Evaluating Smartphone App Based Dietary Tracking in a Military Population: Validation Study (Preprint)
BACKGROUND Collecting dietary intake data is a key component for a majority of nutritional epidemiology studies. Smartphone technology advancements allow researchers to use health and nutrition apps as alternatives to currently available tools (food frequency questionnaires, 24-hour recalls, and food diaries). Service Members (SM) can greatly benefit from the always-available information and easily accessible nature of smartphones to track their intake. Clinicians working with military units can help provide these SM with the skills to evaluate their intake for performance benefits. Understanding the accuracy of these apps is important to determine their effectiveness for use in clinical and research settings. OBJECTIVE This study evaluated the relative validity of self-reported intake with the HealthWatch 360 (HW 360) app compared to the Automated Self-Administered 24-hour Dietary Assessment (ASA24). METHODS Recruitment targeted Army and Air Force SM from Joint Base Lewis-McChord, WA and Joint Base San Antonio-Lackland, TX who currently or previously failed to meet body composition standards. Participants (n=53) completed a demographic questionnaire, baseline anthropometric measurements, and recorded daily intake on the HW 360 app. They returned approximately two weeks later to complete a 24-hour recall using the ASA24. Agreement and relative validity were evaluated using Bland-Altman plots and two one-sided tests at a ± 10% equivalency range of ASA24 mean nutrient intake values between HW 360 and ASA24 data. Multilinear regressions analyzed relationships between participant demographics and relative validity. RESULTS HW 360 was not significantly equivalent to the ASA24. Large levels of underreporting were found in total energy (Mean Difference (Mdiff) = -503.3 kcal, 90% CI: -649.8 to -356.7 kcal), carbohydrates (Mdiff = -52.2 g, 90% CI: -70.4 to -34.1 g), protein (Mdiff = -20.4 g, 90% CI: -29.4 to -11.3 g), and fat (Mdiff = -24.6 g, 90% CI: -32.5 to -16.7 g). Bland-Altman plots failed to illustrate agreement. No significant correlations existed for demographic variables and relative validity. CONCLUSIONS Differences between all variables tested were above clinically significant values and limit the usage of this application in research and clinical settings. Further research is needed to determine the potential causes of underreporting and evaluate methods to minimize this effect. Understanding these effects allows the implementation of a tailored app for use with SM. It has the potential to be an invaluable asset for this population due the unpredictable nature of deployments and training exercises. CLINICALTRIAL ClinicalTrials.gov NCT04959318; https://clinicaltrials.gov/ct2/show/NCT04959318