Psychometric properties and targeting of the psychosocial impact of assistive devices scale in seniors with low vision

Abstract

Purpose : The Psychosocial Impact of Assistive Devices Scale (PIADS) has been widely used to evaluate the effect of assistive devices on quality of life, as well as to gain a better understanding of device acceptance, utilization, and retention rates. The purpose of this research was to investigate item targeting for older low vision patients in a clinical trial of visual assistive smartphone apps.

Methods : Participants were 55+ years of age and had a score of ≥20 on the Telephone Interview of Cognitive Status (TICS). Their visual acuity was between 20/40-20/800, or had less than a 20°visual field in the better eye. They were naïve to study apps upon enrollment. Participants were randomized to one of three visual assistive apps (SuperVision+, Aira, or Seeing AI) and were provided training on a loaner iPhone. Participants completed the PIADS questionnaire for their most frequently used conventional low vision device and the visual assistive app at 3-months (n=88) and 6-months (n=76). Item and person measures were estimated using the method of successive dichotomizations (MSD) Rasch model. A separate “person” measure was estimated for each participant-device-time point combination to assess the effect of each assistive device on each participant at a single point in time. Targeting of PIADS items to our study cohort was assessed by comparing the estimated item and person measure distributions. Parameter standard errors and infit mean square statistics (infits) were evaluated to assess PIADS psychometric properties.

Results : There were 228 person-device combinations. PIADS items were not well targeted to our sample of low vision patients, with items having a bimodal distribution that brackets the person measure distribution (see Figure 1). The mean (SD) of estimated PIADS person measures was 1.57 (1.39) logit. The mean item measure in Rasch analysis is zero (by convention), and the standard deviations were 1.42 logit. Item measure standard errors (mean=0.24, SD=0.06) and person measure standard errors (mean=0.41, SD= 0.12) were relatively small (i.e., the estimates were relatively precise) and did not vary much. 23 out of 26 PIADS items had infits between 0.5-1.5. Mean (SD) PIADS person measure infits were 1.12 logit (0.81).

Conclusions : PIADS item measures were poorly targeted to this sample of low vision patients; items either had too much or too little impact on quality of life.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.