Enhancing a plant-based diet can reduce the risk of multimorbidity in older adults

Abstract

Background

Multimorbidity has become widespread among older adults, and it is influenced by diet. However, research has yet to systematically explore the association of the overall plant-based diet index (PDI), healthful plant-based diet index (hPDI), and the unhealthful plant-based diet index (uPDI) on multimorbidity among older adults in China.

Methods

In this research, we acquired data from 2018 cross-sectional survey of China Longitudinal Health and Longevity Survey (CLHLS). We use binary logistic regression to explore the association between PDI, hPDI, uPDI, and multimorbidity. Besides, restrictive cubic splines (RCS) were employed to explore whether the relationship between them is non-linear, and the effect of the interaction between different PDIs and some variables on multimorbidity was analyzed by subgroups. By imputing the missing values and excluding those with diabetes, we conducted two sensitivity analysis to prove the reliability of our research conclusions.

Results

This analysis encompassed 7,551 participants in total. After correcting for some confounding factors, we discovered that PDI (OR = 0.987, 95% CI: 0.978–0.997, P < 0.05) and hPDI (OR = 0.980, 95% CI: 0.970–0.991, P < 0.001) associated with multimorbidity negatively. In contrast, uPDI (OR = 1.016, 95% CI: 1.007–1.025, P < 0.001) associated with multimorbidity positively. RCS revealed a significant nonlinear relationship between PDI and multimorbidity, while there exists significant linear relationship between hPDI, uPDI, and multimorbidity. The results of interaction analysis indicate a significant interaction between place of residence and physical activity on the above results. We found similar results to the main findings of this study in two different sensitivity analyses.

Conclusion

PDI and hPDI negatively associated with multimorbidity, while uPDI positively associated with multimorbidity. The conclusion of this research provides an inspiring dietary perspective for the future prevention of multimorbidity.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-026-07444-4.

Introduction

With the arrival of aging, the global population of individuals aged 65 and above is escalating rapidly and is anticipated to attain 1.6 billion by 2050 [1]. Extensive research has demonstrated a positive association between age and the prevalence of multimorbidity [2, 3].Multimorbidity is generally defined as the coexistence of at least two chronic diseases within an individual at the same time [4]. Several studies have shown that multimorbidity has an impact on individuals, their families, healthcare systems, and society significantly, especially in environments with scarce resources [5, 6], leading to reduced productivity, limited functioning, poor quality of life and increased mortality [7]. Individuals suffering from multimorbidity are prone to premature mortality and increased hospitalization risks, as well as longer hospital stays compared to individuals with a single chronic disease [8, 9]. Nearly three-quarters of patients aged 65 and above suffer from multimorbidity around world [10]. In the UK and the US, the incidence rate of multimorbidity is at the same level, accounting for about a quarter of the population [11, 12], as well as the elderly individuals in Northeast China [13]. A recent national survey estimated that 69.3% of elderly inpatients in China have multimorbidity [14].

Diet is considered one of the main determinants of health [15]. Many scholars have explored the impact of diet on some common diseases among elderly individuals [16, 17]. For example, scholars have shown that anti-inflammatory diets and protein rich diets can reduce the risk of cognitive impairment in elderly people [16], and plant-based diet (PBD)is associated with frailty in the elderly [18]: the overall plant-based diet index (PDI), the healthful plant-based diet index (hPDI), and frailty are negatively associated, while the unhealthful plant-based diet index (uPDI) is positively associated with frailty. In addition, PBD have gradually become a focus of attention for scholars around the world in recent years. PBD features a high intake of plant-based foods and minimal or no inclusion of animal-source food (come from animal tissues or their metabolites.) [19], which has been proven to prevent some chronic diseases [20]. It is reported that PBD can mitigate the probability of diabetes, cardiovascular diseases [21–23], and metabolic syndrome [24]. Besides, studies have indicated that consuming more plant-based foods and less animal products could be advantageous for health and can mitigate the probability of developing some major chronic illnesses [25].

To take the dietary quality of a PBD into consideration, researchers have further delineated three indices: hPDI, uPDI, and PDI. The hPDI highlights a substantial intake of healthy plant-based foods and minimal intake of unhealthy plant-based foods. On the contrary, uPDI is emphasizing higher intake of unhealthy plant-based foods within the overall PBD framework [26, 27]. PDI evaluates diets with increased plant-based food and decreased animal-source food. Previous studies have found that food in hPDI, such as fresh vegetables and fruits, is correlated with better neurological health, while uPDI harms neurological health [28]. In addition, hPDI has been widely proven to have a positive impact on cardiovascular disease [25, 29], type 2 diabetes [30], hypertension [31] and chronic lung disease [32]. In contrast, an increased intake of uPDI is linked to a higher risk of developing coronary heart disease [33]. Plant-based foods represent a pivotal modifiable lifestyle aspect that effectively safeguards against age-related illnesses and sustains overall well-being [34, 35]. but there is no research systematically exploring the relationship between different PBD and multimorbidity in Chinese older adults population. Therefore, the purpose and significance of our study are to identify the potential association between different PBDs and multimorbidity and to provide better suggestions to promote healthy aging.

Based on the above analysis, the association between different PDIs and multimorbidity still needs to be clarified, so there is an increasing need to explore the relationship between them. Thus, this study will use different PDI patterns (PDI, hPDI, uPDI) and investigate their impacts on multimorbidity among elderly individuals in China.

Methods and materials

Participants and process

In this research, we acquired data from the 2018 waves of China Longitudinal Health and Longevity Survey (CLHLS). CLHLS is one of the most extensive national longitudinal studies to investigate the health status of older adults in China, which randomly recruited 23 Chinese provinces, with a sampling range of approximately 85% of the whole population in China. Therefore, using this database to study older adults in China is highly representative. Besides, this research project was approved by the Biomedical Ethics Review Committee of Peking University in China under the identification code IRB00001052-13074. All participants provided informed consent for the initial and subsequent assessments.

There were 15,874 participants in 2018, and we excluded 5,767 participants with missing values in covariates, then 458 participants with missing values in PDl, hPDI and uPDI were excluded. Finally, 2,056 participants with missing data on the dependent variable and 42 older adults aged < 65 years were excluded. The remaining 7,551 participants were incorporated in the final analysis.

The study’s sample size was determined based on the formula used for calculating sample sizes in cross-sectional research studies [n = (Z²_α/2p q) / δ²] [36], (1) n represents the sample size required for the study; (2) p denotes the prevalence rate of multimorbidity in Chinese older adults; (3) q = (1-p); (4) Zα/2 was set at 1.96, and α was set at 0.05 for a two-sided test; and (5) δ denotes the permissible error, calculated at 0.1p. According to a previous study, the prevalence rate of multimorbidity in Chinese older adults was 25.4% [37], and it was calculated that at least 1,128 participants were needed for this study to reach the required sample size. We ultimately obtained 7,551 valid data, which is sufficient for statistical analysis. The specific data-cleaning process is detailed in Fig. 1.

Fig. 1.

Data-cleaning flow chart

Assessment of PDI, hPDI, and uPDI

The dietary data is sourced from the simplified food frequency questionnaire in the CLHLS database. In this study, we evaluated plant-based dietary patterns by constructing the PDI, an improved method used by Satija et al. [30]. We calculated the modified score of PDI, hPDI, and uPDI based on dividing 16 food groups into three types according to their possible health impacts: ①healthful plant-based foods (fresh fruits, legumes, garlic, whole grains, vegetable oils, fresh vegetables, tea, and nuts); ② unhealthful plant-based foods (pickled vegetables (such as Kimchi from Korea, Sauerkraut from Germany, Zha Cai and Suan Cai from China, these dishes are characterized by high sodium content [38]), refined grains, and sugar: white sugar or candy); ③ animal-source foods (meat, eggs, animal fats, dairy and dairy products, fish and aquatic products). As for overall PDI, we assigned positive numerical values to various plant-based food categories, scaling from 1 for infrequent intake to 5 for the most frequent intake. In contrast, the animal-source food group was assigned inverse scores, where 5 represented the least frequent consumption, whereas 1 denoted the most frequent intake. Regarding hPDI, positive scores are accorded to healthful plant-based foods, while unhealthful plant-based and animal-source foods are allocated inverse scores. Regarding uPDI, reverse scores are accorded to healthful plant-based and animal-source foods, while unhealthful plant-based foods are encoded as positive scores. Finally, the total PDI, hPDI, and uPDI modified scores were calculated by adding up the scores of 16 food groups, ranging from 16 to 80.

Assessment of multimorbidity

In our research, we covered 15 chronic illnesses, including diabetes, hypertension, heart disease, cerebrovascular disease, emphysema/pneumonia/asthma/ bronchitis (respiratory disease), gastrointestinal ulcer, dyslipidemia, malignant tumor, pressure sore, Parkinson’s disease, arthritis, dementia, chronic nephritis, cholecystitis/cholelithiasis (biliary disease) and rheumatoid diseases. Chronic diseases are self-reported based on whether you have this disease. Multimorbidity is conceptualized as the coexistence of two or more chronic illnesses concurrently [39, 40].

Covariates

We controlled many covariates to minimize the impact of possible confounders, including sex (male, female), age ( < = 80 years, > 80years), Body Mass Index (BMI) (< 18.5 kg/m², 18.5 ~ 24.9 kg/m², > 25 kg/m²), marital status (married, other), residence (urban, rural), years of schooling (illiteracy, 0 ~ 6 years, above 6 years), living arrangement (living with family, living alone, living in institutions), occupation before the age of 60 (physical labor, no physical labor), smoking (yes, no), drinking (yes, no), exercise (yes, no), sleep quality (bad, moderate, good), sleep duration (< 7 h, 7 ~ 9 h, > 9 h), life satisfaction (bad, moderate, good), self-reported health (bad, moderate, good), and activities of daily living (ADL) (normal, disability).

Statistical analysis

In our study, categorical variables were denoted by frequency and n (%), and we used mean ± standard deviation (M ± SD) to represent continuous variables that obey normal distribution. The differences between continuous variables with and without multimorbidity were tested by two independent sample t-tests, and Chi-square analysis was introduced to assess the differences between frequency data. We used a binary logistic regression model to evaluate the association of PDI, hPDI, uPDI and multimorbidity, respectively. Thress logistic regression models were constructed: the foundational model (Model 1) excluded any confounding variables; Model 2 controlled sex, age, drinking, smoking, marital status, living arrangement, BMI, exercise, years of schooling, occupation before the age of 60, and residence; Model 3 further controlled for covariates including life satisfaction, self-reported health, sleep quality, sleep duration, and ADL based on Model 2. In addition, we used restrictive cubic splines (RCS) to decide if the relationship between three different PDIs and multimorbidity is nonlinear. We conducted subgroup analyses and further examined the interactions between subgroups. Moreover, we further conducted two sensitivity analyses to determine the robustness of the research results: (1) Using multiple chain equation interpolation method to replace missing data and eliminate the impact of direct deletion of a large number of missing values on the research results; (2) Considering that diabetes patients usually need to control their diet strictly, we excluded the samples with diabetes and re-evaluated the multimorbidity index and the effect of PDI on them. E-value was used to assess the potential impact of unmeasured confounding on the study results. All statistical analyses were conducted utilizing SPSS 26.0 and R 4.3.0. P < 0.05 suggests a statistical significance in our research.

Results

Basic demographic characteristics of participants

The sex ratio among older adults was relatively balanced in this study, with 45.32% of males and 54.68% of females. Among them, there were more people aged over 80 years old, occupying 57.73%. Most older adults had a BMI of 18.5 to 23.9 (59.46%). Moreover, a vast majority resided in rural communities (75.54%) and had comparatively short years of schooling, with a mere 20.37% attaining over six years of formal education. Among the participants, 70.63% of individuals engaged in physical labor before the age of 60, 30.90% of them smoked, 26.62% drank alcohol, and most people had good sleep quality (53.52%). The vast majority of them were living with family (80.94%) and satisfied with their lives, with only 2.97% having poor life satisfaction and 29.11% of older adults suffering from disabilities. In addition, PDI, hPDI, and uPDI scores had significant differences in whether older adults had multimorbidity or not. There exists significant difference in whether participants suffer from multimorbidity under different variables: age, residence, living arrangement, marital status, years of schooling, occupation before the age of 60, exercise, sleep quality, sleep duration, life satisfaction, self-reported health, and ADL. More specific information is presented in Table 1.

Table 1.

Characteristics of participants under different variables

Variables	Total (n = 7,551)	No multimorbidity (n = 2,466)	Multimorbidity (n = 5,085)	P-value
PDI, Mean ± SD	48.15 ± 5.37	48.49 ± 5.27	47.98 ± 5.42	< 0.001
hPDI, Mean ± SD	46.69 ± 5.06	47.19 ± 4.96	46.45 ± 5.10	< 0.001
uPDI, Mean ± SD	49.23 ± 6.60	48.12 ± 6.74	49.77 ± 6.46	< 0.001
Sex, n (%)				0.387
Male	3,422 (45.32)	1,100 (44.61)	2,322 (45.66)
Female	4,129 (54.68)	1,366 (55.39)	2,763 (54.34)
Age, n (%)				< 0.001
<=80	3,192 (42.27)	1,166 (47.28)	2,026 (39.84)
> 80	4,359 (57.73)	1,300 (52.72)	3,059 (60.16)
BMI, n (%)				0.455
< 18.5 kg/m2	1,107 (14.66)	361 (14.64)	746 (14.67)
18.5 ~ 24.9 kg/m2	4,490 (59.46)	1,445 (58.60)	3,045 (59.88)
> 25 kg/m2	1,954 (25.88)	660 (26.76)	1,294 (25.45)
Residence, n (%)				< 0.001
Urban	1,847 (24.46)	930 (37.71)	917 (18.03)
Rural	5,704 (75.54)	1,536 (62.29)	4,168 (81.97)
Living arrangement, n (%)				< 0.001
living with family	6,112 (80.94)	1,965 (79.68)	4,147 (81.55)
living alone	1,217 (16.12)	395 (16.02)	822 (16.17)
living in institutions	222 (2.94)	106 (4.30)	116 (2.28)
Marital status, n (%)				< 0.001
Other	4,271 (56.56)	1,281 (51.95)	2,990 (58.80)
Married	3,280 (43.44)	1,185 (48.05)	2,095 (41.20)
Years of schooling, n (%)				< 0.001
Illiteracy	3,450 (45.69)	930 (37.71)	2,520 (49.56)
0 ~ 6 years	2,563 (33.94)	835 (33.86)	1,728 (33.98)
Above 6 years	1,538 (20.37)	701 (28.43)	837 (16.46)
Occupation before the age of 60, n (%)				< 0.001
Physical labor	5,333 (70.63)	1,431 (58.03)	3,902 (76.74)
No Physical labor	2,218 (29.37)	1,035 (41.97)	1,183 (23.26)
Smoking, n (%)				0.220
Yes	2,333 (30.90)	785 (31.83)	1,548 (30.44)
No	5,218 (69.10)	1,681 (68.17)	3,537 (69.56)
Drinking, n (%)				0.128
Yes	2,010 (26.62)	629 (25.51)	1,381 (27.16)
No	5,541 (73.38)	1,837 (74.49)	3,704 (72.84)
Exercise, n (%)				< 0.001
Yes	2,544 (33.69)	987 (40.02)	1,557 (30.62)
No	5,007 (66.31)	1,479 (59.98)	3,528 (69.38)
Sleep quality, n (%)				< 0.001
Bad	1,152 (15.26)	523 (21.21)	629 (12.37)
Moderate	2,358 (31.23)	775 (31.43)	1,583 (31.13)
Good	4,041 (53.52)	1,168 (47.36)	2,873 (56.50)
Sleep duration, n (%)				< 0.001
< 7 h	2,767 (36.64)	1,026 (41.61)	1,741 (34.24)
7 ~ 9 h	3,337 (44.19)	1,063 (43.11)	2,274 (44.72)
> 9 h	1,447 (19.16)	377 (15.29)	1,070 (21.04)
Life satisfaction, n (%)				0.002
Bad	224 (2.97)	82 (3.33)	142 (2.79)
Moderate	1,960 (25.96)	696 (28.22)	1,264 (24.86)
Good	5,367 (71.08)	1,688 (68.45)	3,679 (72.35)
Self-reported health, n (%)				< 0.001
Bad	1,034 (13.69)	540 (21.90)	494 (9.71)
Moderate	2,850 (37.74)	1,048 (42.50)	1,802 (35.44)
Good	3,667 (48.56)	878 (35.60)	2,789 (54.85)
ADL, n (%)				< 0.001
Normal	5,353 (70.89)	1,681 (68.17)	3,672 (72.21)
Disability	2,198 (29.11)	785 (31.83)	1,413 (27.79)

t t-test, χ² Chi-square test, SD standard deviation, hPDI the healthful plant-based diet index, uPDI the unhealthful plant-based diet index, PDI the overall plant-based diet index, BMI body mass index, ADL activities of daily living

Association between PDI, hPDI, uPDI and multimorbidity

We found that PDI and hPDI associated with multimorbidity negatively in Model 1 (no covariates controlled in this model) in this study. In addition, we found that uPDI associated with multimorbidity positively. The association was still significant after controlling different variables in Model 2 and Model 3. We can observe that in Model 3, life satisfaction, self-reported health, sleep quality, sleep duration, and ADL, which are health-related covariates, were added on the basis of Model 2. The effects are still significant and the values are not significantly different. Therefore, the inclusion of health-related variables was not excessively adjusted. Besides, the E-value of 1.63 indicates that unmeasured confounders are unlikely to substantially affect the association between different PDIs and multimorbidity. More detailed information can be found in Table 2.

Table 2.

Associations of PDI, hPDI, and uPDI with multimorbidity among Chinese older adults

Model	PDI	hPDI	uPDI
Model 1	0.982 (0.974, 0.991) ***	0.972 (0.962, 0.981) ***	1.039 (1.031, 1.047) ***
Model 2	0.985 (0.972, 0.989) ***	0.975 (0.960,0.988) ***	1.020 (1.018, 1.026) ***
Model 3	0.987 (0.978, 0.997) *	0.980 (0.970, 0.991) ***	1.016 (1.007, 1.025) ***

hPDI the healthful plant-based diet index, uPDI the unhealthful plant-based diet index, PDI the overall plant-based diet index

Model 1 was unadjusted

Model 2 controlled s sex, age, drinking, smoking, marital status, living arrangement, BMI, exercise, years of schooling, occupation before the age of 60, and residence

Model 3 further controlled for covariates including life satisfaction, self-reported health, sleep quality, sleep duration, and ADL based on Model 2

* P < 0.05

*** P < 0.001

Restricted cubic splines in the regression model

We can see from part A of Fig. 2 that RCS analysis was performed to explore the relationship between PDI and multimorbidity, with 2 knots set at the 35th, 65th percentiles of PDI (a common setting to balance flexibility and robustness). The results confirmed a significant nonlinear association between PDI and multimorbidity (P overall = 0.001, P non-linear = 0.016; adjusted for sex, age, self-reported health, sleep quality, sleep duration, drinking, smoking, marital status, living arrangement, BMI, exercise, years of schooling, ADL, occupation before the age of 60, and residence). Moreover, both hPDI and uPDI have significant linear relationship with multimorbidity (hPDI: P overall < 0.001, P non-linear = 0.731; uPDI: P overall = 0.004, P non-linear = 0.939). Part B demonstrates that multimorbidity associated with hPDI negatively, and part C exhibits multimorbidity positively associated with uPDI.

Fig. 2.

Restricted cubic spline for testing the hypothesis of non-linear association between PDI, hPDI, uPDI and multimorbidity. (panel A: PDI, panel B: hPDI, panel C: uPDI)

Subgroup analysis

Figures 3, 4 and 5 are findings of subgroup analysis, we found these associations are significant in subgroups of covariates: age, sex, residence, smoking, drinking, exercise, and sleep duration. Statistically significant interactions between residence, exercise and PDI on multimorbidity were observed (P-Interaction < 0.01). In contrast, the P-interaction of hPDI for multimorbidity is only significant under residence. Moreover, the P-interaction of uPDI for multimorbidity is significant in the age and exercise subgroups. We can see from Figs. 3 and 4 that both PDI and hPDI still negatively associated with multimorbidity in different subgroups, with all OR values below 1, and uPDI positively associated with multimorbidity under different subgroups whose OR values are greater than 1. According to the findings of the subgroup analysis, we should advocate for ensuring the supply of healthy food based on urban-rural differentiation, and synchronously adjusting diet and exercise for people with insufficient exercise. In addition, optimize the staple food structure for the elderly and control the risks related to uPDI.

Fig. 3.

Associations of PDI with multimorbidity among subgroups. Notes: PDI the overall plant-based diet index, CI, confidence interval, *** P < 0.001, ** P < 0.01, * P < 0.05

Fig. 4.

Associations of hPDI with multimorbidity among subgroups. Notes: hPDI healthful plant-based diet index, CI, confidence interval, *** P < 0.001, ** P < 0.01, * P < 0.05

Fig. 5.

Associations of uPDI with multimorbidity among subgroups. Notes: uPDI unhealthful plant-based diet index, CI, confidence interval, *** P < 0.001, ** P < 0.01, * P < 0.05

Sensitivity analysis

To test the robustness of the results, we further considered two sensitivity analyses based on the Model 2, and results of the two sensitivity analyses are consistent with our main research findings. Firstly, in order to eliminate the impact of directly deleting a large number of missing values on the research results, we used a logistic regression model to test the randomness of missing values [41]. Due to the significant correlation between missing probability and observed variables, we believe that missing data conforms to the mechanism of random missing. Based on the above conclusion, we further adopt the multiple chain equation interpolation method to interpolate the missing values and conduct sensitivity analysis.The detailed results are shown in Table 3. Secondly, considering that diabetes patients usually need to strictly control their diet, which may have an impact on the choice of diet, we deleted the data of participants with diabetes and conducted a second sensitivity analysis. The results of this sensitivity analysis are presented in Table 4.

Table 3.

Sensitivity analysis by multiple chain equation interpolation

Variables	Multimorbidity(vs.no)
	OR (95%CI)
PDI	0.986(0.979,0.993)***
hPDI	0.974(0.967,0.982)***
uPDI	1.038(1.033,1.044)**

hPDI the healthful plant-based diet index, uPDI the unhealthful plant-based diet index, PDI the overall plant-based diet index, CI confidence interval

** P < 0.01

*** P < 0.001

Table 4.

Sensitivity analysis after removing participants with diabetes

Variables	Multimorbidity(vs.no)
	OR (95%CI)
PDI	0.989(0.981,0.997)**
hPDI	0.985(0.977,0.994)**
uPDI	1.026(1.020,1.032)***

hPDI the healthful plant-based diet index, uPDI the unhealthful plant-based diet index, PDI the overall plant-based diet index, CI confidence interval

** P < 0.01

*** P < 0.001

Discussion

Our research observed higher PDI and hPDI indices associated with lower predisposition to multimorbidity. In contrast, an elevated uPDI score among Chinese older adult associated with multimorbidity negatively. The results of RCS suggest a significant non-linear association between PDI and multimorbidity in Model 3 that controls for all covariates, while there exists significant linear relationship between hPDI, uPDI, and multimorbidity. In addition, two sensitivity analyses have demonstrated the robustness of the research results.

Multimorbidity is composed of many chronic diseases, such as diabetes, hypertension, cardiovascular disease, Parkinson’s disease, arthritis, dementia, etc. Previous extensive research has shown the significant impact of PBD on adverse health outcomes in older adults. For example, PBD can reduce the risk of diabetes, cardiovascular disease and mortality [27, 30, 33]. Research also shows that PBD can significantly reduce the risk of Parkinson’s disease and dementia [42, 43]. The reasons may be as follows: firstly, a large amount of high fiber and low calories in PBD can decrease the risk of diabetes through facilitating weight loss/maintenance [44], the mechanism may be due to dietary fiber lowering plasma cholesterol by binding with bile acids and dietary cholesterol in the intestinal lumen, thereby reducing cholesterol absorption [45]. Secondly, high content of certain elements such as vitamins, minerals, and antioxidants in some PBDs are significantly associated with a reduced risk of type-2 diabetes, insulin resistance, metabolic syndrome and other metabolic diseases [46, 47]. Vitamin C and vitamin E are both powerful antioxidants that play an important role in glucose metabolism, and consuming more mineral zinc may assist in regulating blood sugar levels [1]. Dietary carotenoids are also powerful antioxidant nutrients that can alter the composition of lipoproteins and affect the components of metabolic syndrome [48], which are associated with multimorbidity positively. What’s more, Omega-3 fatty acids in PBD diet can improve obesity-induced metabolic syndrome characteristics, including insulin resistance, hypertension, and dyslipidemia, by reducing plasma triglycerides [49]. A large amount of polyphenol in healthy PBD can prevent the development and progression of metabolic syndrome by reducing weight, blood pressure, and blood sugar, as well as improving abnormal lipid metabolism [50]. Plant-based food components, including polyunsaturated fatty acids, micronutrients, and natural bioactive ingredients, can reduce inflammation, endothelial dysfunction, and oxidative stress, thereby reducing the incidence of age-related chronic diseases [51]. A longitudinal study on the UK Biobank cohort found that higher adherence to dietary patterns rich in vegetables, fruits, and grains, and low in processed meat, was negatively associated with the risk of multimorbidity [52], and a prospective cohort study found that the higher the quality of diet, the lower the probability of multimorbidity [53]. As for animal-source food, consuming poultry and processed meat instead of red meat is associated with a higher risk of multimorbidity [52]. All of these fully demonstrate the significance of systematic research on PBD in preventing and intervening in the occurrence of multimorbidity.

The characteristic of hPDI is that it is rich in fresh fruits, vegetables, and legumes, which have been proven to be beneficial for cardiovascular health [35, 54]. Strict adherence to hPDI will contribute to high levels of antioxidants, unsaturated fats, dietary fiber, and micronutrients in the diet, as well as low levels of saturated fats and heme iron [33]. These substances can help with improving lipid status, strengthen blood sugar control and insulin regulation, improve vascular health, lower blood pressure, reduce inflammation, and thus reduce the occurrence of some chronic diseases, which positively impact multimorbidity. In this research, we discovered that hPDI was significantly negatively associated with multimorbidity than PDI (the OR values of hPDI are less than that of PDI). It has been found that the higher the score of hPDI, the higher the content of vegetables, fruits, plant protein, and whole grains, the lower the content of animal foods and refined carbohydrates, which is more negatively related to the incidence rate of coronary heart disease, chronic kidney disease, and type-2 diabetes than PDI [55, 56]. However, studies have also shown that uPDI has a negative impact on the occurrence of some chronic diseases. uPDI is positively correlated with the occurrence of diabetes [30].The South Korean prospective cohort Study found that uPDI was associated with an increased risk of metabolic syndrome [57], and another research has found that uPDI elevates the risk of hypertension among adults [58]. Therefore, it is recommended that individuals consume more healthful plant-based foods.

Statistically significant interactions on multimorbidity were found among residence, exercise and PDI. In contrast, hPDI’s P-interaction for multimorbidity was only significant with residence, while uPDI’s was significant only in age and exercise subgroups. The possible reasons for these modified effects are as follows: on the one hand, previous studies have shown that there are significant differences in medical services, economy and infrastructure between urban and rural areas, leading to significant differences in the incidence of multimorbidity between urban and rural areas [59, 60]. On the other hand, studies have shown that there are differences in diet between urban and rural areas, so residence may have a modified effect on the relationship between PDI and multimorbidity from these two aspects. In addition, a study shows that physical activity can reduce the incidence rate and mortality of multimorbidity [61], and it can help improve life satisfaction among individuals with multimorbidity [62]. Many previous research has demonstrated that the prevalence of multimorbidity elevates with age [63, 64], namely that elderly individuals tend to be at a greater risk of experiencing multimorbidity as the population ages [65]. Therefore, as the aging population in developing countries expands, multimorbidity may become increasingly common [66].

Limitations

Some limitations still exist that require improvement in this study. Firstly, these data are self-reported and may lead to unavoidable recall bias and social expectation bias during the investigation process. Secondly, the nationwide study questionaire only measured the types of food consumed by participants based on the characteristics of the food. It did not precisely measure the specific antioxidants or micronutrients, such as simple carbohydrates and fiber in the food. Therefore, further precise measurement of the element content in future diets is needed to better confirm its relationship with multimorbidity. Thirdly, due to the cross-sectional nature of this study, we can only inspire insights for future research, and cannot verify the order and causal relationships between these variables. Thus, we can use cohort studies to confirm their possible causal relationship in the future research.

Conclusion

In Chinese older adults, PDI and hPDI are negatively associated with multimorbidity, while uPDI is positively associated with it. Notably, hPDI and uPDI show a significant linear relationship with multimorbidity. Additionally, residence, exercise, and age can modify the association between PBD and multimorbidity. This study re-emphasizes the importance of dietary and nutritional interventions in addressing multimorbidity among the elderly. Accordingly, we advocate increasing the intake of healthy plant-based foods and reducing unhealthy ones-an approach that not only empowers individual health by lowering health risks and improving health status but also provides empirical support for public health policy-making, offering direction for implementing dietary guidance and health interventions.

Authors’ contributions

Each author has met the authorship requirements. T.W. : Methodology, Software, Writing - Original Draft, Writing - Review & Editing, Visualization; C.L.: Software, Writing - Review & Editing, Visualization; L.Y.: Methodology, Conceptualization, Writing - Review & Editing. H.A.: Project administration, Supervision, Writing - Review & Editing. All authors have read and agreed to the published version of the manuscript.

Funding

No funding.

Data availability

The data of CLHLS are available at https://opendata.pku.edu.cn/dataset.xhtml?persistentId=doi:10.18170/DVN/WBO7LK&version=2.0.

Declarations

Ethics approval and consent to participate

The study was conducted according to the guidelines of the Declaration of Helsinki and approved by Peking University’s Ethics Committee. Informed consent was obtained from all subjects involved in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.