Equating Visual Function Scales to Facilitate Reporting of Medicare Functional G-Code Severity/Complexity Modifiers for Low-Vision Patients

Abstract

Objective: To present a method of estimating and equating scales across functional assessment instruments that appropriately represents changes in a patient's functional ability and can be meaningfully mapped to changes in Medicare G-code severity modifiers.

Design: Previously published measures of patients' overall visual ability, estimated from low-vision patient responses to 7 different visual function rating scale questionnaires, are equated and mapped onto Medicare G-code severity modifiers.

Setting: Outpatient low-vision rehabilitation clinics.

Participants: The analyses presented in this article were performed on raw or summarized low-vision patient ratings of visual function questionnaire (VFQ) items obtained from previously published research studies.

Interventions: Previously published visual ability measures from Rasch analysis of low-vision patient ratings of items in different VFQs (National Eye Institute Visual Functioning Questionnaire, Index of Visual Functioning, Activities of Daily Vision Scale, Visual Activities Questionnaire) were equated with the Activity Inventory (AI) scale. The 39 items in the Self-Report Assessment of Functional Visual Performance (SRAFVP) and the 48 items in the Veterans Affairs Low Vision Visual Functioning Questionnaire (VA LV VFQ) were paired with similar items in the AI in order to equate the scales.

Main outcome measures: Tests using different observation methods and indicators cannot be directly compared on the same scale. All test results would have to be transformed to measures of the same functional ability variable on a common scale as described here, before a single measure could be estimated from the multiple measures.

Results: Bivariate regression analysis was performed to linearly transform the SRAFVP and VA LV VFQ item measures to the AI item measure scale. The nonlinear relationship between person measures of visual ability on a logit scale and item response raw scores was approximated with a logistic function, and the 2 regression coefficients were estimated for each of the 7 VFQs. These coefficients can be used with the logistic function to estimate functional ability on the same interval scale for each VFQ and for transforming raw VFQ responses to Medicare's G-code severity modifier categories.

Conclusions: The principle of using equated interval scales allows for comparison across measurement instruments of low-vision functional status and outcomes, but can be applied to any area of rehabilitation.

Keywords: Medicare; Occupational therapy; Outcome assessment (health care); Rehabilitation; Vision, low.

Figure 1

Item measure scatter plots with orthogonal regression lines for a) SRAFVP vs. AI and b) VA LV VFQ vs. AI. The r² values for the regression lines are 0.936 for the SRAFVP and 0.933 for the VA LV VFQ.

Figure 2

Mapping of relative SRAFVP raw scores to the equated visual ability scale (solid line) and paired relative raw score/functional ability measures for 102 visually impaired patients (points). The departures from the curve reflect different patients responding to different subsets of the 38 items in the SRAFVP, which effectively changes the shape of the mapping function across patients.

Figure 3

Illustration of empirical relationships between visual ability and the relative raw scores (i.e., raw score rescaled to range from minimum = 0 to maximum = 1) with an equated visual ability scale on the ordinate for the 7 different visual function rating scales. Shallow slopes indicate a small range of functional ability measures for those instruments and steep slopes indicate a large range of functional ability measures. The arrows indicate the visual ability measure corresponding to a normalized raw score of 0.2 for each of the 7 instruments.

Figure 4

Visual ability measure/normalized raw score pairs estimated from AI responses of 3177 (combined LVROS and legacy data sets) visually impaired patients (points) along with the average mapping function for the sample. The G-code severity modifier categories represent divisions of the visual ability axis rather than raw scores. A “ceiling” and “floor” for the data set are created 2 standard deviations from the mean.

Figure 5

The definition of G-code severity modifiers can be applied to any VFQ score or other indicator variable that has been transformed to an interval visual ability scale equated with the AI visual ability scale.