Evaluating the feasibility of the AI-powered Calcount smart diet smartphone application among healthcare workers in a pilot study

Nutritional management is essential in cardiac rehabilitation, particularly for dietary monitoring, as obesity is a major risk factor for cardiovascular disease. Digital health solutions, such as smart diet applications, hold significant potential to improve dietary habits and weight management among patients with cardiovascular disease (CVD). This study evaluated the feasibility of the AI-powered CalCount app for dietary monitoring among healthcare workers as a preparatory step toward its potential use in cardiac rehabilitation settings.

This pilot study aimed to assess the feasibility, usability, and accuracy of the CalCount smartphone application for dietary assessment among healthcare workers in a cardiology department.

This prospective, single-center feasibility study was conducted in a hospital in Belgium. Twenty healthcare workers from the cardiac rehabilitation center volunteered to participate. Participants documented their dietary intake by taking photos of all consumed meals over a one-month period. The CalCount application analyzed each image, providing detailed assessments of caloric intake along with comprehensive macronutrient and micronutrient profiles. Accuracy was evaluated by comparing app-generated calorie estimates with those calculated by experienced dietitians using the Wilcoxon signed-rank test. Usability was assessed using the System Usability Scale (SUS), where scores above 68 indicate above-average usability. The quality of the mobile health app was assessed using the User Mobile Application Rating Scale (uMARS), which employs 5-point rating scale.

The mean difference between dietitian-calculated and app-generated calorie estimates was 271.5 kcal (P = 0.43), indicating no statistically significant difference. The percentage difference in calorie estimates was less than 20%, supporting the app's accuracy. Usability assessments revealed a SUS score of 47.8 (out of 100), reflecting below-average usability; however, participants rated learnability positively.

The uMARS objective quality score had good reliability (Cronbach alpha = 0.8), overall quality score was 3.2 out of 5, with functionality scoring highest (3.6) and engagement lowest (2.7). Despite its moderate quality rating, certain areas, such as user engagement, require improvement.

This pilot study demonstrates that CalCount is feasible for calorie estimation, with accuracy comparable to dietitian evaluations. Furthermore, a percent difference in calorie estimates below 20% suggests viability for randomized trials. Although the app's usability is below average, it is easy to learn and achieves a moderate overall quality score. These findings highlight the need for enhancements in user engagement and functionality. Larger studies with extended monitoring are necessary to validate these results and further explore the app’s potential in dietary management, particularly for cardiac rehabilitation.

SUS mean scores for individual volunteer