Reliability, utility and construct validity of measurements using a mobile app to rapidly appraise the in-store food environment (Preprint)
2019
BACKGROUND Consumer food environments are increasingly being recognized as influential determinants of food purchasing and subsequent intake and health. We developed a tool to enable efficient, but relatively comprehensive, appraisal of the in-store food environment. The Store Scout mobile app facilitates the evaluation of product (availability and range), placement (visibility, accessibility, proximity to high-traffic areas, and location relative to other products), price (price promotion), and promotion (displays and advertising) across 7 categories of food products, with appraisal given immediately as scores (0-100, where a higher score is more in line with best practice). Primary end users are public health nutritionists and nutritionists employed by store organizations; however, store managers and staff are also potential end users. OBJECTIVE This study aims to evaluate the reliability (interrater reliability and internal consistency), utility (distribution of scores), and construct validity (score by store type) of measurements using the Store Scout mobile app. METHODS The Store Scout mobile app was used independently by 2 surveyors to evaluate the store environment in 54 stores: 34 metropolitan stores (9 small and 11 large supermarkets, 10 convenience stores, and 4 petrol stations) in Brisbane, Australia, and 20 remote stores (19 small supermarkets and 1 petrol station) in Indigenous Australian communities in Northern Australia. The agreement between surveyors in the overall and category scores was evaluated using intraclass correlation coefficients (ICCs). Interrater reliability of measurement items was assessed using percentage agreement and the Gwet agreement coefficient (AC). Internal consistency was assessed by comparing the responses of items measuring similar aspects of the store environment. We examined the distribution of score values using boxplots and differences by store type using the Kruskal-Wallis test. RESULTS The median difference in the overall score between surveyors was 4.4 (range 0.0-11.1), with an ICC of 0.954 (95% CI 0.914-0.975). Most measurement items had very good (n=74/196, 37.8%) or good (n=81/196, 41.3%) interrater reliability using the Gwet AC. A minimal inconsistency of measurement was found. Overall scores ranged from 19.2 to 81.6. There was a significant difference in score by store type (P<.001). Large Brisbane supermarkets scored highest (median 77.4, range 53.2-81.6), whereas small Brisbane supermarkets (median 63.9, range 41.0-71.3) and small remote supermarkets (median 63.8, range 56.5-74.9) scored significantly higher than Brisbane petrol stations (median 33.1, range 19.2-37.8) and convenience stores (median 39.0, range 22.4-63.8). CONCLUSIONS These findings suggest good reliability and internal consistency of food environment measurements using the Store Scout mobile app. We identified specific aspects that can be improved to further increase the reliability of this tool. We found a good distribution of score values and evidence that scoring could capture differences by store type in line with previous evidence, which gives an indication of construct validity. The Store Scout mobile app shows promise in its capability to measure and track the health-enabling characteristics of store environments.
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
20
References
3
Citations
NaN
KQI