Abstract
Purpose:
Outline the association between age-related macular degeneration (AMD) and functional difficulty using novel item response theory (IRT) psychometric techniques, and highlight populations particularly at risk of functional impairment.
Methods:
This cross-sectional study included 5604 US adults. Primary outcomes were item response theory–adjusted visual and physical difficulty scores. Secondary analyses of AMD populations at highest risk of reporting greater functional difficulty were undertaken.
Results:
In total, there were 386 participants with early AMD (mean presenting visual acuity [pVA], 0.12) and 55 with late AMD (mean pVA, 0.35). Those with late AMD reported substantially higher item visual difficulty, whereas those with both early/late AMD reported significantly higher item physical difficulty versus those with no AMD (P < .05). In univariate regression, only those with late AMD reported significantly higher visual difficulty versus those with no AMD (10.1 points [95% confidence interval (CI), 8.2–12.1 points] vs 7.1 points [95% CI, 7.0–7.2 points]; P = .003). Both early/late AMD reported higher physical difficulty versus those with no AMD (11.6 points [95% CI, 11.1–12.1 points; P = .005]; 13.4 points [95% CI, 11.8–15.0 points; P = .03], respectively, versus 11.0 points [95% CI, 10.9–11.1 points]. After adjustment for sociodemographic and medical variables (excluding pVA), only those with late AMD reported significantly greater visual and physical difficulty versus those with no AMD (10.0 points [95% CI, 8.2–11.9 points] vs 7.1 [95% CI, 7.0–7.2 points; P = .002]; and 12.7 points [95% CI, 11.3–14.0 points] vs 11.0 [95% CI, 10.9–11.1 points; P = .02], respectively); greater visual difficulty in those with late AMD persisted after additionally adjusting for pVA versus those with no AMD (9.1 points [95% CI, 7.6–10.6 points] vs 7.1 points [95% CI, 7.0–7.2 points; P = .01]). Among individuals with AMD, lower income, higher medical comorbidities, depression, and pVA predicted greater visual and physical difficulties.
Conclusions:
AMD confers significant functional difficulty among US adults with sociodemographic characteristics influencing dysfunction; highlighting the value of alternatives to Snellen visual acuity in assessing visual characteristics. With aging populations and the increasing prevalence of AMD, health care professionals should be aware of the functional burden of AMD and recognize those at higher risk of functional difficulty.
Translational Relevance:
Contemporary psychometric validation techniques can be effective in accurately describing the level of functional impairment for those with visual impairment.
Functional visual difficulty was assessed by self-report using the following 6 questions: (1) reading small print, (2) up-close housework, (3) seeing curbs or stairs in dim light, (4) noticing objects in peripheries, (5) finding things on a crowded shelf, and (6) daytime driving in a familiar place. Each of the 6 questions was reported on a 5-point ordinal Likert scale ranging from (1) no difficulty, (2) a little difficulty, (3) moderate difficulty, (4) extreme difficulty and (5) unable to do because of eyesight. Options (6) don't do this for other reasons, (77) refused, and (99) don't know, were considered as missing variables.
Functional physical difficulty was assessed by self-report using 13 questions pertaining to problems with (1) managing money, (2) walking a quarter mile, (3) walking up 10 steps, (4) crouching, stooping or kneeling, (5) performing household chores, (6) preparing meals, (7) walking between rooms (same floor), (8) using cutlery, (9) getting dressed, (10) grasping small objects, (11) going to the movies, (12) attending social events, and (13) doing leisure activities at home. Each of the 13 questions was reported on a 4-point ordinal Likert scale ranging from (1) no difficulty, to (2) some difficulty, (3) much difficulty, and (4) unable to do. Options (5) do not do this activity, (7) refused, and (9) don't know, were considered as missing variables.
In this study, we found that US adults with early and late AMD experience significantly greater visual and physical difficulty across a variety of daily tasks. The association between late AMD and self-reported visual difficulty persisted even after adjustment for sociodemographic characteristics, medical comorbidities, and pVA. However, the association between late AMD and self-reported physical difficulty appeared to be driven primarily by level of VA impairment. Among participants with AMD, those with lower income, medical comorbidities, or depression represented a subset of the population more likely to report higher levels of visual and physical difficulty. Our results reinforce the negative impact of AMD-related vision impairment on not only visual functioning but also physical functioning in older US adults and identify groups particularly at risk of functional difficulty.
AMD-related vision loss has been associated with poor visual function and vision-related quality of life,
17,39 consistent with present findings. As expected, late AMD in particular was associated with substantially worse pVA (compared with early or no AMD) and conferred greater visual difficulty, consistent with previous findings outlining impaired pVA as a risk factor for visual difficulty.
23–26,40 However, although impaired pVA may in part explain the association between late AMD and vision-related difficulty, the association persisted even after accounting for pVA. This finding may highlight the importance of using multiple visual assessment tools to fully capture the visual symptoms experienced by patients with late AMD. For example, visual symptoms like poor contrast sensitivity, low luminesce VA, and critical print size are not necessarily well-captured by a Snellen chart alone, but are demonstrably influential in determining the severity of visual functional difficulty for those with AMD.
25,26 A lack of other measurement tools in the NHANES data was a limitation to the present study.
The association between AMD and poor physical function has also been previously described, although inconsistently. Although some studies have found that those with late AMD had significantly lower mobility versus those with early to intermediate AMD
23 and that patients with AMD (late AMD in particular) were more likely to experience difficulty performing activities of daily living, even when adjusting for pVA,
14 other investigators have found no significant association between AMD and physical difficulty when adjusting for pVA.
13 We similarly found that, although the association between self-reported physical difficulty and late AMD persisted after adjustment for known covariates like age and comorbid illness, this effect seemed to be driven mostly by pVA. This finding highlights that VA may be an independent driver of physical difficulty experienced by those with AMD.
In a secondary analysis restricted to participants with AMD, we found that those living below the poverty line, with greater medical comorbidities, or with depression experienced greater visual and physical difficulties. This analysis highlights a group of patients with AMD particularly susceptible to experiencing functional difficulty, and it is important to monitor them closely. There have also been strong links between poverty and functional difficulty demonstrated elsewhere, particularly in low to middle income countries.
41 However, associations remain relatively unstudied in high-income countries like the United States. Poverty and functional difficulty are believed to operate in a cycle, each reinforcing the other.
41 Indeed, we presently demonstrate that, among those with any AMD, poverty is independently associated with greater visual and physical difficulties, even when accounting for sociodemographic features and medical comorbidities. Our findings from a nationally representative population in a high-income countries add to the current body of literature, outlining a group of AMD patients susceptible to experiencing greater functional difficulty.
The association between medical comorbidities and functional impairment is also well-described
42,43; however, it is usually attributed to older age. Here we found that, even when accounting for age, medical comorbidities conferred greater visual and physical difficulties among those with AMD. This finding highlights that specific consideration for functional impairment should be given when assessing AMD patients with medical comorbidities, irrespective of age. Although we were unable to analyze early and late AMD groups separately, we found that those with late AMD were more likely to have more medical comorbidities, as has been described elsewhere,
44–46 suggesting that this group may be at even greater risk of functional difficulty. Equally, there is a well-described association between depression and functional impairment,
47,48 although this association was not previously described for AMD patients to our knowledge. Interestingly, the prevalence of depression in our cohort was substantially lower than has been previously reported for those with AMD and visual impairment.
49,50 This finding may be incidental or indicative of a sampling limitation (those with depression may be less likely to agree to NHANES participation), which may be representative of the significance of nonparticipation as a defining characteristic of depression.
51 Either way, the association between depression and AMD and visual impairment is well-described, and an integrative management approach for those with low vision should always be considered. Additionally, although we did not find higher rates of depression among our AMD cohort in contrast with prior studies,
50,52,53 depression was independently associated with functional difficulty among those with AMD. With the potential additive and multiplicative effects of medical comorbidities and depression on functional difficulty and quality of life,
54,55 our findings suggest that patients with AMD with either medical comorbidities and/or depression are at particular risk of experiencing functional difficulty.
Here we applied IRT to allow for more realistic reflection of what each questionnaire is measuring, and how reliably each question contributes to that measurement.
56 Although only 1 dimension was found in factor analysis for both questionnaires here, making the use of alternative psychometric validation techniques like Rasch possible, we believe IRT is particularly suited for a number of reasons. IRT is especially well-suited for analyzing questionnaire data measuring clinical latent traits, such as visual difficulty and physical difficulty.
57 The GRM (the IRT model chosen for the current study) is further suited for validation of ordinal-scaled Likert-type questionnaires for a number of reasons. GRMs are less constrained than other psychometric validation techniques used in previous studies (i.e., Rasch models or partial credit models) in that they have fewer assumptions (such as assuming equal question discrimination [imposed by Rasch models]), impose fewer restrictions on the data, and are appropriately robust to handle larger datasets with distributions slightly deviated from normal if required.
58,59 For example, whereas Rasch models may be mathematically similar, they constrain item discrimination (assuming equality between items), whereas the IRT GRM used presently does not assume equal discrimination between items, and strives to model data to better fit each item.
35,60 By subsequently imputing modified questionnaire scores, we believe the psychometric models used here provides more accurate comparisons between groups. Further, the GRM is also unique in that it has been demonstrated as particularly appropriate as a robust model for large datasets with distributions deviated from normal.
58,59 Importantly, here we found that only 9 of the original 13 physical difficulty questions reliably measured that latent trait. And notably, by ascribing unique weights to the remaining visual and physical difficulty questions (based on imputed discrimination and difficulty properties) and calculating rescaled, modified scores using GRM validation techniques, we could demonstrate participants’ true level of experienced visual and physical difficulties with a greater degree of accuracy than using nonpsychometrically validated questionnaire scores.
The strengths of the present study include its use of a large, nationally representative sample and the objective measurement of AMD severity. Sociodemographic factors, medical comorbidities, and VA were well-controlled for because of the comprehensive and standardized questionnaire and examination techniques used to collect NHANES data, which allowed for associations to be described in well-adjusted models. The strength of the potential association between AMD and visual and physical difficulties has been highlighted here, given the significance of the results despite the relatively small effect size. Additionally, psychometric validation methods allowed for more accurate, rescaled, modified visual and physical difficulty scores to be compared between patient groups and by AMD severity, which, to the knowledge of the authors, has not been previously used for comparable studies.
Our study is subject to a number of limitations. First, our results may have been influenced by nonparticipation. There were significant demographic differences between subjects with and without missing retinal imaging data (e.g., those with missing data were more likely to have impaired pVA). Because some of these factors are known to (and shown here to) influence visual and physical difficulty, this may have led to over or underestimation of functional difficulty in these subpopulations. For example, the surprisingly low individual item scores (even for those with late AMD) may be reflective of the fact that those with impaired pVA were more likely to have missing retinal imaging data and thus not were included in mean item difficulty score calculations. Such low item scores may also be a reflection of the fact NHANES only documented AMD grading of the worse eye, even though the better-seeing eye (with less severe AMD) usually determines functional vision; as participants compensate with it. We adjusted for this factor presently by using pVA in the better-seeing eye. The low scores may also reflect an under-reporting of symptoms in such a population, although the reasons for this pattern are not able to be determined presently. Interestingly, responses for “noticing objects in peripheries” (presumably associated with peripheral vision) loaded strongly onto the same factor as the other (central vision) visual items. This finding may outline a limitation of using so few questions (i.e., only 6 total visual questions in the dataset); a more comprehensive visual difficulty questionnaire including items relating to peripheral vision may have yielded a second strong factor. Moreover, we relied on self-reported data, and reporter bias may not have affected all demographics equally. Subjective measures of function may not correlate well with objective outcomes such as reading speed or falls and other adverse outcomes. Finally, the relatively small number of participants with late AMD meant that exploratory analyses of subpopulations at increased risk of reporting visual and physical functional difficulty were underpowered when stratified by AMD severity, resulting in us combining early and late AMD categories for these analyses. To mitigate this problem, we used pVA in these models to partially account for AMD severity.
In conclusion, we found that US adults over 40 years of age with late AMD experience substantially greater visual and physical difficulties compared with those without AMD, independent of sociodemographic characteristics and medical comorbidity status. Although self-reported visual difficulty seems to be independent of pVA, the association between AMD and physical difficulty seems to be driven largely by the level of VA impairment. Individuals with AMD with a lower income, more medical comorbidities, and depression experienced significantly greater visual and physical functional difficulties. With an aging US population and the increasing global prevalence of AMD, an awareness of the functional burden of AMD and those particularly at risk, as well as an assessment of visual characteristics other than Snellen pVA alone, will facilitate a comprehensive approach to the assessment and management of patients with AMD.