Small-area mapping of premature mortality and its association with area socio-economic characteristics on the small island of Cyprus
METHODS: Standardised mortality ratios (SMRs) of overall and gender-specific premature mortality were calculated for the period 2004-11 to ensure sufficient numbers. Bayesian hierarchical Poisson models with spatially unstructured and/or structured random effects were used to identify the magnitude of the “social gradient” in terms of each of twenty-six area-level 2001 census indicators.
RESULTS: SMRs (range 0 to 5) were unreliable at this small level of aggregation (N=370, median population 233, IQR: 75-784, 10% of areas >3000). Up to two-fold differences remained in smoothed maps (range: 0.76-1.35), with 39% of the variation explained locally. A striking southeast to northwest spatial pattern was revealed with higher rates in less dense and remote areas. Strong negative associations were observed with proportion of households with persons aged less than 14 (0.89, 95%CI= 0.85-0.94 per SD increase) and proportion of houses constructed after 1990 (0.90, 95%CI= 0.86-0.94). Surprisingly, traditional indicators of deprivation, such as unemployment, did not exhibit associations with premature mortality. The top five indicators associated were increased proportion of retired people, disabled or chronically ill, illiterate, single-person households and households with no PC, all with an increase of about 10% per SD increase.
CONCLUSIONS: There appeared to be a clear urban-rural divide in premature mortality, with rural and remote areas at a disadvantage. While the indicators which demonstrated significant associations with mortality are commonly used as proxy measures of deprivation, they are also characteristic of rural life. Since deprivation may take different meaning depending on context, this should be incorporated into the development of an index of deprivation.