Content Analysis of Diet-Related Mobile Apps: A Self-Regulation Perspective

ABSTRACT

Diet-related mobile apps hold promise in helping individuals self-regulate their eating behaviors. Nevertheless, little is known about the extent to which diet-related mobile apps incorporate the established behavior change theories and evidence-based practices that promote dietary self-regulation. Guided by the self-regulation aspect of Bandura’s social cognitive theory and the Dietary Guidelines for Americans 2010 of the U.S. Department of Agriculture, this study conducts a content analysis of diet-related mobile apps for iPhone (N = 400). In terms of the adherence to the self-regulation aspect of the social cognitive theory, results show that although 72.5% of the apps incorporate at least one theoretical construct, few apps tap all three processes of self-regulation (i.e., self-observation/monitoring, judgment process, and self-reaction). Additionally, outcome expectation is manifested in a majority of the diet-related apps. In terms of the Dietary Guidelines for Americans 2010, while the diet-related apps equally emphasize setting goals for calorie intake or nutrient consumption, more apps feature nutrient tracking than calorie tracking. Implications and limitations are discussed.

Notes

¹ Due to page limit, the flow diagram of the selection process of the diet-related iPhone apps is not included but is available upon request to the corresponding author.

² The two coders both agreed that no eligible app featured distal goal. Because the coding for this feature is invariant, Cohen’s kappa cannot calculate the reliability of distal goal.

³ In the following Results section, the remainder of the apps that are not reported either did not contain the focal feature (coded as “no”), were unclear to judge the level of feature they contain (coded as “unclear”), or lacked the overarching feature (coded as “not applicable”). The coding scheme is available upon request.

⁴ As a conservative measure, Cohen’s kappa over .40 may be considered acceptable (Fleiss, 1981; Landis & Koch, 1977). However, typically kappa over .60 or .70 is considered as good reliability. In our coding, the percentage of agreement was relatively high (82.93% for friends and family and 95.12% for expert). However, the majority of the agreements between two coders were based on the concurrence on a feature/category that was either absent or present. When a large number of absent cases and a small number of present cases exist, even very few disagreements for the coding of presence can bring down the value of kappa dramatically (for an example, see Figure 4 in Krippendorff, 2004).