A health technology assessment submission asks a payer or assessment body to accept that a technology delivers enough additional benefit to justify its cost in that health system. The workflow that produces it runs from evidence planning through dossier assembly, economic modelling, submission, assessment dialogue and decision — typically over months, and typically in parallel across markets with different requirements.
Understanding where the time actually goes is what determines whether automation helps. For the wider picture, see AI for market access and HEOR.
The workflow, stage by stage
1. Evidence planning and gap analysis. What will this HTA body require, what evidence exists, what is missing. Bottleneck: knowing what exists — evidence is scattered without a single view. Automatable substantially; the judgement about whether a gap is worth filling is not.
2. Systematic literature review. Screening, extraction, quality assessment. Bottleneck: volume — screening thousands of abstracts to find dozens of relevant studies. One of the clearest cases for automation in the whole workflow. See AI for systematic literature review.
3. Evidence synthesis and indirect comparison. Bottleneck: methodological judgement and data availability. Automatable: the data assembly feeding the analysis, not the methodological choices.
4. Economic modelling. Bottleneck: parameter sourcing and justification — every input needs a source, and assessment bodies scrutinise these closely. Automatable: parameter identification and source documentation, not the model structure.
5. Dossier assembly. Frequently the single largest time sink, and almost none of it is professional judgement. Substantially automatable, given a governed evidence base and an authored template. See AI value dossier automation.
6. Internal review and approval. Bottleneck: reviewer capacity and rework when claims cannot be substantiated on demand. When every claim arrives with its source passage displayed, review becomes verification rather than investigation.
7. Submission and assessment dialogue. Bottleneck: turnaround under time pressure, with the original analysts frequently reassigned. A submission where every claim's provenance is recorded is dramatically faster to defend than one living in a spreadsheet and someone's memory.
8. Decision, and reuse. Bottleneck: reuse rarely happens well, because each submission tends to be rebuilt from a finished document rather than structured evidence. This is the structural fix a governed base provides.
The bottleneck nobody automates away
Across the eight stages, the genuinely automatable work concentrates in assembly, extraction, substantiation and reuse. The judgement — comparator selection, methodological choices, positioning against precedent, deciding how to present a weakness — is not automatable, and it is what determines whether the submission succeeds. A vendor promising end-to-end submission generation is describing a demonstration, not a deployment.
Where Eclypse sits: one governed evidence base generates the global dossier, regional dossiers and local value stories on demand, with provenance tracked on every claim. See the market access evidence base case study this is drawn from.
The useful diagnostic is which of the eight stages actually consumes your programme's time — it is rarely the one people assume.
Common questions about the HTA submission workflow.
Where does most of the time go in an HTA submission?
Usually dossier assembly and evidence substantiation — reformatting evidence, checking cross-references, verifying sources. A large share of elapsed time and almost none of the professional judgement.
Can AI write an HTA submission end to end?
No. It can produce drafts of the assembly-heavy sections grounded in approved evidence. Methodological decisions and strategic positioning require the health economist.
Why do HTA submissions get rebuilt for each market?
Because the previous submission exists as a finished document rather than structured evidence. A governed evidence base addresses this structural inefficiency.
How does AI help during the assessment dialogue?
By making provenance immediately available, so a query about a claim can be answered in minutes rather than by reconstructing the original analysis.
Bring us your hardest submission — we'll map where the time is going.
Book a working session
ECLYPSE
AI
Our proprietary AI orchestration platform for healthcare: one engine, a registry of reusable task modules and domain agents, and a governed knowledge base — deployed inside your walls and run by your team.