Scientific foundations and validation strategy
The scientific programme starts with an existing randomised intervention dataset and proceeds through transparent secondary analysis, model comparison, independent validation and clinical utility testing
Source research
Prospectively collected randomised intervention data
The source study includes 60 adults allocated across four conditions: CBT, computerised cognitive rehabilitation, combined intervention and waiting-list control. It includes clinical outcomes and multimodal measures relevant to cognitive, personality and neurophysiological profiling.
The website must not publish efficacy results until the analyses and manuscript reporting are complete. The current dataset supports feasibility and hypothesis development, not definitive treatment-response prediction.
Model development logic
Test whether complexity earns its place
The analysis should compare progressively richer models to determine whether each data domain provides incremental value.
1. Clinical measures only
2. Clinical plus cognitive measures
3. Clinical plus cognitive plus personality measures
4. All eligible domains including QEEG
If QEEG does not provide meaningful incremental value, it should not be retained simply because it appears technologically advanced.
Validation ladder
Question
Level
Are data quality, preprocessing and outputs reliable and reproducible?
Analytical validity
Does the exploratory model perform under leakage-safe resampling and uncertainty analysis?
Internal validity
Does performance generalise to an independent Dutch population?
External validity
Does the workflow improve decisions, documentation, efficiency or outcomes in practice?
Clinical utility
Can the system fit Dutch care, governance, procurement and professional responsibility?
Implementation
Scientific transparency
Scientific transparency
Pre-specify primary and exploratory analyses where possible.
Separate training, validation and independent testing.
Report uncertainty and subgroup limitations.
Document missing-data and QEEG quality-control procedures.
Avoid participant-level examples that could enable re-identification.
Publish negative as well as positive incremental-value findings.
Collaborate on independent validation
We are interested in research partnerships that strengthen reproducibility, Dutch clinical relevance and responsible translation.
