TruePositive

Home › Training

Training & courses

Courses for people who have to do the work, not just approve it. Every course is built around the participants' own data, their assessment body and their tooling — and ends with something they can use the following week.

In house Delivered for a single organisation, using your own examples
Open Mixed-audience sessions for individual participants
Hands on Participants leave with runnable code and templates, not slides alone

Course 01

Real-world evidence for health technology assessment

The course for market access, HEOR and medical teams who keep being told to "add some real-world evidence" and need to know what would actually count.

We work through how assessment bodies read an RWE package: where real-world data is accepted without argument, where it is treated as supportive only, and where it is rejected outright — and why those lines sit where they do. Participants then plan an evidence package for a case of their own, and defend it to the room.

Covered

  • What RWE is used for in HTA: burden of disease, comparative effectiveness, external controls, managed entry
  • The EU HTA Regulation and the Joint Clinical Assessment — what changed, and what it means for evidence planning
  • Nordic and national assessment practice, and how much it still varies
  • Data sources: registries, EHR, claims, cohorts — strengths and known traps of each
  • Why RWE submissions fail: unclear estimand, unfixable confounding, undocumented decisions
  • Building an evidence generation plan that survives contact with a reviewer

Course 02

Observational study design and causal inference in health data

The technical companion to the first course, for the people who will design and run the studies.

The organising idea is the target trial: write down the randomised trial you would run if you could, then be explicit about every way your observational study departs from it. It is a discipline that catches most of the classic errors before any code is written.

Covered

  • Estimands: saying precisely what quantity you are trying to estimate
  • Target trial emulation, start to finish, on a real example
  • Immortal time, prevalent user and depletion-of-susceptibles bias — how each one is designed away
  • Propensity scores, weighting and overlap; when matching is the wrong tool
  • Time-varying exposure and confounding: when standard adjustment silently fails
  • Quantitative bias analysis and negative controls: proving you looked
  • Reporting so the study is reproducible and reviewable

Course 03

Analysing health data in secure research environments

Getting productive inside a controlled research platform — where you cannot download the data, cannot install what you like, and every output that leaves is checked.

The course is taught on a live environment. The current standing example is the NIH All of Us Researcher Workbench, which since its 2.0 release runs on a different platform to the one most published tutorials describe — a good illustration of why environment-specific training beats generic documentation. The same session can be taught against a Danish register environment or a client's own platform instead.

Covered

  • How access tiers, approvals and data-use agreements actually constrain analysis
  • Workspaces, compute environments and cost control
  • Controlled versus referenced resources — the concept that explains deletion, sharing and permissions all at once
  • Reproducible analysis in R inside the environment, with version control
  • Cohort building, and why a rebuilt cohort may not reproduce an old count
  • What may and may not leave the environment, and how to prepare outputs for review

Course 04

AI assistants in regulated life-science work

A practical, non-hyped session on where AI assistants genuinely save time in evidence work — and where using one would be a compliance incident.

Aimed at teams that have been given a copilot licence and little else. We cover what these tools do well, the specific ways they fail in technical and regulated work, and the handling rules that keep confidential and personal data out of them. Participants practise on their own realistic tasks.

Covered

  • What the tools are actually good at: drafting, summarising, restructuring, code scaffolding
  • Failure modes that matter here: confident wrong citations, invented figures, stale facts
  • Confidentiality, personal data and what must never be pasted into a prompt
  • Verification habits — treating output as a draft from a fast, unreliable colleague
  • Documentation and audit trail: recording what was AI-assisted and how it was checked
  • Writing prompts and reusable instructions for recurring team tasks

Booking

How a course booking works

  1. A short scoping call

    Who is in the room, what they already know, and what should be different afterwards.

  2. A tailored outline

    You get the agenda and learning outcomes in writing before committing — including which of your own cases will be used as examples.

  3. Delivery

    On site or online. Materials, exercises and code are included and are yours to keep and reuse internally.

  4. Follow-up

    A written summary of open questions raised in the room, and an optional clinic a few weeks later once people have tried to apply it.

Fees are per delivery rather than per participant, so the cost does not rise with the size of the room. Ask for a quote with your dates and group size.

Want a course built around your own cases?

Tell us who is in the room and what they need to be able to do afterwards. You will get a proposed outline before anything is booked.

Request an outline