Design of a mobile phone-based Artificial Intelligence (AI) application to assess dietary intake and provide nudges to improve healthy eating choices: Formative research in Ghana and Vietnam

Background: Low quality diets are a public health problem affecting individuals of all ages worldwide. Nudging for Good (NFG) is a new research project aimed at developing, validating, and examining the feasibility of using artificial intelligence (AI)-based technology to improve adolescent girls' diets in urban Ghana and Vietnam. Objectives: Provide evidence to support the design of a new mobile phone intervention including: a) identifying the demand for mobile app to improve diets in adolescent girls; b) defining the intervention objectives and activities to be delivered via a mobile app; and c) assessing the potential for nudging functionality to be incorporated in the mobile app. Methods: This study used mixed methods including both literature and nutrition-related app reviews, as well as focus group discussions. A literature review was conducted using PubMed and Google Scholar databases, and intervention studies using technology to improve nutrition outcomes of adolescents were included. Nine focus group discussions (five in Ghana and four in Vietnam) were undertaken with 61 girls to gauge smartphone and internet access and use and to obtain feedback on the prototype of a mobile app. Results: Nine studies met the selection criteria for the literature review, including seven randomized control trials (RCTs) and two systematic reviews. The evidence from the literature on how technology-based nutrition interventions should be conducted and what should be the best outcomes of success was mixed. Most of the 22 apps reviewed required manual entry of dietary information, and recommended diets based on motivation to change body weight. In the focus groups discussions, the adolescents suggested modifications on the prototype of the mobile app we presented, and indicated possibility to regularly take pictures of foods and beverages during meal time. Conclusion: We did not find an app available in either the Ghana or Vietnam markets that could improve dietary quality of adolescents without focusing on weight. Most apps available only had manual features to log food intake, which is time consuming. Moreover, adolescents expressed interest in using a new, modified version of the app we presented. We aim at developing a new mobile phone application based on AI technology that gives personalized and reliable nutrition advice to improve adolescent girls’ diets. However, the literature review was not conclusive on what should be the characteristics of an app and how interventions to measure an app’s impact on dietary quality should be conducted.