Reactivity and repeatability of hygiene behaviour: Structured observations from Burkina Faso

If interventions promoting improved hygiene behaviour to prevent childhood diarrhoea are to be implemented and evaluated, valid methods for measuring this behaviour will be required. This paper presents findings from a study to investigate the use of structured observations to measure hygiene behaviour in Burkina Faso. Two hundred mothers with young children (2-36 months) were observed on several occasions, with particular attention focused on events/behaviour surrounding defaecation. Child defaecation occurred most often in a potty (67% of occasions). Stools were most often disposed of into a latrine (79%). Following defaecation the child's bottom was usually rinsed using water alone with a bare hand (76%). Subsequent hand washing by the mother/caretaker was much rarer (29%). None of these behaviours appeared "reactive" to the presence of the observer. Less common behaviours showed some evidence of reactivity. The frequency of child defaecation in the yard increased over the course of three observations (5% to 16%; P = 0.01) and the proportion of occasions on which the child was observed to be cleaned after defaecation declined (95% to 85%; P = 0.01). Mothers usually took with them to the latrine a water recipient (91%). Hand washing after leaving the latrine was observed on 30% of occasions. This proportion declined from 36% to 22% over three observations (P = 0.05). Defaecation by older siblings (aged 3-5 years) was usually into a potty (48%) or directly into a latrine (30%). There was no evidence that this behaviour was reactive. The repeatability of behaviours at the individual level was generally low. The site of index child defaecation ([kappa] = 0.27), how the child's bottom was cleaned ([kappa] = -0.01) and whether the caretaker washed her hands afterwards ([kappa] = 0.26) all showed low repeatability. The method of stool disposal was more repeatable ([kappa] = 0.73). Hand washing by mothers after using the latrine showed moderate repeatability ([kappa] = 0.40). Older sibling's defaecation behaviour had excellent repeatability ([kappa] = 0.90). Our findings suggest that, in studies which aim to measure behaviour at the population level, structured observations may provide a useful tool. Studies which investigate links between hygiene behaviour and diarrhoea incidence at the individual level will require repeated observations of mothers and children since measuring behaviour during a single observation will lead to misclassification of exposure status, resulting in bias which could mask any underlying association. This is likely to be very costly.