Problems encountered with the use of simulation in an attempt to enhance interpretation of a secondary data source in epidemiologic mental health research
Department of Community Health Sciences, University of Calgary, 4thFloor TRW Building, 3280 Hospital Drive N.W., Calgary, Alberta, T2N 4N1, Canada
Hotchkiss Brain Institute, Health Research & Innovation Centre, Room 1A10, 3330 Hospital Dr. N.W. Calgary, Alberta, T2N 4N1, Canada
BMC Research Notes 2010, 3:231 doi:10.1186/1756-0500-3-231Published: 26 August 2010
The longitudinal epidemiology of major depressive episodes (MDE) is poorly characterized in most countries. Some potentially relevant data sources may be underutilized because they are not conducive to estimating the most salient epidemiologic parameters. An available data source in Canada provides estimates that are potentially valuable, but that are difficult to apply in clinical or public health practice. For example, weeks depressed in the past year is assessed in this data source whereas episode duration would be of more interest. The goal of this project was to derive, using simulation, more readily interpretable parameter values from the available data.
The data source was a Canadian longitudinal study called the National Population Health Survey (NPHS). A simulation model representing the course of depressive episodes was used to reshape estimates deriving from binary and ordinal logistic models (fit to the NPHS data) into equations more capable of informing clinical and public health decisions. Discrete event simulation was used for this purpose. Whereas the intention was to clarify a complex epidemiology, the models themselves needed to become excessively complex in order to provide an accurate description of the data.
Simulation methods are useful in circumstances where a representation of a real-world system has practical value. In this particular scenario, the usefulness of simulation was limited both by problems with the data source and by inherent complexity of the underlying epidemiology.