Formalization and acquisition of temporal knowledge for decision support in medical processes

Area of Science:

• Medical Informatics
• Clinical Decision Support
• Health Data Science

Background:

• Long-term medical interventions require precise temporal planning, yet time-related evidence is scarce in Clinical Practice Guidelines (CPGs).
• Clinical data repositories contain valuable temporal information on patient encounters and actions, offering a potential source for temporal knowledge discovery.

Purpose of the Study:

• To classify and represent clinical time constraints using novel formalisms.
• To develop algorithms for automatic generation of temporal models from clinical data.
• To assess the adherence of these data-driven models against CPG recommendations.

Main Methods:

• Analysis of temporal aspects in patient procedures and CPGs for chronic diseases.
• Development of 'micro-temporality' and 'macro-temporality' formalisms.
• Implementation and testing of three algorithms for automated temporal constraint generation from clinical databases.

Main Results:

• A comprehensive classification and representation of medical time constraints were established.
• Automated generation of temporal constraints from data for 8781 Arterial Hypertension patients.
• Identified discrepancies between data-driven visit frequencies (1-7 weeks) and CPG recommendations (2-4 weeks).

Conclusions:

• Experience-based temporal knowledge from clinical data complements CPGs, offering detailed temporal representations.
• The developed model effectively captures temporal knowledge in chronic disease management.
• Algorithms facilitate adherence analysis, revealing potential shortcomings in CPG time recommendations.