Abstract
BACKGROUND: Trust is an essential element in engagement with data sharing and underpins efforts to use data to combat health inequalities. However, research into public trust in data sharing and health care settings may rely on oversimplified notions of what trust entails. How trust relationships manifest in this context has not been widely explored. OBJECTIVE: We aimed to establish the primary reasons for the placement of trust and whether these reasons vary by demographics and domain. We also explored the utility of a composite trust score as a predictor for use of technology in the health sphere. METHODS: We conducted a cross-sectional survey using Qualtrics to explore the challenges associated with trust and judgments of trustworthiness in the context of the use of technology to collect health-related data. Participants were recruited using a marketing firm, Dynata, in July 2022 and were UK census matched for population representation. A total of 99.33% (1192/1200) of the target UK-based participants aged ≥18 years (n=605, 50.8% female; n=587, 49.2% male) were asked to rate their level of trust in others generally and in specific entities on an ordinal scale (1-5). We constructed Bayesian cumulative logit models and hierarchical models to evaluate whether demographic characteristics predicted reasons for domain-specific or general trust. We created a composite trust score across health data domains (range 1-15) and developed models to determine whether this score predicted the likelihood of having used or using a device to track health or well-being. We report all credible intervals at 95%. RESULTS: General trust responses were bimodally distributed, with the most frequently chosen answers being "usually not" and "usually." A cumulative logit model suggested that divorced status predicted choosing "almost always not" or "usually not" (β estimate=-0.71, 95% CI -1.17 to -0.28). "They are reliable and keep their promises" and "They behave responsibly" were the most chosen reasons for placing trust. Trust in family, the National Health Service, and technology companies was primarily driven by familiarity, perceived responsibility, and openness and responsible behavior, respectively. A Bayesian hierarchical model suggested that higher general trust was a strong predictor of a higher composite trust score (β estimate=1.93, 95% CI 1.26-2.59). A higher composite trust score also inversely correlated with the likelihood of having used a device to track health or well-being, whereas higher trust in technology companies and the National Health Service predicted a willingness to use such devices. CONCLUSIONS: Unlike with prior works evaluating trust and trustworthiness, we demonstrate that trust must be understood as context-specific and relational. Policymakers should note that self-reported global trust may not correlate with specific health- and technology-related behaviors and, consequently, that domain-specific measurements of trust are essential in health policy work.