Table 2.

Evaluation criteria for assessing LLM-generated clinical reports within the HealthProcessAI framework.

Criterion	Weight	Description	Validation method
Clinical accuracy	25%	Correctness of medical interpretations and terminology usage	Expert clinical review
Process mining understanding	20%	Accurate interpretation of analytical results	Technical validation
Actionable insights	20%	Quality and feasibility of clinical recommendations	Implementation assessment
Statistical interpretation	15%	Correct analysis of quantitative findings	Statistical validation
Report structure & clarity	10%	Organization and readability	Communication analysis
Evidence-based reasoning	10%	Use of clinical evidence and literature	Evidence synthesis evaluation

*Current validation uses automated LLM evaluation rather than clinical validation. Each criterion is assigned a relative weight based on its importance to clinical utility and interpretability. Validation methods involve domain-specific assessments, including clinical expert review, technical accuracy checks, and implementation feasibility testing.