Introduction
Vision plays a major role in maintaining the abilities needed in everyday life, such as learning, working capacity, self-care, and mobility [
1‐
3]. For instance, Gompel et al. report, that children with low visual acuity (VA) need more time for reading and comprehending text, while the comprehension skill itself does not differ from those with normal VA. [
4] Impaired VA is associated with increased risk of accidents, particularly falling [
5]. Along with maintaining the functional ability, there is evidence suggesting poor vision increases the risk of institutionalization [
6]. Declining vision also causes problems in different aspects of daily life, for instance, in social interactions and daily routines [
7‐
9]. In Finland, visually impaired people have a lower level of education and employment on average, compared to the overall population [
10].
Some of the other population-based studies previously conducted focusing on vision and declining VA are The Rotterdam Study [
11], The Blue Mountain Eye Study [
12], Korea National Health and Nutrition Examination Survey [
13], Melbourne Visual Impairment Project [
14], and The Beaver Dam Eye Study [
15]. Being researched widely around the globe, none of these, however, include Health-Related Quality of Life (HRQoL) and analysis of its connection to declining VA.
Most of the previously conducted research on Quality of Life (QoL) and vision is highly focused only on the visually impaired or on specific eye diseases and their impact on QoL [
16‐
18]. Therefore, they tend to be based on relatively small study populations that are not representative on a larger scale. Furthermore, eye diseases may also have an impact on QoL through other factors besides declining VA, such as potential adverse effects of medication, anxiety about the future, lost time and money spent on the diagnostics, therapeutic interventions, and follow-up, which may cause bias [
19,
20]. To our knowledge, there is only one study investigating the impact of declining visual function on HRQoL in general population—a longitudinal population-based study among Latino people [
21], but the generalizability of these results into other ethnicities is currently unknown.
Correlation of declining VA and QoL has been studied to some extent using vision-specific QoL assessments, such as the Vision-Related Quality of Life (VRQoL) questionnaire [
21,
22]. These instruments are considered highly condition specific [
23], not allowing comparison across diseases and/or treatments. For generalizability and comparability, generic preference-based questionnaires are required. Quality of life measured through VRQoL—questionnaire might also be influenced by non-visual factors [
24]. Hence, it is important to contribute information about declining VA regardless of eye diseases.
In this study, we examined the profiles and the correlation of VA and self-reported HRQoL in Finnish adults in two time points in 2000 and 2011 using nationwide population-based health surveys and generic preference-based HRQoL tools. This comprehensive approach allows us to get an overall perspective on QoL and how it is affected by vision loss. Using generic tools instead of vision-specific measures reduces the potential reporting bias associated with overemphasis of visual factors on the responders’ QoL.
Methods
Study population and survey design
We utilized data from two nationwide surveys of health and well-being carried out by the National Institute for Health and Welfare in 2000 and 2011. These Health 2000 and 2011 studies provided a probability-clustered sampling and weighting scheme that estimates health statistics that are representative of Finnish adult population aged 30 and older at the time of sampling. The sampling scheme also accounts for designed oversampling among the elderly people in the 2000 baseline. Our sample inclusively represents the Finnish adult population with respect to main demographics of Finland. The general research methods have previously been described elsewhere in more detail [
25,
26]. Briefly, the study participants were invited to participate in an interview and health examination in 2000, and in a follow-up survey in 2011. The demographics of study participants are summarized in Table
1.
Table 1The demographics of study participants aged 30 years or older
Sample size (% women) | 8028 (54.7%) | 8006 (53.0%) | 4703 (55.5%) |
Mean age (SD) | 54.71 (16.2) | 55.34 (15.6) | 49.6 (12.1)a 60.0 (12.1)b |
EQ-5D Index Score available | 6148 | 4084 | 3131 |
15D index score available | 6166 | 4266 | 3510 |
Distance VA measured | 6674 | 4619 | 3867 |
Near VA measured | 6646 | 4618 | 3860 |
The mean age for those with complete data on VA was 49.6 years in 2000 and 60.1 years in 2011 and the proportion of women was 55.3% (not shown in the table) in both time points, following quite well the general study population distributions
Visual acuity tests
Habitual distance VA was measured binocularly at 4 m, with current visual correction, using the Snellen eye chart. Habitual near VA was measured at the participant’s preferred reading distance, using the near vision chart. Illumination was set to ≥ 350 lx on the vision charts. [
25,
26] All VA values are presented as decimal equivalents. For comparisons, the VA values were further classified based on forthcoming ICD-11 classification [
27] and a priori judgement based on the clinical relevance of VA. VA ≥ 1.0 was classified as good vision, VA of 0.63–0.8 as adequate vision, VA ≤ 0.5 as weak vision, VA ≤ 0.25 as impaired vision, and VA < 0.1 as severe vision loss or blindness.
Habitual distance and near VA were measured at both time points (2000 and 2011). A change of at least two lines on the Snellen eye chart was considered clinically significant improvement or decline, as smaller changes can be caused by numerous other factors, such as state of the tear film on ocular surfaces or preceding fatigue causing accommodative tiredness during the measurement [
28‐
30].
HRQoL was assessed using two internationally established, generic and standardized preference-based questionnaires—EQ-5D and 15D [
31,
32]. Both methods yield a single index score, as well as a multidimensional profile. The EQ-5D has an answer scale ranging from 1 (no difficulties) to 3 (extreme difficulty) and consists of five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The 15D has an answer scale ranging from 1 (no difficulties) to 5 (extreme difficulty) and comprises 15 dimensions: mobility, vision, hearing, breathing, sleeping, eating, speech, excretion, usual activities, mental function, sexual activity, discomfort and symptoms, depression, distress, and vitality. For both methods, weighting the separate dimensions with population-based preference weights yields index scores ranging from 0 to 1 for 15D and from − 0.59 to 1 for EQ-5D, with 1 representing the best possible HRQoL. In the present study, 15D was weighted using Finnish preference weights, whereas EQ-5D was weighted using UK time-trade-off weights in order to achieve the widest possible comparability [
33]. Clinically meaningful difference can be defined as the least change health-care professionals or the study participants themselves may observe. In the present study, we used previously given thresholds of ≥ 0.07 for EQ-5D and ≥ 0.015 for 15D as clinically meaningful differences [
34,
35].
For the examination of separate dimensions of EQ-5D, dimension level 1 was considered as “no difficulties” while reported level 2–3 resulted “difficulties” in such dimension. For 15D, the individual dimensions were converted to 0–1 scale by applying the established Finnish multi-attribute utility weights. Validity, feasibility, and reliability of this method are further discussed in previous publication. [
36] For brevity, only ‘Vision’ dimension of 15 total individual dimensions included in 15D was reported separately here in addition to 15D index scores.
HRQoL was evaluated in both time points (2000 and 2011) to assess the changes during the follow-up period. Eye examination alongside the HRQoL assessment made it possible to evaluate the effect of declining vision on the quality of life.
Data analysis
The overall prevalence for distance and near VA grouped according to age was estimated by using population data at StatFin database (Statistics Finland). The data (collected and reported annually) were applied for both samples according to the investigation year.
Odds ratios for individual EQ-5D dimensions and correlations between measured distance or near VA and HRQoL index scores were conducted implementing the Complex Samples module in IBM SPSS Statistics for Windows, version 24 (IBM Corp., Armonk, N.Y., USA), to account for the complex sampling design. The analyses were adjusted for age, sex, and the most common comorbidities, specified below. P-values were adjusted with Bonferroni correction when making multiple comparisons. Changes in VA during the follow-up period and its impact on HRQoL changes were estimated through linear regression with adjustments for new incident diagnoses of the comorbidities, as well as baseline HRQoL. Variance inflation factors (VIFs) were used to measure multicollinearity in regression analyses.
To utilize the data more effectively, multiple imputations method was used to handle missing comorbidities in the regression analyses concerning the longitudinal changes [
37]. Missing data were predicted using respondent’s non-missing data in five imputations applying iterative Markov chain Monte Carlo method [
38]. After conducting the five imputations, the estimates of the variables with previously missing values were pooled to give single estimates to be utilized in the final analysis. Missing VA changes were not imputed.
When comparing the data between the time points, the weighting scheme calculated by National Institute for Health and Welfare was applied to account for the intentional oversampling in 2000 time point as well as the loss to follow-up. The sampling scheme is based on IPW-method (reverse probability), further discussed in the previous publications. [
39,
40] Subgroups with minimum size of three participants were included in population analysis. Age distributions for both time points, taken from population statistics in the StatFin (Statistics Finland) database, were applied to better represent the impact of declining VA population-wise. Kendall’s tau-B test and/or regression models were used when estimating the associations between continuous and ordinal variables.
Comorbidities
For most of the analyses, common diseases were considered to account for their potential impact on the HRQoL. The diseases were self-reported both in 2000 and 2011 and were classified to major comorbidity groups for robustness. Myocardial infarction, Angina Pectoris, heart failure, rhythm disorders, and “other heart disorder” were considered as “Heart diseases.” Asthma, chronic obstructive pulmonary disease (COPD), chronic bronchitis, and “other pulmonary disease” were categorized as “Pulmonary diseases.” “Vascular diseases” included stroke and varicose veins in lower limbs. “Musculoskeletal conditions” included self-reported rheumatoid arthritis, arthrosis, fractures, and osteoporosis. “Psychiatric diseases” consists of psychotic disorders, depression, anxiety, psychoactive substance abuse, or “other psychiatric disease.” In addition, hypertension, diabetes, Parkinson’s disease, and cancer (unspecified) were included in our model.
Study participant was considered to have comorbidity, if participant reported having any of the conditions included in the comorbidity group. When examining new incident diagnoses during the follow-up period, each condition was scrutinized in 2000 baseline and in 2011 follow-up. If study participant reported one or more new condition included in the given comorbidity group during 2011 follow-up, participant was classified as having incident comorbidity, regardless of the presence of other conditions included in that specific comorbidity group in baseline.
Discussion
Visual acuity plays a major role in self-reported HRQoL, the threshold of 0.5 being notable. Subjects with VA below this threshold have greater and progressive decreases in their HRQoL. Functionally, the threshold of 0.5 in distance VA is relevant; as with current visual correction, it serves as the requirement for standard driving license in many states in the United States of America [
41] and the countries of the European Union (directive 2006/126/EC), thereby affecting individual’s activities of daily living. Moreover, previous publication reports differences in self-care when VA is near the threshold of 0.5 (from 0.4 to 0.63) [
42]. Our results point out that low vision correlates with low HRQoL. The impact of having lower VA is also clinically meaningful and does not impact only single dimension of the assessments applied.
The odds of having difficulties in individual EQ-5D dimensions become greater as VA is lower, which is the most evident in the areas of self-care and usual activities. Those with weak VA are more than twice as likely to experience problems in those two dimensions, possibly increasing the need for daily assistance, compared to those with good VA. With VA being impaired or study participants blind, the odds progressively accumulate to over tenfold.
Visual acuity plays a major role in self-reported HRQoL, especially in the dimensions of usual activities and self-care. The magnitude of association between EQ-5D dimensions associated and the lower VA status decreased between the time points. Especially, for mobility, the impact of low VA is evident in baseline, while in 2011 the impact is no longer statistically significant. During this period, improved mobility aids, such as the Segway Personal Transporter, electric travel aids including GPS-locating, and tactile gloves have been implemented, which might have contributed for those with vision loss not experiencing difficulties with mobility, self-care, and usual activities [
43‐
46]. Moreover, voice-controlled applications and devices have become routinely used. Environmental architecture and paying attention to accessibility may also contribute to the improvement, although there are still many aspects that complicate usual activities for those with visual disability [
46]. Observed improvement in anxiety/depression dimension of EQ-5D between the time points may be partly due to improvements in availability and accessibility of social services between 2000 and 2011.
Low VA was associated with increased Anxiety/depression in the present study, and this has also been shown in previously conducted research [
8,
47]. In our study, the impact of lower VA status in observed HRQoL became prominent only once vision was impaired or severely lost. While not in focus of the present study, in a supplemental analysis we examined the relation between anxiety/depression and measured VA, using Beck Depression Inventory (BDI), which purely concentrates on evaluating depression and mental well-being (see Supplementary Data). Unlike with anxiety/depression dimension of EQ-5D, where declining vision appears to have an impact only once vision is impaired or severely lost, BDI-scores and VA seem to have a linear connection. This promotes the endeavor to maintain visual function even before the onset of visual impairment. Due to the differences in questionnaires, EQ-5D is not as sensitive in identifying the early signs of anxiety/depression, applying also for 15D assessment. There may also be a psychological explanation for this, which is related to the question layout. In EQ-5D, the question about anxiety/depression is more straightforward with a limited answer scale, and study participants might not always identify depression and anxiety correctly. In BDI on the other hand, the subject is presented with a wide range of questions about the various, specific background factors contributing to depression, giving a more thorough analysis of the subject’s mental state.
The longitudinal findings from 15D assessment show similar changes in the index values. An important difference between 15D and EQ-5D assessments is that 15D includes a Vision dimension, which had a strong statistically significant association with declining VA in our analyses. Additionally, the decline is more strongly associated with a decline in distance VA than with a decline in near VA. Overall, our findings emphasize the importance of maintaining good VA to prevent the incremental loss of HRQoL.
Both our results and those of The Los Angeles Latino Eye Study (LALES) indicate that HRQoL, while evaluated with different assessments, shows significant correlation to declining vision, suggesting that aging alone does not account for the decline in HRQoL [
21]. The LALES-study, while being a population-based study, unfortunately had a follow-up time of only 4 years. As a result, there were only 83 individuals whose VA declined in this time. These individuals appeared to have a slightly milder decrease (though statistically insignificant) in their general health (assessed via The National Eye Institute Visual Function Questionnaire (NEI VFQ-25)) when compared to those with stable VA. This is contrary to our findings, where declining VA caused decrease in HRQoL index values except for near VA applying EQ-5D assessment. For declining distance VA, HRQoL change assessed through 15D was also clinically meaningful. These findings may differ due to the differences in methods, follow-up time, study populations, or used HRQoL questionnaires.
The strengths of this study include a relatively long follow-up period of 11 years and a large study sample, which inclusively represents Finnish adult population aged 30 or older. Being widely collected and comprehensive, our study population and design reduced the impact of confounding factors. In comparison to samples collected from health-care units, our data do not consist of specific patient groups, which is a major strength.
Previously conducted research in German population-based sample confirms that VRQoL declines with age, assessed through NEI VFQ-25 questionnaire [
22]. Their approach, however, lacks longitudinal perspective thereby making it difficult to exclude the impact of aging from other changes in society during the lifetime of different age groups. In our study, we took into account the effect of aging and common comorbidities, thus reducing their impact on the results.
As internationally established questionnaires were implemented in assessing the HRQoL, this study is comparable to previous and future research about conditions unrelated to vision. Questionnaire-based data collection represents the study participants’ perception of their everyday QoL. EQ-5D might not always be sensitive enough to show statistically significant changes as reported by Jones et al. [
48]. However, when showing statistically significant differences, the specificity of this finding is evident in various health conditions [
49‐
51].
The data collection was well-executed and a high proportion of the subjects in 2000 study participated also in the follow-up study in 2011. Although previously reported in other studies [
52,
53], there were no significant differences in continuation in present study between males and females. Overall, the adherence to present study (58%) can be regarded good and in line with the previous studies, especially considering the trend of reduced participation in epidemiological studies during the past decades [
54,
55]. Loss to follow-up was also compensated by applying calibrated weighting scheme and the differences in samples were adjusted by weighting the analyses with IPW-method. [
39] Due to the study design, immigration to Finland after year 2000 has not been covered as the baseline sample defines characteristics for the weighting [
26]. Study participants participating in the eye examination in 2000 were similar in their distribution when compared to those who had complete information from both time points.
When examining the association between longitudinal changes in HRQoL and VA, we accounted for incidence of common comorbidities. We found that 15D and EQ-5D were more strongly associated with the declining distance VA than with any of the examined incident comorbidities, except musculoskeletal conditions when examining EQ-5D. Association between HRQoL and declining near was not as large. The absolute impact of Parkinson’s disease on HRQoL was somewhat higher than impact of declining distance VA.
Saarni et al. have previously reported impact of various health conditions on HRQoL [
56]. In their cross-sectional analysis, cataract and glaucoma have only marginal effect while the impact of macular degeneration is statistically significant and 15D-wise clinically important as well. However, in their analysis, health conditions are implemented as ‘averaged’ diseases, regardless of how much their vision or functioning is affected by the disease. In this work, we reported that longitudinal VA decline however has statistically significant and even clinically important impact on HRQoL. Comparing the loss of HRQoL associated with various health problems helps to put in perspective the potential impact achievable by the development of treatments for these conditions.
There are also potential limitations in our study. We were unable to assess the impact of other types of visual impairment, such as diminishing visual field, on HRQoL. The eye examination was carried out by a general practitioner, rather than an ophthalmologist, which can be considered as a weakness. However, it was necessary due to the large sample size and the complex study design. Another limitation was that we had to combine comorbidities into rather large groups (e.g., heart disease or psychiatric disorder) when estimating their impact on HRQoL, as new diagnoses during the 11-year follow-up are scarce for many specific diseases, such as psychosis. Moreover, we did not have the data available to differentiate disease types or severities, for example, specific cancer types, and had to account them as homogenous groups, regardless of the potential heterogeneity of specific diseases. Also, in the present analysis, we examined the general visual function and did not evaluate impact of specific vision-impairing diseases separately. For instance, age-related macular degeneration and glaucoma have been previously associated with declining QoL [
57,
58]. In the future analyses it would be beneficial to evaluate and compare HRQoL impact of disease-specific HRQoL impacts in longitudinal setting. The long duration of the follow-up might also cause problems, as there has been progression on the therapies for various conditions. Also, the diagnostics have developed, meaning that a growing proportion of diseases are being diagnosed even symptomless.
Concerning EQ-5D and 15D index scores, we included only study participants with complete information collected in each assessment. The proportion of study participants with complete HRQoL data and insufficient VA-data was quite low, ranging from 0.4 to 0.8%, and therefore potential bias resulted by it is also likely to be low. Taking the loss to follow-up into account, it needs to be acknowledged that the proportion of study population with complete information from both time points is somewhat lower. Even so, the study sample remains large (4703 participants) and representative. Additionally, the study population was predominantly Finnish, and the results may not be applicable to other countries and ethnicities. However, the comparability is somewhat improved by our use of UK time-trade-off weights for EQ-5D instead of the Finnish preference weights.
Although the findings of the present study are based on large, representative Finnish population-based samples, more detailed analyses with large population-based study samples are required to validate the generalizability of these results into other settings as well. Moreover, additional studies with over 10 years of follow-up are necessary to ascertain the impact of declining vision on HRQoL.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.