Table 3.

Efficiencies and challenges of using Geo-Sampler tool and protocols for population-level data collection in Guatemala

Steps in sampling protocol	Efficiencies	Challenges and considerations for future work
1. Training of study staff on Geo-Sampler Tool	Professional contacts facilitated access to Epicentre staff	Currently, limited technical documentation When available, technical documentation is in limited languages Formal training on the tool is not currently available
2. Selecting samples and digitizing sampled structure coordinates	No special software required Geo-Sampler tool using very up-to-date, high resolution imagery, so we were able to identify recently constructed structures Sample can be selected at one time, by one person Sample of structures behind walls can be easily selected Geo-Sampler does not retain the list of selected structure coordinates after the program is shut down, which limits privacy concerns	Sample selection required significant expert time If updated samples are needed, extra time from someone trained in the tool will be required, unless multiple people are trained If multiple people are trained and are selecting the sample, significant coordination and oversight would be required to ensure quality control, consistency, and to eliminate repetition of structures, unless adaptations are made to Geo-Sampler to allow for simultaneous use of multiple users
3. Locating selected structures in the field	Use of the Android/Google Maps platform was intuitive and well-known to local study staff, cost-saving Satellite overlay on Google Maps useful not only for finding tagged structures but also identifying principal entrances, alleyways, etc. when attempting to approach structures (many located in walled compounds, etc.) Selecting structures rather than relying upon investigators’ concepts of what a “residence” looked like allowed us to include non-traditional living situations	Initial version of the Geo-Sampler tool provided.kml files but without identifiable latitude and longitude, which then required a 2 stage process to determine. This was changed during the course of the study by Epicentre Saved efficiencies of multi-stage sampling somewhat offset by inefficiency of inevitably many tagged structures not being residences. This method does not allow for people who do not live in structures. People living on the street or in cars would still be left out of these surveys A few coordinates selected were close enough that 2 different structures were given the same study ID by different data collectors. This was discovered and addressed in the data cleaning phase by recoding one of the residences of each pair Connectivity in our sites in Guatemala was generally good, but we experienced frequent signal drop-outs, requiring large-format printed physical map back-up at all times. This would likely be the case in many LMIC settings, especially rural areas. Drop-outs in connectivity also caused rapid phone battery drain (due to searching signal), and required staff to carry multiple recharging packets when in the field in order to keep phones charged