Assessing the Validity and Usability of a Speech Recognition Dietary Assessment Tool for Older Adults (Preprint)

BACKGROUND With the global population ageing rapidly, achieving healthy ageing is crucial, especially in countries like South Korea, where over 14.2% of citizens are aged 65 and older, qualifying it as an ‘aged society’. By 2050, the Asia-Pacific region is expected to see unprecedented ageing, with 10 countries classified as 'ageing societies' (over 7% of its population aged 65 and over), five as 'aged societies' (over 14% aged 65 and over), and 11 as 'super-aged societies' (over 21% aged 65 and over) with South Korea as anticipated to be among the latter, projected to have over 40% of its population aged 65 and older. Effective nutritional assessment for this demographic remains challenging due to limitations in conventional methods, exacerbated by factors like cognitive decline and low literacy. Despite efforts to address these challenges with new technological solutions like web, scanners, and mobile-based tools, issues of accuracy, usability, and cost persist. OBJECTIVE This study aims to introduce and evaluate a speech-based dietary assessment tool as a promising alternative for improving the accuracy and usability of nutritional assessment in older adults. METHODS In a randomised cross-over design, 18 participants aged 65 and older were divided into two groups, each using both a pen-and-paper food diary (FD) method and a speech recognition dietary assessment (SR) method over two weeks. Dietary intake was recorded and analysed over non-consecutive 3-day periods each week while usability was assessed via the Systems Usability Scale (SUS). RESULTS The results of the nutrient intake analysis indicated small mean differences between methods, with a 0.49% difference in energy and an average of 5.02% in essential nutrients. Significant correlation coefficients for nutrient intake between methods were observed, with the highest for fat (0.740), protein (0.738), and dietary fibre (0.671). Bland-Altman analysis indicated no substantial bias in energy or essential nutrient intakes, except for cholesterol which showed a discernible proportional bias. SUS scores between the SR and FD methods showed no statistical significance; however, the SR method received a higher standardised rating, characterised as 'acceptable' with a score of 72.77, compared to the FD method, rated as 'marginally acceptable’ with a score of 66.25. CONCLUSIONS These findings suggest that the SR method is adequately comparable to the traditional FD method for dietary intake assessment in older adults and is on par in terms of usability, indicating its potential as a valid and acceptable alternative. CLINICALTRIAL KCT0009677