Assessing the validity and reliability of geotracking devices in urban settings of Nairobi, Kenya

Spatial methods are critical to the understanding of infectious disease transmission and accurate exposure assessments. It is widely acknowledged that a One-Health approach is required to understand infectious disease processes, implying that knowledge of spatial patterns of both humans and animals is necessary. This is particularly so in urban areas and informal settlements where human-animal interactions are increasing. Towards this, we have described in prior work the feasibility of geotracker deployments using Tractive Global Positioning System (GPS) devices. This study extends that work by examining the validity (difference between true and reported locations) and reliability (reported locations remaining constant while in the same location) of data obtained from the deployment of tractive GPS devices in urban environments in Nairobi, Kenya. We recorded GPS data in three different types of locations - open spaces, alleyways, and inside homes - in each of two neighborhoods characterized by different socioeconomic statuses (SES). Our results from a set of generalized linear models indicated that the median distance between true and reported locations ranged from 9 m to 28 m, depending on housing density and location type, with the upper bound of 95% prediction intervals ranging from 16 m to 53 m, again depending on housing density and location type. In addition, Tractive devices showed very strong reliability in both our lower- and higher-housing density neighborhoods, with median distances ranging from 0-1 m between sequential reported locations while not moving. This implies that these geotracking devices are useful for understanding where humans and animals spend their time, but only up to a certain level of spatial granularity.