Explaining differential socioeconomic effects in population health interventions: development and application of a new tool to classify intervention agentic demand.

BACKGROUND: The agentic demand of population health interventions (PHIs) might influence how interventions work. Highly agentic interventions (eg, information campaigns) rely on recipients noticing and responding to the intervention. Resources required for individuals to benefit from highly agentic interventions have a socioeconomical pattern, thus agentic demand might affect intervention effectiveness and equity. Systematic evidence exploring these associations is missing due to the absence of adequate tools to classify agentic demands. We aimed to develop such a tool and test its application. METHODS: Our iterative development process involved: (1) systematic identification of diet and physical activity PHIs; (2) coding of intervention actors and actions; (3) data synthesis; (4) expert qualitative feedback; and (5) reliability assessment. We searched nine databases for articles published between Jan 1, 2010, and Aug 17, 2020. For all included articles, we coded the actors (people required to act within an intervention) and their actions (what they were required to do for the intervention to have its intended effects). We combined these codes for similar intervention types to develop overarching schematic flow chart diagrams used to identify concepts, and we organised these into a draft tool. After expert feedback, and we assessed inter-rater reliability of the final version. We applied the final tool in a proof-of-concept review, extracting studies from three existing equity-focused systematic reviews on tool category, overall intervention effect, and differential socioeconomic effects and visualised findings. FINDINGS: We identified three concepts affecting agentic demands of intervention components: exposure, two levels (how recipients encounter the intervention); mechanism of action, five levels; and engagement, two levels (how recipients respond to the intervention). We then combined these concepts to form 20 categories that grouped together interventions with similar agentic demands. In the review, we applied the tool to 26 PHIs that included 163 components. Intervention components were concentrated in a small number of categories, and their categorisation was related to intervention equity but not to effectiveness. INTERPRETATION: We present a novel tool to classify the agentic demand of PHIs and demonstrate its feasibility within a systematic review. Linking intervention types to their effect on inequalities enables these factors to be considered when designing or selecting interventions. Users of the tool can avoid implementing intervention types that are likely to widen inequalities or implement them alongside counter-strategies to minimise any adverse equity effects. Applying this tool within future research, policy, and practice to design, select, evaluate, and synthesise evidence from PHIs has the potential to advance our understanding of how interventions work and their effect on socioeconomic inequalities. FUNDING: Public Health Policy Research Unit (PH-PRU), National Institute for Health and Care Research (NIHR) Policy Research Programme.