Predicting population dental disease experience at a small area level using Census and health service data

Background Information on the dental disease patterns of child populations is required at a small area level. At present, this can be provided only by expensive whole population surveys. The aim of this study was to evaluate the ability of Census data combined with health service information to prov ide estimates of population dental disease experience at the small area lev el. Method Clinical dental data were collected from a large cross-sectional sur vey of 5-year-old children. A preliminary series of bivariate linear regres sion analyses were undertaken at ward level with the mean number of decayed , missing or filled teeth per child (dmft) as the dependent variable, and t he Census and health service and lifestyle variables suspected of having a strong relationship with dmft as independent variables. This was followed b y fitting a multiple linear regression model using a stepwise procedure to include independent variables that explain most of the variability in the d ependent variable dmft. Results All deprivation indicators derived from the Census showed a highly significant (p<0.001) bivariate linear relationship with ward dmft. The Jar man deprivation score gave the highest R-2 value (0.45), but the Townsend i ndex (R-2=0.43) and the single Census variable 'percentage of households wi th no car' (R-2 = 0.42) gave very similar results. The health and lifestyle indicators also showed highly significant (p<0.001) linear relationships w ith dmft. The R2 values were generally much lower than the deprivation-rela ted Census variables, with the exception of the percentage of residents who smoked (R2 = 0.42). None of the health or lifestyle variables was included in the final dental disadvantage model. This model explained 51 per cent o f the variability of ward dmft. Conclusions The results demonstrate the strong relationship between dental decay and deprivation, and all of the commonly used measures of deprivation exhibited a similar performance. For this population of young children hea lth and health services shelf data did not improve on the ability of depriv ation-related Census variables to predict population dental caries experien ce at a small area level.