Inter- and intra-rater reliability of digitally captured images of plate waste

Aim: Digital intake estimation has been shown to be strongly correlated with real-time visual estimation of dietary intake in cafeterias and long term care settings. The objectives of the present study were to estimate intra- and interrater agreement for digitally captured plate waste in hospitalised adults. Methods: Lunch plates (n = 106) of hospitalised adults were photographed immediately prior to tray collection. Two raters used the modified Comstock method to assess food waste from each of the digital images. Additionally, a randomly selected subset of photographs was re-estimated 72hours following the first estimation. Inter- and intrarater reliability were estimated by calculating the intraclass correlation coefficients (ICC). Results: In the total population, the ICC for interrater reliability was 0.95 (95% CI 0.93-0.97, P < 0.0001) for the entire meal, 0.94 (95% CI 0.9-0.96, P < 0.0001) for vegetables, 0.96 (95% CI 0.9-0.96, P < 0.0001) for starch and 0.96 (95% CI 0.94-0.97, P < 0.0001) for the main course. ICC for intrarater reliability was 0.99 (95% CI 0.97-1.0, P < 0.0001) for the entire meal, 0.95 (95% CI 0.86-0.98, P < 0.0001) for vegetables, 0.97 (95% CI 0.92-0.99, P < 0.0001) for starch and 0.99 (0.98-1.0, P < 0.0001) for the main course. Findings were similar for regular diets and somewhat weaker for soft diets. Conclusions: Dietary intake estimated from digital images has a high degree of both inter- and intrarater reliability in hospitalised patient populations.