Abstract
Occupational therapy practitioners who provide low vision rehabilitation require assessment tools that focus on vision-dependent activities of daily living (ADLs) to accurately identify the limitations clients are experiencing because of their vision loss and to provide effective interventions. The Self-Report Assessment of Functional Visual Performance (SRAFVP) was developed in 1995 as a standardized self-report occupational therapy assessment to identify limitations in completing vision-dependent ADLs in older adults with low vision (Gilbert & Baker, 2011). Some of the 38 task items on the form have become dated, necessitating a two-phase project to revise the form to accurately represent tasks that older adults with low vision complete in today’s age of digital technology.
In Phase 1 of the project, we updated and validated the content of the revised form (Zemina, Warren, & Yuen, in press) using occupational therapy experts in low vision rehabilitation and consumers with age-related eye disease and low vision to produce a 50-item draft of the Revised SRAFVP (R–SRAFVP). The purpose of this Phase 2 study was to evaluate the measurement properties of the R–SRAFVP.
Method
Study Design
In this study, we used a cross-sectional survey research design to establish the construct validity of the R–SRAFVP for older adults with low vision. The University of Alabama at Birmingham institutional review board approved the study.
Participants
One hundred two participants were recruited via convenience sampling from three occupational therapy low vision clinics in three geographic regions (Florida, Oklahoma, and Pennsylvania) between January and October 2016. Eighty-seven people met the eligibility criteria. Inclusion criteria included age 60 yr or older and diagnosed age-related eye disease resulting in low vision. Exclusion criteria included vision impairment from brain injury, congenital eye disease, or moderate risk for cognitive impairment defined as having a score of 5 or less on the Short Portable Mental Status Questionnaire (Pfeiffer, 1975). Participant characteristics are described in Table 1.
Sample Characteristics (N = 87)
Note. M = mean; SD = standard deviation.
Participants could have multiple eye disease diagnoses.
Based on the classification of visual impairment used by the American Optometric Association (2017).
Based on the Mars Numeral Contrast Sensitivity Test.
Procedures
Four licensed occupational therapists with a mean of 12 yr work experience in low vision rehabilitation (range = 2–28 yr) administered the 50-item draft of the R–SRAFVP to the participants. All had formal postprofessional graduate education in low vision rehabilitation, and three were credentialed as Specialty Certified in Low Vision by the American Occupational Therapy Association. The therapists provided low vision rehabilitation services to older adults for a minimum of 20–30 hr/wk and routinely administered the original 38-item SRAFVP. Before data collection, the occupational therapists completed a 90-min training session on administering the study protocol and assessments via the GoToTraining online platform (LogMeIn, Inc., Boston).
After obtaining informed consent, each occupational therapist administered the R–SRAFVP as part of the initial assessment before starting low vision rehabilitation. The occupational therapist read aloud each R–SRAFVP item and description and asked the participant to verbally rate his or her ability to complete the task using the following 4-point rating scale: 1 (great difficulty: may perform some aspect of the task but requires assistance for the majority of the task, or cannot perform the task in a safe and efficient manner), 2 (moderate difficulty: performs the task with difficulty even under optimal conditions—i.e., difficulty performing the task in a timely manner or questionable safety and efficiency—or makes errors, or assistance is required for a substantial part of the task), 3 (minimal difficulty: performs the task with some difficulty—i.e., can perform the task only under optimal conditions—or may require assistance for a small part of the task), and 4 (independent: experiences no difficulty performing the task safely, accurately, and efficiently); the final option was not applicable (NA; does not perform, or is no longer performing, the task for various reasons).
The occupational therapists recorded participants’ responses on the R–SRAFVP form. Information on age, gender, race, marital status, living arrangement, education level, number of years since primary eye disease diagnosis, and self-reported comorbid chronic conditions was also collected. They also evaluated the participant’s contrast sensitivity level using the Mars Letter Contrast Sensitivity Chart (Mars Perceptrix, Chappaqua, NY) according to manual directions (Arditi, 2010). The Mars Chart produces a linear log contrast value between 0.04 and 1.92 that converts to four impairment categories for people older than age 60: normal (1.52–1.76), moderate impairment (1.04–1.48), severe impairment (0.52–1.00), and profound impairment (<0.48; Arditi, 2010).
Participants’ primary eye disease diagnosis and best corrected Snellen distance visual acuity were extracted from the medical record. We adopted the classification of visual impairment used by the World Health Organization (American Optometric Association, 2017) to categorize the participants’ visual acuity as mild impairment (20/30 to 20/60), moderate impairment (20/70 to 20/160), severe impairment (20/200 to 20/400), or profound or blind impairment (20/500 to 20/1,000).
Data Analysis
The principal-components analysis (PCA) factor-extraction method was used to extract the minimum number of components required to adequately describe the underlying factor structure of the R–SRAFVP (Hair, Tatham, Anderson, & Black, 1998). Nine items (3, 11, 21, 23, 26, 29, 37, 38, and 44) containing 30% or more NA responses were eliminated before PCA to improve the psychometric properties of the instrument (Moret et al., 2007). For the remaining 41 test items, the mean (M) ± standard deviation (SD) percentage of NA responses was 7.6 ± 5.7 (range = 0%–24.1% for individual items). NA responses were coded as “great difficulty” because it was likely that participants provided this rating if they either required considerable assistance or could not perform the task in a safe and efficient manner. Moreover, there is little difference in parameter estimates when missing questionnaire data are less than 10%, regardless of the missing data patterns (Roth, 1994).
Subsequently, factorability of the 41 items was examined to ensure that the item correlation matrix was appropriate to identify distinct and reliable components. Several criteria were used, including diagonal values in an anti-image correlation matrix (a measure of sampling adequacy of individual items), the Kaiser–Meyer–Olkin (KMO) statistic (a measure of sampling adequacy or shared variance in the items), and Bartlett’s test of sphericity for testing that all correlations are zero (Hair et al., 1998). Item 4 on the diagonal of the anti-image correlation matrix was less than 0.5 and was eliminated. The remaining 40 items showed good sampling adequacy with a KMO value of 0.76 (recommended value 0.6), and Bartlett’s test of sphericity was significant (p < .001), indicating that the data were suitable for factor analysis (Hair et al., 1998).
The 40 items were then refined and validated through PCA and item analysis. We determined the number of components extracted using the Guttman–Kaiser eigenvalue >1 rule (Hair et al., 1998). Varimax orthogonal rotation was set to achieve a simple structure and to facilitate interpretation of the factor solution. An item was retained if its factor loading was ≥0.4 (in absolute value) because this value strongly explains the amount of variance of a factor (Hair et al., 1998). Cronbach’s α was used to estimate the internal consistency reliability of each component. We computed the scale content validity index (scale-CVI) for the final R–SRAFVP version on the basis of the percentage of agreement among the eight occupational therapy experts’ ratings of the relevance of each item in Phase 1 of the project (Lynn, 1986).
To provide additional evidence of construct validity, we used a known-groups method to support that the R–SRAFVP scale score was able to distinguish between (1) participants with mild impairment in distance visual acuity (including central or ring scotoma) and those with moderate, severe, or profound impairment and (2) participants with normal visual contrast sensitivity and those with moderate, severe, or profound impairment. Parametric or nonparametric statistics were used to test the hypotheses depending on data normality. We tested the hypotheses using two-sided tests with α set at .05. IBM SPSS Statistics (Version 24; IBM Corp., Armonk, NY) was used to conduct data analysis.
Results
Preliminary PCA of the 40 items yielded 11 components, which was more than expected. The initial eigenvalues of the 11 components were [11.43, 3.47, 2.47, . . . , 1.26, 1.19, and 1.01]; each explained [28.6%, 8.7%, 6.2%, . . . , 3.2%, 3.0%, and 2.5%] of the variance, respectively. We examined solutions for 11, 10, 9, and subsequent components using varimax rotation of the factor loading matrix. The nine-factor solution, which explained 67% of the variance, was preferred because it preserved a high number of items with significant primary loadings on distinct components and produced a largely logical item grouping that conceptually aligned with the original categories and item grouping envisioned by the research team. The extracted communalities of the 40 items (i.e., proportion of variance in each item explained by the retained components) were all more than 0.50 except Item 5 (M = 0.67, range = 0.44–0.81), which are well represented in the common factor space (Field, 2005). We dismissed factor solutions with eight and fewer components because they were difficult to interpret (e.g., the grouping of items into components was odd and lacked a coherent structure).
Having set the nine components to be extracted, we eliminated seven items (9, 13, 19, 20, 34, 43, and 46) after varimax rotation because they had factor loadings of ≥0.4 on two components. At the end of the item reduction phase, 33/40 items were retained and formed nine components: Personal Care (6 items), Oral Care (1 item), Meal/Laundry Preparation (6 items), Financial Management (3 items), Telephone Usage (2 items), Personal Preference Activities (4 items), Reading (3 items), Writing (4 items), and Functional Mobility (4 items). The final 33 item solution is presented in Table 2.
Principal-Components Analysis of the Revised SRAFVP, Including Component Structure, Item Loading Coefficients, Eigenvalues, Percentage of Variance Accounted for by Each Factor, and Reliability Coefficients
Note. Item loading coefficients > |0.40| that are not cross-loading are
Item description included digital technology.
Item 4 with an anti-image correlation coefficient of <.5 was deleted.
The internal consistency reliability coefficients as estimated by the Cronbach’s α of each component ranged from .72 to .84, which is considered acceptable to good in the early stage of instrument development (George & Mallery, 2003). No substantial increases in the α coefficients could have been achieved by eliminating an item in its respective component. The corrected item-to-total correlations between scores of an individual item and the summation score of the remaining items in each component were all more than 0.3 (moderate correlation). The α coefficient of the R–SRAFVP was .92 (95% confidence interval [.89, .94]), which is excellent. The scale-CVI for the final R–SRAFVP version, which is the average of the item-CVI of the 33 items (Polit, Beck, & Owen, 2007), was .94: 20 items achieved unanimous agreement (CVI = 1), 9 items had a CVI of .88, and 4 items had a CVI of .75.
Mean composite scores were created for each of the nine components on the basis of the item scores that had primary loadings on a distinct component and for the R–SRAFVP scale. Higher scores indicated less assistance required to complete the tasks on the R–SRAFVP. All subscales significantly correlated with one another (except correlations between Oral Care and Telephone Usage and between Oral Care and Functional Mobility) and with the R–SRAFVP scale score (ps < .05).
For the known-group comparisons, independent-samples t tests revealed that participants with mild impairment in visual acuity (better than 20/60, or central or ring scotoma; M = 2.73, SD = 0.46) scored significantly higher on the R–SRAFVP than participants with moderate to profound visual acuity impairment (M = 2.45, SD = 0.57), t(85) = 2.18, p = .032, Cohen’s d = 0.54. R–SRAFVP scores were also significantly higher for participants with normal contrast sensitivity (M = 2.69, SD = 0.37) than participants with impaired contrast sensitivity (M = 2.47, SD = 0.60), t(60) = 2.03, p = .047, d = 0.44. Levene’s test for equality of variances indicated unequal variances scores between the two groups, and degrees of freedom were adjusted from 85 to 60.
Discussion
We found that nine distinct components encompassed the underlying responses of the participants on the R–SRAFVP items and that these components were moderately internally consistent. The finding of moderate internal consistency indicates that the items on each subscale measured aspects of the same construct (vision-dependent ADLs). The significant correlations among the subscales indicate that they are related and measure vision-dependent ADLs. Participants with moderate to profound loss in visual acuity or contrast sensitivity reported greater difficulty completing the tasks on the R–SRAFVP than those with milder losses, which supports the construct validity of the assessment. Although 17 items were eliminated, the original nine-component structure was retained. The items forming each component were largely consistent with those originally envisaged in the content validation phase. Six of the nine categories were almost identical to the original categories, including Personal Care, Financial Management, Telephone Usage, Reading, Writing, and Functional Mobility.
The R–SRAFVP categories form a logical structure. Item 45 (shopping—locating item/price) grouped under Personal Care made sense when considered in the context of shopping to procure items to complete health management and personal grooming such as medications and shampoo. A new category—Meal/Laundry Preparation—emerged from the separate categories of meal preparation and clothing care on the original SRAFVP. These important instrumental ADLs require similar visual performance skills: ability to use vision to read instructions, identify settings, and manipulate devices (e.g., fill measuring cups, or set a stove or washing machine dial). The items in the new single category match those originally envisaged for the separate categories. Regrouping of two items from the original leisure category with shaving (Item 6) and reading a timepiece (Item 36) seems incongruous until one considers them as personal preference activities that do not absolutely need to be completed.
Oral Care (Item 7) significantly loaded on a single distinct component but could be conceptually included with Personal Care because the α coefficient for Personal Care remained at .72 after including the item. Item 39, reading—credit/debit cards (which was grouped under the Writing category rather than the Reading category), was not as envisioned or conceptually accurate. Instead of deleting this item, which would have reduced the representativeness of the content, we moved it to the Reading category. Removing Item 39 from the Writing category increased its α coefficient slightly from .78 to .81, indicating that the item did not fit well with the underlying construct of Writing. Moving Item 39 to the Reading category decreased the α coefficient slightly from .78 to .73, but it improved the clinical utility of the assessment. After this minor regrouping, all eight subscales significantly correlated with one another and with the R–SRAFVP scale score.
Several items added during the content validation phase were not included in the final R–SRAFVP version, including Items 23, 26, 37, 38, and 44 that focused on the use of personal computer and portable electronic devices (see Table 2). Although occupational therapy experts in the content validation phase encountered clients using these devices and thus recommended their inclusion on the R–SRAFVP form, more than 40% of participants rated them as NA (range = 41.4%–62%) for Financial Management, Reading, and Writing. The participants in this study reported that digital technologies were largely relevant for using the telephone, leisure, or media entertainment.
Limitations
The smaller sample size was a limitation. For moderately overdetermined factors with 3–7 items per factor, a sample size of 100–200 is preferred for stable and precise estimates of population loadings, and a minimum sample size of 300 is needed for exploratory factor analysis to obtain a good recovery of population factors (MacCallum, Widaman, Zhang, & Hong, 1999). However, the nine distinct components from the PCA along with moderate internal consistency reliability for component items were congruent with that originally envisaged. Therefore, the results of this preliminary study indicated that the R–SRAFVP demonstrated sufficient reliability and validity to evaluate the ability of older adults with low vision from age-related eye disease to complete vision-dependent ADLs.
Implications for Occupational Therapy Practice
The findings of this study have the following implications for occupational therapy practice:
The R–SRAFVP addresses the key ADLs that older adults with low vision most frequently report having difficulty completing and as such are the tasks that will most likely require intervention.
Focusing on the 33 highly relevant tasks on the SRAFVP increases assessment efficiency by enabling occupational therapy practitioners to more quickly identify the client’s most likely areas of performance limitation in vision-dependent ADLs.
Conclusion
The R–SRAFVP is a standardized self-care assessment that focuses on vision-dependent ADLs. It measures the tasks that older adults with age-related eye disease most often report as being difficult to perform because of their vision loss. This revised version of the original SRAFVP published in 1995 consists of 33 items that are grouped into nine components: Personal Care, Oral Care, Meal/Laundry Preparation, Financial Management, Telephone Usage, Personal Preference Activities, Reading, Writing, and Functional Mobility. New items included in the R–SRAFVP reflect the increased use of digital technology by older adults for leisure, media entertainment, and communication. Construct validity of the R–SRAFVP was supported by the finding that participants with moderate to profound loss of visual acuity or contrast sensitivity reported greater difficulty completing the tasks on the R–SRAFVP than those with milder visual loss.
Footnotes
Acknowledgments
We thank Deann Bayerl, Yolanda Cate, and Sarah Rodenburg for their invaluable assistance with data collection for this study. We also thank the adults with low vision who generously participated in the study.