Real-world evidence · Health AI · Epidemiology

Real-world evidence for drug development and health AI.

Dry-lab science for pharma, biotech and AI teams: we design and run studies in the Danish national registers, from disease burden and comparative effectiveness to the evaluation of clinical prediction models.

Capabilities

What we generate.

Three areas of work. The question changes between them; the standard for designing a study does not.

Health AI

Cohort design, model evaluation and hypothesis studies for teams building clinical models.

Working with AI teams →

Biostatistics & epidemiology

Survival, competing risks, causal inference and prediction, underneath both and in the randomised trials we analyse.

Selected work →

Six register study types

  1. 01

    Disease Epidemiology & Burden

    The size and shape of the patient population.

  2. 02

    Treatment Patterns & Patient Journeys

    What actually happens after diagnosis.

  3. 04

    External Control Arms

    A comparator for a single-arm study, drawn from patients who would have qualified.

  4. 05

    Post-Authorisation Safety & Effectiveness (PASS)

    After launch, in the patients who were never enrolled; to GVP Module VIII where it is a regulatory commitment.

  5. 06

    Feasibility & Data Strategy

    Whether the question can be answered at all, before anyone commits to a full study.

Track record

Selected work.

The team's published record spans register epidemiology, clinical machine learning and human genetics. Client work includes register and health-economic studies, and randomised trials published in NEJM Evidence and elsewhere.

See all →
npj Aging2026
National cancer burden

Disentangling shifting demographic and treatment effects on years of life lost to cancer in Denmark

View Article →
Clinical Epidemiology2025
Registry data resource

The Danish Lymphoid Cancer Research (DALY-CARE) Data Resource: The Basis for Developing Data-Driven Hematology

View Article →
+54 more studies →

Why Frameshift

Design before data.

No sample size fixes a cohort defined wrong, a time zero that leaks information from the future, or an endpoint chosen after someone has seen the curves.

Ólafur B. Davíðsson

Ólafur B. Davíðsson

Co-Founder & Scientific Director
Ragnar P. Kristjánsson

Ragnar P. Kristjánsson

Co-Founder & Managing Director
Mikkel Werling

Mikkel Werling

AI Lead

Get in touch

Planning a register study
or evaluating a model?

Tell us what you are trying to establish. We usually reply the same or next working day.

> plan(study)
#> 1  15–20 min scoping call
#> 2  Feasibility read, in writing
#> 3  Proposal
> reply_time(study)
#> [1] "same or next working day"
> contact()
#> [1] "info@frameshift.dk"
> scope()

This form is processed by Formspree. Your name, email, and message are forwarded to us and stored by Formspree until deleted. Formspree does not sell or share your data. See our privacy policy. Please do not include confidential or patient-level data in an initial enquiry.