These three data sets comprise each a provincial-level representative rural survey of adults in Chiang Rai and Salavan, and a two-round census survey with a three-month interval in five villages across the two sites (3 in Chiang Rai, 2 in Salavan). The surveys were implemented by 10-member survey teams in each country between November 2017 and April 2018. Antimicrobial resistance (AMR) is a global health threat that endangers the achievement of the Sustainable Development Goals, especially Goal 3 on 'Good Health and Well-Being'; Leading UK and global strategy papers aiming at improving people's antibiotic usage to fight and prevent AMR thereby focus exclusively on awareness-raising campaigns, but this narrow approach suffers from conceptual, methodological, and empirical weaknesses. In response, our study intends to improve the understanding of patients' antibiotic-related health behaviour to inspire more targeted and unconventional interventions in low- and middle-income countries (LMICs). Speaking to the themes of "awareness and engagement" and "informal markets and access to antibiotics" we will investigate three research questions: (1) What are the manifestations and determinants of problematic antibiotic use in patients' healthcare-seeking pathways? (2) Will people's exposure to a behavioural health systems intervention diffuse or dissipate within a network of competing healthcare practices? (3) Which proxy indicators facilitate the detection of problematic antibiotic behaviours across and within communities? Our interdisciplinary approach frames behaviour within a shared activity space. By drawing on theories and tools from public health, medical anthropology, sociology, and development economics, and by focusing on vulnerable rural dwellers in the DAC countries Thailand and Laos, we will be able to generate innovative and unprecedentedly detailed open-access survey data on antibiotic-related behaviour and its social, economic, and spatial determinants. We aim to maximise complementarities with other ongoing projects in the region that (1) implement biomarker testing and education campaigns in clinical settings, (2) generate mixed-method evidence on cross-cultural patterns of antibiotic use, and (3) engage with the general public to improve global health awareness. We will apply a rigorous three-stage stratified cluster random sampling design to produce district-level representative survey data of the antibiotic use of 2,400 villagers; and we will carry out social network censuses in four communities with a total of 2,400 villagers. Using satellite imagery and digital data collection tools, we can realise these sample sizes at 75% of the cost of conventional survey approaches. Pursuant to our research questions, we will generate novel insights into the nature and variability of Thai and Lao antibiotic usage and health behaviours using the following methods: We will (1) use event sequence analysis and multilevel regression to investigate the impact of technology and digital media as well as economic, social, and spatial characteristics of patients on adverse antibiotic usage, (2) apply social network analysis to understand how knowledge and practice diffuse from clinical interventions into village communities, and (3) use latent class analysis to detect problematic conditions for antibiotic use through easy-to-collect proxy indicators. Under the umbrella of the Oxford Tropical Network-an inspiring and enabling research environment-this project will be made possible through collaboration across world-leading researchers and groups in health behaviour research (KEMRI Wellcome Trust Research Programme; Kenya), health economics and public engagement (Mahidol Oxford Tropical Medicine Research Unit in Thailand; LOMWRU in Laos), evidence-based antibiotic policy (Oxford University Clinical Research Unit; Viet Nam), social network analysis (CABDyN Complexity Centre; Oxford), development economics (Technology and Management Centre for Development; Oxford), and global health training (Centre for Tropical Medicine and Global Health; Oxford). ODA relevance follows from our partnerships, capacity building activities, and research interest in vulnerable groups in LMICs.

1) Provincial-level representative survey: Three-stage stratified cluster random sampling design. The first stage involved the random selection of 30 primary sampling units (clusters) across five purposively selected districts in each site, stratified by their distance to the nearest urban centre (using data from the US National Geospatial Intelligence Agency). The second stage enumerated all residential buildings within the selected villages using satellite imagery from Google Maps and Bing Maps, of which we sampled 5% of the buildings (but at least 30 houses) in a stratified interval sampling approach to ensure spatial representativeness. During the survey implementation, the third sampling stage involved selecting randomly one respondent for every five adults in each chosen house. 2) Two-round village census: community-level social network census surveys in five purposively selected villages across the two field sites (3 in Chiang Rai, 2 in Salavan). The villages were selected in consultation with local stakeholders; guiding criteria for selection were (1) village size and structure, (2) remoteness and road accessibility, (3) economic status as approximated by village-level infrastructure and facilities, (4) ethnic composition and (5) number and location of health facilities within a 2 km radius. The villages had between 300 and 1,500 residents. Within the selected communities, all households were approached, their adult members enumerated and invited to participate.