Visual impairment is associated with physical and mental comorbidities in older adults: a cross-sectional study

We identified 5,348 (1.8% of the sample) patients with a Read code for visual impairment (Table 1). Women were over-represented in the visual impairment group compared to controls (63.2% versus 57.1% for controls; P <0.001). Individuals with recorded visual impairment tended to be older (mean age 80.9 years versus 74.9 years for controls; P <0.001), with 37.1% 80 years old or older compared to 12.1% of controls and were also more likely to be socially deprived (visual impairment Carstairs score 0.08 versus controls -0.35; P <0.001) although differences between quintiles were small.

Overall 95.0% of individuals with visual impairment had at least one other condition compared to 84.5% of controls (P <0.001) (Table 2). Whilst controls were more likely to have one (visually impaired 9.9% versus controls 19.7%; OR 0.55 95% CI (0.50 to 0.60)) or two (visually impaired 14.7% versus controls 19.9%; OR 0.72 95% CI (0.67 to 0.78)) conditions, those in the visual impairment group were significantly more likely to have four conditions (visually impaired 15.4% versus controls 11.7%; OR 1.17 95% CI 1.08 to 1.26) and five or more conditions (visually impaired 37.4% versus controls 17.8%; OR 2.05 95% CI 1.94 to 2.18) (Table 2) (even after standardising for age, sex and social deprivation).

For the visual impairment group, 27 out of 29 physical conditions were significantly more prevalent relative to controls, with only bronchiectasis and migraine showing no significant differences (Table 3). The largest differences after standardisation for age, sex and deprivation were for multiple sclerosis (0R 3.31, 95% CI 2.43 to 4.51) and diabetes (OR 2.76, 95% CI 2.60 to 2.93), both of which are recognised antecedents of visual impairment. The most commonly diagnosed condition for those with visual impairment was hypertension, recorded for 55.6% compared to 46.7% of controls (OR 1.40, 95% CI 1.33 to 1.48).

Depression had the largest prevalence for the individual mental health conditions (visually impaired 18.2% versus controls 12.0%; OR 1.53, 95% CI 1.43 to 1.65) and Table 4 highlights that those with visual impairment had significantly higher prevalence for all eight mental health conditions. The biggest difference after standardisation for age, sex and deprivation was for learning disability (visually impaired 0.5% versus controls 0.2%; OR 3.31, 95% CI 2.20 to 4.97), followed by anorexia or bulimia (visually impaired 0.5% versus controls 0.2%; OR 2.23, 95% CI 1.51 to 3.28) (although clearly the absolute prevalences of these conditions were low) and for any `other psychoactive substance misuse’ which is a heterogeneous grouping of problematic use of non-prescription and prescription drugs ever recorded (visually impaired 12.9% versus controls 4.8%; OR 1.79, 95% CI 1.65 to 1.95).

Causality cannot be directly inferred from these cross-sectional data, but it is known that many chronic conditions have visual consequences. For example, optic neuritis occurs in 70% of MS cases [21], the ocular complications of diabetes account for the majority of blind registrations in the UK among working age adults [22] and stroke causes visual impairment in 30% of cases [23]. Other conditions, such as learning disability (more than three times more likely in the visual impairment group) are associated with a high prevalence of visual defects from childhood [24]. Notably, hypertension (55.6%), coronary heart disease (28.8%) and diabetes (25.9%) were the most prevalent conditions in the visually impaired group, and visual impairment may pose significant barriers to encouraging a healthy lifestyle in this population. Visual impairment may contribute to a compromised diet (due to difficulty in preparing food), increased difficulties engaging in exercise, and high levels of isolation and loneliness. Management of these conditions in patients with visual impairment will require active integration of clinical, social and rehabilitation services to help improve health.

Clinicians also need to be alert to depression and hearing loss which will be more common in their patients with visual impairment compared to those without, and which are likely to complicate management of other comorbid conditions they are dealing with. Visual impairment is a recognised risk factor for depression which was identified as the most prevalent mental health condition (18.2%); this is broadly consistent with previous prevalence estimates for this group (13.5% to 33% [25]-[27]). Comorbid hearing loss was also identified in 17.7% of people in this study, compared to only 9.3% in controls. Both depression and hearing loss are commonly underdiagnosed within primary care services [28],[29] and have been related to loss of functional independence, increased risk of falls, hospitalisation and mortality [30]-[32]. Identifying and managing depression and hearing loss will require increased integration between services (including rehabilitation and social services) and clear referral pathways. Integration of depression screening and integrated depression treatment within eye care services is currently being evaluated [33],[34]. Recommendations have also been made to integrate hearing screening tests into the eye care and rehabilitation services, which could substantially increase the numbers of people identified with hearing loss [28],[35]. There is also recent evidence of Governmental commitment to tackling comorbid depression and dual sensory loss at national levels [36]. These advancements represent a move towards integrating services, but continuing integration to provide holistic care for patients will require more evidence based recommendations and commitment from the healthcare community.

Conversely, checking for and taking account of visual impairment is important for those providing care for other conditions. Clinicians need to be mindful of the increased falls risk in visual impairment when prescribing drugs for comorbid physical/mental health conditions which could also contribute to falls (for example, hypotensive and hypoglycaemic agents, bladder anticholinergics, opioid analgesics, sedative and psychotropic drugs), or in relation to safe medicines use especially when regimens are complex.

The study is a cross sectional analysis of patients in primary care using diagnostic criteria based on records within a primary care database. Recording of disease will, therefore, not be as accurate or consistent as it would be in a more formal epidemiological survey. Compared to the relatively few studies which have examined comorbidity in visual impairment [16],[17], strengths of our study include the large sample size and the assessment of a much wider range of comorbid physical and mental conditions. Further, this study does not rely on patient self-report of comorbidities and takes account of confounders, such as age, gender and social deprivation. An important limitation is that visual impairment is based on GP recording of one or more Read Codes rather than formal measurement of visual function. The Read Codes used were a broad set which do not always clearly define the exact nature of the visual impairment, and visual impairment is under-recorded compared to epidemiological estimates [1]. However, our expectation is that most people with very severe impairment will have been coded. From this perspective, the finding that 1.8% of the sample was defined as having visual impairment is consistent with 0.5% of Scottish residents being formally registered as visually impaired, which is known to include only about half of those eligible for registration [37]. Registration can only be done by a consultant ophthalmologist with clear visual acuity and visual field criteria defining `partially-sighted’ or `blind’ registration. The study reported by Van Nispen [17] had clear clinical inclusion criteria for patients with visual impairment, including assessment by an ophthalmologist, but it is important to recognise that this requirement in itself is likely systematically to exclude the very frail and people who are housebound or resident in care-homes. All studies in this area are, therefore, likely to be biased in some way, and the limitations of our study are inevitable in all studies relying on routine clinical coding, even in Scottish primary care where use of electronic medical records is longstanding and coding reasonably reliable. However, these limitations have to be balanced against the ability to analyse data from very large representative samples, and we believe the findings are valid.

Additional limitations are that there may be reporting bias, where people who attend frequently are more likely to have all clinical conditions recorded, in which case people with visual impairment recorded will also be more likely to have other conditions recorded, and that we were only able to control for three confounding variables. Reporting bias is likely to be present to some extent, but we believe the impact of this is likely to be small because patients in the UK are required to register with a single practice to access National Health Service care, and their clinical record is automatically transferred when they change practice. All specialists write to the GP and, therefore, populate the primary care record. Reporting bias in cradle-to-grave record is less likely than in one where records are created anew every time patients register, although may be present for conditions where patients may significantly self-care (for example constipation) than ones where registers are more complete (for example, ischaemic heart disease). If present, reporting bias would be expected to somewhat inflate the odds ratios of having comorbidity in people with visual impairment compared to those without, but would have less influence on the estimates of comorbidity rates in the visually impaired population. With regard to other confounders, we were restricted by what data were available. Clinical records do not reliably record marital status, income, occupation or education, but do reliably record address/postcode. We, therefore, used postcode assigned socioeconomic status, gender and age because we had near complete data on them. However, this represents the trade-off between large and representative routine data analyses and more detailed epidemiological studies which are likely to be smaller and likely less representative (because the sickest and frailest will not take part). Moreover, indicators such as income, occupation or education are likely to be strongly correlated with each and, therefore, at least partly accounted for by measures of socioeconomic status such as the Carstairs score used in this study.

The current study is a cross-sectional analysis. Cohort studies are required to enable causality of visual impairment and comorbidities to be explored further. Little is known about how visual impairment interacts with the genesis or progression of other diseases. For example, it is not clear how poor lifestyle factors (diet, exercise) which are compounded by visual impairment might contribute to hypertension and coronary heart disease. This is an important area for policy and research. Future research should seek to determine if there are cost-effective interventions which may help manage or prevent comorbidities, for example, nutritional and/or exercise interventions for elderly patients with visual impairment. Additionally, although there are incentives for GPs to record more accurate and complete registers of some comorbidities, this is not the case for visual impairment. Therefore, in accordance with recommendations made in the recently published Chief Medical Officer’s annual report [38], improvement in the quantity and quality of visual (and other sensory) impairment data would help inform future developments to local and national services for people with visual impairment.