Comparing the performance of indicators of hand-washing practices in rural Indian households

To compare the results obtained from 26 proxy indicators of domestic hand-washing practices with those obtained from direct, 'structured' observation of hand-washing in a sample of 387 households and to assess the potential of these indicators for use in the evaluation of hygiene promotion campaigns.

Fieldwork in rural India between February 2005 and April 2006. Household-level data on hand-washing practices and the availability of soap and water were collected by structured observation, questionnaire survey, pocket voting, hand-wash demonstration and environmental check. Between them these techniques produced 27 binary indicators of hand-washing practices, each of which was used to classify households as 'hand-washing' or 'non-hand-washing. To assess the extent to which household classification based on each of 26 proxy indicators concurred with classification based on observation, we used the kappa statistic. The prevalence of households defined as 'hand-washing' according to each indicator was compared statistically with the prevalence according to structured observations by testing for a significant difference between two proportions.

Agreement between all the proxy indicators and the observation data was poor and all but two of the indicators produced estimates of hand-washing prevalence that were significantly different from that resulting from observation.

Although some interventions may be able to use proxy indicators as a guide to the magnitude and direction of their impact, these indicators do not provide an accurate guide to the actual practice or prevalence of hand-washing. Structured observation remains the best indicator of those tested.