What the programme covers
Adding hospitals can improve diversity, but it can also create shortcuts when disease prevalence, acquisition protocol or label source is strongly tied to site. A defensible design measures these distributions and assigns sources according to the intended claim.
What a useful specification includes
A defensible request defines the clinical task, source evidence and acceptance criteria before patient-level data moves. The exact fields and thresholds depend on the intended model claim.
- Role of each source site in the study design
- Patient, site, scanner and time separation
- Protocol and device-distribution targets
- Comparable inclusion and ground-truth rules
- Population, prevalence and subgroup requirements
- Per-site minimums and uncertainty thresholds
Quality and validation controls
Multi-site quality review separates true clinical variation from source-specific formatting, workflow or label artifacts. Site identifiers remain available for governed analysis when necessary and permitted.
- Cross-site patient and duplicate controls
- Scanner and protocol distributions
- Site-specific missingness and label sources
- Prevalence and population differences
- Holdout integrity and site-shortcut analysis
Availability, rights and delivery
A multi-site release proceeds only for sources with verified inventory and programme-specific authorization. Additional countries or sites in a sourcing scope remain under feasibility until those conditions are met.
Public pages describe a sourcing and engineering capability, not guaranteed ready inventory. Each release remains subject to verified programme inventory, programme-specific authorization, privacy review, technical acceptance and buyer licence terms.
Questions, answered directly.
Is more sites always better?+
No. Source relevance, label comparability, permissions and study design matter more than an unqualified site count.
Can one hospital be held out for validation?+
Yes when it provides an appropriate independent population and sufficient eligible cases.
How is patient overlap prevented?+
Identity-safe resolution and patient-level split rules are applied where the programme can reliably detect overlap.
Can sites from different countries be combined?+
Potentially, subject to buyer-specific feasibility, applicable permissions, privacy controls and compatible cohort definitions.