Analysis of the relationship between early adverse childhood experiences and late-life health poverty vulnerability among rural middle-aged and elderly people in Ningxia: the mediating effect of chronic diseases and their comorbidities

Abstract

Background

Previous research has suggested that there is a certain degree of relationship between adverse childhood experiences (ACEs) and chronic diseases and their comorbidities. However, the association between ACEs and health poverty vulnerability, along with the role played by chronic diseases and their comorbidities in this context, has been scarcely investigated. Therefore, this study aims to explore whether ACEs are related to the health poverty vulnerability of middle-aged and elderly people in rural Ningxia and whether chronic diseases and their comorbidities mediate the association between them.

Methods

A total of 5,533 valid samples were included from the 2024 Ningxia Rural Residents’ Health Status and Health Service Utilization Survey data. Pearson’s test was applied to investigate the relationships among the key variables. Regression models were used to examine the mediating effects of chronic diseases and their comorbidities, and the Sobel test and the bootstrap method were used to determine the final path effects. Following the mediation analysis, a moderated mediation model was further examined to investigate the moderating effect of depression on the mediating mechanism. Additionally, to assess the robustness of the association between ACEs and health poverty vulnerability, we conducted multiple sensitivity analyses. These included E-values, missing data simulation, boundary analysis, and bias factors.

Results

There was a significant correlation between ACEs, chronic diseases and their comorbidities and health poverty vulnerability (p < 0.001). ACEs are directly associated with health poverty vulnerability (β = 0.0177, p < 0.001). ACEs are significantly associated with chronic diseases and their comorbidities (β = 0.0904, p < 0.001), whereas chronic diseases and their comorbidities have a significant association with health poverty vulnerability (β = 0.0256, p < 0.001). Chronic diseases and their comorbidities partially mediated the correlation of ACEs on health poverty vulnerability (β = 0.0023, p < 0.001). The Sobel test and bootstrap method validated the robustness of the regression analysis results. The moderated mediation model analysis revealed that while the moderating effect was not significant, the mediating effect remained significant across different levels of depression. Multiple sensitivity analyses consistently indicate that the positive correlation between ACEs and health poverty vulnerability exhibits robust stability.

Conclusion

ACEs are significantly positively associated with health poverty vulnerability, and chronic diseases and their comorbidities mediate this relationship. Our research findings indicate that prevention strategies for the health poverty vulnerability of middle-aged and elderly individuals should focus on creating a positive childhood environment and developing comprehensive strategies for integrating medical and preventive care for chronic diseases.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26298-4.

Introduction

Poverty is a persistent global issue that requires constant reflection and prevention by nations, society, and individuals. With the comprehensive victory in the poverty alleviation battle achieved by the end of 2020 and the steady advancement and implementation of the comprehensive well-off strategy, the country has made a great leap from absolute poverty to relative poverty, bridging this historical gap. Importantly, although we have currently addressed the issue of absolute poverty at the material level, poverty is a dynamic process, and its meaning goes beyond mere material deprivation to include dimensions such as health and education. Therefore, there is still a possibility of falling back into poverty in the future [1]. Historically, health poverty was frequently incorporated into the measurement of poverty. This metric, however, somewhat lags behind that of China, which has already accomplished complete poverty reduction. As early as 2000, the World Bank introduced the concept of “poverty vulnerability” to reflect the probability of future welfare decline due to various risk shocks [2]. On this basis, Liu Yue and others proposed the dynamic and forwards-looking concept of “health poverty vulnerability,” defining it as the probability of a household or individual falling below the poverty line due to health-related risk factors, reflecting the risk probability of this household or individual falling into poverty or falling back into poverty [3]. To determine the likelihood that formerly underprivileged residents will relapse into poverty because of health problems, this study used the health poverty vulnerability indicator. The World Health Organization (WHO) defines multimorbidity as an individual simultaneously suffering from two or more chronic diseases. The relationship between health poverty vulnerability and chronic illnesses and their comorbidities has been the main focus of research in recent years [4]. The increasing vulnerability to poverty among middle-aged and elderly people in rural areas is influenced and constrained by various factors, including economic income, health issues, social support, and other unexpected shocks [5, 6]. After China’s overall success in reducing poverty in 2020, one of the main concerns in poverty governance was the vulnerability to health poverty caused by disease conditions. For example, Miao Xiaoping noted in China Daily that chronic diseases and their comorbidities not only severely reduce patients’ quality of life but also increase the risk of repeated hospitalizations. The accompanying medical expenses add to the economic burden on patients. He suggested that the country and locality should place great importance on the management of chronic disease comorbidities [7]. Therefore, chronic diseases and their comorbidities may be associated with health poverty vulnerability among the rural elderly population. ACEs or childhood adversities refer to various potential harms that individuals may suffer in terms of health, development, psychology, and social aspects before the age of 18, mainly consisting of abuse, neglect, and family dysfunction [8]. An increasing amount of evidence suggests that childhood abuse, neglect, and social psychological stress are significant and difficult-to-change risk factors for diseases and disabilities in adulthood [9]. Therefore, many researchers have explored the relationships between adverse childhood experiences and health outcomes in different populations. However, these studies merely indicate the relationship between adverse childhood experiences and health outcomes such as chronic diseases, without elucidating the nature of the relationship among these three factors following the emergence of health poverty vulnerability [10–14]. In other words, they fail to examine the mediating effect of chronic diseases and their comorbidities in the association between adverse childhood experiences and health poverty vulnerability. Therefore, this study aims to investigate the relationship among ACEs, chronic diseases and their comorbidities, and health poverty vulnerability in middle-aged and elderly individuals. Based on the aforementioned theoretical framework and literature, we propose the following hypothesis:

• H1: Health poverty vulnerability is associated with ACEs.

• H2: Chronic diseases and their comorbidities are associated with ACEs.

• H3: Health poverty vulnerability correlates with chronic diseases and their comorbidities.

Methods

Study population and data source

The data for this study came from the latest survey data of 2024 on health status and utilization of health services among rural residents in Ningxia in a natural population cohort established in 2009. This queue has been followed up 6 times over the course of 15 years. The respondents were rural residents aged 45 years and above from Haiyuan, Yanchi, Pengyang and Xiji Counties in Ningxia. The researcher obtained the respondents’ informed consent before conducting this study. This study was approved by the Ethics Committee of Ningxia Medical University (approval number: 2022-G165).

Questionnaire and survey methods

During the questionnaire design phase, we refined and optimized the questionnaire content through expert consultations. Before the formal investigation, we identified existing issues through a preliminary survey and made the necessary adjustments. After this series of revisions, the quality of the questionnaire significantly improved.

To ensure good scientific rigor and regional representativeness, we used a multistage stratified cluster random sampling method to select samples from four counties. The sampling process is separated into three stages. As shown in Fig. 1. Finally, trained investigators conducted face-to-face surveys with farmers, covering topics such as personal basic information, illness and medical treatment within the past two weeks, hospitalization within the past year, chronic disease prevalence, health status and healthy lifestyle of members aged 15 and above, health status of residents aged 45 and above, and household economic conditions. Notably, during the survey, only household members who had lived in the area for more than six months were included in the study. A total of 5,533 respondents were included in this study after samples with participants under 45 years of age and missing information were eliminated. The absence of relevant important variables is shown in Fig. 2.

Fig. 1.

Flowchart of the multistage stratified cluster random sampling method

Fig. 2.

Absence of relevant important variables

Measures

Measurement of adverse childhood experiences

On the basis of previous research on ACEs, these documents highlight the significant impact of ACEs on health and social outcomes [10, 15–18]. Considering the data available for this study, in conjunction with the internationally recognized ACE evaluation scale and previous related research, we selected personal childhood information from this survey as ACE evaluation indicators, including childhood hunger, delayed medical treatment for illnesses, poor physical condition, incomplete family structure (not both parents being alive), poor physical condition of the father, poor physical condition of the mother, father not having attended school, and mother not having attended school, for a total of 8 variables. To specifically express the exposure to ACEs, the above ACE items were dichotomized during the analysis, and on this basis, each individual’s cumulative ACE score was obtained, ranging from 0 to 8. The higher the score is, the greater the intensity of childhood adverse experience exposure.

The content validity of this questionnaire is primarily demonstrated through a literature foundation. The design of questionnaire items strictly adheres to the theoretical framework of life course epidemiology and directly references internationally recognized childhood adversity scales for middle-aged and elderly populations. Specifically, the above eight variables exhibit consistency with domestic and international childhood adversity modules; Measurements of adversities such as childhood hunger and healthcare accessibility align with key indicators of physical neglect in ACEs research. Parental education indirectly reflects parental investment in offspring from a cultural cognitive perspective. Assessments of parental health indirectly reflect family structural integrity and parental caregiving capacity. The absence of both parents or prolonged parental illness inherently substantially increases children’s risk of neglect and emotional deprivation.

Additionally, compared to authoritative childhood adversity scales both domestically and internationally, this questionnaire lacks dimensions measuring abuse and neglect for the following reasons: Residents in rural Ningxia exhibit a certain degree of cultural sensitivity, viewing family matters as highly private. Incorporating terms like “abuse” (e.g., physical or emotional abuse) and “neglect”—which carry strong negative connotations—directly into the questionnaire could easily trigger defensive reactions and resistance among respondents. Secondly, if interviewers directly ask parents whether they have abused or neglected their children, this may be culturally perceived as challenging parental authority and tarnishing family honor. This could lead respondents to refuse to answer or provide inaccurate information out of shame or to protect the family, severely compromising data authenticity and validity. Finally, within Ningxia’s traditional cultural context, severe corporal punishment might not be perceived as “abuse” but rather as normal “family discipline”; similarly, neglecting children due to family livelihood concerns might not be recognized as “neglect.” Directly applying modern psychological scale terminology may raise issues of cultural validity. Therefore, this questionnaire indirectly captures experiences related to neglect and emotional deprivation through multiple proxy indicators and contextual questions. While these items avoid explicit terms like “abuse” or “neglect,” they effectively reflect analogous psychological states.

Measurement of chronic diseases and their comorbidities

The prevalence of chronic diseases is determined by asking, “Have you ever been diagnosed with a chronic disease by a doctor?” The chronic diseases included in this study are hypertension, diabetes, intervertebral disc disease, cerebrovascular disease, chronic gastroenteritis, coronary heart disease, rheumatoid arthritis, chronic obstructive pulmonary disease, and other chronic diseases. Other chronic diseases are characterized by a combination of less prevalent conditions, such as bronchial asthma and chronic kidney disease. Additionally, chronic disease comorbidity is defined as the presence of two or more chronic diseases in the same individual.

Measurement of health poverty vulnerability

Health poverty vulnerability refers to the probability that a household’s or individual’s welfare will fall below the poverty line due to health-related risks, reflecting the risk of falling into poverty in the future. Current domestic and international methods for measuring health poverty vulnerability include vulnerability as expected poverty (VEP) theory, vulnerability as expected utility (VEU) theory, and vulnerability as uninsured exposure to risk (VER) theory [19]. Given that this study uses cross-sectional data to compensate for the shortcomings of it, this research is based on the theory of VEP. The theoretical framework is underpinned by two principles: Firstly, consumption as the core welfare indicator. Compared to income, consumption better smooths short-term fluctuations for individuals and households, thereby more stably and accurately reflecting their long-term welfare levels and resource constraints. Consequently, the distribution of future consumption serves as an ideal proxy variable for measuring future welfare status and its associated risks. Secondly, the uncertainty of future welfare status. The uncertainty of future welfare states. The future consumption levels of individuals and households are uncertain. On the one hand, observable heterogeneous characteristics such as education, health, age, and household assets determine the expected level of future consumption (i.e., the conditional expectation). On the other hand, unpredictable random shocks such as unemployment, illness, and natural disasters introduce uncertainty into future welfare, ultimately determining the volatility of consumption around its expected level (i.e., the conditional variance). Finally, future welfare risks for individuals and households can be modelled and quantified. By estimating consumption determination equations using historical data, we can reconstruct the probability distribution of their future consumption. This distribution centres on the conditional expectation, with its dispersion measured by the conditional variance. This transformation renders the calculation of the probability that future consumption falls below a given poverty threshold a solvable statistical problem. This method uses a three-stage feasible generalized least squares (FGLS) approach to measure the health poverty vulnerability of middle-aged and elderly people in Ningxia [20, 21]. The specific steps are as follows:

The first step is to estimate the future consumption equation. We establish the basis for the following specific analysis of vulnerability by estimating the consumption equation via ordinary least squares (OLS). Equation (1) can be used to represent it:

1

where represents the consumption level of individual i in period t + 1, and X_it is a series of characteristic variables that affect the future consumption level of individuals. Combining literature data and field survey results, the health social determinants model [22, 23] is used to extract explanatory variables from five dimensions: biological and genetic factors (gender, age), individual lifestyles (self-rated health status, nighttime sleep duration, napping duration, exercise frequency, smoking and drinking habits), social and community networks (marital status, number of relatives to whom one can turn for help, number of friends to whom one can turn for help, and borrowing situation), structural social factors (education level, housing type, water source type, toilet type, cooking fuel type, the time required to reach the nearest county hospital and the logarithm of per capita annual income) and macro social conditions (type of health insurance and whether they are low-income households). Considering the heterogeneity among different individuals, the square of the residuals is used as an approximation of the consumption variance . The squared residuals are used as the dependent variable to construct a regression model of the squared residuals against the characteristic variables of rural middle-aged and elderly people, represented by Eq. (2):  

2

The second step is to estimate the expected value and variance of the logarithm of future consumption for the rural middle-aged and elderly populations. Weighted regression is performed on the logarithm of consumption and the squared residuals to obtain the FGLS estimates and . The expected value and variance of future consumption logs are then estimated, as shown in Eqs. (3) and (4):

3

4

Finally, health poverty vulnerability is estimated via the poverty line. The choice of the poverty line can affect the accuracy of health poverty vulnerability. Given that China has now entered the stage of comprehensive poverty eradication, this study employs the World Bank’s 2022 poverty line standard for lower-middle-income countries—US$3.65 per day—to align China’s poverty reduction achievements with international benchmarks. This approach facilitates more effective international comparisons and trend analysis within a multidimensional poverty framework, enabling the assessment and consolidation of these accomplishments within an international context. Since the focus of this study is on rural middle-aged and elderly individuals aged 45 and above, the log-normal distribution is a perfect match for describing the consumption levels of such low-income groups compared with other distribution types. Therefore, the probability of individual i being in poverty at time t can be expressed as Eq. (5):

5

where represents the expected health poverty vulnerability index. Φ (•) represents the cumulative density function under the standard normal distribution. Z represents the poverty line. represents the expected value of the logarithm of per capita consumption. represents the expected variance in the logarithm of per capita consumption. In this study, the vulnerability threshold is set to 0.5, meaning that if an individual has a probability greater than 0.5 of falling into poverty in the future, that individual is considered to have poverty vulnerability.

Statistical analysis

In our research process, we used EpiData 3.1 to code the data and conducted a comprehensive treatment of missing values for the collected survey data before analysis. Finally, all statistical analyses were performed using STATA 17.0 and R 4.5.1.

Descriptive analysis

We started the analysis procedure by performing statistical descriptions of the important variables and sociodemographic traits. The frequency (percentage) is used to describe categorical variables, whereas the mean (standard deviation, SD) is used for continuous variables.

Differential analysis

The differences in characteristics between different subgroups of health poverty vulnerability were compared via the chi-square test or t test.

Correlation analysis

Second, Pearson correlation analysis was used to measure the bivariate correlation between ACEs, chronic diseases and their comorbidities and health poverty vulnerability.

Mediation analysis

We adopt the strategy of Judd and Kenny and analyse the relationships among ACEs, chronic diseases and their comorbidities with health poverty vulnerability among middle-aged and elderly individuals on the basis of the causal step method of Baron and Kenny [24]. The following three regression equations estimate the effects of the following different effects: (1) Regress the independent variable (adverse childhood experiences) against the dependent variable (health poverty vulnerability) to verify their correlation; (2) Regress the independent variable against the mediating variable (chronic diseases and their comorbidities) to verify their correlation; (3) Regress both the independent variable and the mediating variable against the dependent variable. This simultaneously verifies whether the mediating variable correlates with the dependent variable and assesses whether the association between the independent variable and the dependent variable is attenuated. If the regression coefficient of the independent variable diminishes after introducing the mediating variable, this indicates partial mediation. Should the association between the independent variable and the dependent variable ultimately cease to be statistically significant, this signifies full mediation. In this study, We employed a directed acyclic graph (DAG) to determine the minimal sufficient adjusted set of covariates for estimating the total impact of ACEs on health poverty vulnerability without confounding [25–27]. As shown in Fig. 3. These sets include gender, education level, number of relatives who can rely on help, the number of friends who can rely on help, and the time needed to reach the nearest county hospital. To increase the precision of the model estimates, the aforementioned factors are incorporated into all regression analyses. After the regression, we performed rigorous verification via the Sobel test to determine whether the indirect effect was significantly nonzero. If the test shows that the Z value exceeds 1.96, the indirect effect is statistically significant. To determine the specific value of the effect and establish a 95% confidence interval, we evaluated the direct, indirect, and total effects via the bootstrap method with 5000 iterations. When the 95% confidence interval does not include 0, it indicates a significant association. All the statistical tests were two-tailed, and only those with P < 0.05 were considered statistically significant. Finally, after validating the mediating effect of adverse childhood experiences on health poverty vulnerability through chronic diseases and their comorbidities, this study examined a moderated mediation model to further explore the moderating role of depressive mood in this mechanism [28]. Specifically: The post-mediation model examined whether depression moderates the latter segment of the mediating pathway, specifically the strength of the association between “chronic diseases and their comorbidities and health poverty vulnerability.” For each model, we calculated conditional indirect effects at low, medium, and high depression levels using the Bootstrap method (5,000 resamples) and tested the moderated mediation indices to determine the statistical significance of the moderating effects [29].

Fig. 3.

Directed acyclic graph (DAG). In the DAG, each node (circle) represents a variable, and each unidirectional arrow represents a direct association. The absence of an arrow assumes that there is no direct association. The exposure variable (ACEs) is represented by a play button, and the outcome variable (health poverty vulnerability) is represented by a stop button. The pink circles represent potential confounding variables

Sensitivity analysis

To assess the robustness of the association between ACEs and health poverty vulnerability, we conducted multiple sensitivity analyses. First, to evaluate the impact of unmeasured confounders on the observed effect, we applied the E-value analysis method proposed by VanderWeele and Ding (2017). This value is defined as the minimum strength of association between an unmeasured confounder and the exposure, and between the unmeasured confounder and the outcome, required to explain the observed association between the exposure and the outcome [30]. Second, to examine the potential impact of different missing data mechanisms on the study’s primary findings, we selected three methods for sensitivity analysis. First, we simulated an additional 10% data loss based on the existing dataset [31]. The simulation design utilized the current effective sample size (n = 5533), setting the additional missing proportion at 10%. Using the risk indicator “health poverty vulnerability” identified in prior analyses as a stratification variable, the sample was divided into high-risk (vulnerable) and low-risk (vulnerable) groups. The mean characteristics of Adverse Childhood Experiences (ACEs) were calculated for each group. The missing data mechanism simulation included three scenarios: High-risk missing: Assuming 80% of missing cases belong to the high-risk group, their ACEs values were drawn from a normal distribution centered on the high-risk group mean; Low-risk missing: Assuming only 20% of missing cases belong to the high-risk group, their ACEs values are drawn from a normal distribution centered on the low-risk group mean; Random missing: Assuming missing data occur completely at random, with ACEs and health poverty vulnerability distributions consistent with the population. Second, we re-assign missing values through boundary analysis based on a quantitative model of selection bias under three assumptions [32]: (1) Optimistic bound: Assuming missing data reduces observed effects by 20%; (2) Conservative bound: Assuming missing data increases observed effects by 30%; (3) Extreme bound: Assuming missing data increases observed effects by 60%. Finally, to quantify the bias intensity required to overturn conclusions, we calculated the bias reversal factor [33]. Specifically, we calculated the bias factor required to attenuate the strength of the association between ACEs and health poverty vulnerability (OR value) to a meaningful threshold (set at OR = 1.10 in this study).

Results

Descriptive analyses of variables

Figure 4 shows that 9% of participants may experience future poverty as a result of health problems. Table 1 lists the descriptive statistics of the research subjects. This study included a total of 5,533 participants, with an average age of 61.36 years, comprising 3,306 men and 2,227 women. Additionally, there are significant differences between those with health poverty vulnerability and those with nonhealth poverty vulnerability in terms of age, self-rated health status, nighttime sleep duration, weekly exercise frequency, smoking, drinking, marital status, number of relatives they can seek help from, number of friends they can seek help from, whether the family has any borrowing situation, education level, type of drinking water, type of toilet, type of cooking fuel, time required to reach the nearest county hospital, logarithm of per capita annual, whether to participate in the urban–rural integrated resident medical insurance or the urban employee medical insurance, and whether to be a low-income household.

Fig. 4.

Distribution of the health poverty vulnerability index

Table 1.

Sociodemographic characteristics and critical variables of the participants

Variable	Overall (N= 5533)	Health poverty vulnerability			P
		No (N =5033)	Yes (N = 500)
Biological and genetic factors dimension
Gender				0.069	0.792
Male	3306(59.75)	3010(59.81)	296(59.20)
Female	2227(40.25)	2023(40.19)	204(40.80)
Age(years)	61.36(9.88)	60.02(9.10)	74.83(6.82)	-35.395	< 0.001
Individual lifestyle dimension
Self-assessment of health status				81.119	< 0.001
Very good	209(3.78)	192(3.81)	17(3.40)
good	1006(18.18)	955(18.97)	51(10.20)
fair	1692(30.58)	1582(31.43)	110(22.00)
average	1799(32.51)	1605(31.89)	194(38.80)
poor	827(14.95)	699(13.89)	128(25.60)
Nighttime sleep duration				33.935	< 0.001
< 7 h	1912(34.56)	1691(33.60)	221(44.20)
7 ~ 8 h	3252(58.77)	3019(59.98)	233(46.60)
> 8 h	369(6.67)	323(6.42)	46(9.20)
Nap duration				0.503	0.778
< 30 min	982(17.75)	899(17.86)	83(16.60)
30 ~ 60 min	3737(67.54)	3394(67.43)	343(68.60)
> 60 min	814(14.71)	740(14.70)	74(14.80)
Exercise frequency				82.813	< 0.001
6 times or more	1332(24.07)	1273(25.29)	59(11.80)
3–5 times	893(16.14)	839(16.67)	54(10.80)
1 ~ 2 times	639(11.55)	585(11.62)	54(10.80)
Not even once	229(4.14)	201(3.99)	28(5.60)
never	2440(44.10)	2135(42.42)	305(61.00)
Smoking				62.631	< 0.001
Never smokes	4191(75.75)	3740(74.31)	451(90.20)
smokes	1172(21.18)	1128(22.41)	44(8.80)
Quit smoking	170(3.07)	165(3.28)	5(1.00)
Drinking alcohol				37.856	< 0.001
Never drinks alcohol	5061(91.47)	4567(90.74)	494(98.80)
Drinking alcohol	388(7.01)	383(7.61)	5(1.00)
Quit drinking	84(1.52)	83(1.65)	1(0.20)
Social and community network dimensions
Marital status				188.730	< 0.001
Not married	643(11.43)	491(9.76)	152(30.40)
married	4890(88.38)	4542(90.24)	348(69.60)
Number of relatives they can seek help from				212.506	< 0.001
0	610(11.02)	486(9.66)	124(24.80)
1	1255(22.68)	1094(21.74)	161(32.20)
2	1449(26.19)	1307(25.97)	142(28.40)
3 ~ 4	1326(23.97)	1267(25.17)	59(11.80)
5 ~ 8	611(11.04)	599(11.90)	12(2.40)
9 or more	282(5.10)	280(5.56)	2(0.40)
Number of friends they can seek help from				212.258	< 0.001
0	1617(29.22)	1348(26.78)	269(53.80)
1	1391(25.14)	1252(24.88)	139(27.80)
2	1156(20.89)	1096(21.78)	60(12.00)
3 ~ 4	852(15.40)	826(16.41)	26(5.20)
5 ~ 8	319(5.77)	314(6.24)	5(1.00)
9 or more	198(3.58)	197(3.91)	1(0.20)
Borrowing situation				327.749	< 0.001
No	3115(56.30)	2642(52.49)	473(94.60)
Yes	2418(43.70)	2391(47.51)	27(5.40)
Dimension of structural social factors
literacy levels				236.823	< 0.001
illiterate	2158(39.00)	1815(36.06)	343(68.60)
elementary school	2073(37.47)	1929(38.33)	144(28.80)
Middle school	981(17.73)	971(19.29)	10(2.00)
high school and above	321(5.80)	318(6.32)	3(0.60)
Housing type				0.621	0.431
Nonsanitary housing	948(17.13)	856(17.01)	92(18.40)
Sanitary housing	4585(82.87)	4177(82.99)	408(81.60)
Type of drinking water				17.802	< 0.001
Decentralized water supply	232(4.19)	193(3.83)	39(7.880)
Tap water	5301(95.81)	4840(96.17)	461(92.20)
Type of toilet				21.808	< 0.001
Nonsanitary toilet	4992(88.96)	4446(88.34)	476(95.20)
Sanitary toilet	611(11.04)	587(11.66)	24(4.80)
Type of cooking fuel				84.710	< 0.001
Nonclean energy	278(5.02)	210(4.17)	68(13.60)
Clean energy	5255(94.98)	4823(95.83)	432(86.40)
Time required to reach the nearest county hospital	58.19(34.34)	58.53(34.83)	54.77(28.87)	2.332	0.020
Per capita annual income (logarithmic)	7.34(6.98)	7.77(6.09)	3.04(12.08)	14.708	< 0.001
Macro social conditions dimension
urban‒rural integrated medical insurance or the urban employee medical insurance				10.015	0.002
No	64(1.16)	51(1.01)	13(2.60)
Yes	5469(98.84)	4982(98.99)	487(97.40)
Low-income household				692.794	< 0.001
Yes	1741(31.47)	1323(26.29)	418(83.60)
No	3792(68.53)	3710(73.71)	82(16.40)

Correlation analysis

As shown in Table 2, the results of the Pearson’s correlation analysis revealed a significant correlation between ACEs, chronic diseases and their comorbidities, and health poverty vulnerability (P < 0.001). Specifically, ACEs are positively correlated with chronic diseases and their comorbidities (r = 0.127, p < 0.001), and they are also positively correlated with health poverty vulnerability (r = 0.233, p < 0.001).

Table 2.

Bivariate correlation matrix for ACEs, chronic diseases and their comorbidities, and health poverty vulnerability

Variable	ACEs	Chronic diseases and their comorbidities	Health poverty vulnerability
ACEs	1.000
chronic diseases and their comorbidities	0.127***	1.000
health poverty vulnerability	0.222***	0.233***	1.000

^***P < 0.001

Analysis of the mediating effect of the action path

Regression analysis was used to explore the mediating role of chronic diseases and their comorbidities on the relationship between ACEs and health poverty vulnerability. To avoid the problem of multicollinearity, we calculate the variance inflation factor (VIF) for the covariates in the three regression equations separately. The results show that the maximum VIF value is 1.51, which is far less than 10, indicating that there is no serious multicollinearity problem. Table 3; Fig. 5 show the results of the mediation analysis of chronic diseases and their comorbidities between ACEs and health poverty vulnerability after controlling for covariates. In the total effects regression, ACEs are significantly associated with health poverty vulnerability (). When the effects of chronic diseases and their comorbidities are considered, this relationship is still significant (). Furthermore, ACEs are significantly associated with chronic diseases and their comorbidities. (). Chronic diseases and their comorbidities also have a significant association with health poverty vulnerability. ().

Table 3.

The mediating role of chronic diseases and their comorbidities in the relationship between aces and health poverty vulnerability

Variable	Health poverty vulnerability	Chronic diseases and their comorbidities	Health poverty vulnerability
ACEs	0.0200*** (0.0159,0.0241)	0.0904*** (0.0664,0.1144)	0.0177*** (0.0136,0.0217)
Gender	-0.0161** (-0.0253, -0.0071)	0.1255*** (0.0720,0.1791)	-0.0194*** (-0.0284, -0.0104)
literacy levels	-0.0567*** (-0.0618, -0.0515)	-0.0711*** (-0.1014-0.0408)	-0.0549*** (-0.0600, -0.0497)
number of relatives they can seek help from	-0.0264*** (-0.0304, -0.0225)	-0.0200* (-0.0431,0.0030)	-0.0259*** (-0.0298, -0.0220)
number of friends they can seek help from	-0.0192*** (-0.0231, -0.0154)	-0.0114 (-0.0338,0.0111)	-0.0189*** (-0.0227, -0.0152)
Time required to reach the nearest county hospital	-0.0003*** (-0.0005, -0.0002)	-0.0006 (-0.0013,0.0001)	-0.0003*** (-0.0004, -0.0002)
chronic diseases and their comorbidities			0.0256*** (0.0212,0.0301)
Constant	0.3979*** (0.3699,0.4259)	0.4956*** (0.3311,0.6601)	0.3852*** (0.3575,0.4130)
Observations	5533	5533	5533
R-squared	0.2149	0.0268	0.2328

***P < 0.001, ^**P < 0.05

Fig. 5.

Relationships among ACEs, chronic diseases and their comorbidities, and health poverty vulnerability

Sobel and bootstrap tests were conducted to check for indirect, direct, and total effects [34]. As shown in Table 4, Sobel’s test indicated that the indirect effect (Z = 6.186, P < 0.001), direct effect (Z = 0.0177, P < 0.001), and total effect (Z = 0.0200, P < 0.001) were significant. This indicates that chronic diseases and their comorbidities partially mediate the association between ACEs and health poverty vulnerability. The bootstrap method indicated that the direct effect of ACEs on health poverty vulnerability was 0.0177 (95% CI: 0.0139, 0.0215), whereas the total effect was 0.0200 (95% CI: 0.0161, 0.0238). The indirect effect of healthy lifestyles on cognitive function mediated by depressive symptoms was 0.0023 (95% CI: 0.0016, 0.0031). All the paths of action were significant, indicating the robustness of the mediation model results [35].

Table 4.

Sobel and bootstrap tests for the mediation models

Paths	Observed Coefficient	Sobel test			Bootstrap test
		Z Value	P		Bootstrap Standard Error	P	LLCI	ULCI
Indirect effect	0.0023	6.186	< 0.001		0.0004	< 0.001	0.0016	0.0031
Direct effect	0.0177	8.550	< 0.001		0.0019	< 0.001	0.0139	0.0215
Total effect	0.0200	9.608	< 0.001		0.0020	< 0.001	0.0161	0.0238

Mediational analysis with moderation

Table 5 presents the core path coefficients and their statistical significance in the research model. The results indicate that the mediating path is valid, but the moderating effect is not.

Table 5.

Path coefficients and significance tests

Path /effect	β	SE	t	p
ACEs → D115	0.1045	0.0129	8.109	< 0.001
D115 → vul2	0.0280	0.0027	10.273	< 0.001
ACEs → vul2	0.0260	0.0024	10.896	< 0.001
Post-processing adjustment	–0.0002	0.0005	–0.292	0.770
Front-end adjustment	–0.0031	0.0029	–1.071	0.284

Table 6 shows the magnitude and stability of the mediating effects across different levels of depression. In both models, the indirect effect values for chronic diseases and their comorbidities at all levels were non-zero and statistically significant.

Table 6.

Analysis of conditional indirect effects

Model		Post-processing adjustment model			Front-end adjustment model
Level of depression		Indirect effect value	95% CI		Indirect effect value	95% CI
Mild depression		0.0030	[0.0019,0.0044]		0.0027	[0.0016,0.0039]
Average depression		0.0029	[0.0021,0.0040]		0.0023	[0.0014,0.0032]
High depression		0.0029	[0.0019,0.0040]		0.0018	[0.0005,0.0032]

Table 7 presents the results of the two moderated mediation index tests. The index values for both models are extremely small, and their confidence intervals include zero, indicating that depression does not significantly moderate the mediating effect.

Table 7.

Test of the moderated Mdiating index

Adjustment type	Index	95% CI
Post-processing adjustment index	0.0000	[-0.0014,0.0010]
Front-end adjustment index	-0.0001	[-0.0027,0.0009]

Sensitivity analysis

Analysis based on 5,533 complete samples indicates that the association between ACEs and health poverty vulnerability remains robust across various sensitivity analyses. Logistic regression results indicate a significant positive correlation between ACEs and health poverty vulnerability. Each additional unit of ACEs increases the risk of health poverty vulnerability by 42%. After adjusting for chronic diseases and their comorbidities, the association weakened (OR = 1.38), suggesting chronic diseases partially mediate the relationship. Finally, after controlling for gender, After controlling for gender, educational attainment, number of supportive relatives, number of supportive friends, travel time to the nearest county hospital, and chronic disease status, the association remained significant. The point estimate from the E-value analysis was 2.19, with a lower 95% CI bound of 1.92. Simulations with 10% data missing yielded significant results across all three missingness patterns. Boundary analysis tested hypotheses about ACE missingness. Only under highly optimistic ACE missingness assumptions did results become non-significant; otherwise, the association remained strong. Threshold analysis indicated that a bias factor of 1.29 would be required to reduce the observed ACE effect to a weak association (OR = 1.10). Detailed results are presented in Table 8.

Table 8.

Sensitivity analysis results

Analysis type	Effect value	P
Binary logistic regression	OR(95%CI)
Unadjusted model	1.42(1.30 ~ 1.55)	< 0.001
Adjusted chronic conditions	1.38 (1.26 ~ 1.51)	< 0.001
Fully adjusted model	1.22 (1.11 ~ 1.35)	< 0.001
Sensitivity analysis
E-value analysis	E-value (lower bound of 95% CI)
E-value analysis	2.19 (1.92)	-
Missing data simulation	Simulation of OR(Simulation of 95CI)
High-risk missing	1.42 (1.32,1.51)	< 0.001
Low-risk missing	1.33 (1.23,1.24)	< 0.001
Random missing	1.37 (1.26, 1.49)	< 0.001
Boundary analysis	OR
Highly optimistic of ACEs	0.85	> 0.05
Optimistic boundary of ACEs	1.14	< 0.05
Baseline estimate of ACEs	1.42	< 0.05
Conservative boundary of ACEs	1.77	< 0.05
Extreme boundaries of ACEs	2.13	< 0.05
Bias factor	1.42	-

Discussion

This study utilized data from the 2024 survey on the health status and utilization of health services among rural residents in Ningxia and revealed a significant association between ACEs and chronic diseases, their comorbidities, and health poverty vulnerability among middle-aged and elderly people in Ningxia. Chronic diseases and their comorbidities partially mediate the relationship between ACEs and health poverty vulnerability. The results of the Sobel and bootstrap tests support the results of the stepwise regression method, indicating that the results are robust.

The positive association between ACEs and health poverty vulnerability

This study supported Hypothesis 1 by demonstrating that ACEs positively impact health poverty vulnerability. Several potential mechanisms support this relationship. For example, parental divorce during childhood can reduce family income, affect physical health during childhood, and increase the risk of health poverty vulnerability in adulthood [36]. During childhood, parental disabilities cause family members to endure a greater risk of health poverty for a long period due to the burden of disability and population support [37]. The education level of parents also affects their children’s income; the lower the parents’ education level during childhood is, the lower the children’s future income is to some extent [38, 39]. Additionally, 15.9% of the global disability-adjusted life years can be attributed to childhood malnutrition caused by hunger, which also results in significant economic losses [40]. For example, the total health cost associated with ACEs in Europe is $581 billion annually, and in North America, it is $741 billion, which is approximately 3% of each region’s GDP [41].

The positive association between aces and chronic diseases and their comorbidities

ACEs are closely associated with the occurrence of chronic diseases and their comorbidities, confirming Hypothesis 2. The relationships between ACEs and chronic diseases and their comorbidities can be explained through various biological, psychological, and behavioural mechanisms. According to research, for example, ACEs encourage overactivation of the hypothalamic‒pituitary‒adrenal (HPA) axis in the neuroendocrine system, which results in the release of more adrenocorticotropic hormones. This, in turn, triggers the release of excessive cortisol, which ultimately leads to metabolic dysfunction, an increase in body fat, and the promotion of obesity [42]. Prtrov et al. reported that childhood abuse can promote the development of hypertension in adulthood through increased levels of inflammatory markers (C-reactive protein (CRP), IL-6, and fibrinogen), which are important mediating factors [43]. Notably, ACEs have an intergenerational transmission effect. Mothers with a history of childhood abuse or neglect have a greater risk of their children becoming obese, and there is a dose‒response relationship between the two. Additionally, these cases often occur in families with low social status and high levels of family conflict [44, 45]. In contrast, even in the context of childhood adversity, positive childhood experiences can reduce the levels of biomarkers associated with chronic diseases in individuals [46]. Research has indicated that insufficient quality and quantity of the childhood diet is an important cause of gastrointestinal immune and digestive function deficiencies in children, leading to gastrointestinal diseases in adulthood [47]. In contrast, a sufficient and well-rounded food supply during childhood can provide individuals with a solid nutritional foundation, thereby preventing stunted physical and cognitive development in children due to malnutrition, as well as the onset of various chronic diseases such as diabetes and obesity in adulthood [48]. Psychologically, ACEs can lead to the onset of depression, and depression significantly increases the risk of cardiovascular diseases [49]. ACEs and associated early adverse emotions can alter original psychological development, leading to psychological disorders, triggering biological stress responses, and prompting the HPA axis to secrete excessive cortisol, thereby increasing the risk of individuals developing depression and anxiety [50]. In terms of behaviour, ACEs increase the likelihood of these individuals engaging in negative behaviours, such as smoking, drinking, and staying up late. Some studies have indicated that there is a correlation between ACEs and adult smoking behaviour, possibly because the psychoactive properties of nicotine can alleviate the stress caused by ACEs. Additionally, ACEs have altered the traditional Chinese diet, leading to insufficient intake of fruits, vegetables, and whole grains while increasing the consumption of high-salt and high-sugar foods [51]. These unhealthy eating habits, which lead to poor weight control and low dietary quality, are significant risk factors for obesity, hypertension, and cardiovascular diseases. In addition, sleep problems closely related to health are also affected by ACEs. Research has shown that ACEs are closely related to the deterioration of sleep quality, and this impact persists into old age [52]. The various pieces of evidence above demonstrate that the occurrence of multiple ACEs increases the incidence of chronic diseases such as obesity, hypertension, cardiovascular diseases, and diabetes [53]. Notably, the prevalence of chronic diseases and the quantity of ACEs to which individuals are exposed exhibit a strong dose‒response relationship [54]. Research has shown that, compared with individuals without ACEs, those who have experienced four or more ACEs have a 2–4 times greater risk of developing chronic health problems [9].

The positive association between chronic diseases and their comorbidities and health poverty vulnerability

Chronic diseases and their comorbidities are related to health poverty vulnerability, which supports Hypothesis 3. In fact, this is indeed the case. Chronic diseases and their comorbidities are closely related to the health care expenditures of patients and their families. As a result, these groups and their families tend to have high utilization rates of health care services and a high proportion of medical expenses relative to their overall household expenditures [55]. Notably, as the number of chronic diseases increases, the frequency of health care services also increases. Iranian scholars have reported that in low- to middle-income economic environments, the more chronic diseases individuals aged 45–75 have, the greater their hospitalization risk and the more frequent their hospitalizations are [56]. Similarly, in high-income countries such as the UK, the number of comorbidities is closely related to the use of medical services such as hospitalization [57, 58]. Long and difficult-to-cure sickness cycles in patients with chronic diseases and comorbidities necessitate many medical visits and may even result in unnecessary medical treatment and polypharmacy, which further increases the financial burden of health care costs. A nationwide survey in China revealed that in individuals with diabetes, the likelihood of a household incurring catastrophic health expenditure (CHE) increases by 39% with each additional chronic disease [59]. A study on elderly individuals aged 60 years and above in Beijing, China, who participated in the Urban Employee Basic Medical Insurance (UEBMI) programme revealed that, compared with spending on a single chronic disease, health care expenditures increased by 3.4 times and 5.3 times for those with two and three chronic diseases, respectively. Additionally, health care expenditures for individuals with multiple chronic conditions account for 95% of the total. Chronic diseases and their comorbidities create a high demand for medical and health services among rural elderly individuals, but owing to the overall low economic level in rural areas, these individuals lack the ability to pay, ultimately leading them into a vicious cycle of poverty and illness [60]. On the other hand, this study hypothesizes that when chronic diseases and comorbidities among middle-aged and elderly individuals in a household worsen, the household labour force decreases or weakens, leading to a reduction in total household income. The most worrisome aspect is that, within a family, the costs of health care for middle-aged and older people unavoidably impede investments in the health and education of children, which impedes the next generation’s advancement in terms of health and personal growth and ultimately results in the intergenerational transfer of poverty [61, 62]. In summary, chronic diseases and the economic burden they create are the main causes of health poverty [63].

The mediating role of chronic diseases and their comorbidities

The results of the Sobel and bootstrap tests indicate that chronic diseases and their comorbidities partially mediate the relationship between ACEs and health poverty vulnerability. Specifically, the direct effect of ACEs on health poverty vulnerability was 0.0177, the total effect was 0.0200, and the indirect effect mediated by chronic diseases and their comorbidities was 0.0023. The mediation analysis reveals that chronic diseases and their comorbidities significantly mediated the relationship between ACEs and health poverty vulnerability, though the effect size was small. This finding suggests that while this mediating pathway exists, it is not the primary mediating route. The reasons for this may be related to several factors. First, the small effect size indicates that other more significant mediating pathways are at play. The pathway of chronic diseases and their comorbidities represent only the tip of the iceberg. ACEs may collectively influence health poverty vulnerability through multiple channels, including psychological factors (such as anxiety, depression, and chronic stress) [16, 64], behavioural and lifestyle factors (such as smoking and alcohol consumption) [65], and socioeconomic status (such as fragile social networks) [66]. Secondly, the relatively minor effects of chronic disease comorbidity further confirm that health poverty vulnerability is a complex state determined by multiple factors [67–69]. While the impact of a single chronic disease comorbidity may be small, when combined with other coexisting risks, the cumulative effects can exceed a certain threshold, paradoxically heightening sensitivity to health poverty. Therefore, from the perspective of public health and population prevention, identifying and intervening in each “small-effect pathway” represents a long-term strategy for reducing population vulnerability. Finally, the study design and measurement tools may have limited the estimation of effect sizes. First, the cross-sectional design cannot definitively establish the temporal sequence between adverse childhood experiences and chronic disease onset, lacking conclusive evidence for causality. Additionally, both adverse childhood experiences and chronic disease status were obtained through respondents’ recollections and self-reports, introducing potential information bias. In summary, while the mediating effect of chronic diseases and their comorbidities is small, it reveals the complexity and multifactorial nature of health poverty vulnerability stemming from adverse childhood experiences. This finding also reminds healthcare professionals that, in addition to effective clinical interventions, comprehensive, life-course-spanning public health strategies must be implemented to effectively curb the intergenerational transmission of health poverty vulnerability.

Modulating effects of depression

The results of the moderated mediation model indicate that ACEs exert a mediating effect on health poverty vulnerability through chronic diseases and their comorbidities, but this pathway is not moderated by depression. Chronic diseases and their comorbidities play a stable mediating role in the association between ACEs and health poverty vulnerability, while depression may function as an independent risk factor rather than a moderator. This finding may be attributed to the following reasons. First, ACEs cause intrinsic physiological attrition through bio-embedded effects that accumulate over time, independent of subsequent psychological states. As a form of childhood life injury, ACEs may exert long-term effects on biological aging processes and health status through overactivation of the hypothalamic-pituitary-adrenal axis and elevated inflammation levels [70]. Once established, this biologically embedded mechanism produces adverse health outcomes with high persistence and stability, making them resistant to modification by adult psychological states [71]. The chronic diseases and their comorbidities examined in this study may represent one manifestation of such progressive damage, with their association to ACEs remaining consistent regardless of fluctuating depression levels. Furthermore, depression and chronic disease comorbidity may both be independent outcomes of ACEs, ultimately converging to influence health poverty vulnerability rather than acting as moderators within this pathway. Research findings indicate that ACEs independently predict the risk of developing depression and chronic conditions such as heart disease [10].

Sensitivity analysis

Multiple sensitivity analyses consistently indicate that the positive association between ACEs and health poverty vulnerability exhibits robust stability. E-value analysis demonstrates that to completely invalidate our findings, an unmeasured confounder must exist that correlates with both ACEs and health poverty vulnerability, with an association strength of at least RR = 2.19. However, such confounders are rare in practical scenarios. Subsequent missing data simulations and boundary analyses further confirmed this, showing the association remained statistically significant under various plausible missing data scenarios. However, the final threshold analysis provided a more rigorous perspective. It revealed that a moderately strong unmeasured confounder (bias factor of 1.29) could reduce the observed effect to 1.10. At the public health level, such unmeasured factors are plausible. In summary, while the association between ACEs and health poverty requires a strong unmeasured confound to be statistically negated, some moderately strong unmeasured factors in practice would weaken this association. Therefore, these findings suggest the necessity of using more comprehensive and complete data to control for potential confounders in future research.

Limitations

The following limitations should be noted for this study. First and foremost, the cross-sectional design restricts inferences about causal relationships between variables. The analysis of pathways—specifically whether adverse childhood experiences precede chronic disease onset, thereby further influencing health poverty vulnerability—is primarily grounded in life course theory and prior longitudinal research evidence [72–74]. While this temporal sequence has biological and epidemiological plausibility, the study’s data cannot rule out reverse causality or complex bidirectional relationships. For instance, chronic illness during childhood may impose financial burdens on families due to medical needs, thereby increasing vulnerability to adversity. Future prospective cohort studies tracking participants from childhood are thus needed to clarify temporal sequences and causal relationships. Second, core variables—including adverse childhood experiences and chronic disease status—rely on self-reporting, introducing measurement error. Regarding childhood adversity, recall bias exists due to the passage of time and interference from current health status, emotional state, and social expectations. Chronic disease status may also be misclassified due to access to medical care and daily disease awareness. These non-differential biases may underestimate the true strength of the association between variables. To mitigate this limitation, future research should integrate multiple data sources—such as outpatient records and physiological indicators—to conduct multi-source validation of key variables. Thirdly, although this study subsequently used the DAG method to filter out covariates, there may still be potential confounding variables that could interfere with the results. Fourth, we excluded some samples with missing key variables during the data cleaning process, which may also introduce some estimation bias. Finally, a significant limitation of this study lies in the adaptation of the ACEs scale. To better suit the research participants and context, we modified several items through translation and cultural adaptation adjustments. Although these modifications were guided by experts, the revised scale has not yet been empirically validated within the sample of this study. This may introduce measurement error and potentially weaken relationships associated with this variable. Therefore, future research should prioritize formally validating the adapted ACEs scale within our studied population. This process should include: (1) Confirming the hypothesized factor structure through confirmatory factor analysis; (2) Assessing internal consistency (e.g., Cronbach’s alpha) and test-retest reliability; and (3) Evaluating construct validity by examining correlations with other established scales (e.g., discriminant validity). This will ensure robust and accurate measurement in future research.

Conclusion

In this study, the health poverty vulnerability, chronic diseases and their comorbidities, and ACEs of middle-aged and older adults in rural Ningxia were examined. Our findings indicate that ACEs are significantly correlated with chronic diseases, comorbidities, and health poverty vulnerability among middle-aged and elderly individuals in rural Ningxia. Chronic diseases and their comorbidities mediate the relationship between ACEs and health poverty vulnerability. Sensitivity analysis also supports these two points. Therefore, relevant governments and personnel should consider the mechanisms of alleviating adverse childhood experiences and chronic diseases and their comorbidities when formulating strategies to improve the health poverty situation among rural middle-aged and elderly populations.

Authors’ contributions

All the authors contributed to the study conception and design. Y X: Funding acquisition, Investigation, Methodology, Writing-review and editing and Project administration. H Q: Funding acquisition, Investigation, Methodology, Writing-review and editing, supervision and Project administration. J C: Conceptualization, Writing-original draft, Writing-review and editing and Software. F Z: Conceptualization, Writing-original draft, Writing-review and editing and Software. J Y: Formal analysis, Writing-review and editing and Methodology. H M: Supervision, Writing-review and editing. All authors contributed to the article and approved the submitted version. The first draft of the manuscript was written by J C and F Z, and all the authors commented on previous versions of the manuscript. All the authors have read and approved the final manuscript.

Funding

This research was supported by the National Natural Science Foundation of China (grant number 72264032,72164033),the Ningxia Medical University Institutional Research Project (grant number XT2022012),the Open competition mechanism to select the best candidates for key research projects of Ningxia Medical University (grant number XJKF240314) and the 2024 Autonomous Region Youth Science and Technology Talent Cultivation Project(grant number Ningxia Association for Science and Technology Document No. 2 (2025)).

Data availability

General correspondence and requests for source data and materials should be addressed to J. Chang. Requests for access to data should be addressed to J. Chang ( changjianhua0210@163.com).

Declarations

Ethics approval and consent to participate

The researcher obtained the respondents’ informed consent before conducting this study. Our research follows the Declaration of Helsinki.This study was approved by the Ethics Committee of Ningxia Medical University (approval number: 2022-G165).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.