(P22-10-26) Protocol, Technological Innovation, and Process Results for a Large-Scale, Student-Led Participatory Mapathon of Diverse Food Environments in Bangkok, Thailand

Objectives: To collect data and images from diverse food outlets, and characteristics of consumer-outlet interactions, for multifaceted spatial analysis of food environments (FEs) in Bangkok, Thailand.

Methods: Eligible students at selected university campuses in Inner Bangkok attended a 1-hour orientation on a rolling basis, after which they filled out a baseline survey on perceived FEs, collected data on 10 unique food outlets of their choice using SurveyCTO, and filled out an endline food-frequency questionnaire (FFQ). The SurveyCTO form, which incorporated co-design from two workshops with 26 students, flexibly accommodates diverse interaction types with built and digital FEs to capture key objective and subjective FE parameters. To capture food availability, enumerators upload images or screenshots of all parts of menus following a protocol that facilitates accurate extraction of food names and prices using computer vision and language models (if menus are incomplete/unavailable, data are entered manually, as occurred in 9.4% of cases). All data feed into a centralized Google Suite backend that continuously updates a public map of unique outlets assessed to-date (to prevent duplicates), a tracking system for students to check and compare their and their peers’ progress, and provides automation for most data quality checks and administrative tasks such as onboarding and active follow-up.

Results: From April 2 2025 to May 31 2026, 243 enumerators mainly from Mahidol Phayathai, Mahidol Bangkok Noi, Chulalongkorn, and Thammasat Tha Phra Chan campuses completed orientation and 210 completed the project (target: 200). Median days from orientation to FFQ completion: 25 (IQR: 12.3-48.8). Data on 2229 restaurants, vendors, markets, food courts, and other food outlets were recorded (5.3 cases/day; median 12.8 mins./case; 5.1 out of 7 cases were on weekdays) over a polygon spanning 1900 km2. Data include 5025 foods purchased, 12858 images of menus, and diverse interaction types (36% ate on the premises, 26% takeaway, 36% ordered remotely). 2130 (95.6%) of cases met passing quality requirements (target: 2000).

Conclusions: High quality information on large and dynamic urban FEs can be efficiently collected by integrating a participatory approach and technological innovation in data collection and monitoring.

Funding Sources: Mahidol University Strategic Research Fund