(P17-008-25) Evaluating Acceptability and Utilization of an AI-Assisted Dietary Assessment Tool in Middle-Aged Adults

Objectives: Despite advances in dietary assessment methods, high participant burden remains a challenge. We evaluated the acceptability and completion rates of Keenoa, an AI-assisted smartphone application that integrates image recognition, compared to the web-based Automated Self-Administered 24-hour (ASA24) tool among middle-aged and older adults (NCT06455111).

Methods: During an 8-week block randomization within the fourth examination of the Framingham Heart Study Third Generation cohorts, participants were assigned weekly to complete a 3-day dietary record using either ASA24 or Keenoa. Acceptability was assessed by the proportion agreeing to complete the dietary assessment, those completing all 3 days, and a satisfaction survey.

Results: We included 115 FHS participants (54% female, mean age 58 years, mean body mass index 29 kg/m²). Among 51 participants randomized to ASA24, 47 agreed to participate. Among 64 participants randomized to Keenoa, 55 agreed to participate, with 2 excluded due to lack of access to Google or Apple IDs. For both ASA24 and Keenoa, 38 participants logged at least one food item. After excluding days with energy intake < 600 kcal, 27 (53%) of those assigned to ASA24 and 23 (36%) of those assigned to Keenoa completed all 3 days. Of completers, the average daily record for Keenoa was 1779 kcal (carbohydrate: 187.8g, protein: 82g, total fat: 74.8g, food count: 12.6, photos taken: 7), and for ASA24 was 2037 kcal (carbohydrate: 208.1g, protein: 85g, total fat: 89.2g, food count: 15.5, photos taken: NA). Responses were similar to a question about satisfaction with the tool (Keenoa: 3.3 and ASA24: 3.1 out of 5-point Likert scale).

Conclusions: Keenoa's acceptability was similar to ASA24, but ASA24 showed higher mean intakes and food counts. Integrating smartphone-compatible digital dietary assessment tools is essential for advancing research; however, barriers to use and validation in diverse populations, particularly among middle-aged and older adults, require further study.

Funding Sources: NHLBI, NIDDK