López-Abente Gonzalo, Aragonés Nuria, García-Pérez Javier, Fernández-Navarro Pablo
Environmental and Cancer Epidemiology Unit, National Centre for Epidemiology, Carlos III Institute of Health, Madrid, Spain.
Geospat Health. 2014 Nov;9(1):27-35. doi: 10.4081/gh.2014.3.
The municipal, spatial pattern of male stomach cancer mortality in Spain, spanning the period 1989-2008, was studied, comparing the results of depicting mortality using different expected-case computation methods in a spatial and spatio- temporal modelling context. Expected cases for each municipality were first calculated by two methods: (i) using reference rates for each 5-year period; and (ii) using average reference rates for the overall period. This was visualised by two types of models: (i) independent maps for each period based on the model proposed by Besag, York and Mollié; and (ii) a series of maps over time based on a model with spatio-temporal interaction terms. An additional model, based on mortality rate ratios as an alternative to the traditional use of standardised mortality ratios, was also fitted. Integrated nested Laplace approximations were used as the Bayesian inference tool. The results show that, in general, the geographical pattern was maintained across the study period, and that the maps differed appreciably according to the method used to obtain the expected number of cases. While the use of average reference rates appears to be the most suitable choice where the aim is to study time trends by area, it may nevertheless mask the spatial pattern in situations where the time trend is very marked and the study period is long. When it comes to studying changes in the spatial pattern of stomach cancer mortality, we feel that it is most useful to plot independent maps by period and use the "local" rates for each period as reference in the computation of expected cases.
研究了1989 - 2008年期间西班牙男性胃癌死亡率的城市空间格局，在空间和时空建模背景下比较了使用不同预期病例计算方法描绘死亡率的结果。首先通过两种方法计算每个城市的预期病例：(i) 使用每5年期间的参考率；(ii) 使用整个期间的平均参考率。这通过两种类型的模型进行可视化：(i) 基于Besag、York和Mollié提出的模型为每个时期绘制独立地图；(ii) 基于具有时空交互项的模型绘制一系列随时间变化的地图。还拟合了一个基于死亡率比的额外模型，作为传统标准化死亡率比的替代方法。使用集成嵌套拉普拉斯近似作为贝叶斯推断工具。结果表明，总体而言，地理格局在整个研究期间保持不变，并且根据用于获得预期病例数的方法，地图存在明显差异。虽然使用平均参考率似乎是按地区研究时间趋势的最合适选择，但在时间趋势非常明显且研究期较长的情况下，它可能会掩盖空间格局。在研究胃癌死亡率空间格局的变化时，我们认为按时期绘制独立地图并在计算预期病例时使用每个时期的“局部”率作为参考最为有用。
