Methodological approaches to conducting pooled cross-sectional time series analysis: the example of the association between all-cause mortality and per capita alcohol consumption for men in 15 European states

Abstract

Aim: To compare different statistical models in order to estimate the association of alcohol consumption and total mortality when time series data stem from different regions.

Data and methods: Data on per capita consumption in 15 European countries were combined with standardized mortality rates covering different periods between 1950 and 1995. An indicator of region-specific drinking patterns was measured without reference to a concrete time point, thus generating a hierarchical data structure. Two groups of models were compared: pooled cross-sectional time series models with different error structures and hierarchical linear models (random coefficient models).

Results: If historical time is not controlled for in cross-sectional models, this might result in estimating a negative association between alcohol consumption and total mortality. Hierarchical linear models or cross-sectional models controlling for historical time, however, resulted in the expected positive association. Only hierarchical linear models were able to adequately estimate the moderating effect of drinking patterns on the association between alcohol consumption and total mortality.

Conclusion: For pooled cross-sectional time series data, control for the potential impact of historical time is of utmost importance. Hierarchical linear models constitute a superior alternative to analyze such complex data sets, especially as time-independent characteristics of regions can be implemented in the model.