Combining discrepant diagnostic information from multiple sources: are complex algorithms better than simple ones?

Abstract

Optimizing methods of combining discrepant diagnostic information from multiple sources is one of the more daunting tasks facing the field of child psychopathology. Several researchers have asserted that complex information-combining schemes, those in which certain information or sources of information are weighted differently from others, are preferable. This study provides theoretical and empirical evidence that simple information-combining schemes, those in which all information from all sources is weighted equally, will as a rule work as well as complex schemes and may even work better. The implications of these findings for diagnostic instrument and algorithm design are discussed. A two-step diagnostic procedure utilizing a simple information-combining scheme is presented.