Assessment of artificial neural networks in health futures research

Although brief, the overview of literature indicates that the application of artificial neural networks (ANN) in futures research is still very much in the exploration phase, and less common than might be expected based upon theoretical concepts. Results of studies are for the most part encouraging, but it remains to determine real, sustainable value since well controlled and repeated studies have not been completed. More time appears to be required for the application to mature. One is forced to characterize the ANN field, as it relates to futures research, as "emerging with significant promise". The use of ANN as a tool in futures research has therefore at least 2 practical issues to face: (i) a lack of handbook guidance from the literature for the uninitiated; and (ii) prospects of significant effort and uncertain return on investment. Concerning guidance, the problems in which ANN perform well tend to be very nonlinear, but such situations are also least likely to comply with any generalized prescription. ANN might be considered as complex solutions to complex problems. Based upon these factors, it may be argued that the preconditions for practical use of ANN in futures research should be: (i) availability of team-based research (access to existing ANN expertise); and (ii) that the research should include other more conventional methods to duplicate estimates.(ABSTRACT TRUNCATED AT 250 WORDS)