Evidens

Pull the inputs.
Skip the PDFs.

AI agents read every Medicinrådet decision for you — comparators, utility weights, survival models and assumptions extracted, cited, and organised by sygdomsområde. Know the precedent before you design the analysis.

Explore the demo Private beta · no credit card.
evidens.app/demo
Hæmatologi
Venetoclax
Anbefalet
Oral azacitidin
Ikke anbefalet
Abemaciclib
Anbefalet
Health-economic inputs
EQ-5D-3L0,889relapsfri → 5L
OS · Spline Gen. gamma
ICER · mio. kr./QALY · ved AIP
4,1
Why Engine

The 0.09 gap is an instrument artefact: VIALE-A used EQ-5D-5L, QUAZAR the older 3L…

Built for the analysts at HTA consultancies and pharma market-access teams who build Danish submissions.

Updated daily — agents re-scrape Medicinrådet overnight
43
Appraisals
10
Sygdomsområder
215
Documents indexed
3
Flagship datasets
Why now

Record volume. Zero structured access.

Medicinrådet decides more than ever — and every decision adds precedent your next submission is judged against.

86
applications to Medicinrådet in 2025
a record year — volume keeps climbing
63%
anbefalet or delvist anbefalet
53 of 84 medicines assessed in 2025
~19
weeks of actual processing
against the 18-week target track

The submitted Excel model is never published. The vurderingsrapport is the only public trace of the inputs the council accepted — and it is a Danish PDF, built for reading, not for reuse.

There is no structured database to fall back on: the only free cross-agency HTA database missed 75% of sampled appraisals in a 2025 peer-reviewed audit. The status quo is agency websites, folders of PDFs, and re-reading work someone on your team already did.

Sources: Medicinrådets årsberetninger 2024–2025 · Lif's Medicinrådstracker · INAHTA database completeness study (2025).

How it works

From dossier to model input, in three moves

The workflow a health economist actually runs — no re-reading required.

01

Browse by sygdomsområde

Start where a health economist thinks — pick a disease area and see every appraisal, with its Medicinrådet status.

Hæmatologi8
Lungekræft6
Nyrekræft5
02

Pull the cited inputs

Open any appraisal and read the utilities, survival models and assumptions — each as applicant → council, cited to the passage.

Utility · relapse-free remission
EQ-5D-3L0,730,71
03

Reconcile with the Why Engine

When two dossiers disagree on a value, ask why. Retrieval over the full texts explains the gap, with sources.

Why do VIALE-A and QUAZAR disagree on the RFS utility?
Vurdering · VIALE-A · s.44
AI agents

Ask the corpus. Agents do the reading.

Pose the question a submission raises — agents retrieve the passages, reconcile the numbers, and answer with citations you can open.

The agents

Evidence you can defend in a submission

Three agents work the corpus — everything they claim is anchored to a citation you can open, not a black box.

Extraction agent

Every number, traced applicant → council

Every utility weight, survival model and assumption pulled from the PDF as an applicant → council value, traced to the exact passage. No more re-reading 90-page vurderinger.

Health-economic inputs
EQ-5D-3LUtility · RFS remission0,730,71
EQ-5D-5LUtility · progressed0,61accepteret
Konfidentiel nettoprisXXXX
Synthesis agent

The methodology, read for you

AI reads the methodology, not just the numbers: risk-of-bias on each source study, and EQ-5D-3L → 5L normalisation so values from different eras are comparable.

OS extrapolation
Spline Gen. gamma
Low risk of bias 3L → 5L normalised
Why Engine agent

Ask why two dossiers disagree

When two appraisals disagree on a utility or an extrapolation, ask why. Retrieval over the full texts explains the gap — instrument, population, or council adjustment — with citations.

Why Engine
Why do VIALE-A and QUAZAR disagree on the relapse-free utility?
Retrieved passages
Vurdering · VIALE-A · s.44
Vurdering · QUAZAR · s.38

The 0.09 gap is an instrument artefact: VIALE-A elicited utilities with EQ-5D-5L, QUAZAR with the older 3L, which scores remission systematically higher. Normalised to 5L, the values converge.

Trust

Compliance-grade by default

The guarantees a pharma legal review asks about first — answered up front.

Never trained on your data

Your questions, files and models stay yours. Nothing you do in Evidens is used to train a model — ours or anyone else's.

GDPR-first, EU-hosted

EU data residency, SSO, and an audit trail on every extracted value — built to pass a pharma compliance review.

Redactions stay redacted

Values printed XXXX in the source are never inferred, estimated or reconstructed. Public documents only.

Fresh every day

Agents watch Medicinrådet and re-scrape daily, so new anbefalinger and vurderinger land the day they publish.

"Half my model-building time goes into finding and re-justifying inputs I've already seen in an old vurdering. This is that time back."
— Senior health economist, Nordic market access (paraphrased)

See it on your own sygdomsområde

Explore the working demo, then request access to bring it to your team.

FAQ

Questions a health economist asks first

Is the Danish data live?

The demo runs on a snapshot of public Medicinrådet documents — 43 appraisals across 10 sygdomsområder. In the live product, agents re-scrape Medicinrådet daily, so new anbefalinger and vurderinger are extracted and searchable the day they publish.

Can it check whether my comparator has precedent?

Yes — that is the first question Evidens is built to answer. Medicinrådets metodevejledning (§2.4.2) makes it a formal gate: an analysis against a comparator generally suffices only if the council has previously assessed and recommended it for the indication. Ask the assistant and you get the matching appraisals with status, year and the citations to check.

How accurate is the extraction — can I trust the numbers?

Every value is shown as the applicant → council pair and cited to the exact passage (e.g. Vurdering · §7.3 · s.44), so you verify against the source PDF in one click. Flagship datasets are hand-verified; nothing is shown without a citation you can check.

Do you train AI models on my data?

No — never. Your questions, uploaded files and models are not used to train any model, ours or a third party's. Prompts and outputs stay inside your workspace, hosted in the EU under GDPR.

Can I export to my own model?

Yes. Pull utilities, survival choices and assumptions into the Excel or R model you already build in — citations attached. Evidens meets your model where it is; it is not another silo to migrate to.

How is confidential or redacted data handled?

Values printed as XXXX in the source stay redacted — commercially confidential figures are never inferred, estimated or reconstructed. Evidens only processes public Medicinrådet documents.

Which markets are covered?

Denmark, deep — every Medicinrådet appraisal, down to the extracted inputs. On the roadmap: TLV (Sweden), NoMA (Norway), the Joint Nordic HTA-Bodies (JNHB), and EU joint clinical assessment (JCA) reports as they publish. Tell us your priority when you request access — it shapes the order we build.

Is this validated for reimbursement decisions?

No. Evidens is decision-support that makes building and justifying a model faster; it is not a source of record. Always verify against the primary Medicinrådet document before it informs a submission.

Private beta

Built for Nordic HTA teams

We're onboarding a first cohort of health economists working on Danish submissions — with Sweden, Norway and EU JCA reports on the roadmap. Tell us what you'd reuse most and we'll prioritise it.

  • Denmark (Medicinrådet) live first
  • Your inputs, cited back to the source PDF
  • Export to your model, not another silo

No spam. We use this only to prioritise Nordic HTA teams for the private beta.