Comparison of metrics for assessing face washing behaviour for trachoma control

There is currently no single, easy-to-use, reliable indicator to assess whether a face has been washed with soap in the context of trachoma elimination. This study aimed to compare survey report, script-based pictorial recall and facial cleanliness indicators as alternatives to structured observation for measuring face washing behaviour. This method validation study was nested in the Stronger-SAFE trial, Oromia Region, Ethiopia. Structured observation was conducted in randomly selected households for three hours from dawn. The primary caregiver in each household participated in a survey to capture (self)-reported behaviour and/or script-based pictorial recall, a routine-based diary activity to covertly capture information on face washing behaviour of themself and any children aged 1-12. Children 4-12 years old directly participated in the survey and pictorial recall in a subset of households. The facial cleanliness of children aged 1-12 was assessed qualitatively and using the quantitative Personal Hygiene Assessment Tool (qPHAT). Prevalence estimates, sensitivity, specificity and predictive values were computed for each behavioural indicator with observation data as the gold standard. The appropriateness of script-based pictorial recall was assessed using baseline and 3-month follow-up data. Baseline data were collected from 204 households in 68 clusters. Survey estimates of face washing and face washing with soap among caregivers and children were 32% to 60% and 5% to 31% higher than observed behaviour, respectively. Face washing prevalence estimates from pictorial recall were lower than survey estimates and comparable with observations for some face washing with soap indicators (0.3% to 13% higher than observations). Specificity of pictorial recall indicators was high (85% to 99%), but the sensitivity was low (0% to 67%), resulting in a low positive predictive value for all indicators. Both qualitative facial cleanliness indicators and qPHAT scores were poorly correlated with observed face washing earlier that morning. Pictorial recall overestimated face washing with soap among both caregivers and children following intervention delivery but not at baseline. Survey (self)-reported data on face washing is highly inaccurate. Script-based pictorial recall does not correctly classify those who wash their face with soap, and is subject to differential bias following intervention exposure, and facial cleanliness is a poor indicator of recent face washing in settings where faces become rapidly dirty again after washing. Alternatives to structured observation cannot be recommended to monitor the effectiveness of face washing interventions in community settings. Trial Registration ISRCTN registry ISRCTN40760473, https://doi.org/10.1186/ISRCTN40760473.