A physician-based architecture for the construction and use of statistical models

Area of Science:

• Medical Informatics
• Biostatistics
• Clinical Research

Background:

• Physicians require specialized tools to effectively utilize published research findings.
• Interpreting complex statistical data from clinical trials poses a significant challenge for medical practitioners.

Purpose of the Study:

• To introduce a novel knowledge-based computer framework, the physician-based (PB) architecture, designed to create tailored tools for physicians.
• To demonstrate the utility of the PB architecture using the interpretation of two-arm parallel randomized clinical trials (TAPRCT) as a case study.

Main Methods:

• Statistical models represented using influence diagrams.
• A physician-based user interface (patient-flow diagram) translating influence diagram elements into user-friendly language.
• A cohort-state diagram encoding clinical-epidemiological knowledge for statistical model transformations.
• A Bayesian system (THOMAS) implementing the PB architecture to support physician decision-making.

Main Results:

• The PB architecture successfully maps complex statistical concepts into a physician-understandable format.
• The THOMAS system aids physicians in assessing clinical significance, prior beliefs, and methodological concerns related to TAPRCT data.
• The developed framework facilitates better interpretation of randomized clinical trial results.

Conclusions:

• The physician-based architecture provides a viable solution for developing specialized computer tools for physicians.
• This framework enhances physicians' ability to leverage published research, particularly from TAPRCT.
• The PB architecture has broad applicability for various statistical tools and user groups in healthcare.