Bayes Theorem and Conditional Dependence of Symptoms: Different Models Applied to Data of Upper Gastrointestinal Bleeding

Summary:

In computer-aided prognosis theoretical arguments in favour of a sophisticated model often do not result in a markedly better performance. In this study, straightforward extensions of the simple and comprehensive independent Bayes model, taking interactions between variables into consideration, are investigated, using a clinical data set of upper gastrointestinal bleeding, four sets of variables and different measures of performance (discriminatory ability, sharpness, reliability). For all criteria of performance, the differences between the models were small in the sets with few variables. In the sets with many variables there were marked differences between the models; however, no model was superior in all aspects of performance. Incorporation of interactions in models based upon Bayes’ theorem are worthwhile if many variables are used and the discriminatory ability is considered.

Bei der computerunterstützten Prognose führen theoretische Argumente für ein anspruchsvolles Modell nicht unbedingt zu besseren Ergebnissen. In dieser Studie wurden naheliegende Erweiterungen des einfachen und leicht verständlichen Unabhängigkeits-Bayes-Modells unter Berücksichtigung von Interaktionen zwischen Variablen mit Hilfe eines klinischen Datensatzes zur oberen Gastrointestinalblutung, vier Variablensätzen und verschiedenen Erfolgskriterien (Diskriminierungsfähigkeit, Schärfe, Zuverlässigkeit) untersucht. Die Unterschiede zwischen den Modellen waren gering bei den Datensätzen mit wenigen Variablen und allen Erfolgskriterien. Größere Unterschiede wurden bei den Datensätzen mit vielen Variablen beobachtet. Jedoch zeigte sich kein Modell hinsichtlich aller Kriterien überlegen. Die Berücksichtigung von Interaktionen im Bayes-Modell erscheint sinnvoll bei Datensätzen mit vielen Variablen und bei Betrachtung der Diskriminierungsfähigkeit.

Key-Words:

Computer-aided Prognosis - Independent Bayes Model - Conditional Dependence - Linear Logistic Regression - Performance Criteria - Upper Gastrointestinal Bleeding

Schlüssel-Wörter:

Computerunterstützte Prognose - Unabhängigkeits-Bayes-Modell - bedingte Abhängigkeit - lineare logistische Regression - Leistungskriterien - obere Gastrointestinalblutung

• REFERENCES

• 1 Anderson J. A. Separate sample logistic discrimination. Biometrika 1972; 59: 19-36.
• 2 Bishop Y. M. M, Fienberg S. E, Holland P. W. Discrete Multivariate Analy sis. Cambridge: MIT; 1978
• 3 Chow C. K, Liu C. N. Approximating discrete probability distributions with dependence trees. IEEE Trans. Inform. Theor 1968; 14: 462-467.
• 4 Coomans D, Broeckaert I, Jonckheer M, Massart D. L. Comparison of multivariate discrimination techniques for clinical data application to the thyroid functional state. Meth. Inform. Med 1983; 22: 93-101.
• 5 Copas J. B. Regression, prediction and shrinkage (with discussion). J. roy. Statist. Soc B 1983; 45: 311-354.
• 6 Croft D. J, Machol R. E. Mathematical models in medical diagnosis. Ann. biomed. Engin 1974; 2: 69-89.
• 7 Dawid A. P. Properties of diagnostic data distributions. Biometrics 1976; 32: 647-658.
• 8 Engelman L. Stepwise logistic regression. In Dixon W. J. (Edit.) BMDP Statistical Software. Berkeley: University of California; 1983
• 9 Fryback D. G. Bayes’ theorem and conditional nonindependence of data in medical diagnosis. Comp. biomed. Res 1978; 11: 423-434.
• 10 Habbema J. D. F, Hilden J, Bjerregaard B. The measurement of performance in probabilistic diagnosis: I. The problem, descriptive tools, and measures based on classification matrices. Meth. Inform. Med 1978; 17: 217-226.
• 11 Habbema J. D. F, Gelpke G. F. A computer program for selection of variables in diagnostic and prognostic problems. Comp. Prog. Biomed 1981; 13: 251-270.
• 12 Hall G. H. The clinical application of Bayes’s theorem. Lancet 1967; II: 555-557.
• 13 Hartigan J. Cluster analysis of variables. In Dixon W. J. (Edit.) BMDP Statistical Software. Berkeley: University of California; 1983
• 14 Hilden J, Habbema J. D. F, Bjerregaard B. The measurement of performance in probabilistic diagnosis. II: Trust-worthiness of the exact values of the diagnostic probabilities. Meth. Inform. Med 1978; 17: 227-237.
• 15 Hilden J, Habbema J. D. F, Bjerregaard B. The measurement of performance in probabilistic diagnosis. III: Methods based on continuous functions of the diagnostic probabilities. Meth. Inform. Med 1978; 17: 238-246.
• 16 Lincoln T. L, Parker R. D. Medical diagnosis using Bayes theorem. Health Serv. Res 1967; 2: 34-45.
• 17 Morgan A. G, Clamp S. E. O.M.G.E. International upper gastrointestinal bleeding survey 1978–1982 Scand. J. Gastroenterol 1984; 19 Suppl (Suppl. 95) 41-58.
• 18 Norusis M. J, Jacquez J. A. Diagnosis. I. Symptom nonindependence in mathematical models for diagnosis. Cornput. Biomed. Res 1975; 8: 156-172.
• 19 Norusis M. J, Jacquez J. A. Diagnosis. II. Diagnostic models based on attribute clusters: a proposal and comparisons. Comput. Biomed. Res 1975; 8: 173-188.
• 20 Ohmann C, Künneke M, Zaczyk R, Thon K, Lorenz W. Selection of variables using “independence Bayes” in computeraided diagnosis of upper gastrointestinal bleeding. Stat. Med 1986; 5: 503-515.
• 21 Ohmann C, Thon K, Stöltzing H, List E, Zaczyk R, Ennis M, Lorenz W, Röher H. D. The personal computer as an aid to documentation of upper gastrointestinal endoscopy. Theor. Surg 1986; 1: 69-83.
• 22 Ohmann C, Thon K, Stöltzing H Qin Yang, Lorenz W. Upper gastrointestinal tract bleeding: assessing the diagnostic contributions of the history and clinical findings. Med. Dec. Mak 1986; 6: 208-215.
• 23 Press S. J, Wilson S. Choosing between logistic regression and discriminant analysis. J. Amer. Stat. Ass 1978; 73: 699-705.
• 24 Roy S. N, Mitra S. K. An introduction to some non-parametric generalizations of analysis of variance and multivariate analysis. Biometrika 1956; 43: 361-376.
• 25 Russek E, Kronmal R. A, Fisher L. D. The effect of assuming independence in applying Bayes theorem to risk estimation and classification in diagnosis. Comput. biomed. Res 1983; 16: 537-552.
• 26 Thon K, Ohmann C, Stöltzing H, Lorenz W. Medical decision making in upper gastrointestinal bleeding: the impact of clinical criteria for the diagnosis of recurrent hemorrhage. Theor. Surg 1986; 1: 32-39.
• 27 Titterington D. M, Murray G. D, Murray L. S, Spiegelhalter D. J, Skene A. M, Habbema J. D. F, Gelpke G. J. Comparison of discrimination techniques applied to a complex data set of head injured patients. J. roy. Statist. Soc. A 1981; 144: 145-175.
• 28 Toussaint G. T, Sharpe P. M. An efficient method for estimating the probability of misclassification applied to a problem in medical diagnosis. Comp. Biol. Med 1975; 4: 269-278.
• 29 Trampisch H. I. Zuordnungsprobleme in der Medizin: Anwendung des Lokations-modells. In Überla K, Reichertz P. L, Victor N. (Eds) Medizinische Informatik und Statistik, Vol. 63. Berlin: Springer; 1986