number of body pain sites, Sleep duration.

3.3 Non-linear association between labor intensity and frailty: restricted cubic spline (RCS) analysis

Due to violation of the proportional odds assumption, frailty status was dichotomized as frail versus non-frail (including prefrail) for all subsequent modeling. Restricted cubic spline analysis revealed a significant J-shaped non-linear association between labor intensity and frailty (P for overall association < 0.001; P for non-linearity = 0.001). Frailty risk declined at low-to-moderate labor intensity but increased markedly at higher levels, as illustrated in Figure 2.

Non-linear association between labor intensity and frailty risk in older farmers.

3.4 Multivariable logistic regression analysis of factors associated with frailty in older farmers

After full adjustment, higher labor intensity remained independently associated with increased odds of frailty in both continuous and categorical models. Other significant independent risk factors included older age, widowhood or divorce, polypharmacy, malnutrition, depressive symptoms, and multiple pain sites. Model fit was satisfactory (Hosmer–Lemeshow test: χ2 = 12.321, df = 8, P = 0.137) and no severe multicollinearity was detected (all VIF < 5) (Figures 3, 4.).

Association between categorized labor intensity and frailty in older farmers.

Association between continuous labor intensity and frailty in older farmers.

3.5 Effect modification and subgroup analyses

No significant interaction between age and labor intensity was observed (P for interaction = 0.407), indicating a stable association across age groups. In stratified analyses, the positive association between high labor intensity and frailty persisted in all subgroups. The association was stronger among women (OR = 5.23, 95% CI: 2.89–9.48) than men (OR = 2.70, 95% CI: 1.49–4.91), and varied by region, with the strongest effect observed in Zunyi (OR = 7.10, 95% CI: 2.38–21.16). Full subgroup estimates are presented in Table 3.

SubgroupLabor intensityOR (95% CI)PBy sexMaleQ22.076 (1.135–3.796)0.018MaleQ32.703 (1.490–4.906)0.001FemaleQ22.357 (1.422–3.906)0.001FemaleQ35.230 (2.885–9.481)< 0.001By nutritional statusWell-nourishedQ21.868(1.089–3.205)0.023Well-nourishedQ33.880 (2.247–6.699)< 0.001Malnourished/at riskQ22.160 (1.256–3.714)0.005Malnourished/at riskQ32.864 (1.492–5.500)0.002By regionZunyiQ23.510 (1.196–10.300)0.022ZunyiQ37.101 (2.382–21.164)< 0.001GuiyangQ23.083 (1.245–7.635)0.015GuiyangQ33.367 (1.230–9.219)0.018TongrenQ21.367 (0.745–2.509)0.312TongrenQ33.135 (1.599–6.147)0.001BijieQ22.395 (