An architecture for a continuous, user-driven, and data-driven application of clinical guidelines and its evaluation

Area of Science:

• Medical Informatics
• Clinical Decision Support Systems
• Artificial Intelligence in Healthcare

Background:

• Continuous care requires sophisticated decision support systems capable of handling complex, time-sensitive patient data.
• Existing guideline-based systems often lack the flexibility for continuous care, multiple task types, and adaptive modes.
• There is a need for architectures that support both data-driven and user-driven decision-making in real-time.

Purpose of the Study:

• To design, implement, and evaluate a novel architecture for continuous guideline (GL)-based decision support.
• To address requirements for continuous care, multiple task types, and data/user-driven modes.
• To assess the technical feasibility and functional validity of the proposed architecture.

Main Methods:

• Developed the PICARD architecture, integrating a temporal reasoning engine and diverse application interfaces.
• Conducted a technical evaluation with 19 simulated preeclampsia scenarios.
• Performed functional evaluations with domain experts and a clinical evaluation involving 36 clinicians comparing manual decisions with PICARD-assisted decisions.

Main Results:

• Technical evaluation showed high correctness (1.0) and completeness (0.96).
• Functional evaluation validated system recommendations with minor style suggestions.
• Clinician use of PICARD increased guideline action completeness from 41% to 93% and reduced redundant actions from 68% to 3%.

Conclusions:

• The PICARD architecture is technically feasible and functionally valid for continuous GL-based decision support.
• It effectively meets requirements for continuous care over extended periods.
• PICARD demonstrates potential benefits for managing chronic patients by enhancing decision completeness and reducing redundancy.