**3. Results**

#### *3.1. Dietary Diversity Score and Baseline Characteristics*

The average DDS among participants was 3.78 ± 1.04 (ranged from 1.33 to 8.00). Table 1 presented the baseline characteristics of participants according to DDS tertiles. Compared with individuals in the lowest tertile of DDS, those scoring higher in DDS were older when they entered the survey, had a higher proportion of men, had higher BMI, were more likely to live in southern China and urban areas, had higher education level and income, were less likely to be a smoker, and had lower physical activity. The distribution of DDS was similar in regular alcohol drinkers and the others.


**Table 1.** Baseline characteristics of participants according to dietary diversity score (DDS) tertiles.

MET, metabolic equivalent of task. Continuous variables were presented as means and standard deviations, and categorical variables were presented as percentages. Continuous variables were tested between groups by one-way analysis of variance, and categorical variables were tested by chi-square test. a DDSs were grouped into tertiles from low to high (T1, T2, T3). b Per capita household incomes were grouped into tertiles at each phase and was labeled as low, middle, and high, respectively.

#### *3.2. Dietary Diversity Score and Nutrient Intakes*

DDS was positively associated with intakes of protein, fat, dietary fiber, vitamin A, riboflavin, niacin, vitamin C, vitamin E, calcium, phosphorus, potassium, magnesium, iron, zinc, and selenium. Meanwhile, higher DDS was inversely associated with intakes of carbohydrate, sodium, and manganese (Table 2).

> **Table 2.** Nutrient intakes across dietary diversity score (DDS) tertiles.



**Table 2.** *Cont*.

Values are medians and quartiles. Tests for linear trend across DDS tertiles were conducted by assigning the midpoint values of DDS and treating the variables as continuous in a linear regression model, prior to that, the values of nutrient intakes were transformed to log to reach normality. All nutrients were associated with DDS tertiles with *p*-trend < 0.001, except for dietary fiber (*p*-trend = 0.041), thiamin (*p*-trend = 0.260), magnesium (*p*-trend = 0.033), and copper (*p*-trend = 0.138). a DDSs were grouped into tertiles from low to high (T1, T2, T3).

#### *3.3. Dietary Diversity Score and Disability in Activities of Daily Living*

During a median follow-up of 9 years (ranged from 2 to 18; total person years: 52,297), 601 ADL was reported. Absolute ADL rates according to DDS tertiles (from low to high: T1, T2, T3) were 14.4, 10.5, and 8.7 per 1000 person years, respectively. After the adjustment of covariates, higher DDS was associated with decreased risk of ADL disability (Table 3).

**Table 3.** Association between dietary diversity score (DDS) and disability in activities of daily living (ADL).


Values were hazard ratios and 95% confidence intervals unless specified. Hazard ratios were estimated by Cox proportional regression models. Multivariate models were adjusted for: Model 1: age at entry (continuous), gender (men or women), living region (southern or northern China), residency (urban or rural), income (low, middle, or high), and education level (primary school and below or middle school and higher); Model 2: additionally included smoking status (smoker or not), physical activity (≤100 or >100 metabolic equivalent of task-hours/week), body mass index (continuous), and comorbidities (no or yes). Tests for trend were performed by assigning the midpoints of each DDS tertiles and treating the value as continuous in a separate regression model. a DDSs were grouped into tertiles from low to high (T1, T2, T3).

#### *3.4. Sensitivity Analyses*

In the sensitivity analysis, the association of DDS with ADL disability did not change by excluding individuals whose follow-up time was less than 5 years, or additional adjustment of alcohol consumption and phase at entry (Table S1).

#### *3.5. Subgroup Analyses*

The significant association between DDS and ADL disability was observed in all subgroups based on baseline characteristics, including gender, living region, age at entry, and comorbidity. However, the association was more pronounced in participants living in southern China, participants aged over 65 years at entry, and participants without comorbidities than the others. A significant interaction term between DDS and comorbidity was observed (Figure 2).


**Figure 2.** Subgroup analysis of association of continuous dietary diversity score (DDS) with disability in activities of daily living (ADL). Hazard ratios were estimated by Cox proportional regression models. Multivariate models were adjusted for age at entry (continuous), gender (men or women), living region (southern or northern China), residency (urban or rural), income (low, middle, or high), education level (primary school and below or middle school and higher), smoking status (smoker or not), physical activity (≤100 or >100 metabolic equivalent of task-hours/week), body mass index (continuous), and comorbidities (no or yes). Analyses within subgroups were adjusted for all other covariates.

#### *3.6. Dietary Exposure Measures*

We found both the average DDS across phases and the baseline DDS were associated with lower odds of ADL disability. The association was more pronounced for the average DDS. The association between the recent DDS prior to the end of the survey and ADL disability was insignificant (Figure 3).

**Figure 3.** Dietary diversity score (DDS) at different surveys and disability in activities of daily living (ADL). Participants involved in three or more dietary surveys were included (*n* = 2756). Average DDS was the cumulative mean DDS from baseline to the phase prior to the end of the survey (report of disability in activities of daily living, loss to follow-up, the phase of 2015, whichever occurred first). Baseline DDS was obtained from the phase at entry. Recent DDS was obtained from the phase before the end of the survey. All DDSs were continuous. Odds ratios were estimated by logistic regression models. Models were adjusted for age at entry (continuous), gender (men or women), living region (southern or northern China), residency (urban or rural), income (low, middle, or high), education level (primary school and below or middle school and higher), smoking status (smoker or not), physical activity (≤100 or >100 metabolic equivalent of task-hours/week), body mass index (continuous), and comorbidities (no or yes).

## **4. Discussion**

This study found that higher DDS was inversely associated with the risk of ADL disability in Chinese adults. The association was stable after the adjustment of physical activity, BMI, and comorbidity. To our best knowledge, the present study is the first one revealing the beneficial effect of dietary diversity on ADL disability.

In the western population, the Mediterranean diet is most frequently investigated when addressing the effect of dietary patterns on functional capacity. A meta-analysis showed that higher adherence to the Mediterranean diet is associated with decreased risk of frailty and functional disability in the elderly [25]. Besides, adherence to the healthy eating index may be also associated with better physical performance among elderly people [26]. In the eastern population, prospective studies found the Japanese diet was inversely associated with functional disability in Japanese individuals aged 65 years and above [27,28]. In contrast, evidence on DDS and ADL or functional capacity is relatively limited. We only observed one study that reported an insignificant relationship between DDS and higher-level functional capacity [29]. Our study found an inverse association between DDS and the risk of ADL disability based on a population-based cohort study. We believe this work will make contributions to literature on healthy aging. Besides, compared with the approach of dietary pattern assessment, the DDS approach is easier to compute, more suitable for comparison across different populations, and more appropriate to use for guiding people to follow a healthy diet. This work will contribute to the prevention of ADL disability.

Our study observed that each additional point on DDS was associated with a nearly thirty percent reduction in the risk of ADL disability. The benefits of DDS on ADL might be related to the following mechanisms. First, aging is associated with a loss of muscle mass, leading to frailty, sarcopenia, and functional disability [30]. More dietary protein is needed for the maintenance of good muscle function in the elderly [30,31]. Epidemiological studies showed that higher dietary protein may slow down the process of muscle mass loss [32–34]. Our study found individuals with higher DDS had higher intakes of protein, which may be a benefit for ADL independence. Second, it has been widely recognized that inflammation and oxidative stress played important roles in the process of aging [35,36]. We observed a positive trend between DDS and intakes of antioxidants (e.g., vitamin E, vitamin C, selenium) among the study population. Higher DDS may promote healthier aging by countering inflammation and oxidative stress. Third, aging is also caused by the loss of bone mass [37], which may increase the risk of frailty and fracture and eventually accelerate the loss of ADL independence [38–40]. In this study, we observed that participants with higher DDS enjoyed high intakes of protein, calcium, phosphorus, potassium, which were beneficial to bone health [41]. Fourth, a diverse diet has a positive effect on gu<sup>t</sup> microbiota [42]. Healthy gu<sup>t</sup> microbiota may promote the absorption of micronutrients and modulate individual response to dietary protein [31,43].

An interaction between DDS and comorbidity on the risk of ADL disability was observed in this study. The association of DDS with ADL disability was more pronounced among participants without comorbidity at baseline than those with comorbidity. We assumed that this might be because, among participants with comorbidity, the progression of disease played a more important role in the loss of ADL independence than the impact of diet. However, we should note that, even in individuals with comorbidity, the beneficial effect of DDS still existed.

In this study, we calculated the cumulative average DDS to address participants' long-term dietary exposure. To compare the difference between different dietary exposure measures, we estimated the OR for the average DDS across phases, the baseline DDS, and the recent DDS prior to the end of the survey in relation to ADL disability. The results indicate that when estimating the effect of DDS on ADL disability, the average DDS is the most pronounced of the three approaches, which is consistent with previous studies [44,45]. The result indicates that, in estimating the effect of dietary factors on ADL, long-term dietary exposure is more important than the recent exposure. Findings also implicated that the measure that could address the dynamic dietary exposure is more preferable to the single baseline data.

The strengths of this study include prospective design and population-based samples, which provided advantages for causal inference. Repeated dietary surveys allowed us to capture participants' long-term dietary exposure. There are several limitations. First, although comorbidities were considered in our analysis, other factors that may influence ADL (e.g., dementia, accident) were not included because of a lack of data. Second, the present study was based on a dynamic cohort study, participants joined the survey at a wide age range; however, in the subgroup and sensitivity analyses, we took participants' age at entry and phase at entry into consideration, which could partly mark up for this defect. Third, the ADL disability was self-reported. Participants might overestimate or underestimate their abilities to finish some tasks because of social desirability or misunderstanding.
