Purchase this article with an account.
Ecosse Luc Lamoureux, Ryan Man, Alfred Gan, Preeti Gupta, Eva Fenwick; Using Patient-Reported Outcome Measures (PROMs) to predict Vision Impairment in Individuals with Type 2 Diabetes (T2DM). Invest. Ophthalmol. Vis. Sci. 2020;61(7):5147.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In high volume and low-resourced settings, assessing vision impairment (VI) in those with T2DM can be challenging. We evaluated the feasibility of utilizing a diabetic retinopathy (DR) quality of life PROM (RetBank) to detect VI in those with T2DM.
500 T2DM individuals, comprising 108 and 392 with mild-moderate and severe DR/DME [diabetic macular edema], respectively, were recruited from eye clinics in Australia and administered the RetBank (256 items over 10 domains). Any; mild-moderate; and severe VI were defined as a logarithm of the minimum angle of resolution (LogMAR) >0.3; >0.3 to <1.0; and ≥1.0, respectively. Participants’ responses were randomly divided into training (N=400) and validation (N=100) sets after multiple imputation of missing responses. Lasso regression with the 1-standard-error penalty was used to select items that were predictive of VI across ≥80% of 10 imputed training datasets, in order to construct a prediction model that was then tested on the validation sets. Average model discriminative accuracy was assessed using the area under the receiving operating characteristic curve (AUC).
The sample mean (SD) age was 60.4 (12.4) years and 36% were female. Of these, 138 (27.6%); 362 (72.4%); 211 (42.2%); and 151 (30.2%) had none; any; mild-moderate; and severe VI in any eye, respectively. The model identified responses from 2 items from the Symptoms (SY) domain, and 1 item each from the Activity Limitation (AL) and Convenience (CV) domains that could detect the presence of any VI with a good discriminative accuracy of 83% (AUC: 0.83, 95% CI [confidence interval]: 0.69, 0.96). Similarly, 2 items from the SY domain, and 1 AL item could detect mild-moderate VI with a discriminative accuracy of 75% (AUC: 0.75, 95% CI: 0.60, 0.91); and 2 SY, 1 AL, 1 DV and 1 CV items were able to detect severe VI with a very good discriminative accuracy of 86% (AUC: 0.86, 95% CI: 0.75, 0.97).
Answering 3-5 DR-specific items can identify any; mild-moderate; or severe VI in those with T2DM with good accuracy. While further work is needed to improve our model, our data are the first to show that PROMs can be used as rapid, easy-to-use and self-administered screening tools for VI, especially in populations facing barriers to accessing healthcare; and those living in underprivileged and underserved areas.
This is a 2020 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only