Using remote, spatial techniques to select a random household sample in a dispersed, semi-nomadic pastoral community: utility for a longitudinal health and demographic surveillance system

BACKGROUND: Obtaining a random household sample can be expensive and challenging. In a dispersed community of semi-nomadic households in rural Tanzania, this study aimed to test an alternative method utilizing freely available aerial imagery. METHODS: We pinned every single-standing structure or boma (compound) in Naitolia, Tanzania using a ‘placemark’ in Google Earth Pro (version 7.1.2.2041). Next, a local expert assisted in removing misclassified placemarks. A random sample was then selected using a random number generator. The random sample points were mapped and used by survey enumerators to navigate. RESULTS: We created a spatial sample frame and a random sample in 34.5 student working hours, 3 local expert hours and 1.5 academic working hours. Challenges included determining whether homes were occupied or abandoned, developing a protocol for placemark inclusion and quality issues with the aerial imagery itself. In the field, 175 sample points were visited and 170 of these (97 %) were actual households. The primary advantages of this method were the: (a) ability to generate a robust random sample in a rural and remote area; (b) lack of reliance on existing, external population data sources; and (c) relatively low levels of funding and time required. CONCLUSIONS: This method to develop a spatial sample frame was efficient and cost-effective when compared to in-field generation of a household inventory or GPS tracking of households. Utilizing a local expert to review the sample frame prior to field testing greatly increased accuracy. Overall, this method is a promising alternative to expensive and possibly biased household inventories or in-field GPS data collection for all households.