Critical appraisal and evidence-based medicine, taught the way a clinician actually thinks — starting with a patient, not a p-value.
This course grew out of a session originally taught to a mix of medical students and non-medical learners. Each lesson builds on the last: we start with the patient, move to asking a well-formed clinical question, then learn to judge the evidence itself — and finally, how to use LLMs responsibly as a research assistant along the way.
The patient at the centre: history, examination, SOAP notes, PaJR, spotting fake news, and a first look at the evidence pyramid.
AvailableTurning a vague clinical worry into a PICO question, matching question type to study design, and building a real PubMed search.
AvailableThe three-question framework — validity, results, applicability — plus ARR, RRR, NNT, and confidence intervals worked through by hand.
AvailableHow trials get pooled, reading a forest plot, heterogeneity (I²), funnel plots for publication bias, and appraising a review with AMSTAR-2.
AvailableSensitivity, specificity, PPV/NPV, likelihood ratios, ROC curves, and why prevalence changes how useful a test is for your patient.
AvailableCohort and case-control designs, hazard ratios, confounding, and why RCTs are the wrong tool for long-term risk or rare harms.
AvailableHow guidelines are built (GRADE), spotting industry influence, and reconciling a guideline with a patient who doesn't fit the trial average.
Availablep-values vs. confidence intervals, what "significant" really claims, p-hacking, surrogate endpoints, and reading limitations sections critically.
AvailableA real, de-identified patient taken end to end — history, PICO question, search, appraisal, decision — tying the whole course back to Lesson 1.
AvailableA fast, jargon-translated map of the whole course — PICO as a search schema, RCTs as A/B tests, likelihood ratios as Bayesian updates, GRADE as test-coverage vs. ship-decision. Read this first if you have no clinical background, or keep it open as a cheat sheet while working through the lessons.
Open the Techie Summary →A companion prompt library for using an LLM at each stage of the EBM workflow — formulating PICO questions, building search strategies, screening abstracts, extracting trial methodology, calculating ARR/RRR/NNT, and stress-testing your own appraisal. Framed with one hard rule throughout: the LLM drafts, you verify every number and citation against the source.
Open the Prompt Library →A shortlist of the most counterintuitive results across the course — the same test giving an 8% predictive value in one population and 90% in another, a "40% risk reduction" that turns out to mean treating 25 people to help one, five underpowered trials adding up to a confident pooled answer, and more. Each one links back to the lesson where it's worked through in full.
Browse the Insights →