Abstract
Purpose :
To assess agreement between Spectacle Wear Time (SWT) recorded by a temperature logger (TSWT) and manually recorded wear logs (RSWT) , and determine if human review of time vs. temperature plots (TVT)) improve agreement.
Methods :
Subjects were 12 children aged 4.1 to 39.7 months and 1 adult (EMH) who wore spectacles (Dilli-Dalli, Hauppauge, NY) with a temperature logger (TheraMon, Hargelsburg Austria) attached to the spectacle headband for 14 consecutive days. Temperature was recorded every 15 minutes. A log of wear times was recorded by the parent/adult subject. The optimum temperature to dichotomize TSWT data (wear/non-wear) based on RSWT data was determined by maximizing (Sensitivity+Specificity-1) within and across child subjects. Four human raters reviewed sample TSWT TVT plots along with RSWT data to learn the characteristics of TSWT. Raters then scored deidentified subject TVT plots. These human assessments of SWT (HSWT) were then compared for agreement between raters (for reliability / validity) and agreement with the adult dataset (for accuracy).
Results :
The 12 children averaged of 8.1 hours per day by RSWT, and 7.4 hours per day by TSWT. Across all subjects, a temperature >27.88° C threshold for indicating wear resulted in optimal agreement between TSWT and RSWT. Within subject optimal thresholds varied from 23° to 29°C. Summary of SWT results by method are provided in Table. TSWT was unable to differentiate glasses left in a warm environment and not being worn from glasses being worn by a child. The inter-rater reliability was 97.2% for the four human raters SWT estimates, with pairwise comparisons ranging from 95.3% to 98.4%. Accuracy, assessed by comparing the 4 human raters SWT estimates to the wear log of the adult subject yielded pairwise comparisons ranging from 99.7% to 99.9%.
Conclusions :
Determining SWT by a temperature sensor attached to the spectacle headband is feasible in children and correlates with parent wear logs. As ambient temperature can be above body temperature, misattribution of wear is possible but can be minimized by human inspection of the TVT profiles.
This is a 2020 ARVO Annual Meeting abstract.