Constructing a Minimal Diagnostic Decision Tree

Abstract:

Classification trees and discriminant function analysis were employed in order to ascertain whether a small number of diagnostic decision rules could be extracted from a large inventory of items. Several models, involving up to 17 symptoms, that led to a broad psychiatric diagnosis were then tested on a small validation sample of 53 patients. All methods, with the exception of CART used without any pruning, generated identical trees involving four items. Almost 90% of the validation sample was able to be correctly classified by all methods although poor classification performance was noted in the case of one particular diagnosis, Schizoaffective Psychosis. In contrast, stepwise linear discriminant analysis originally selected 17 items, although three out of the first four items selected were identical to those chosen by the tree-building methods. Although more research is required, there are indications that the latter methods may be usefully employed in constructing parsimonious decision trees.

Key-Words

Classification Trees - Computer-assisted Diagnosis - Discriminant Analysis

• REFERENCES

• 1 Berwick DM, Murphy JM, Goldman MPH. et al. Performance of a five-item mental health screening test. Med Care 1991; 29: 169-76.
• 2 McKenzie DP, Clarke DM, Low LH. A method of constructing parsimonious diagnostic and screening tests. Int J Meth Psychiatr Res 1992; 02: 71-9.
• 3 Edworthy SM, Zatarain E, McShane DJ, Bloch DA. Analysis of the 1982 ARA lupus criteria data set by recursive partitioning methodology; new insights into the relative merit of individual criteria. J Rheumatol 1988; 15: 1493-8.
• 4 McGorry PD, Copolov DL, Singh BS. Royal Park Multidiagnostic Instrument for Psychosis: part I. rationale and review. Schizophr Bull 1990; 16: 501-15.
• 5 McGorry PD, Singh BS, Copolov DL. et al. Royal Park Multidiagnostic Instrument for Psychosis: part II. development, reliability and validity. Schizophr Bull 1990; 16: 517-36.
• 6 Hand DJ. Artificial Intelligence and Psychiatry. Cambridge: Cambridge University Press; 1985
• 7 de Ville B. Applying statistical knowledge to database analysis and knowledge base construction. Proceedings of the Sixth Conference on Artificial Intelligence Applications. Santa Barbara, Cal. 1990: 30-60.
• 8 McKenzie DP, Low LH. The construction of computerized classification systems using machine learning methods: an overview. Comput Human Behav 1992; 08: 155-67.
• 9 Cook EF, Goldman L. Empiric comparison of multivariate analytic techniques: advantages and disadvantages of recursive partitioning analysis. J Chronic Dis 1984; 37: 721-31.
• 10 Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and Regression Trees. Belmont, Cal: Wadsworth; 1984
• 11 Kass GV. An exploratory technique for investigating large quantities of categorical data. Appl Stat 1980; 29: 119-27.
• 12 Biggs D, de Ville B, Suen E. A method of choosing multi-way partitions for classification and decision trees. J Appl Stat 1991; 18: 49-62.
• 13 Quinlan JR. Simplifying decision trees. Int J Man-Machine Stud 1987; 27: 221-34.
• 14 Weiss SM, Kulikowski CA. Computer System that Learn: Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning and Expert Systems. San Mateo, Cal: Morgan Kaufmann; 1991
• 15 Gammerman A, Thatcher AR. Bayesian diagnostic probabilities without assuming independence of symptoms. Meth Inf Med 1991; 30: 15-22.
• 16 Hughes CA, Gose EE, Roseman DL. Overcoming deficiencies of the rule-based medical expert system. Comput Meth Progr Biomed 1990; 32: 63-72.
• 17 Stewart PW, Stamm JW. Classification tree prediction models for dental caries from clinical, microbiological, and interview data. J Dent Res 1991; 70: 1239-51.
• 18 Wallace CS, Patrick JD. Coding Decision Trees. Technical Report 91/153. Melbourne, Australia: Monash University; 1991
• 19 Wallace CS, Freeman PR. Estimation and inference by compact coding. J Royal Statist Soc, Series B 1987; 49: 240-65.
• 20 Shannon CE, Weaver W. The Mathematical Theory of Communication. Urbana: University of Illinois Press; 1949
• 21 Patrick JD. SNOB: a Program for Discriminating between Classes. Technical Report 91/151. Melbourne, Australia: Monash University; 1991
• 22 Piiowsky I, Levine S, Boulton DM. The classification of depression by numerical taxonomy. Br J Psychiatry 1969; 115: 937-45.
• 23 Forsyth RS, Clarke DD, Wright RL. Over-fitting Revisited: an Information-Theoretic Approach to Decision-Tree Simplification. Technical Report AAG9101. Nottingham University; UK: 1991
• 24 Brink S, Van Schalkwyk DJ. Serum ferritin and mean corpuscular volume as predictors of bone marrow iron stores. S Afr Med J 1982; 61: 432-4.
• 25 Keks NA, McKenzie DP, Low LH. et al. Multidiagnostic evaluation of prolactin response to haloperidol challenge in schizophrenia: maximal blunting in Kraepelinian patients. Biol Psychiatr 1992; 32: 426-37.
• 26 Melrose JP, Stroebel CF, Glueck BC. Diagnosis of psychopathology using stepwise multiple discriminant analysis. Comp Psychiatry 1970; 11: 43-50.
• 27 SPSS Inc. SPSS-X User’s Guide. (3rd ed).. Chicago: 1988
• 28 FirstMark Technologies. KnowledgeSEEK-ER User’s Guide. Ottawa, Ont; 1990
• 29 Jennrich R, Sampson P. Stepwise discriminant analysis. In: BMDP Statistical Software Manual (voll). Dixon WJ. (ed). Berkeley, Cal: University of California Press; 1988: 337-56.
• 30 Huberty CJ. Issues in the use and interpretation of discriminant analysis. Psychol Bull 1984; 95: 156-71.
• 31 Tabachnick BG, Fidell LS. Using Multivariate Statistics. (2nd ed).. New York: Harper Collins; 1989
• 32 Hauck WW, Miike R. A proposal for examining and reporting stepwise regressions. Stat Med 1991; 10: 711-5.
• 33 Oliver J, Wallace CS. Inferring decision graphs using the minimum message length principle. Proceedings of the Fifth Joint Conference on Artificial Intelligence; Hobart, Australia: 1992: 361-7.
• 34 Feinstein AR. Clinical Biostatistics XXXIX. The haze of Bayes, the aerial palaces of decision analysis, and the computerized Ouija board. Clin Pharmacol Ther 1977; 21: 482-96.