Rapid consensus building to strengthen AMR surveillance in LMICs using One Health approach: a modified table-top exercise coupled with a Delphi study protocol.

Snacks—foods and beverages consumed between meals—are a significant part of our modern diets. In the Netherlands, 28% of daily energy intake occurs outside breakfast, lunch, and dinner [], and similar percentages are reported in other countries, such as the United States (23%) [], Canada (24%) [], Greece (34%) [], France (19%) [], and Brazil (21%) []. While the relationship between snacking and BMI remains inconclusive [-], individuals with overweight or obesity tend to consume more energy-dense snacks high in sugar and fat, and their snack consumption is positively associated with adiposity []. In the Netherlands, snacks also contribute significantly to sugar intake, accounting for half of daily mono- and disaccharide consumption []. Understanding snacking behaviors, snack choices, and their underlying factors is essential for improving diet quality [,]. Various factors influence snacking, including culture [], social and physical environment [,], hunger [], physical activity (PA) [], sleep [], distraction [], and affect []. However, research into snacking behavior has primarily relied on 2 distinct methodologies: self-reported assessments and controlled laboratory experiments. Each approach has advantages and limitations. Self-reports enable assessment in naturalistic environments; nevertheless, their completion is perceived as burdensome [] and is susceptible to inaccuracies from both intentional and unintentional response biases, as well as challenges in estimating portion sizes. Underestimation of self-reported energy intake can reach 47% [], and snack consumption, in particular, is prone to being underreported [,]. In contrast, controlled laboratory environments facilitate precise monitoring of food intake but are less suitable for prolonged data acquisition. Furthermore, artificial laboratory environments may not reflect real-life eating behaviors. For instance, the impact of portion size on intake is less pronounced in the laboratory compared with at home [], different physiological responses are observed [], eating rates might differ [,] and real-life stressors are known to exert a more pronounced impact than artificial stressors []. To address these challenges, we developed the SnackBox, an automated dietary assessment device designed to enable structured measurement of snacking behavior in seminaturalistic settings []. The SnackBox is a rectangular platform containing electronics with 3 sensing coasters, each storing a snack container or beverage bottle, and automatically recording the amount consumed (in grams) and the time of consumption. A prior validation study demonstrated that the SnackBox estimated portion sizes with an intraclass correlation coefficient of 0.80 and a mean percentage error of −6%, outperforming self-report methods, which yielded an intraclass correlation coefficient of 0.60 and a mean percentage error of −40% []. Detection of snack events showed 81% agreement with an observational reference, compared with 54% for self-report []. The device operates with minimal researcher or user intervention, and gathered data are processed in real time, enabling the triggering of just-in-time adaptive interventions or ecological momentary assessment (EMA) surveys upon snack event detection to capture, for example, eating motives or contextual factors.