Study population
The vast majority of all children (around 90%) in the region are listed as regular patients at the Public Dental Service that provides free dental care between 1 and 19years with recall intervals varying from 3 to 24months depending on the individual need. Data on the experience of manifest (dentin) caries is registered according to the WHO-criteria [
11] and annually reported to the community dentistry unit. The fluoride concentration in piped water supply is generally low (<0.3ppm) except in the northern part (the municipality of Kungsbacka) where the natural fluoride content in the drinking water is approximately 1.0ppm.
The present study included 9,973 children between 3-6years of age for whom caries data were reported in 2006 and 10,927 children with corresponding data from 2010. The age distributions of the children examined in 2006 and 2010, respectively, did only differ marginally (proportions of children aged 5years: 49.3% in 2006; 50.3% in 2010). The overall coverage of the total 3-6-year population of the Halland region was 76% in 2006 and 77% in 2010. The remaining children were not recalled for a regular check-up that particular year or visited a private dentist at a clinic not among the in-reporting ones or located outside the region. The study was approved by the Halland Hospital Ethical committee as well as The Swedish Data Inspection Board.
The Halland region consists of six municipalities that are subdivided into 66 parishes. Geo-maps were produced by using the ESRI ArcGIS system (Environmental Systems Research Institute, Inc., USA). Each child was geo-coded with respect to his/her residence area (parish).
Socio-economic characteristics
Statistics Sweden provided parish-level data from years 2006 and 2010, respectively, on three socio-economic indicators: (i) the proportion with post-secondary education (any schooling beyond the high-school level) among all residents; (ii) the proportion of immigrants (more specifically, individuals born outside Sweden and individuals born in Sweden with both parents born outside Sweden) among all residents; and (iii) the proportion of families with low purchasing power (according to Swedish standard; 19,500 USD annual income) among all residing families with at least one child (19 years old; family with the same residence address).
Epidemiological and statistical methods
Reported dmfs>0 for a child was considered as the primary caries outcome. The method for estimating caries risks at a given year of reporting has been explained previously [
10]. Briefly, a parish-level relative risk (RR) was calculated as the observed-to-expected ratio, where the expected number was obtained from the sex-specific caries (dmfs>0) rates for the total study population of 3-6year old children residing in the Halland region. Moreover,
smoothed RRs (SmRR) for each parish were obtained by running a Bayesian hierarchical mapping model, which allow parish-specific RRs to be smoothed towards global and local average risk levels across the study region [
12]. We underline that such Bayesian smoothing yielded shrinkage of the conventional observed-to-expected ratios. The corresponding statistical certainty geo-maps were obtained by calculating the posterior probabilities of a parish-specific relative risks above 1 given the data, denoted Pr(RR>1|data), using 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 lowered caries risk, Pr(RR<1|data)=1 - Pr(RR>1|data)>0.95, was colored
green. The remaining parishes were colored
yellow, which indicates a weaker statistical evidence for an elevated/lowered relative risk.
Evidential, positional changes in the caries risk estimated for the children living in a specific parish in 2006 and 2010, respectively, were assessed by comparing the certainty geo-maps along with the posterior probabilities, Pr(RR>1|data): if the certainty color changed and Pr(RR>1|data) differed at least 25% for a specific parish, when comparing the results for the study years 2006 and 2010, the positional change was considered evidential. Hence, the rationale for classifying a positional change as evidential is based on the change over time in the statistical evidence for an elevated (or lowered) relative risk. Clearly, it seems extraordinary to observe a parish with a positional change from an evidently elevated caries risk (red) in 2006 to an evidently lowered caries risk (green) in 2010, or vice versa. Nevertheless, there were more modest, yet substantial, positional changes (from red/green to yellow, or vice versa) that we considered evidential. Also notice that a parish could change certainty color without being indentified to have an evidential positional change (provided Pr(RR>1|data) differed less than 25%).
The statistical analyses were performed using the free software Rapid Inquiry Facility [
13], which provides an extension to ESRI ArcGIS functions [
14], along with free software for Bayesian data analyses, WinBUGS [
15].