Mediating effects of social and intellectual activities on cognitive decline caused by hearing loss: insights from the China health and retirement longitudinal study (CHARLS)

Abstract

Background

Hearing impairment is increasingly recognized as a pivotal determinant of cognitive function. As the global population ages, the urgency to pinpoint and leverage modifiable risk factors to alleviate cognitive deterioration intensifies.

Objective

This investigation seeks to delineate the correlation between auditory deficits and cognitive capabilities, especially examining the intermediary influence of social and intellectual engagements on this dynamic.

Methods

This research utilized cross-sectional data from the 2015 iteration of the China Health and Retirement Longitudinal Study (CHARLS). A multiple linear regression model was utilized to assess the relationship between hearing loss and cognitive function. Furthermore, mediation analysis was performed to investigate the mediating role of social and intellectual activities in this association.

Results

The analysis revealed a pronounced negative association between hearing loss and cognitive function (B = -0.531, 95% CI: -0.658 to -0.390). After adjusting for social and intellectual activities, hearing loss was found to be 92.15% of the total observed effect on cognitive decline. In the mediation analysis, social and intellectual activities were found to mediate approximately 7.85% of the relationship between hearing loss and cognitive decline.

Conclusion

The findings highlight the association between hearing loss and reduced cognitive function and suggest that greater engagement in social and intellectual activities may be related to better cognitive outcomes. These results support the potential value of enhancing hearing screening and promoting activity engagement among older adults.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12883-025-04353-8.

1. Introduction

Hearing loss is one of the most common sensory impairments in the elderly, with profound impacts on auditory and linguistic functions [1]. Beyond its frequent co-occurrence with dementia in aging populations, hearing loss has been established as a significant risk factor for the onset of cognitive decline [2–4]. The risk of dementia is notably elevated in individuals with hearing loss, with severity correlating with increased risk [4]. The Lancet has highlighted hearing loss as the foremost modifiable risk factor for dementia, accounting for 9% of all preventable dementia risks [5]. While the association between hearing loss and an elevated risk of dementia is well-established, the underlying mechanisms remain insufficiently understood. It is posited that social and intellectual engagement may act as crucial mediators in this relationship, potentially mitigating the cognitive decline associated with hearing loss.

Age-related cognitive decline is a common feature of the aging process, often accompanied by other sensory impairments such as presbycusis, or age-related high-frequency hearing loss [6]. Presbycusis typically begins in middle age and becomes more pronounced with advancing age, affecting a substantial proportion of the elderly population [7]. Studies estimate that approximately two-thirds of adults over the age of 70 experience some degree of hearing loss, many of whom are unaware or undiagnosed [8]. This age-related hearing impairment has been associated with reduced auditory input and diminished cognitive stimulation, contributing to an increased risk of cognitive decline [5]. Importantly, the severity of hearing loss has been shown to correlate with the level of social disengagement, as individuals with more pronounced hearing difficulties are more likely to withdraw from conversation and group settings [9]. Moreover, hearing aids, which can partially restore auditory input, may play a mitigating role in this relationship by facilitating communication and enabling greater social and intellectual engagement [10]. However, hearing aid adoption remains low, especially in low- and middle-income countries, due to cost, stigma, and access barriers [11].

Social and intellectual activities, which are known to promote cognitive health by enhancing cognitive reserve [12], are often compromised in individuals with hearing impairment due to communication difficulties. This reduction in engagement may contribute to accelerated cognitive decline [13]. Older adults with hearing impairments frequently experience anxiety and embarrassment stemming from communication challenges, which often result in social withdrawal and reduced participation in social and intellectual activities—such as meeting with friends, attending volunteer events, playing mahjong or chess, and engaging in educational or training programs—as documented in the CHARLS dataset and supported by prior literature [14–16]. Furthermore, engagement in such activities is universally acknowledged as a crucial determinant of cognitive reserve, playing a significant role in mitigating the risk of dementia [17, 18]. Consequently, social and intellectual engagement is viewed as a critical mediator in the relationship between hearing loss and cognitive function.

Although the present study does not include participants diagnosed with dementia, it is situated within a broader public health context where cognitive decline is a major concern. Dementia currently affects more than 30 million individuals globally and is projected to rise to over 150 million by 2050, particularly in low- and middle-income countries [19, 20]. It is characterized by a profound decline in cognitive abilities, manifesting as a deterioration in key cognitive functions such as memory, attention, processing speed, and reasoning [17]. This cognitive decline has become a significant public health issue, burdening individuals, families, and societies [17]. Identifying modifiable factors like hearing loss and understanding their mechanisms could play a pivotal role in developing preventive strategies for cognitive decline, particularly in aging populations.

Currently, the intermediary role of social and intellectual activities in modulating the impact of hearing loss on cognitive function remains insufficiently examined. This study aims to clarify the link between hearing impairment and cognitive function and to assess the mediating role of social and intellectual activities. By doing so, we hope to provide actionable insights for designing effective interventions to mitigate cognitive decline among the elderly.

Methods

Participants

The China Health and Retirement Longitudinal Study (CHARLS) is a comprehensive, nationally representative survey designed to assess the health, socioeconomic status, and demographic characteristics of Chinese residents aged 45 and above. Employing a multistage, stratified, probability-proportional-to-size sampling technique, CHARLS has successfully gathered a representative cross-section of the general population. Biennial follow-up assessments provide ongoing insights into a variety of dimensions including health metrics and economic conditions. This study uses cross-sectional data from 21,095 participants collected in 2015, chosen for its larger sample size and broader variable coverage compared to other years. Although CHARLS is a longitudinal study, we utilized only the 2015 wave for a cross-sectional analysis. Thus, each participant appears only once in the dataset, and no repeated measures were included .Participation in CHARLS was entirely voluntary, and respondents provided informed consent with the option to withdraw at any time.

Participants and key variables were rigorously selected in accordance with established research protocols [5, 21]. Initially, participants lacking data on hearing loss were excluded (n = 1,413). Subsequently, those missing cognitive function assessments were also removed (n = 8,693). Moreover, individuals with missing covariate data, as detailed by the variables presented in Table 1, were excluded to maintain a dataset comprising only complete records (n = 4,773). For inclusion in the final analysis, all data points for each participant required completeness without any omissions. Ultimately, the analysis incorporated a total of 6,216 participants. The methodology for participant selection is depicted in Fig. 1. All interviews were conducted by trained field staff using standardized computer-assisted personal interview (CAPI) procedures, in accordance with CHARLS protocols. Cognitive assessments, self-reported hearing questions, and activity engagement measures were collected in face-to-face home interviews during the same structured session, typically on the same day.

Table 1.

Baseline characteristics by hearing status and univariate associations with cognitive Function(N = 6216)

Variables	Hearing status, Mean (SD), or N (%)			P-value (Group comparison)	β (95% CI)	P-value (Regression)
	Total(N = 6216)	Good(N = 2276)	Poor(N = 3940)
Age, (years)	59.70 (8.42)	58.18 (8.10)	60.58 (8.47)	< 0.001	-0.06(-0.07 - -0.05)	< 0.001
Social and intellectual activities	1.85 (2.27)	2.18 (2.50)	1.66 (2.10)	< 0.001	0.22(0.18–0.25)	< 0.001
BMI	24.19 (3.92)	24.36 (3.93)	24.09 (3.91)	0.012	0.05(0.03–0.07)	< 0.001
Cognitive function	12.74 (2.95)	13.25 (2.88)	12.45 (2.95)	< 0.001	-	-
Hearing status				-		< 0.001
Poor	–	–	–		–0.81(–0.95– –0.65)
Normal	–	–	–		Reference
Gender				0.524		< 0.001
Male	3273 (52.7)	1211 (53.2)	2062 (52.3)		0.48(0.34–0.63)
Female	2943 (47.3)	1065 (46.8)	1878 (47.7)		Reference
Education				< 0.001		< 0.001
Above HS	367 ( 5.9)	163 ( 7.2)	204 ( 5.2)		4.51(4.15–4.86)
HS	685 (11.0)	333 (14.6)	352 ( 8.9)		2.42(2.18–2.65)
Blow HS	4543 (73.1)	1586 (69.7)	2957 (75.1)		3.85(3.55–4.15)
Illiterate	621 (10.0)	194 ( 8.5)	427 (10.8)		Reference
Marital				0.494		< 0.001
Married	5623 (90.5)	2067 (90.8)	3556 (90.3)		0.87(0.62–1.12)
Single	593 ( 9.5)	209 ( 9.2)	384 ( 9.7)		Reference
Smoking				0.613		0.186
Yes	2805 (45.1)	1017 (44.7)	1788 (45.4)		0.09(–0.04– 0.24)
No	3411 (54.9)	1259 (55.3)	2152 (54.6)		Reference
Alcohol				0.006		< 0.001
Yes	2462 (39.6)	953 (41.9)	1509 (38.3)		0.41(0.26–0.56)
No	3754 (60.4)	1323 (58.1)	2431 (61.7)		Reference
Hypertension				< 0.001		0.257
Yes	1685 (27.1)	536 (23.6)	1149 (29.2)		–0.09(− 0.26– 0.06)
No	4531 (72.9)	1740 (76.4)	2791 (70.8)		Reference
Diabetes				0.097		0.096
Yes	493 ( 7.9)	163 ( 7.2)	330 ( 8.4)		–0.22(–0.50–0.04)
No	5723 (92.1)	2113 (92.8)	3610 (91.6)		Reference

Note: Continuous variables are presented as mean (SD); categorical variables as number and percentage, N (%). Group comparisons between good and poor hearing groups were conducted using Student’s t–tests or chi–square tests, as appropriate. The β coefficients and corresponding 95% confidence intervals (CIs) were derived from univariate linear regression models, representing the association of each variable with cognitive function (as a continuous outcome).Bold font indicates statistical significance (P < 0.05)

Abbreviations: N, number; SD, standard deviation; HS, high school; BMI, body mass index; –, not available

Fig. 1.

Flowchart of Participant Inclusion for the China Health and Retirement Longitudinal Study (CHARLS)

Measures

Dependent variables

Cognitive function was evaluated in the 2015 survey through in-person assessments spanning four domains: orientation, computation, memory, and drawing. Orientation was gauged with a maximum score of 5 points, assessing participants’ awareness of the current year, month, day, weekday, and season, awarding 1 point for each correct response. The computation section required participants to subtract 7 from 100 consecutively five times, with each correct subtraction earning 1 point, culminating in a possible total of 5 points. For the memory test, participants were initially asked to recall as many words as possible from a list of 10 read aloud by the interviewer (immediate recall). Following the orientation, computation, and drawing tests, they were prompted to recall the words again (delayed recall). Memory scores were calculated as the average of the immediate and delayed recall scores, on a scale from 0 to 10. In the drawing test, participants were tasked with accurately replicating two overlapping pentagons, with a correct depiction scoring 1 point. The composite cognitive function score, encompassing all domains, ranged from 0 to 21, with higher scores indicating superior cognitive abilities [22].

The cognitive assessments used in CHARLS are adapted from the Health and Retirement Study (HRS) and have been widely used and validated in large-scale aging research, both in Western and Chinese populations. Prior studies have demonstrated acceptable internal consistency, test–retest reliability, and construct validity for these measures in middle-aged and older adults [23, 24]. These cognitive domains—orientation, memory, computation, and drawing—are considered effective proxies for screening general cognitive function in epidemiological surveys.

Independent variables

Objective audiometric measures such as pure-tone audiometry were not collected in the 2015 wave of CHARLS. Therefore, hearing status in this study was based on self-reported data. Participants were asked: “Would you say your hearing is excellent, very good, good, fair, or poor?” Those who reported their hearing as ‘Fair’ or ‘Poor’ were categorized as experiencing hearing loss, whereas ‘Excellent,’ ‘Very good,’ and ‘Good’ were considered indicative of normal hearing.

While self-reported hearing status offers a practical means of assessment in large-scale surveys, it is important to acknowledge its limitations. Studies have shown that self-assessments may either underestimate or overestimate actual hearing loss compared to objective audiometric testing. For example, Curti et al. found that the correlation between self-reported hearing and pure-tone audiometry ranged from 43 to 71%, with variation influenced by age and sex [25]. Similarly, Choi et al. reported that among 14,345 participants, 13.1% underestimated and 5.1% overestimated their hearing impairment [26]. These discrepancies indicate a risk of misclassification, which may introduce measurement bias into the analysis.

Mediator variables

Social and intellectual engagement was assessed through closed-ended questions, where participants responded to a predefined list of specific activities included in the CHARLS questionnaire. Within the Health Status and Function module of the CHARLS survey, social and intellectual engagements over the preceding month were evaluated across four distinct categories [18]. Social activities encompassed interactions such as meeting with friends, participating in sports or social clubs (e.g., dancing, fitness), engaging in community organizations, and volunteer or charity work. Intellectual activities included playing strategic games like mahjong or chess, attending educational or training courses, making stock investments, and using the Internet. The frequency of each activity was quantified with a scoring system: ‘Never’ (score = 0), ‘Occasionally’ (score = 1), ‘Almost weekly’ (score = 2), and ‘Almost daily’ (score = 3). The scores were aggregated based on the frequency of activities, resulting in total scores that ranged from 0 to 24 for both social and intellectual engagements. The complete list of social and intellectual activities assessed in CHARLS is provided in Supplementary Table S2.

Covariates

Potential confounders and mediators were delineated from existing research and integrated into a Directed Acyclic Graph (DAG), which informed the analytical strategy depicted in Fig. 2 [5, 18, 21]. The full list of covariates included in all regression and mediation models were: age, gender, marital status (married or not), educational level (illiterate, below high school, high school, and above high school), smoking status (no/yes), alcohol consumption (no/yes), body mass index (BMI), hypertension (no/yes), and diabetes (no/yes). These variables were selected based on their established associations with both hearing and cognitive function and are detailed in Table 3.

Fig. 2.

Directed Acyclic Graph. The directed acyclic graph (DAG) delineates the interrelations among covariates, primary exposure factors, and outcomes. Exposure factors are highlighted in yellow boxes, while orange boxes identify confounders that influence both exposure and outcomes. Blue boxes mark the causal determinants of the outcome variable. Causal pathways are represented by green lines, and purple lines depict pathways indicative of bias

Table 3.

Relationship between hearing loss, social and intellectual activities, and cognitive function

	Model 1					Model 2 (with mediator)
	B	SE	β	P–value		B	SE	β	P–value
Hearing(reference = Normal)	–0.522	0.073	–0.086	< 0.001		–0.489	0.073	–0.079	< 0.001
Social and intellectual activities	–	–	–	–		0.104	0.016	0.080	< 0.001
Age	–0.041	0.005	–0.118	< 0.001		–0.039	0.004	–0.111	< 0.001
Gender(reference = female)	0.308	0.105	0.052	0.003		0.349	0.105	0.059	< 0.001
Marital status(reference = single)	0.302	0.122	0.030	0.013		0.307	0.121	0.031	0.011
Education(reference = illiterate)				< 0.001					< 0.001
Above HS	4.226	0.183	0.337			3.989	0.185	0.318
HS	3.373	0.156	0.358			3.242	0.156	0.344
Blow HS	2.187	0.119	0.329			2.136	0.119	0.321
Smoking(reference = no)	–0.205	0.099	–0.034	0.038		–0.231	0.099	–0.038	0.019
Alcohol(reference = no)	0.089	0.079	0.014	0.258		0.034	0.079	0.005	0.668
BMI	0.039	0.009	0.053	< 0.001		0.036	0.009	0.048	< 0.001
Diabetes(reference = no)	–0.166	0.130	–0.015	0.200		–0.177	0.129	–0.016	0.171
Hypertension(reference = no)	0.018	0.082	0.002	0.824		0.020	0.082	0.003	0.807

Note: All models were adjusted for age, gender, BMI, marital status, education, alcohol consumption, smoking, hypertension, diabetes, and (where applicable) social and intellectual activities. Model 1 examines the relationship between hearing loss and cognitive function after adjusting for these covariates. Model 2 explores this relationship with additional adjustments for mediating factors

Abbreviations: B, unstandardized coefficient; SE, standard error; β, standardized coefficient; –, not available

Statistical analysis

Statistical analyses were executed utilizing R version 4.4.0 in the RStudio computing environment. The Directed Acyclic Graph (DAG) was constructed with the aid of DAGitty (DAGitty). All continuous variables were expressed as means with standard deviations (SDs), and categorical variables as percentages.

Group comparisons of baseline characteristics between participants with good and poor self-reported hearing were conducted using Student’s t-tests or Chi-square tests, as appropriate. Univariate linear regression analyses were then used to examine the association between each independent variable and cognitive function, with results presented in Table 1.

To further assess the relationships among key study variables, Spearman correlation coefficients were calculated and reported in Table 2. To address potential Type I error due to multiple comparisons, we applied the Benjamini–Hochberg false discovery rate (FDR) correction to the corresponding p-values. A mediation model, as proposed by Baron and Kenny [27] was utilized to examine the mediating effects of social interaction and intellectual activities on the nexus between hearing loss and cognitive function. Furthermore, a non-parametric bootstrap approach with 1,000 resamples was used to evaluate the total, indirect, and direct effects [28]. Statistical significance of the indirect (mediating) effect was determined using the mediate() function from the R mediation package. A mediation effect was considered significant if the 95% bootstrap confidence interval for the indirect effect (a × b) did not include zero. This approach provides a robust and distribution-free test of mediation and is widely recommended over traditional stepwise methods. All regression models were adjusted for age, gender, marital status, education level, smoking, alcohol consumption, BMI, hypertension, and diabetes. The results for all covariates are presented in Table 3. Regression analysis outcomes are presented as unstandardized (B-values) and standardized coefficients (β-values). B-values represent the true impact of independent variables on the dependent variable in their original units, whereas β-values facilitate the comparison of variables across different scales, reflecting their relative importance in the model.

Table 2.

Pairwise spearman correlations between hearing status, social and intellectual activities, and cognitive function

Comparison	Spearman’s ρ	P–value	FDR–adjusted p–value
Hearing status ~ Cognitive function	–0.150	< 0.001	< 0.001
Hearing status ~ Social and intellectual activities	–0.117	< 0.001	< 0.001
Social and intellectual activities ~ Cognitive function	0.188	< 0.001	< 0.001

Note: Spearman correlation coefficients (ρ) are reported for pairwise associations among hearing status, social and intellectual activities, and cognitive function. P-values were adjusted using the Benjamini–Hochberg method to control the false discovery rate (FDR). All correlations were statistically significant at P < 0.001 (two-tailed)

To evaluate whether the inclusion of social and intellectual activities improved the overall explanatory power of the model, a model comparison using an F-test was conducted. The results demonstrated that incorporating the mediators significantly improved model fit (F = 43.295, p < 0.001), and full results are presented in Supplementary Table S4. Multicollinearity among predictors was assessed using Variance Inflation Factors (VIF). All VIF values were below 2.5, indicating no serious multicollinearity. Detailed VIF results are provided in Supplementary Table S3.To ensure the robustness of the results, subgroup analyses and sensitivity analyses were also conducted. Statistical significance was established at a p-value of < 0.05 (two-tailed).

Results

Participant characteristics

At baseline, 6,216 participants were assessed in the study, with an average age of 59.70 ± 8.42 years (Table 1). Of these, 63.4% were identified as having poor hearing. The average cognitive function score among all participants was 12.74 ± 2.95, while the average score for social and intellectual activities stood at 1.85 ± 2.27. Males comprised 52.7% of the sample, and 83.1% of the participants had received an education level below high school. Relative to those with normal hearing, individuals with poor hearing were more likely to have hypertension, a lower BMI, and a lesser propensity for alcohol consumption (all P < 0.05). No significant differences were observed between the groups concerning diabetes, gender, marital status, or smoking status (all P > 0.05).

Univariate linear regression analyses indicated that all of the following variables—age, BMI, gender, educational level, marital status, alcohol consumption, hearing status, and engagement in social and intellectual activities—were significantly associated with cognitive function (P < 0.05), as shown in Table 1. All of these covariates were subsequently included in the multivariable regression models to control for potential confounding effects.

Correlation among key variables

The correlation analysis presented in Table 2 delineates the relationships between hearing loss, social and intellectual activities, and cognitive function. The analysis indicates that hearing loss is inversely associated with both social and intellectual activities (Spearman’s ρ = –0.117, FDR–adjusted P < 0.001) and cognitive function (ρ = –0.150, FDR–adjusted P < 0.001). Conversely, social and intellectual activities demonstrate a positive correlation with cognitive function (ρ = 0.188, FDR–adjusted P < 0.001).

Association between hearing and cognitive function

Based on the data presented in Table 3, findings from Model 1 highlight a substantial negative association between hearing loss and cognitive function (β = –0.086, p < 0.001), after adjusting for age, gender, marital status, education level, smoking, alcohol consumption, BMI, hypertension, and diabetes. Model 2 further incorporates social and intellectual activities as mediators, and the association remains robust (β = –0.079, p < 0.001), indicating a partial mediating effect. Notably, the association between hearing loss and cognitive function remained statistically significant even after controlling for all sociodemographic and health–related covariates. Bootstrap analysis underscores that the overall impact of hearing loss on cognitive function is significant, with a total effect of –0.531 (95% CI: –0.658 to –0.390) as depicted in Fig. 3a. After adjusting for social and intellectual engagements, the effect size of hearing loss on cognitive function diminishes slightly to –0.489 (95% CI: –0.616 to –0.350), shown in Fig. 3b, accounting for 92.15% of the total effect. The indirect mediation effect via social and intellectual activities, calculated at –0.042 (95% CI: –0.059 to –0.030), confirms significant mediation, contributing to 7.85% of the total effect, as detailed in Table 4.

Fig. 3.

Mediation model of hearing loss, social and intellectual activities, and cognition. Note: a, represents the path coefficient of hearing loss on social and intellectual activities; b, represents the path coefficient of social and intellectual activities on cognitive function. Path coefficients are presented as unstandardized regression coefficients. * P < 0.05 (two–tailed); ** P < 0.01 (two–tailed); *** P < 0.001 (two–tailed)

Table 4.

Mediation hypothesis model of social and intellectual activities in the relationship between cognition and hearing loss

Pathway	Effect	BootLLCI	BootULCI	p–value	Proportion of effect (%)
Total effect (c)	–0.531	–0.658	–0.390	< 0.001	100%
Direct effect (ca.)	–0.489	–0.616	–0.350	< 0.001	92.15%
a	–0.400	–0.514	–0.287	< 0.001	–
b	0.104	0.073	0.135	< 0.001	–
Indirect effects (a*b)	–0.042	–0.059	–0.030	< 0.001	7.85%

Note: The model controls for age, gender, education level, marital status, smoking status, alcohol consumption, BMI, hypertension, and diabetes

Abbreviations: Effect, unstandardized regression coefficient; BootLLCI, bootstrap lower limit confidence interval; BootULCI, bootstrap upper limit confidence interval, –, not available

Additionally, further subgroup analysis by age, detailed in Table S1, reevaluates the mediating role of social and intellectual activities in the interplay between hearing loss and cognitive function. The findings remain significant for participants under the age of 75, underscoring the potential critical role of these activities in reducing the adverse effects of hearing loss on cognitive function, particularly among the elderly demographic.

Discussion

This research established a marked correlation between self–reported hearing loss and cognitive decline, revealing that individuals who reported experiencing hearing difficulties consistently exhibited lower cognitive function scores than those who reported normal hearing. Moreover, the analysis indicated that social and intellectual activities serve as partial mediators in this relationship. These findings highlight the observed associations between self–reported hearing loss and lower cognitive function, and suggest that participation in social and intellectual activities may attenuate this association. Although all correlations among the key variables were statistically significant (P < 0.001), the correlation coefficients were modest in magnitude (ρ ranging from − 0.12 to 0.19). This is likely due to the large sample size (N = 6,216), which increases statistical power and can lead to the detection of small effects as statistically significant. Therefore, the strength and direction of the associations, rather than the p–values alone, should be the focus when interpreting these results.

The findings of this study reinforce the significant link between hearing loss and cognitive function, suggesting that hearing loss may act as an important correlate of cognitive decline in older adults. Even after adjusting for covariates, the association with cognitive impairment remained robust, aligning with previous research [29–31]. For instance, Michalowsky et al. [32], demonstrated that hearing loss of any kind is strongly associated with heightened cognitive decline risk, potentially mediated through specific mechanisms. Another cross–sectional study [33], which analyzed data from 1,335 participants, demonstrated that after adjusting for confounding factors such as age, gender, hypertension, diabetes, and other factors, the risk of dementia in individuals with hearing loss was 2.66–fold higher than in those with normal hearing. While variations in sample characteristics and methodologies exist among studies, the evidence consistently points to hearing loss as a factor that accelerates cognitive decline. This study strengthens this hypothesis by demonstrating that even with multiple adjustments, the link between hearing loss and cognitive function persists. Furthermore, these results not only corroborate prior research but also expand the evidence base using a distinct population, underscoring the pivotal role of hearing loss in cognitive health.

Social and intellectual activities serve as critical mediators in the relationship between hearing loss and cognitive function. This finding aligns with prior research, which has established links between hearing loss and diminished social and intellectual engagement [20, 34, 35]. and between these activities and cognitive function [12]. However, earlier studies have inadequately explored their mediating mechanisms. This study offers fresh insights, suggesting that hearing loss may indirectly impair cognitive function in older adults by reducing engagement in these activities. Our hypothesized model (DAG) illustrates that hearing impairment induces communication difficulties, leading to reduced social participation, which may further result in social isolation and diminished intellectual activity [35, 36]. Engagement in social activities may enhance cognitive function by providing increased cognitive stimulation, maintaining neuroplasticity, and reducing feelings of isolation, which are known contributors to cognitive decline [18]. Encouraging participation in social and intellectual activities may serve as a promising avenue for supporting cognitive health in individuals with hearing loss, although causality remains to be confirmed through longitudinal studies. Importantly, the mediating effect of social and intellectual activities was not inferred solely from changes in regression coefficients. Rather, we employed a formal mediation analysis using the mediate() function from the R mediation package with 1,000 bootstrap simulations. The significance of the mediating (indirect) effect was determined by whether the 95% confidence interval for the product of the indirect paths (a × b) excluded zero. This non–parametric approach is robust and widely recommended in mediation research, particularly when assumptions of normality may not hold [28, 37].

Communication challenges stemming from hearing impairment can create discomfort or anxiety in social contexts, thereby reducing interaction opportunities and, subsequently, participation in social and intellectual pursuits [38, 39]. Lower levels of social engagement are recognized as a risk factor for cognitive decline; prolonged social isolation can disrupt the brain’s social–cognitive networks, ultimately contributing to cognitive deterioration [40]. Research demonstrates that social and intellectual activities offer diverse and enriching environments, fostering engagement in complex, cognitively stimulating activities that help guard against cognitive decline, particularly in domains such as language, executive function, and overall cognition [18, 41]. Studies further indicate that animals in unstimulating environments may undergo structural alterations in brain components, including synapses, dendrites, axons, and glial cells [42]. These findings collectively support the view that social and intellectual disengagement may represent a behavioral pathway linking hearing loss to cognitive decline. Our findings reinforce the essential mediating role of participation in social and intellectual activities in the association between hearing loss and cognitive function.

This study, grounded in a large, nationally representative sample, strengthens the generalizability and relevance of the findings. By rigorously controlling for confounding variables such as age, gender, and education level, it provides a clearer insight into the relationships between hearing loss, social and intellectual activities, and cognitive function. Furthermore, by combining multiple forms of social and intellectual engagement into a single composite score, this study highlights the overall relevance of activity participation to cognitive health. From a public health standpoint, the findings highlight the crucial role of fostering social and intellectual engagement, especially for individuals with hearing impairments. To counteract cognitive decline, public health initiatives might focus on enhancing accessible social environments and expanding interaction opportunities for those with hearing loss. Moreover, the results advocate for incorporating diverse intellectual activities into elderly care services to decelerate cognitive deterioration. Encouraging and supporting older adults’ engagement in a broad array of social and intellectual activities could thus emerge as an effective strategy for maintaining cognitive health, helping to reduce the public health burden of cognitive decline in an aging population.

Nevertheless, it is important to emphasize that the present study was based on cross–sectional data. As such, the observed associations should not be interpreted as causal relationships. The observed relationships reflect correlations only, and the temporal direction between variables cannot be determined. Therefore, the potential bidirectionality of associations—such as whether cognitive decline affects self–reported hearing or activity engagement—must be acknowledged. Mediation analysis within cross–sectional data is limited by constraints related to time sequence and causality, underscoring the need for validation in future longitudinal studies. Additionally, hearing loss data primarily rely on participants’ self–reports. While prior studies suggest a degree of reliability in self–reported hearing status [43], this approach may introduce information bias, especially in the underreporting of mild hearing loss. Although several studies have validated the general reliability of self–reported hearing loss in large–scale surveys [15, 16], it is equally important to acknowledge empirical evidence showing significant discrepancies between subjective assessments and objective audiometric testing. For instance, Curti et al. and Choi et al. both demonstrated that a substantial proportion of participants misclassified their hearing ability [25, 26]. Such misclassification may lead to attenuation or inflation of the observed associations in this study, particularly in the case of mild hearing loss. While our large sample and robust analytical strategy help mitigate this issue to some extent, future research should aim to incorporate objective hearing assessments to enhance precision. Furthermore, as this study’s sample comprises exclusively middle–aged and older adults in China, its external validity may be limited due to geographical and cultural influences, complicating direct generalization to other age groups or regions. Although multiple confounding factors were controlled for in the analysis, constraints in the data source (CHARLS database) mean that some unmeasured confounders, such as lifestyle and dietary habits, may still impact the results. Future studies could integrate additional variables and adjust for a broader range of confounders to strengthen the robustness and generalizability of these findings.

Conclusion

The study investigates the interplay between hearing loss, social and intellectual activities, and cognitive function. The findings highlight a robust association between self–reported hearing loss and cognitive function, with social and intellectual activities serving as partial mediators. However, due to the cross–sectional nature of the data, causality cannot be inferred. These insights underscore the importance of further investigating the potential mediating role of social and intellectual engagement in this association. Future studies using longitudinal or experimental designs are necessary to establish temporal direction and explore causal pathways. While our findings suggest possible intervention targets, their effectiveness in preventing cognitive decline should be verified through prospective research. Future programs should prioritize hearing screening for middle–aged and older adults and actively promote their participation in these activities to preserve cognitive health, offering critical direction for advancing healthy aging and shaping effective public health initiatives.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CHARLS

China health and retirement longitudinal study

DAG

Directed acyclic graph

BMI

Body mass index

HS

High school

SE

Standard error

BootLLCI

Bootstrap lower limit confidence interval

BootULCI

Bootstrap upper limit confidence interval

Author contributions

This study was conceived by PJ and HY. Data curation and analysis were conducted by PJ and YL. HY oversaw the review, overall supervision, and management of the manuscript. All authors contributed to the drafting of the manuscript and have reviewed and approved the final version for submission.

Funding

This work was supported by the 2023 Fudan Clinical Scientist Training Program, the Excellent Doctors–Excellent Clinical Researchers Program (SYB202008), the Shanghai Science and Technology Committee (STCSM) Science and Technology Innovation Program (21Y31900500), and the National Natural Science Foundation of Shanghai (21ZR1411800).

Data availability

More details on accessing research data can be found on the CHARLS database website (https://charls.charlsdata.com/pages/Data/2015-charls-wave4/en.html).

Ethics approval and consent to participate

All procedures performed in studies involving human participants were in accordance with the Declaration of Helsinki. The original CHARLS was approved by the Ethical Review Committee of Peking University, and all the participants from CHARLS provided signed informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.