Objectives: To assess the quantity and quality of studies using an observational measure of behaviour during the COVID-19 pandemic, and to narratively describe the association between self-report and observational data for behaviours relevant to controlling an infectious disease outbreak. Design: Systematic review and narrative synthesis of observational studies. Data sources: We searched Medline, Embase, PsychInfo, Publons, Scopus and the UK Health Security Agency behavioural science LitRep database from inception to 17th September 2021 for relevant studies. Study selection: We included studies which collected observational data of at least one of three health protective behaviours (hand hygiene, face covering use and maintaining physical distance from others (‘social distancing’) during the COVID-19 pandemic. Studies where observational data were compared to self-report data in relation to any infectious disease were also included. Data extraction and synthesis: We evaluated the quality of studies using the NIH quality assessment scale for observational studies, extracted data on sample size, setting and adherence to health protective behaviours, and synthesized results narratively. Results: Of 27,279 published papers on COVID-19 relevant health protective behaviours that included one or more terms relating to hand hygiene, face covering and social distancing, we identified 48 studies that included an objective observational measure. Of these, 35 assessed face covering use, 17 assessed hand hygiene behaviour and seven assessed physical distancing. The general quality of these studies was good. When expanding the search to all infectious diseases, we included 21 studies that compared observational versus self-report data. These almost exclusively studied hand hygiene. The difference in outcomes was striking, with self-report over-estimating observed adherence by up to a factor of five in some settings. In only four papers did self-report match observational data in any domains. Conclusions: Despite their importance in controlling the pandemic, we found remarkably few studies assessing protective behaviours by observation, rather than self-report, though these studies tended to be of reasonably good quality. Observed adherence tends to be substantially lower than estimates obtained via self-report. Accurate assessment of levels of personal protective behaviour, and evaluation of interventions to increase this, would benefit from the use of observational methods.
BMC Public Health