Accuracy of seven GPS-based geolocation techniques <i>versus</i> a surveyor map: a cross-sectional study in a village in the Comoros

Objective To assess the accuracy of seven commonly used Global Positioning System (GPS)-based geolocation techniques, including a handheld GPS device and smartphones using Google Maps, Open Data Kit (ODK) and Research Electronic Data Capture (REDCap) with and without internet connectivity, for measuring distances between houses in a field epidemiology setting. Design Cross-sectional study conducted in February–March 2024. Setting Field study in a village in the Comoros where leprosy is endemic. Participants 55 randomly selected house pairs, of which 50 were included in the analysis. Primary outcome measure Measurement error, defined as the difference in straight-line distance between house pairs measured by each index test (GPS-based geolocation technique) and the reference standard (surveyor map). Results Three index tests showed substantial shortcomings, including indeterminate results (one technique), outliers (two techniques), systematic underestimation (mean measurement error −3.0 m for one technique) and high variability (SD of measurement error ranging from 9.4 m to 16.6 m). The remaining four index tests showed little bias (mean measurement error ranging from −1.7 m to 0.4 m across techniques) and low variability (SD of measurement error ranging from 5.7 m to 7.4 m). For the most precise technique, 95% of measurements were estimated to fall between an underestimation of 11.9 m (95% CI 8.9 to 17.5 m) and an overestimation of 10.5 m (95% CI 8.4 to 13.7 m). Even this narrowest range was two to three times wider than the claimed accuracy. Conclusions GPS-based geolocation techniques may show substantial variability under field conditions and may give rise to misplaced confidence in their accuracy when this is assumed rather than empirically assessed. Careful selection and testing of techniques in context, along with transparent data handling and consideration of uncertainty, are needed to improve the reliability of geospatial data in public health research and practice.