A Physician-Based Architecture for the Construction and Use of Statistical Models

Abstract:

Physicians need specially tailored computer tools to take advantage of published research results. We present a knowledge-based computer framework -the physician-based (PB) architecture - for constructing such tools, and we use the problem of physicians’ interpretation of two-arm parallel randomized clinical trials (TAPRCT) as a working example. Statistical models are represented by influence diagrams. The interpretation of influence-diagram elements are mapped into users’ language in a domain-specific, physician-based user interface, called a patient-flow diagram. Statistical-model transformations that maintain the semantic relationships of the model and that embody clinical-epidemiological knowledge are encoded in a mediating structure called the cohort-state diagram. The algorithm that coordinates the interactions among the knowledge representations uses modular actions called construction steps. This architecture has been implemented in a Bayesian system, called THOMAS, that supports physician decision making in light of TAPRCT data. This support entails assessing clinical significance, prior beliefs, and methodological concerns. We suggest that the PB architecture applies to a wide range of statistical tools and users.

Keywords

Computer-aided Therapy - Knowledge Representation - Statistical Programs - Probability Theory - Clinical Epidemiology

• REFERENCES

• 1 Feinstein AR. Clinical Epidemiology: The Architecture of Clinical Research. Philadelphia: W. B. Saunders Company; 1985
• 2 Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology: The Essentials. 2nd ed. Baltimore: Williams & Wilkins; 1988
• 3 Sackett DL. Clinical Epidemiology: A Basic Science for Clinical Medicine. 2nd ed. Boston: Little, Brown; 1991
• 4 Hjalmarson AA. Effect on mortality of metoprolol in acute myocardial infarction. Lancet 1981; 02: 823-7.
• 5 Greenland S, Robins JM. Identifiability, exchangeability, and epidemiological confounding. Int J Epid 1986; 15: 413-9.
• 6 Howard RA, Matheson JE. Readings on the Principles and Applications of Decision Analysis. Menlo Park, CA: Strategic Decision Group; 1981
• 7 Oliver RM, Smith JQ. Influence Diagrams, Belief Nets, and Decision Analysis. Chichester: Wiley; 1990
• 8 Verma T, Pearl J. Causal networks: Semantics and expressiveness. In: Shachter RD, Levitt TS. eds. Proceedings of the Fourth Annual Workshop on Uncertainty in Artificial Intelligence. University of Minnesota, August 19-20: AAAI: American Association for Artificial Intelligence; 1988: 352-9.
• 9 Smith JQ. Influence diagrams for statistical modelling. Ann Stat 1989; 17: 654-72.
• 10 Eddy DM, Hasselblad V, Shachter R. The Statistical Synthesis of Evidence: Meta-analysis by the Confidence Profile Method. Boston: Academic Press; 1992
• 11 Wulff HR. et al. What do doctors know about statistics?. Stat Med 1987; 06: 3-10.
• 12 Musen MA. Automated Generation of Model-based Knowledge-Acquisition Tools. London: Pittman; 1989
• 13 Shachter RD, Heckerman DE. Thinking backward for knowledge acquisition. AI Magazine 1987; a; 08: 55-61.
• 14 Lehmann HP. A Bayesian Computer-based Approach to the Physician’s Use of the Clinical Research Literature [Dissertation]. Stanford University; 1991
• 15 Shachter RD. Posterior mode analysis. In: Oliver RM, Smith JQ. eds. Influence Diagrams, Belief Nets, and Decision Analysis. Wiley: London; 1990
• 16 Berndt EK. Estimation and inference in nonlinear structural models. Ann Econ Soc Meas 1974; 3/4: 653-65.
• 17 Nelder JA, Wolstenhome D. A front-end for GLIM. In: De Antoni F, Lauro N, Rizzi A. eds. COMPSTAT-86. Fort Collins CO: American Statistical Association; 1986: 113-7.
• 18 Oldford RW, Peters SC. DINDE: Towards more sophisticated software environments for statistics. SIAM J Sci Stat Comp 1988; 09: 191-211.
• 19 Lubinsky DJ. TESS: A tree-based environment for statistical strategies. In: De Antoni F, Lauro N, Rizzi A. eds. COMPSTAT-87. 1987: 565-7.
• 20 Huber PJ. Environments for supporting statistical strategy. In: Gale WA. eds. Artificial Intelligence and Statistics. Reading MA: Addison-Wesley; 1986: 285-94.
• 21 Goldman RP, Charniak E. Dynamic construction of belief networks. Proceedings of the Sixth Annual Workshop on Uncertainty in Artificial Intelligence; Cambridge MA: 1990: 90-7.
• 22 Breese J. Knowledge Representation and Inference in Intelligent Decision Systems [Dissertation]. Department of Engineering-Economic Systems, Stanford University; 1987
• 23 Bacchus F. Using first-order probability logic for the construction of Bayesian networks. In: Heckerman D, Mamdani A. eds. Uncertainty in Artificial Intelligence. Proceedings of the Ninth Conference. Washington DC: Morgan Kaufmann Publ; 1993: 219-26.
• 24 Egar JW, Musem MA. Graph-grammar assistance for automated generation of influence diagrams. In: Heckerman D, Mamdani A. eds. Uncertainty in Artificial Intelligence: Proceedings of the Ninth Conference. Washington, DC: Morgan Kaufmann Publ; 1993: 235-42.
• 25 Lehmann HP. A decision-theoretic model for using scientific data. In: Henrion M et al. eds. Uncertainty in Artificial Intelligence 5. Amsterdam: North-Holland Publ Comp; 1990: 309-394.
• 26 Weiner JM, Horowitz RS, Bauer M. Clinical trial expert system. In: De Antoni F, Lauro N, Rizzi A. eds. COMPSTAT-87. 1987: 117-22.
• 27 Chytil MK. Metadata and its role in medical informatics. In Salomon R, Blum B, Jørgensen M. eds. MEDINFO 86. Washington DC: North-Holland Publ Comp; 1986: 149-54.
• 28 Louis TA. et al. Crossover and self-controlled designs in clinical research. N Engl J Med 1984; 210: 24-31.
• 29 Cantor AB. Sample size calculations for the log rank test: A Gompertz model approach. J Clin Epid 1992; 45: 1131-6.