Reported dmfs/DMFS >0 for an individual was considered as the primary caries outcome. A parish-specific relative risk (RR) was calculated as the observed-to-expected ratio, where the expected number of individuals with dmfs/DMFS >0 was obtained from the age- and sex-specific caries (dmfs/DMFS >0) rates for the whole region of Halland or, more precisely, for the total study population. The following age strata were used: 3-6, 7-11,12-18, and 19 years. Thus, the expected number for a parish equals the sum of the products
n
i
×
r
i
across the age- and sex-strata
i(3-6 year old girls; 3-6 year old boys; 7-11 year old girls; etc.), where
n
i
denotes the stratum-specific number of study individuals residing in the parish and
r
i
denotes the corresponding caries rate observed in the total study population. The computations of the RRs were performed using the free software Rapid Inquiry Facility [
12], which provides an extension to ESRI
® ArcGIS functions [
13]. The Rapid Inquiry Facility (RIF) along with free software for Bayesian data analyses, WinBUGS [
14], provides a powerful tool for geo-mapping based on epidemiological data. The caries risk maps show the smoothed RRs (SmRR) for each parish, which were obtained by running the Bayesian hierarchical mapping model in RIF/WinBUGS. We underline that such Bayesian smoothing yields pronounced downward adjustment of a (conventional) RR for a parish with few study persons, estimated with relatively high uncertainty, if that RR turns out notably elevated. Hence, by presenting smoothed caries risk geo-maps, rational adjustments of the conventional (parish-specific) RRs are taken into account [
15,
16].
We present separate caries risk geo-maps for preschool children (3-6 years), schoolchildren (7-11 years) and adolescents [12-19 years; based on age-stratified (12-18 and 19 years, respectively) analysis]. Along with each caries risk geo-map, we provide the corresponding statistical certainty geo-map. A posterior probability of a parish-specific relative risk above one given the data, denoted Pr(RR>1|data), was obtained by the Bayesian approach. A parish with data yielding strong statistical evidence of an elevated caries risk, more precisely Pr(RR>1|data) > 0.95, was colored red in the certainty geo-map. By contrast, a parish with evidently low caries risk, Pr(RR<1|data) > 0.95, was colored green. Analogously, parish-specific 90% credibility intervals for the relative risk were obtained; and each parish with a 90% credibility interval that covers 1 was colored yellow in the certainty geo-map indicating a weaker statistical evidence for a high or low relative risk in such parishes.
We addressed geographical co-variations between caries risk and residents'' level of education by calculating Spearman''s correlations (rS) between the SmRRs and the proportions with post-secondary educational level among the residents (considered as a group-level indicator of socio-economy) across the 66 parishes.