Comparison of respondent-reported and sensor-recorded latrine utilization measures in rural Bangladesh: a cross-sectional study.

Background: Health improvements realized through sanitation are likely achieved through high levels of facilities utilization by all household members. However, measurements of sanitation often rely on either the presence of latrines, which does not guarantee use, or respondent-reported utilization of sanitation facilities, which is prone to response bias. Overstatement of sanitation metrics limits the accuracy of program outcome measures, and has implications for the interpretation of related health impact data. Methods: We conducted a cross-sectional study of 213 households in 14 village water, sanitation and hygiene committee clusters throughout rural Bangladesh and used a combined data- and relationship-scale approach to assess agreement between respondent-reported latrine utilization and sensor-recorded measurement. Results: Four-day household-level respondent-reported defecation averaged 28 events (inter-quartile range [IQR] 20-40), while sensor-recorded defecation averaged 17 events (IQR 11-29). Comparative analyses suggest moderately high accuracy (bias correction factor=0.84), but imprecision in the data (broad scatter of data, Pearson's r=0.35) and thus only weak concordance between measures (ρc=0.29 [95% BCa CI 0.15 to 0.43]). Conclusions: Respondent-reported latrine utilization data should be interpreted with caution, as evidence suggests use is exaggerated. Coupling reported utilization data with objective measures of use may aid in the estimation of latrine use