STRENGTHS AND LIMITATIONS OF THIS STUDY

One of the primary strengths of this research lies in the integration of multiple logistic regression and propensity score matching (PSM) for analytical purposes.

A limitation of the study is that it does not conclusively establish a causal link between visual impairment and subjective cognitive decline (SCD).

While PSM aids in balancing differences among covariates, it does not account for unobservable systemic variances.

The use of pension data as a proxy for the economic status of the elderly, alongside living areas and daily exercise routines to infer living conditions, may not provide a comprehensive reflection of their actual circumstances.

Introduction

Visual impairment constitutes a prevalent functional limitation among the elderly, significantly impacting their quality of life and overall well-being. With ageing, the visual system experiences a structural and functional decline, leading to adverse changes in visual acuity,1 motion perception2 and the temporal resolution of vision.3 The elderly demographic is expanding rapidly, forecasting a substantial increase in age-related ocular diseases and visual impairments.4 Conditions such as glaucoma, cataracts, corneal diseases, age-related macular degeneration and diabetic retinopathy contribute to visual impairment in this age group. Consequently, older adults with visual impairments encounter greater challenges in performing daily activities (eg, eating and bathing) and face a heightened risk of adverse events (eg, falls, fractures and disability) as well as various psychosomatic illnesses.5 Concurrently, concerns about the impact of visual impairment on cognitive function are escalating. Evidence from both cross-sectional and longitudinal studies indicates a link between visual impairment and cognitive decline in the elderly.6–10 According to the Health and Retirement Study cohort, individuals reporting visual impairments were 35% and 25% more likely to develop possible cognitive impairment without dementia and probable dementia, respectively, compared with those without visual impairments.11 The coexistence of visual and cognitive impairments significantly burdens patients and their families, underscoring the importance of identifying and pre-emptively addressing potential cognitive decline risk factors at an early, cognitively unimpaired stage.

Subjective cognitive decline (SCD) is characterised by self-perceived continuous decline abilities relative to a previously normal condition, with no observable deficits on age-adjusted, gender-adjusted or education-adjusted standardised cognitive assessments,12 marking a pivotal phase in cognitive preservation.13 Neuroimaging studies have revealed that SCD is associated with white matter degeneration,14 15 reduced volume of the hippocampus,16–19 grey matter atrophy,17–22 reduced entorhinal cortex,17 23 amyloid deposition,24 altered spontaneous functional activity19 25 and hypometabolism in the brain.20 26 27 Moreover, subtle retinal thickness alterations in the macular region, correlating with Aβ deposition, have been identified in individuals with SCD.28 Compared with cognitively intact individuals, those experiencing subjective cognitive decline exhibit an accelerated cognitive deterioration pace and are more likely to progress to dementia.29–32 A longitudinal study spanning 7 years demonstrated that the SCD cohort was 4.5 times more likely to transition to dementia than their non-SCD counterparts.32 Collectively, these findings highlight the critical importance of proactive SCD prevention, risk factor management and deterrence of irreversible cognitive decline to enhance the quality of life in the elderly.

Despite extensive research on the association between visual impairment and cognitive dysfunction, studies specifically examining the link between visual impairment and SCD remain scarce. Zhaoyuan, a city located in the northwest of Shandong Province, China, is characterised by a notably ageing population. As of 2021, individuals aged 60 and above comprised 27.97% of the city’s total population (556,813), amounting to 155 740 people. This demographic trend prompted a cross-sectional study conducted in Zhaoyuan City to investigate the determinants of SCD, with a particular focus on the impact of visual impairment on SCD among the elderly.

MethodsStudy population

A cross-sectional study was conducted among residents aged 60 and above from July to October 2021 in Zhaoyuan, China. Random sampling, supplemented by considerations of the field environment, selected three research sites (Linglong Township, Daqinjia Subdistrict Office and Mengzhi Subdistrict Office) from the nine towns and five subdistrict offices within Zhaoyuan. The study included 428 older adults who exhibited normal objective cognitive test results and completed questionnaires. was used to calculate the sample size.33 p is the prevalence; deff is the design effect, and e is the absolute error, with the latter estimated as the product of relative error and prevalence. The calculation of the sample size necessary for the formal survey was predicated on an estimated prevalence of SCD among older adults of 33.5%, as reported in a prior Australian study.34 This analysis determined that a total of 382 participants were required, assuming a permissible error margin of the prevalence rate (0.2 p), a confidence level of 95% and a design effect coefficient of 2. The permissible error margin, d can be derived from the equation d=r*p, where r represents the relative error.33 The relative error is defined as the ratio of the absolute difference between the sample mean and the population mean to the population mean, which is inherently a positive value. In this context, with a relative error r of 20%, the resulting permissible error margin d was calculated to be 0.067 rather than adopting a direct value of 0.05. Given that cluster sampling’s efficiency is lower than that of simple random sampling, it necessitates a larger sample size to achieve comparable accuracy. This sampling approach assumes homogeneity within sample units, reducing sampling efficiency. A design effect greater than 1, typically valued at 2, was accounted for in this study.33 The exclusion criteria were: (1) dementia from various causes; (2) severe physical or mental illnesses; and (3) cognitive dysfunction due to vascular, traumatic, psychiatric, internal medical diseases, infections, poisoning, drug side effects, substance abuse and other neurodegenerative diseases. The objective cognition was tested by the Chinese version of Montreal Cognitive Assessment Basic (MoCA-BC: ≤19 for <7 years of education, ≤22 for 7–12 years of education and ≤24 for >12 years of education).35

Assessment of SCD

The subjective cognition of the elderly was assessed using the Chinese version of the SCD Questionnaire (SCD-Q9). The SCD-Q9 was developed by Gifford et al in the USA in 201536 and was adapted into a Chinese version of the scale by Hao et al in 2019.37 It has two main dimensions, including overall memory ability and daily activity ability. The Chinese version of SCD-Q9 has good internal validity and reliability with Cronbach’s α coefficient (0.886) and Spearman-Brown split-half coefficient (0.892).37 This tool was employed to differentiate individuals with SCD from cognitively normal elderly, requiring 3–5 min for respondents to complete. The SCD-Q9, comprising nine questions related to memory, scores responses as follows: ‘yes’ and ‘often’ receive one point, ‘occasionally’ earns 0.5 point and ‘no’ scores 0 point, totalling up to nine points. A cut-off value of five distinguishes normal cognition and SCD, with higher scores indicating a greater likelihood of cognitive impairment. The conceptual framework is detailed in online supplemental file 1.

Sociodemographic data

Sociodemographic information, including age, sex, education level, living area, marital status, self-reported health status and life satisfaction, was gathered using a specifically designed questionnaire. A notable reluctance was observed among urban elderly to disclose their economic status, while rural counterparts often reported having no monthly income. Consequently, the economic condition of participants was approximated through pension information. Additionally, inquiries were made regarding participants' engagement in exercise activities (ie, walking, running, climbing, boxing, dancing, playing ball games and swimming) for more than 30 min weekly, smoking (defined as continuous or cumulative smoking for over 6 months) and drinking (defined as drinking in the past year such as beer, wine or liquor) habits. Questions also covered visual, hearing and olfactory impairments. The detailed questionnaire is available in online supplemental file 2.

Statistical analyses

All statistical analyses were conducted using Stata statistical software (V. 16.0), with p<0.05 for significance. Descriptive statistics summarised the sociodemographic variables and subjective cognitive status. χ2 test was used to compare group differences of categorical variables. The Mann-Whitney test was used to compare independent groups on continuous variables lacking a normal distribution. A preregression analysis with SCD as the dependent variable and visual impairment as the independent variable preceded the multivariate logistic regression analysis. A multivariate logistic regression analysis aimed to determine the independent impact of visual impairment on SCD, considering subjective cognitive status as the dependent variable and including all covariates.

Propensity score matching (PSM) was used to mitigate the confounding biases concerning the effect of visual impairment on SCD. PSM, suitable for observational studies where randomised clinical trials are unfeasible,38 simulates randomisation in non-random data within a counterfactual framework, first proposed by Rosenbaum and Rubin in 1983.39 For individuals experiencing visual impairment, we can only observe their current subjective cognitive state, but not their subjective cognitive state without visual impairments under the same physical conditions. Hence, a multivariate logistic regression model calculated the PMS based on characteristics such as age, sex, education level, living area, marital status, pension status, physical activity, smoking and drinking habits, self-reported health status, life satisfaction and hearing and olfactory impairments. Controls similar to those with visual impairments were matched among subjects without visual impairments based on these scores, creating approximately randomised data sets. Visual impairment was thus treated as a treatment factor in this quasi-experiment, with the visually impaired group as the treatment group and the non-visually impaired group as the control group. Comparing subjective cognitive status between these groups, using the matched non-visually impaired group as a counterfactual, allowed for the determination of the average treatment effect of the treated (ATT).

Various matching techniques are employed in PSM, with consensus suggesting no single method suits all scenarios. Typically, outcomes from different matching methods are compared and analysed. In this study, nearest neighbour matching was initially applied, followed by assessments of result, robustness through kernel and radius matching. The k-nearest neighbour matching was set at k=2, and the radius for radius matching was designated as 0.05. Matched variables included potential confounders such as age, sex, education level, living area, marital status, pension status, physical activity, smoking and drinking habits, self-reported health status, life satisfaction and hearing and olfactory impairments. Kernel matching served as a representative to evaluate the matching efficacy through a balance test, comparing the visual impairment group with the non-visual impairment group. Effective matching was indicated by an absolute standardised difference in covariates after matching of less than 10%, or when covariate comparisons revealed no significant statistical difference.40 41 The goal was to minimise the standardised difference postmatching to enhance matching quality.

Results

This study included 428 participants, including 290 individuals with normal cognition and 138 with SCD (table 1). Of these participants, 63.1% were affected by visual impairment, while 32.2% exhibited SCD. 52.1% of the participants were women, 59.8% had over 9 years of education, 45.6% were from rural areas, 85.3% are currently married, 68.7% are receiving a pension, and 85.7% are engaging in exercise. Tobacco use was reported by 12.9% of the elderly and alcohol consumption by 28.7%. Participants with normal cognition reported a higher incidence of good self-reported health status and life satisfaction (P<0.001). Visual impairment was observed in 58.6% of individuals with normal cognition, increasing to 72.5% among those with SCD (P=0.006). Hearing impairment was noted in 40.9% of subjects, while olfactory impairment was present in only 7.9%.

Table 1

Sociodemographic characteristics of study participants

Multivariate logistic regression analysis (table 2) showed that bad self-reported health status (OR, 2.84; 95% CI 1.73 to 4.68), lack of physical exercise (OR, 2.21; 95% CI 1.18 to 4.11) and visual impairment (OR, 1.83; 95% CI 1.10 to 3.05) were risk factors for SCD in the elderly. The education level of more than 9 years (OR, 0.46; 95% CI 0.28 to 0.76) served as a protective factor.

Table 2

Multivariate logistic regression for subjective cognitive decline risk

The standardised differences of covariates before and after matching were shown in figure 1 and table 3. Postmatching analysis showed that the absolute standardised difference for all covariates was well within the 10% threshold, indicating a successful match. Furthermore, postmatching analysis revealed no significant differences in characteristics between the visual impairment and non-visual impairment groups (P>0.05), demonstrating the effectiveness of the PSM method in minimising differences between these groups. Before matching, the kernel density functions of propensity scores for the visual impairment and non-visual impairment groups differed significantly (figure 2). However, postmatching, these functions were similar, with most propensity scores falling within a common range, suggesting high-quality matching (figure 3). As shown in figure 4, there were few systematic differences between the matched groups, confirming the balance and overall efficacy of the matching process.

Table 3

Balance test results of the study on the effect of visual impairment on the subjective cognition of the elderly

Figure 1

Standardisation percent difference of covariates before and after matching.

Figure 2

Nuclear density diagrams of the treatment group and the control group before matching.

Figure 3

Nuclear density diagrams of the treatment group and the control group after matching.

Figure 4

Propensity scores of PSM between the visual impairment and non-visual impairment groups. PSM, propensity score matching.

PSM model (table 4) provided the estimated results of the three matching methods on the impact of visual impairment on the subjective cognition of the elderly. The ATT derived from the nearest neighbour matching method was 0.145, indicating a 14.5% higher incidence of SCD in the visually impaired elderly compared with their non-visually impaired counterparts, with this difference being statistically significant (P<0.05). Radius matching revealed an ATT of 0.136 for visual impairment’s effect on elderly subjective cognition, and kernel matching yielded an ATT of 0.138, both statistically significant (P<0.05). The similarity in ATT estimates and significance across the three matching methods demonstrate the robustness of the findings. After matching potential con-founders, the risk of SCD in older adults with visual impairment was found to increase by 13.6%–14.5% compared with the non-visually impaired group. Furthermore, multivariate logistic regression analysis indicated that the visually impaired were 1.83 times more likely to develop SCD than those without visual impairment (OR, 1.83; 95% CI 1.10 to 3.05; P=0.020), supporting the conclusion from both multivariate logistic regression and PSM analyses that visual impairment elevates the risk of developing SCD.

Table 4

PSM model results of the effect of visual impairment on the subjective cognition of the elderly

Discussion

This study focused on the subjective cognition of the elderly in Zhaoyuan City, analysing the impact of visual impairment on this aspect. It was found that 63.1% of the elderly experienced visual impairment, and 32.2% exhibited a decline in subjective cognitive ability. Notably, the prevalence of visual impairment among individuals with SCD (72.5%) exceeded that in the cognitively normal group (58.6%). Multivariate logistic regression analysis identified bad self-reported health status, lack of physical exercise and visual impairment as risk factors for SCD, whereas an education level of more than 9 years served as a protective factor. The application of PSM to minimise differences between the visually impaired and non-visually impaired groups revealed that visual impairment detrimentally affects elderly subjective cognition, increasing the risk of SCD by 13.6%–14.5%. Consequently, it is advised that the elderly adopt a healthy lifestyle and undergo regular ocular examinations to prevent visual impairment and SCD. Visually impaired older adults should be particularly vigilant about their subjective cognition, actively engage in prevention strategies against SCD and avert its progression to irreversible cognitive impairment.

The study revealed that 32.2% of older adults experienced SCD. This finding aligns with a prior study indicating that 33.5% of the elderly population, aged between 65 and 85 years, reported a decline in subjective cognitive abilities.34 Data analysis from the 2015 and 2016 Behavioural Risk Factor Surveillance System (BRFSS) in the USA identified that 11.4% of adults aged 55–64 years reported SCD, in contrast to 14.3% of those aged 65–74 years.42 Conversely, the Chinese Alzheimer’s Biomarker and Lifestyle Study reported a 51% prevalence of SCD among the northern Chinese Han population over 65 years of age,43 suggesting that regional and population selection factors may influence the variance in reported SCD prevalence. The decline in subjective cognitive ability serves as a clinically significant marker for future cognitive deterioration, with an estimated 2.3% and 6.6% of older adults with SCD progressing to dementia and mild cognitive impairment annually.29 Therefore, older adults are advised to monitor their subjective cognitive status closely and proactively address relevant risk factors during the cognitively unimpaired stage. For those already affected by SCD, implementing targeted interventions can mitigate the severity of symptoms and forestall the advancement of cognitive impairment, thereby enhancing the quality of life and promoting healthy ageing among the elderly.

Our findings revealed that 63.1% of elderly residents in Zhaoyuan City were afflicted with visual impairments. Among those with SCD, 72.5% suffered from visual impairments, a figure significantly higher than the 58.6% observed in the cognitively normal group (p=0.006). Multivariate logistic regression analysis indicated that visual impairment significantly increased the risk of SCD in older adults (OR, 1.83; 95% CI 1.10 to 3.05), underscoring a strong association between visual impairment and diminished subjective cognition in the elderly. Similarly, data from the BRFSS demonstrated that individuals with visual impairments were 3.5 times more likely to report SCD-related functional limitations than those without.44 Prior research has identified visual impairments as a key indicator of cognitive decline risk,45 with a clear independent correlation with cognitive impairment risks.11 46 A cohort study focusing on older women highlighted visual impairment as a predictor of subsequent cognitive deterioration.47 Furthermore, a longitudinal study found that participants with baseline visual impairments experienced a more significant reduction in cognitive function over 8 years, regardless of the impairment severity.8 Research from the UK also revealed that visually impaired individuals face a higher risk of cognitive impairment, potentially experiencing accelerated, faster trajectories of cognitive decline compared with those without visual impairments.45 Elderly individuals with visual impairments are twice as likely to encounter five or more physical or mental health complications.5 The combination of visual and cognitive impairments not only severely impacts patients’ health and life quality but also their overall well-being. Therefore, it is crucial for everyone, especially the elderly with visual impairments, to adopt healthy lifestyle practices, undergo regular ocular examinations and preserve optimal visual function to prevent visual impairment. Additionally, treating relevant eye pathologies in visually impaired older adults may aid in delaying cognitive decline. Implementing effective treatment and intervention strategies for older adults with combined visual impairment and SCD could significantly mitigate the risk of severe cognitive impairment.

The high prevalence of SCD underscores the importance of identifying its risk factors and implementing targeted interventions to prevent its progression into mild cognitive impairment, Alzheimer’s disease and other forms of dementia. Evidence consistently indicates that older adults with lower levels of education are at an increased risk of developing Alzheimer’s disease or other dementias.48–50 Conversely, our findings suggest that individuals with higher educational attainment may have a reduced risk of SCD, reinforcing the idea that education exerts a protective effect on various cognitive domains in the elderly.51 It is vital for the elderly to engage in lifelong learning, enhance their cognitive reserves, and keep their minds active. Furthermore, our study corroborates previous research showing that older adults reporting good health status are less likely to experience a decline in subjective cognitive abilities.52 Additionally, we observed that elderly individuals who do not engage in physical exercise face a greater risk of developing SCD. There is an inverse relationship between physical activity and the risk of cognitive impairment,43 with exercise aiding in the maintenance of cognitive function and reducing the risk of cognitive decline. Therefore, seniors are encouraged to participate in moderate physical activities, such as walking, dancing, playing ball and swimming, and to engage in social and leisure activities, thereby fostering a positive outlook and leading a fulfilling life. Studies have found that people with sensory impairments, including visual impairment and hearing impairment, show a faster trajectory of cognitive decline.8 45 Our study highlights that sensory impairments, specifically visual impairments, are associated with accelerated cognitive decline. However, no significant link was identified between hearing impairment and SCD in our research. Similarly, studies have found an association between vision and cognition, but the connection between auditory function and cognition remains unclear.47 53 Nonetheless, early detection and treatment of visual or hearing impairments in older adults are still crucial. In conclusion, this study identifies four key factors influencing the subjective cognitive abilities of the elderly. Proactive management of these modifiable risk factors is essential for delaying the onset and progression of more severe cognitive impairments.

In the PSM model, a satisfactory match was achieved for all covariates, with the absolute value of the standardised difference remaining within 10% after postmatching.39–41 Visual impairment was found to increase the risk of SCD in the elderly by 13.6%–14.5% after adjusting for potential confounders such as age, sex, education level, living area, marital status, pension status, physical exercise habits, smoking, drinking, self-reported health status, life satisfaction and sensory impairments. A clinical trial has demonstrated notable cognitive function improvements in older adults following cataract surgery, underscoring the link between visual impairment and cognitive decline.54 Those receiving vision correction exhibited enhanced cognitive function compared with their counterparts without such interventions.55 56 There are two primary theories explaining the impact of visual impairment on cognitive function in older adults. The sensory deprivation hypothesis posits that cognitive load increases due to prolonged, inadequate sensory stimulation and diminished sensory input.57 58 Visual impairments lead to reduced engagement in physical and mental activities, contributing to cognitive decline.59 60 Conversely, the common cause hypothesis suggests that both visual impairment and cognitive decline stem from shared factors like ageing, inflammation or neurological conditions.46 58 61 62 At the juncture where Alzheimer’s disease and various dementias manifest, cognitive impairment becomes irreversible, making intervention challenging. Hence, it is imperative for the elderly to undertake early prevention measures against visual impairment and associated conditions to avert SCD and cognitive decline. Promoting primary eye care education can enhance awareness among the elderly, mitigating vision loss and reducing the risk of accidents, such as falls. Timely vision correction for those with visual impairments not only improves their sight and quality of life but also aids in preventing cognitive deterioration. Regular cognitive screenings and proactive management of modifiable risk factors are crucial in preventing or delaying cognitive impairment. Moreover, establishing a comprehensive eye care service education and management system will significantly benefit the cognitive health of the elderly, particularly those with visual impairments.

The study employed a synergistic approach, combining multiple logistic regression analysis and PSM, to analyse the influence of visual impairment on subjective cognition among the elderly. PSM served to mitigate potential confounders, enabling a more precise estimation of visual impairment’s impact on SCD in this demographic. By facilitating the matching of samples, PSM helped to circumvent biases introduced by incorrect functional form specifications, although without fully addressing endogeneity issues arising from selection bias and omitted variables. The PSM procedure necessitates the exclusion of samples outside the common support range, leading to a partial loss of sample data. While PSM effectively balances covariate differences, it cannot account for unobservable systemic variations between the treatment and control groups.

Conclusions

The study revealed a notable prevalence of SCD among the elderly in Zhaoyuan, China (32.2%), with a significantly higher occurrence of visual impairment among SCD sufferers compared with cognitively normal individuals. The study identified several risk factors for SCD, including bad self-reported health status, a lack of physical exercise and visual impairment, whereas education level over 9 years emerged as a protective factor. According to PSM analysis, visual impairment was linked to a 13.6%–14.5% increased risk of SCD in the elderly, following the adjustment for observable heterogeneities between the visually impaired and non-visually impaired groups. Emphasising the importance of lifelong learning, cognitive reserve enhancement, regular physical activity, vigilant eye health monitoring and proactive visual impairment prevention are crucial for mitigating SCD risk in the elderly.

Data availability statement

Data are available upon reasonable request. Data are available upon reasonable request. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

The study involving human participants were reviewed and approved by the Medical Ethics Committee of the Second Hospital of Shandong University (KYLL-2020(KJ) P-0142). All subjects provided signed informed consent to participate in the study.

Acknowledgments

The authors would like to thank all participants for their time and excellent works.