Potential explanatory variables

For each covariate, the geographical scale of interest was the municipality (ADM2) level when available or the departmental (ADM1) level, otherwise. The pool of variables tested related to both human population and environmental conditions (Supp. Table 1).

The Trypanosoma cruzi seroprevalence in public blood banks by year and department was provided by the Pan American Health Organization (PAHO). The presence of Triatoma dimidiata and Rhodnius prolixus at the municipality level was obtained after combining records from a national surveillance report of 2013 [17] and data from [18, 19]. We also extracted data on presence/absence of these two vector species, from which the proportion of municipalities infested for each department was calculated. Data on vector control interventions implemented in Colombia (1998–2014) were extracted from [20]. Census data were obtained from the Colombia’s Department of Statistics (DANE) website [21]. Climate variables were extracted from the Köppen-Geiger climate classification maps at a 1-km resolution [22]. Finally, the map layer used was obtained from Database of Global Administrative Areas (GADM) (https://gadm.org/ [23]).

Other covariates included the setting of the survey (urban, rural, indigenous, or mixed population (including urban, rural and unknown settings); the year when the survey was conducted; an effect for years and decades (full details in Supp. Table 1). Indigenous settings comprised those with Amerindian populations mostly following traditional lifestyles as described in [3]. Definitions for urban and rural populations followed the Colombian governement criteria [21].