The Validity, Time Burden, and User Satisfaction of the FoodImage™ Smartphone App for Food Waste Measurement Versus Diaries: A Randomized Crossover Trial

Abstract

The FoodImage™ smartphone app transmits to researchers users' photographs of food selection and food waste, and includes user-tagged information about waste reasons and destination. Twenty-four participants were trained to record food waste using FoodImage, food waste diaries requiring visual estimation of waste quantities, and diaries requiring scale weights. Participants used each method during three staged food-waste scenarios (food preparation, eating, and clean-out) in a randomized crossover trial. Two participants had extreme values for the weighed diary method; therefore, accuracy results are reported with and without these two participants' data. Error was calculated as waste estimated with the experimental method minus directly weighed waste. Mean absolute error from FoodImage was significantly smaller than or equal to the error from both diary methods in each scenario. Furthermore, the mean values from FoodImage were equivalent to directly weighed values in two out of the three tasks; while weighed diaries were equivalent in two tasks only when the two participants with extreme values were removed. Visually estimated diaries were equivalent for only one task. All 24 participants preferred FoodImage to diaries and all rated FoodImage as less time consuming. Over one week, FoodImage would require ~24 fewer minutes of users' time to record all data. Unlike food waste diaries, FoodImage also transmits data to researchers in real-time and provides detailed data on food selection and intake. ClinicalTrials.gov Identifier: NCT03309306.

Keywords: Food waste; accuracy; food waste diaries; household; measurement; smartphone app; time burden; user satisfaction; validity study.

Figures

Figure 1.

Mean error with 95% confidence interval bars by measurement method and task for all participants (black cross) and adherent participants (red cross). Notes: 20 g equivalency band depicted with green dashed lines. **The 95% confidence interval for diary with scale confidence interval extends in both directions beyond the region depicted on the graph. *The mean and entire 95% confidence interval for diary with scale lies above the region depicted on the graph.

Figure 2.

Mean Absolute Measurement Error by Method and Task. Whiskers are 95% confidence intervals. Bars within the same task cluster that share a letter are not significantly different at the 5% level. Cross-hatched bars extend outside the graph’s range.