The use of classification and regression trees in clinical epidemiology

A critique is presented of the use of tree-based partitioning algorithms to formulate classification rules and identify subgroups from clinical and ep idemiological data. It is argued that the methods have a number of limitati ons, despite their popularity and apparent closeness to clinical reasoning processes. The issue of redundancy in tree-derived decision rules is discus sed. Simple rules may be unlikely to be "discovered" by tree growing. Subgr oups identified by trees are often hard to interpret or believe and net eff ects are not assessed. These problems arise fundamentally because trees are hierarchical. Newer refinements of tree technology seem unlikely to be use ful, wedded as they are to hierarchical structures. (C) 2001 Elsevier Scien ce Inc. Ah rights reserved.