Table 1.
Scrub typhus reporting criteria.
Fig 1.
Trend in the annual number of reported scrub typhus cases (A) from 1980–2018 and annual incidence rate per 100,000 population (B) from 1985–2018.
Fig 2.
Scrub typhus cases by month from 2003–2018.
Fig 3.
Administrative regions of Thailand [National Statistical Office, Ministry of Information and Communication Technology; reprinted from mapchart.net under a CC BY license, with permission from Minas Giannekas original copyright 2019].
Fig 4.
Seasonality of scrub typhus cases by month per region from 2003–2018.
Fig 5.
Map of Thailand depicting: (A) the geography with elevation, provincial boundaries and location of Chiangrai province; (B) total scrub typhus cases from 2003–2018 per district; and (C) interpolation of scrub typhus cases at sub-district level by kriging using semi-variogram based on the centroids of geographical coordinates of each sub-district (for clarity, only provincial boundaries are shown) [created using tmap and tmaptools in R software–R Core Team 2018 [38–40]].
Fig 6.
Chiangrai province with: (A) the geography with elevation, main rivers and sub-district boundaries; (B) scrub typhus cases from 2003–2018 per sub-district; and (C) interpolation of scrub typhus cases at sub-district level by kriging using semi-variogram based on the centroids of geographical coordinates of each sub-district [created using tmap and tmaptools in R software–R Core Team 2018 [38–40]].
Fig 7.
Time series analysis of (A) reported monthly scrub typhus cases, (B) total monthly rainfall in mm and (C) average monthly temperature in°C for Chiangrai province.
Table 2.
Results of the best global GLM with binomial negative link function (p<0.001), with theta = 2.12 (standard deviation = 0.026), explaining the number of scrub typhus cases in Chiangrai province from 2003–2018 by sub-district and by month with explanatory geographical and meteorological variables in the initial model.
Estimate and standard deviation (SD) were given for each selected explanatory variables, with P value and VIF (Variance Inflation Factor). For the best selected model the log likelihood = -8925.35 with degree of freedom (DF) = 12, null deviance = 3497.2, R2 estimated by maximum likelihood (R2ML) = 0.25, wr = 0.36 and Akaike Information Criteria (AIC) = 17875 (see S6 Fig).
Fig 8.
Results of the best General Additive Modeling (GAM) explaining the number of scrub typhus cases in Chiangrai province over 2003–2018 by sub-district and by month, using a binomial negative link function (theta = 2.12).
The smoothed variables selected in the best GAM were (A) the geographical distribution of sub-district (longitude and latitude of the centroid), (B) total monthly rainfall (mm), (C) temperature (°C), (D) population size, (E) rain-fed land cover, (F) flooded land cover, (G) habitat complexity and (H) mosaic land cover.
Table 3.
Results of general additive modelling (GAM) explaining the number of cases of scrub typhus per sub-district in Chiangrai province using a negative binomial link (theta = 2.12), with approximate significance of smooth terms.
For the best selected model, the deviance explained = 50.4%, R2 = 0.37, maximum likelihood = 8868.2, AIC = 17689.8 (see Fig 8).