Hearing difficulties and memory problems: the mediating role of physical health and psychosocial wellbeing

Abstract Objective To investigate whether the association between hearing difficulties and self-reported memory problems is mediated by physical health and psychosocial wellbeing. Design A cross-sectional study. Path analyses were used to test potential theoretical models (psychosocial-cascade, common cause) of the association between hearing difficulties and memory problems, adjusting for age. Study Sample A sample of 479 adults (18–87 years) completed self-reported outcome measures. Results Half the participants reported clinically significant hearing difficulties and 30% self-reported memory problems. In the direct model, reporting hearing difficulties was associated with a greater likelihood of reporting memory problems (β = 0.17, p = 0.007, 95% Confidence Intervals [CI] = 0.00, 0.01). Hearing difficulties were also associated with poorer physical health, but this did not mediate the association with memory. Psychosocial factors, however, fully mediated the relationship between hearing difficulties and memory problems (β = 0.03, p = 0.019, 95% CI = 0.00, 0.01). Conclusions Adults with hearing difficulties may be more likely to self-report memory problems, irrespective of age. This study supports the psychosocial-cascade model, as the association between self-reported hearing and memory problems was explained entirely by psychosocial factors. Future studies should investigate these associations using behavioural measures, as well as explore whether interventions can reduce the risk of developing memory problems in this population.


Introduction
Dementia is a leading cause of disability globally and is growing in prevalence because of an ageing population.An estimated 57.4 million people worldwide were living with dementia in 2019 and this is expected to rise to 152.8 million by 2050 (GBD 2019Dementia Forecasting Collaborators, 2022).As a progressive disease without effective treatment or cure, prevention of dementia is of critical importance, not only for the individual but also to reduce the impact on healthcare and economic systems (Livingston et al., 2020).Untreated hearing loss is a highly prevalent condition, estimated to affect 30-48.7% of people over 50 years of age (Tsimpida et al., 2020), and is potentially one of the greatest modifiable risk factors for dementia in midlife, increasing dementia risk by 30-50%, with 8% of dementia prevalence estimated to be reduced if this single factor was eliminated (Livingston et al., 2020).In addition, hearing loss has been shown to be independently associated with other modifiable risk factors for dementia (Livingston et al., 2020), such as social isolation and loneliness (Maharani et al., 2019;Shukla et al., 2020), depression (Rutherford et al., 2018), and cardiovascular disease risk (CVD) factors (e.g.obesity, diabetes, hypertension, physical inactivity) (Besser et al., 2018;Kuo et al., 2021).However, there is conflicting evidence of the potential mediating and/or moderating role these factors may have on increasing dementia risk.
Several potential explanations have been posited for the association between hearing loss and dementia, which have been discussed in detail elsewhere (e.g.Powell et al., 2021;Wayne & Johnsrude, 2015).For example, the social cascade model proposes that the communication difficulties experienced by people with hearing loss can lead to reduced social interactions (Pichora-Fuller, 2015), increasing the risk of social isolation and loneliness, resulting in depression, insufficient cortical stimulation, and, subsequently, cognitive impairment/decline (Dawes et al., 2015).On the other hand, the information degradation model suggests that as auditory signals are degraded, individuals with hearing loss require additional cognitive effort for correct auditory processing (Peelle , 2018).The added effort is thought to divert resources from other cognitive processes, such as working memory and decision-making (Lin & Albert, 2014;Martini et al., 2014).In addition, this diversion of resources has been hypothesised to lead to impaired cognitive performance and, potentially, permanent changes in neuroplasticity and accelerated brain atrophy (sensory deprivation model) (Davis et al., 2016;Lin et al., 2014).Alternatively, the common-cause model posits that hearing loss and cognitive decline are not necessarily part of a causal pathway, but are both expressions of widespread neurodegeneration, where hearing loss presents first, which may be linked to vascular risk factors and microvascular disease in the cochlea and brain (Mudar & Husain, 2016;Pichora-Fuller et al., 2015).
Evidence for each of these models has been mixed.For example, data from the English longitudinal study of ageing (ELSA) found that social isolation and loneliness partially mediated the association between hearing loss and memory, providing support for the social cascade model (Maharani et al., 2019).By comparison, Dawes et al. (2015) found hearing aids were associated with improved cognitive performance independently of social isolation, providing support for the cognitive load model.Support for the common-cause pathway can be demonstrated by the increased prevalence of cardiovascular risk factors, such as diabetes and hypertension (Besser et al., 2018).However, whether this increase in risk is clinically meaningful is still unclear (Engdahl et al., 2015).Differences in characteristics of these cohorts, for example age, and the methods employed (e.g.statistical analyses) may explain some of the discrepant results found.Furthermore, whilst there has been increasing interest in modelling potential pathways for the association between hearing loss and cognitive decline, these have predominantly been conducted in older adults.Hearing loss and other chronic health conditions are often progressive and likely begin much earlier in life than the onset of dementia.As such, it may be important to investigate these pathways earlier in life to delay, or even prevent, cognitive decline and dementia.A further caveat of existing evidence in this area is that studies employing multiple linear regression or mixed models report weaker associations between hearing loss and cognition, than those using correlation analysis, indicating that the latter may produce more conservative estimates of effect (Loughrey et al., 2018).
On this basis, the current study investigated the potential theoretical mechanisms for the association between hearing loss and self-reported cognitive (memory) impairment, addressing potential confounds of existing research through the inclusion of younger participants, and as well as the use of path-analyses, which is more appropriate for testing theoretical models.

Study design and participants
A cross-sectional, online survey design was employed.Adults !18 years, with or without hearing loss were invited to take part via charity websites (e.g.Hearing Link, Age UK), online support groups (e.g."Living with Hearing Loss"), Loughborough University students, and posts on social media (e.g.Facebook, LinkedIn, Twitter).Ethical approval was received from Loughborough University (Human Participants) Sub-Committee (Ref: 2020-1177-102).

Measures
Respondents completed an anonymous online survey using Jisc Online Surveys between March and June 2020.Prior to completing the survey, respondents were provided with information about the research.They were then asked to complete an informed consent form.All outcomes were assessed using validated self-report measures that have been frequently used in previous research.

Demographics
Included age (in years), gender identity, ethnicity, highest level of occupation measured using UK Office for National Statistics categories (2020), and highest level of education.

Hearing loss-related outcomes
The Revised Hearing Handicap Inventory for the Elderly and Adults (HHI-R; Cassarly et al., 2020) assesses the effects of hearing loss on social/situational and emotional participation restrictions.A score of six or more on the 18-item measure was found to detect hearing impairment with 73.2% sensitivity and 73.8% specificity, with a Cronbach's alpha of 0.98 (Cassarly et al., 2020).Respondents whose scores were equal to or greater than six were therefore classed as having hearing difficulties.
International Outcome Inventory for Hearing Aids (IOI-HA, Cox & Alexander, 2002) was used to measure average daily hearing aid use (hearing aid users only).This was scored on a fivepoint scale from "none" to "more than eight hours a day".

Cognition
The main outcome of interest was cognitive impairment, which was assessed via an affirmative response to "Do you think you have memory problems?".This was deemed suitable since singleitem self-reported memory problems have been shown to be associated with increased dementia risk in later (R€ onnlund et al., 2015;Weber & Maki, 2016).

Cardiovascular disease risk (CVD)
The presence of any of the following was categorised into a dichotomous variable of CVD risk factors ("yes/no"): heart attack, stroke, diagnosis of type one or two diabetes, coronary heart disease, chronic lung disease, hypertension, atrial fibrillation, use of medication for diabetes, high blood pressure, and high cholesterol.Smoking status was recorded as: "never", "past smoker", or "current smoker".

Physical activity
The short form of the International Physical Activity Questionnaire (IPAQ-SF; Craig et al., 2003) was used to identify the frequency of physical activity over the previous seven days for vigorous (e.g.running, cycling) and moderate (e.g.dancing, gardening) physical activities, as well as walking (i.e.mild physical activity).Total physical activity was calculated by summing the total number of days participants reported engaging in vigorous, moderate, and walking activity, with higher scores indicating greater physical activity.
The six-item Lubben Social Network Scale (Lubben et al., 2006) comprised of two equally weighted subscales for friendship and family relationships.Each item was scored on a scale from zero to five.Responses were scored on a continuous scale ranging from 0-30, with lower scores indicating greater social isolation (Cronbach"s alpha ¼ 0.83).
Centre for Epidemiologic Studies -Depression Scale (CES-D; Zivin et al., 2010) consists of eight items that require a "yes" or "no" response and are totalled to give a score between zero and eight.A score of four or greater indicates significant symptoms of depression (Cronbach's alpha ¼ 0.81).

Statistical analysis
The extent to which hearing difficulties, memory problems, and physical activity were associated with demographic, hearing, CVD risk and psychosocial variables were initially assessed using Spearman Rank correlation coefficients.Path analysis was then used to investigate potential mediators of the association between hearing and memory problems.Total score on the HHI-R was used as a continuous exogenous variable and memory problems as a binary outcome (endogenous) variable.Diagonal weighted least squares were used to estimate effects and robust standard errors were applied.All path analyses were completed in R Studio using the "lavaan" package.We report the standardised path coefficients.The models were also used to calculate the bootstrapped 95% confidence interval (CI).A two-tailed p-value <0.05 indicated statistical significance.

Results
Descriptive characteristics can be found in Table 1 for the total sample (N ¼ 479), as well as those classified as reporting no hearing difficulty (n ¼ 237, 49%) or hearing difficulty (n ¼ 241, 50%).
On completion of data collection, due to the low frequencies in some demographic categories, binary variables based on a median split were computed for ethnicity, occupation, and education.For the IPAQ-SF, frequency data (i.e.number of days) were available for 76.4% (n ¼ 366) of the sample.The IPAQ-SF recognises that multiple types of activity can be done on a single day and does not distinguish specific days for each activity (i.e.moderate, vigorous, walking).Therefore, each type of activity could range from 0-7, with total physical activity ranging from 0-21 over a 7-day period.Therefore, data were analysed for total physical activity (0-21), moderate or vigorous physical activity (MVPA) (0-14), and walking (0-7).A Metabolic Equivalent of Task (MET) score could not be calculated because most (86.4%, n ¼ 414) participants failed to report the number of minutes engaged in each physical activity type per day.Nevertheless, Spearman rank correlation coefficients were calculated where both frequency and hours were available (13.6%, n ¼ 65), which demonstrated that the frequency of activities alone was strongly correlated with MET scores (see Supplemental Materials).
Overall, in comparison to participants who did not report hearing difficulties, participants that did were statistically more likely to be older.They were also more likely to report CVD risk and memory problems.Mixed results were found for physical activity, whereby those who reported hearing difficulties were physically active on fewer days overall and walked on less days but did not differ in MVPA.In addition, participants with hearing difficulties reported greater loneliness and social isolation.However, there was no difference between hearing difficulty groups for depression (CES-D total score).
Spearman rank correlation coefficients demonstrated significant associations between hearing difficulties, self-reported memory problems, and physical activity, as well as several other variables (see Supplemental Materials).Age was also significantly associated with each of these variables.
Following the initial exploration of associations between hearing difficulties and memory problems, further investigations were carried out to explore the potential mediating role of physical health (physical activity), CVD risk factors) and psychosocial wellbeing (social isolation, loneliness).As age was the only demographic variable that was significantly different between hearing groups, it was controlled for, by regressing it onto each of the endogenous variables.

Psychosocial wellbeing
Path analysis was used to test the associations between hearing difficulties, psychosocial wellbeing and subjective memory problems.Model A (Figure 1(A)) showed a direct path between hearing score and memory problems, whereby higher scores on the HHI-R (i.e.greater hearing difficulties) were associated with greater odds of reporting memory problems (standardised estimates [b] ¼ 0.17, p ¼ 0.007, 95% CI 0.00, 0.01).Social isolation was entered first as a mediator.Higher scores on the HHI-R were associated with greater social isolation, (b ¼ À0.33, p < 0.001, 95% CI À0.11, À0.06) (Figure 1(B)), however, the indirect effect (not shown on path models) was not statistically significant (p¼.128), and the total effect remained significant (b ¼ 0.15, p ¼ 0.005, 95% CI 0.00, 0.01).The addition of loneliness to the model (Figure 1(C)) removed the direct effect of the hearing score on memory problems (p¼.212), but the indirect effect remained significant (b ¼ 0.03, p ¼ 0.023, 95% CI 0.00, 0.01).

Discussion
The results of the current study suggest that the association between hearing difficulties and subjective memory problems may be fully mediated by psychosocial wellbeing.By comparison.although engaging in less walking was associated with greater hearing difficulties, physical health did not mediate the association with subjective memory problems.On this basis, our findings appear to provide support for the psychosocial cascade but not the common-cause model of the association between hearing difficulties and cognitive (memory) impairment.

Psychosocial wellbeing
In the current sample, the addition of social isolation and loneliness to the model removed the direct association between hearing difficulties and memory problems.
This finding aligns with existing literature, where social isolation and loneliness have been found to partially mediate the association between hearing loss and cognition (Dawes et al., 2015;Maharani et al., 2019).Likewise, an analysis from the ELSA (Maharani et al., 2019) found that, over a 10-year followup period, the association between hearing impairment and poorer episodic memory was partially mediated by social isolation and loneliness.While the current study found the association between hearing and memory was fully mediated by psychosocial factors, differing measures across studies may explain some disparity in findings.For example, although previous studies have used continuous, objective measures of memory/cognition (e.g.episodic memory), in the current study memory was assessed subjectively via a single "yes/no" response to the question: "Do you think you have memory problems?".Critically, binary data can inflate associations, which may be why the current study found full rather than partial mediation.
Furthermore, subjective memory problems have been shown to be highly correlated with poorer psychosocial well-being (Yates et al., 2017).Indeed, in this study, subjective memory problems correlated with all psychosocial variables measured (see Supplemental Material), albeit weakly.A recent study of older adults aged 60-to 75-years found that subjective memory problems were associated with poorer psychosocial well-being, but not diagnostic tests or objective measures (e.g.electroencephalogram recordings) of cognitive impairment (Sheffler et al., 2022).As such, subjective memory problems could more accurately reflect psychosocial well-being rather than cognitive function (Yates et al., 2017).Nevertheless, subjective responses have been shown to predict later life decline in cognitive function (R€ onnlund et al., 2015) and may still be a useful assessment of memory problems.
Taken together, the current body of evidence suggests that psychosocial well-being likely plays a crucial role in the relationship between hearing difficulties and memory.Thus, interventions (e.g.hearing aids) that aim to address psychosocial outcomes in adults with hearing loss may reduce the risk of developing cognitive deficits.However, findings from recent reviews suggest that there is limited high quality evidence to support whether hearing interventions can reduce social isolation and loneliness in adults with hearing loss, suggesting that further research in this area is warranted (Bott & Saunders, 2021;Ellis et al., 2021).

Physical health
Although hearing difficulties were associated with engaging in fewer days walking, this was not demonstrated for other levels of physical activity measured (i.e.moderate, vigorous).Additionally, no mediating effect of physical activity or health was found for the association between hearing difficulties and memory problems.Therefore, these results do not appear to provide support for the common-cause model.
The lack of association of these factors with memory could be due to several reasons.Firstly, we were only able to assess the frequency (i.e.number of days) respondents engaged in specific physical activities.We could not estimate the duration of activity due to a poor response rate on this item of the IPAQ-SF.Consequently, this may explain why no association with subjective memory was observed, as the association with physical activity could be dependent on the time and type of the activity undertaken (Elliott-King et al., 2018).Secondly, the directionality of the association between hearing and physical health is less clear than for other associations.Most of the research in this area has only addressed the association whereby cardiovascular health is a risk factor for or predictor of hearing difficulties, with findings often being inconsistent (e.g.Curhan et al., 2013;Engdahl et al., 2015;Lohi et al., 2022).However, it is possible that there is a bidirectionality to this association, whereby hearing difficulties may lead to adverse health behaviours (e.g. less physical activity), which may increase the risk of cardiovascular ill-health.
It should also be highlighted that, somewhat unexpectedly, the path between hearing difficulties and walking was significant.This was not the case with total or moderate or vigorous physical activity.A potential explanation for these findings could be that walking and mild activities have greater prevalence in middle-and older adults compared to more intense physical activities (Kuo et al., 2021;Loprinzi, 2013).As such, there may be a greater resolution in engagement, which may explain why a significant difference was only detected for walking.Indeed, a similar finding was observed by Kuo et al. (2021), who found the greatest difference in accelerometer measured physical activity between hearing groups was for mild physical activity (e.g.walking).This is particularly noteworthy, as gait-speed and walking duration are well-known risk factors for frailty, falls, and cognitive decline (Peel et al., 2019;Tian et al., 2021), which has been posited to be due to vestibular dysfunction in adults with hearing loss (Martin et al., 2021).Given the growing epidemiological evidence for the associations of physical inactivity with both hearing loss (Kuo et al., 2021;Wells et al., 2020) and dementia (Brasure et al., 2018;Livingston et al., 2020;Soni et al., 2019), further investigations that use validated and robust measures of a wide range and intensity of physical activity are necessary.

Study limitations and future research
We acknowledge that there may be several potential limitations with regard to the design of the current study, suggesting that caution should be exercised when interpreting the results.Firstly, the use of cross-sectional data means that the direction of the associations cannot be determined.For instance, the association between hearing difficulties and physical activity could be because those with hearing difficulties are less inclined to participate in such activities throughout their lives.Further investigation is required to identify the barriers and facilitators to physical activity in adults with hearing loss.
Secondly, the use of self-report data may limit confidence in the conclusions that are drawn.To mitigate this, only validated measures were included in this study.However, to accurately diagnose hearing loss, a pure-tone audiometry assessment is the gold standard, though validated measures of self-reported hearing loss can provide a greater understanding of the impact on daily life (Wayne & Johnsrude, 2015).Thus, the inclusion of both self-report and audiometric measures of hearing loss may be optimal.It should also be noted that memory problems were assessed with a single-item question.In future studies, validated cognitive assessments should be used to confirm the current findings and identify the specific domains associated with hearing difficulties.
Lastly, the recruitment strategy was targeted, but somewhat idiosyncratic, and therefore may be difficult to reproduce.In relation, 41% of the responses were collected during the height of the first UK lockdown (March to May 2020).Subsequently, the responses to physical activity and psychosocial measures may have been influenced by the multi-nationwide lockdowns in response to the COVID-19 pandemic.Even so, research was undertaken prior to the COVID-19 pandemic (e.g.Besser et al., 2018;Kuo et al., 2021;Maharani et al., 2019) has consistently shown that hearing loss is associated with physical and psychological health, suggesting that findings from the current study are unlikely to be entirely accounted for by the pandemic.

Conclusions
The current study provides support for the psychosocial cascade model of the association between hearing difficulties and memory problems, as well as some evidence for the association between hearing difficulties and physical health.On this basis, further high-quality research is needed to investigate whether interventions that aim to improve psychosocial wellbeing in adults with hearing difficulties can also improve cognitive outcomes.However, these associations should first be investigated with behavioural and physiological measures to corroborate the current findings derived from self-report outcome measures.Overall, the association between hearing difficulties and memory problems warrants scrutiny to establish whether hearing loss can lead to cognitive decline, and if the use of hearing intervention can improve cognitive outcomes.

Figure 1 .
Figure 1.Figures represent path models of association for the psychosocial model of association between hearing difficulties and subjective memory problems; (A) represents the direct path between hearing difficulties and memory problems, (B) with social isolation as a mediator, (C) with social isolation and loneliness as mediators.Values represent standardised estimates [b] of direct paths.Ã p < 0.05, ÃÃ p < 0.01.

Figure 2 .
Figure 2. Figures represent path models of association for the physical health model of association between hearing difficulties and subjective memory problems; (A) days of total physical activity included as a mediator, (B) days of moderate and vigorous physical activity included as a mediator, (C) days walking included as a mediator, (D) days walking and presence of cardiovascular risk factors included as mediators.Values represent standardised estimates [b] of direct paths.Ã p < 0.05, ÃÃÃ p < 0.001.

Table 1 .
Characteristics for total sample and those with self-reported normal hearing and self-reported hearing difficulties.
Note.Where sample size is not specified, complete data for the variable were available.Values represent frequencies (and percentages).b Mean (and SD).c Lower scores indicative of more social isolation.MVPA: Moderate or Vigorous Physical Activity.a (n ¼ 370).