Sociodemographic, behavioral, and medical risk factors associated with visual impairment among older adults: a community-based pilot survey in Southern District of Hong Kong

Visual impairment (VI) is significantly associated with poor physical health, limitation, and disability [1, 2]. Previous studies have shown a correlation between the severity of VI and the incidence of falls, fractures, chronic diseases and mortality in the elderly [3‐8]. Visually impaired people are more likely to have poorer social network, lower employment status, and mental health problems [9‐11]. Using different scoring scales to measure and evaluate quality of life (QoL) have suggested that VI severity is negatively correlated with QoL scores [12‐17]. Increased VI is also correlated with substantial health burden [18]. Elderly people with VI are more likely to require social support services and nursing home placement, [19, 20] and their community would also need to bear the associated medical cost and productivity loss [21, 22].

Socioeconomic factors can explain 69.4% of the global variations in the prevalence of moderate to severe VI and 76.3% of the global variations in prevalence of blindness [23]. Unfavorable socioeconomic status is associated with a higher prevalence of low vision and blindness in many countries [24‐28]. Behavioral factors like smoking status and alcohol-drinking pattern were associated with VI among older adults, [29, 30] while biomarkers such as increased serum low-density lipoprotein (LDL), homocysteine, and being underweight were shown to be associated with blindness [31]. Thus, individual health is mutually determined by the biological, behavioral, cultural, social and environmental factors. A broader framework was required for analyzing visual health problems from biosocial and socio-behavioral perspectives [32].

To our knowledge, there was no territory-wide visual health survey of older adults in Hong Kong but had been a district-level survey in 2002, in the Shatin District (a northern region of Hong Kong which is less densely populated than the main urban areas of Hong Kong Island and Kowloon) [33]. We hypothesized that spatial (Shatin and Southern Districts) and temporal (2002 and 2016) differences on the prevalence of VI among older adults would be statistically significant at district-level of Hong Kong when compared our results with the previous survey in 2002.

In addition, the survey in 2002 only covered the sociodemographic characteristics of the subjects in statistical analysis, but did not measure the associations of behavioral and medical risk factors with VI. The Singapore Malays Eye Study was an example of examining the associations of sociodemographic, behavioral and medical risk factors with VI, which shown that older age, lower education, unemployment, being widowed, diabetes and hypertension were independently associated with bilateral VI [34]. We hypothesized that behavioral and medical risk factors would be associated with VI among older adults, other than sociodemographic characteristics.

Our primary objective was to estimate the prevalence of VI among older adults within Southern District (located on the southern part of Hong Kong Island), in order to examine the significance of spatial and temporal differences on the prevalence of VI at district-level of Hong Kong when compared our results with the survey in 2002 and then show the rationale of conducting a territory-wide survey throughout Hong Kong.

The secondary objective was to measure the associations of sociodemographic, behavioral and medical risk factors with VI among older adults, in order to review the visual health problems and identify the vulnerable groups for VI from biosocial and socio-behavioral perspectives. The findings of this pilot survey will be useful in evaluating the feasibility of a forthcoming 5-year eye screening programme, and designing future large-scale eye research surveys.

Ethics approval of this study was granted by the Institutional Review Board (IRB) and Ethics Committees (EC) of the Hong Kong University / Hospital Authority Hong Kong West Cluster (reference number: UW 15–160) on 16th March 2015. All subjects agreeing to participate in the study gave their written, informed consent prior to the administration of any eye assessment or questionnaire.

This community-based, cross-sectional, pilot survey was conducted at Chi Fu, which is a suburb randomly selected from the 17 possible District Council Constituency Areas (DCCAs) of the Southern District of Hong Kong. According to the 2016 Population By-census, Chi Fu had an estimated population of 15,784 and domestic households of 5437. The residents aged 50 and above for this area was 6427 (46.52% males, 53.48% females) [35]. On average, its residents had a higher educational level, household income, and proportion living in private housing (rather than public housing) than the mean for the entire Southern District, Shatin District and Hong Kong. (Supplementary file 1, Table 6).

Invitation letters (in Chinese) for a free eye assessment and visual health survey, to be conducted at the local community hall, were sent to all registered households within the study area (i.e. Chi Fu DCCA) at the same time to ensure that all residents would have the same opportunity to get the information and participate in our survey. On a first-come-first-served basis (i.e. convenience sampling which is common in pilot study), residents who were interested to participate responded by calling back to make an appointment at a designated time for a free eye consultation with an ophthalmologist, during which, trained research assistants explained the details of the study and distributed the questionnaire (Supplementary files 10, 11) for their own completion. The targeted study population were local Chinese residents aged 50 and above, with no extra inclusion and exclusion criteria. The survey fieldwork was carried out from the 11th to the 28th, January 2016.

At the study site, optometrists measured the presenting visual acuity (VA) for each eye of the subject, using a Logarithm of the Minimum Angle of Resolution (logMAR) visual acuity chart at a distance of 3 m. If none of the large optotype on the highest line could be identified, VA was then assessed for counting fingers, hand motion, and presence or absence of light perception, in that order. Refractive error was measured by subjective refraction. Experienced ophthalmic assistants measured intraocular pressure and performed autorefraction using NIDEK Tonoref II. Slip-lamp examination of eyelids, ocular surface and anterior segment using Topcon SL 701 with DC4, as well as optical coherence tomography (OCT) scan using the Cirrus-5000 SD-OCT were performed by ophthalmologists, to examine the fundus without dilating the pupil and measure the macula and peripapillary retinal nerve fiber layer (RNFL) thickness and volume. We did not test the visual field defects, but measured the blood pressure, ankle-brachial index (ABI), body weight, and height of the subjects.

A structured and self-administrated questionnaire in Chinese was used to collect data on sociodemographic, behavioral, and medical characteristics. Sociodemographic and behavioral information covered age, gender, type of housing, educational level, marital status, employment status, monthly household income, smoking and alcohol-drinking habits. Subjects were inquired for any history of chronic medical conditions like diabetes mellitus (DM), hypertension, and hyperlipidemia, as well as three common eye diseases including age-related macular degeneration (AMD), cataract, and glaucoma.

We used the World Health Organization (WHO) criteria to define blindness as presenting VA of < 3/60 (logMAR > 1.3) and low vision as presenting VA of < 6/18 and ≥ 3/60 (logMAR > 0.5 and ≤ 1.3), without considering the visual fields [36]. Bilateral VI included both blindness and low vision in the better-seeing eye. For providing further insight into unilateral VI, we presented the results of blindness and low vision with the same definition in terms of the worse-seeing eye [34, 37]. Myopia and hyperopia were defined as spherical equivalent refractions of <− 0.50 diopter or > + 0.50 diopter, respectively [38]. Body Mass Index (BMI) is equivalent to the weight (in kilograms) over height squared (in centimeters) and was categorized into underweight (< 18.5), normal weight (18.5–22.9), overweight (23–24.9), and obesity (≥25) [39‐41]. Ankle-brachial index (ABI) was categorized into blockage (< 1.0), normal (1.0–1.4), and rigid arteries (> 1.4) [42]. Blood pressure was categorized into low (≤90/60), normal (> 90/60 and < 120/80), pre-high (120/80–139/89), and high (≥140/90) [43, 44].

For descriptive data analysis, median with interquartile range (IQR) was used to summarize continuous variables, whilst frequencies and percentages were used to summarize categorical variables. Crude, age-and-gender-adjusted, age-specific and gender-specific prevalence estimates of low vision, blindness, bilateral and unilateral VI were expressed in percentage with 95% confidence intervals (CIs). Relative changes in odds ratios (ORs) were calculated to measure the magnitude of confounding and examine the confounding effect of age on the associations of VI with obesity, hypertension, hyperlipidemia and educational level. Multiplicative interaction model and stratified analysis were done to examine the effect modification between independent variables on the risk for unilateral and bilateral VI. Univariate and multivariate logistic models with backward stepwise method were conducted to examine the associations between VI and independent variables at subject-level. Generalized linear mixed effects models (GLMM) were used to investigate the associations of VI with eye examination results at eye-level analysis after adjusting the inter-eye correlation. Crude and adjusted odds ratios (CORs and AORs) with 95% CIs were calculated to show the strength of associations. A p-value of < 0.05 was considered statistically significant. We included records with incomplete data in analyses and used multiple imputation method to handle missing data by creating 2000 multiple datasets and combining estimates from imputed datasets to obtain the overall estimate, based on the assumption of missing at random [45, 46]. The database was maintained and managed using Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA, USA). All statistical analysis were performed using R package version 3.5.1 (R Development Core Team, 2018) [47].

A total of 222 subjects, median age 67 years (IQR: 61–72), agreed to participate in the study, including 87 males (39.19%) and 135 females (60.81%). The majority was aged between 60 and 69 (49.55%). Most of them had completed secondary education (51.35%) and lived in private housing (94.14%). They were predominantly married (85.14%) and not working (89.64%). Nearly three-fourths (75.22%) had monthly household income below $25,000. The self-reported prevalence of DM, hypertension, hyperlipidemia, AMD, cataract and glaucoma were 12.61, 39.64, 15.32, 4.05, 22.97 and 2.70%, respectively. The proportions of current and past smokers were 1.35 and 3.6%, respectively, whilst those of current and past alcohol-drinkers were 17.12 and 1.8%, respectively. (Supplementary file 2. Table 7).

The numbers of right eye and left eye with VI (low vision or blindness) were 46 and 48 respectively whilst the cases of bilateral and unilateral VI were 21 and 73 respectively. The crude rates of bilateral and unilateral VI were 9.46 and 32.88% respectively while the age-and-gender-adjusted rates were 6.89 and 30.5% respectively. Age-specific rates of 59–59, 60–69, 70–79 and ≥ 80 were 17.5, 31.82, 37.74 and 57.89% respectively for unilateral VI, as well as 0, 9.09, 13.21 and 21.05% respectively for bilateral VI. Gender-specific rates of males and females were 34.48 and 31.85% respectively for unilateral VI, as well as 12.64 and 7.41% respectively for bilateral VI. (Supplementary file 3. Table 8).

Univariate logistic regression results showed the significant associations of unilateral or bilateral VI with older age, low educational level and obesity. Household income and having a history of hypertension were significantly associated with unilateral VI, while living in temporary housing and having a history of hyperlipidemia were significantly associated with bilateral VI. (Supplementary file 4. Table 9) We did not find any significant crude association of unilateral or bilateral VI with current or past smoking or alcohol-drinking, but found that current or past smoking was associated with being obese and having a history of hypertension. (Supplementary file 5. Table 10).

Multivariate logistic regression results with backward stepwise method in Table 1 noted that subjects with older age (AOR 1.04; 95% CI 1.00–1.09) and educational level higher than secondary with either non-degree (AOR 0.28; 95% CI 0.09–0.80) or degree tertiary education (AOR 0.29; 95% CI 0.08–0.89), were significantly associated with unilateral VI. In addition, subjects with older age (AOR 1.08; 95% CI 1.02–1.15), being obese (AOR 4.00; 95% CI 1.49–11.41), having a history of hyperlipidemia (AOR 3.60; 95% CI 1.13–10.97), and living in temporary housing (AOR 8.26; 95% CI 1.81–34.92), were significantly associated with bilateral VI. Older age was likely to be common variable contributing to unilateral or bilateral VI. Our results did not find any significant adjusted association of smoking and alcohol-drinking status with unilateral or bilateral VI.

The relative changes in ORs of exposures on unilateral or bilateral VI after adjusting for age were calculated to measure the magnitude of confounding in Table 2. After adjusting for age, for unilateral VI, ORs of having a history of hypertension decreased from 1.97 (1.12–3.51) to 1.52 (0.83–2.79), having non-degree tertiary education increased from 0.2 (0.07–0.56) to 0.26 (0.08–0.72), and having degree tertiary education increased from 0.21 (0.06–0.61) to 0.29 (0.08–0.86), with the effect of age remaining significant. Similarly, for bilateral VI, ORs of having secondary education increased from 0.19 (0.06–0.53) to 0.25 (0.08–0.74). Since the corresponding relative changes in ORs were 22.84, 30, 38 and 31.58%, we noted that age was the confounders of the unilateral VI-hypertension, unilateral VI-educational level and bilateral VI-educational level relationships due to a relative change of > 10% in the ORs of having a history of hypertension or higher educational level on unilateral or bilateral VI after adjusting for age.

In Table 3, we examined whether hyperlipidemia interacts with age, gender, obesity, hypertension or cataract on the risk for unilateral or bilateral VI. For unilateral VI, the interaction term between:

Gender and obesity were likely to modify the association between hyperlipidemia and the risk for unilateral VI, so combining them would allow for a more accurate and greater effect than the sum of their individual effects. No significant interaction term was found for bilateral VI. We have summarized the non-significant results of examining the effect modification between variables on the risk for unilateral or bilateral VI. (Supplementary files 6–9. Tables 11, 12, 13, 14).

ORs, stratified by gender, obesity, hypertension, hyperlipidemia and cataract, were shown in Table 4. For unilateral VI, in stratified analysis, we found that:

For bilateral VI, our results of stratified analysis showed that:

On analysis of individual eyes, in Table 5, the proportions of myopia or hyperopia, abnormal or suspicious optic disc, and abnormal or undetermined macular appearance were 78.38, 13.51, 8.56 and 74.77%, 11.71, 9.01%, for the right and left eyes, respectively. VI was significantly associated with myopia or hyperopia (OR 6.69; 95% CI 2.07–21.60) and marginally associated with abnormal or undetermined macular appearance (OR 3.02; 95% CI 0.92–9.94) after adjusting the inter-eye correlation, but was not statistically associated with abnormal or suspicious optic disc. In the GLMM model for the association of VI with myopia or hyperopia, the variance estimates for random effects and intra-class correlation were 4.02 and 0.55 respectively to account for inter-eye correlation in outcome by adding a random effect.