June 2015
Volume 56, Issue 7
Free
ARVO Annual Meeting Abstract  |   June 2015
Twenty Questions: An adaptive version of the PLoVR ultra-low vision (ULV) questionnaire
Author Affiliations & Notes
  • Gislin Dagnelie
    Ophthal-Lions Vision Cntr, Johns Hopkins Univ, Baltimore, MD
  • Michael P Barry
    Biomedical Engineering, Johns Hopkins University, Baltimore, MD
  • Olukemi Adeyemo
    Ophthal-Lions Vision Cntr, Johns Hopkins Univ, Baltimore, MD
  • Pamela E Jeter
    Ophthal-Lions Vision Cntr, Johns Hopkins Univ, Baltimore, MD
  • Robert W Massof
    Ophthal-Lions Vision Cntr, Johns Hopkins Univ, Baltimore, MD
  • Footnotes
    Commercial Relationships Gislin Dagnelie, None; Michael Barry, None; Olukemi Adeyemo, None; Pamela Jeter, None; Robert Massof, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2015, Vol.56, 497. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Gislin Dagnelie, Michael P Barry, Olukemi Adeyemo, Pamela E Jeter, Robert W Massof, PLoVR Study Group; Twenty Questions: An adaptive version of the PLoVR ultra-low vision (ULV) questionnaire. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):497.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: To use the calibrated results of our self-reported visual ability questionnaire in 80 individuals with ultra-low vision (ULV; <20/500) to derive an adaptive version. The survey was developed as part of the Prosthetic Low Vision Rehabilitation (PLoVR) study, was shown to have excellent person (0.99) and item (0.97) reliability (ARVO 2014, #2150), and is available in 150 and 53 item versions. An adaptive version using anchored item measures and Bayesian estimation should allow rapid estimation of person measures.

Methods: Using the 150 item responses for the 80 subjects used in the previous Rasch analysis, a Bayesian estimation procedure was simulated in which for each subject each subsequent item was selected from those not previously administered that was closest to the expected person measure, on the basis of all previously administered items. After each iteration, i.e., item, we determined the distribution of deviations from the known final person measures, and of the standard errors of the estimates. Percentile values of deviations and standard errors were tabulated and used to determine the number of items that should be administered to obtain a pre-determined accuracy and/or precision level in the person measure estimate for a newly enrolled ULV patient.

Results: With increasing item number, the deviation from the previously determined person value and the standard error of the estimate both decrease, with accuracy and precision limited by the inherent measurement noise in the original sample. Based on our original data the lowest achievable 95th percentile of the standard error is 0.3 logits, and it therefore makes sense to terminate the adaptive questionnaire when the deviation reaches that level, with acceptable probability. Acceptable probability for a low vision researcher may be 95% confidence, but this will require administration of 93 items; for 90% confidence, at least 80 items are required. On the other hand, administration of 15 items will bring the estimate for 50% of respondents within 0.3 logits from the 150-item estimate. If the proverbial 20 questions are administered, this accuracy level will be obtained for 56% of respondents.

Conclusions: We continue to expand our database to strengthen the deviation and standard error tables, but even with the current 80 respondents their use allows rational decision criteria for termination of the adaptive version of the PLoVR ULV questionnaire.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×