TECH-VER

TECH-VER standardizes technical verification of health economic decision-analytic models to identify implementation errors and improve calculation transparency and reproducibility.

Key Features:

• Structured domains: Organized into five verification domains: input calculations; event-state (patient flow) calculations; result calculations; uncertainty analysis calculations; and other overall checks (e.g., validity or interface).
• Completeness and consistency checks: Identification and justification of all relevant model calculations as prerequisites for verification tests.
• Black-box testing: Initial checks that compare model outputs against prior expectations or reference values without inspecting internal code.
• White-box testing: Detailed examination of program code or cell-by-cell analysis for crucial calculations when black-box tests yield unexpected results.
• Model replication and parallel programming: Independent replication or parallel implementation of models used when issues persist after white-box testing.
• Transparency and root-cause identification: Systematic procedures to detect implementation errors and trace their root causes in model calculations.
• Fit-for-purpose tests and exemplars: Includes fit-for-purpose tests, practical tips, and demonstrative error examples to guide verification decisions.
• Training and quality control: Serves as an internal framework for training and for quality control during model development.

Scientific Applications:

• Model verification across stakeholders: Applicable to health economic decision-analytic models developed by drug manufacturers, consultancies, and academic institutions across diverse software environments.
• Internal quality assurance: Employed for internal quality control and validation of implementation during model development.
• External model appraisal: Supports external technical appraisal and critical evaluation of model credibility and accuracy.

Methodology:

Perform completeness and consistency checks followed by hierarchical verification tests comprising black-box testing, white-box (code or cell-by-cell) examination, and model replication or parallel programming when required.