Purpose: Sackett’s five-step evidence-based medicine cycle — Ask a focused clinical question, Acquire the evidence, Appraise its quality and relevance, Apply it to the patient in front of you, Assess the outcome and your own performance — is the field’s standard model of how a clinician should relate to evidence over time. It was not the organising structure of this repository (the modules are organised by clinical moment: teaching, documentation, ward rounds, registry analytics, reasoning cognition, social context, evidence search). This document is a second, orthogonal cut through the same material: every module and framework mapped onto which EBM step(s) it actually trains.
How to read the table: A module’s primary EBM step is where its core mechanic lives. Many modules touch more than one step — that overlap is marked, not hidden, because the honest answer is that real clinical reasoning rarely respects clean phase boundaries. Modules that don’t map onto the EBM cycle at all (because they’re about something else entirely — persona design, longitudinal monitoring infrastructure) are listed in their own section at the end rather than forced into a fit that isn’t there.
| Step | What it means | The question being answered |
|---|---|---|
| Ask | Convert a clinical problem into a focused, answerable question (often PICO-structured) | “What exactly do I need to know?” |
| Acquire | Search for and retrieve the relevant evidence or information | “Where do I find the answer?” |
| Appraise | Critically evaluate the evidence’s validity, size of effect, and relevance | “Is this evidence any good, and how good?” |
| Apply | Integrate the evidence with clinical expertise and the specific patient’s circumstances | “Does this evidence fit my actual patient?” |
| Assess | Evaluate the outcome of applying it, and one’s own performance in the cycle | “Did it work, and did I do this well?” |
| Module / Framework | Ask | Acquire | Appraise | Apply | Assess | Primary Step |
|---|---|---|---|---|---|---|
| 1 — Socratic Clinical Reasoning | ● | ● | ● | Ask (Steps 1.0–1.3 force the learner to frame the clinical question before any answer is given) | ||
| 2 — Patient-Advocate Case Documentation | ● | ● | ● | Acquire (structured capture of symptoms, exam, prescriptions is evidence acquisition at the individual-case level) | ||
| 3 — Extended Patient-Advocate Monitoring | ● | ● | ● | Acquire → Assess (ongoing data capture, then Step 3.6–3.7 assess trajectory against baseline) | ||
| 4 — Peer-Level Ward Round Preparation | ● | ● | ● | Apply (the module’s centre of gravity is applying known clinical knowledge under ward-round time pressure) | ||
| 5 — Real-Time Case Review & Data Audit | ● | ● | ● | Appraise (Steps 5.1, 5.8 audit data quality and significance within one case) | ||
| 6 — Registry-Level Analytics | ● | ● | ● | ● | Acquire → Appraise (registry querying is acquisition; the structured-vs-narrative confidence distinction running through every level is appraisal) | |
| 7 — Longitudinal & Cross-Case Learning | ● | ● | ● | ● | Appraise (Step 7.12’s overfitting/replication audit is the most rigorous appraisal step in the entire stack) | |
| 8 — Socratic-Mode Design Specification | ● | ● | Appraise, but appraising prompts, not clinical evidence — a meta-layer (see Note below) | |||
| 9 — N-of-1 Case Research Protocol | ● | ● | ● | ● | Acquire → Appraise (Stage 2 is a literal PRISMA-style search; Stages 3–6 are structured appraisal; the module is organised around the full cycle more explicitly than any other) | |
| 10 — Medical Journal Article Reading | ● | ● | ● | ● | Appraise (the entire module is a deep-dive appraisal engine — methods, statistics, bias, applicability — applied to one already-acquired article) | |
| 11 — Patient Education Query Intelligence | ● | ● | Apply (translating evidence/knowledge into what a specific patient needs to hear) | |||
| 12 — Differential Diagnosis Deepdive | ● | ● | Appraise (adversarially appraising the diagnostic conclusion itself, not external literature — a sibling of Module 8’s meta-appraisal) | |||
| 13 — Medication Reconciliation & Polypharmacy | ● | ● | ● | ● | ● | Touches all five at the single-patient level; closest thing in the stack to a complete EBM cycle run on one clinical problem (medication safety) end to end |
| 14 — Global Health & Resource-Constrained Reasoning | ● | ● | ● | ● | Apply (the entire module is about applying evidence/management under a ceiling the evidence base usually doesn’t account for) | |
| 15 — Illness Script Acquisition | ● | ● | Apply (script retrieval and discrimination is applying compiled prior knowledge to a new presentation) | |||
| 16 — Bidirectional Basic Science ↔ Clinical Integration | ● | ● | Apply (deploying mechanism knowledge onto diagnosis and vice versa) | |||
| 17 — Semantic Qualifiers & Problem Representation | ● | ● | Ask (this is the question-framing step, occurring even earlier than Module 1’s) | |||
| 18 — Causal vs. Probabilistic Network Reasoning | ● | ● | ● | Appraise → Apply (reweighting evidence within a case is a live, miniature appraisal cycle) | ||
| 19 — Community & Social Medicine Insights | ● | ● | ● | ● | Apply (integrating social context is the textbook definition of applying evidence to “this patient’s circumstances”) | |
| 20 — Recognition-Primed Decision Model | ● | ● | Apply (action selection under time pressure — the most compressed, real-time version of Apply in the stack) | |||
| 21 — Evidence Frontier Search | ● | ● | ● | ● | ● | Runs the full cycle explicitly and in order (Steps 21.1 / 21.2–21.6 / 21.7 / 21.8 / 21.9 map almost one-to-one onto Ask / Acquire / Appraise / Apply / Assess) |
Legend: ● = the module meaningfully trains this step. Blank = the module does not meaningfully touch this step, even if a stray prompt brushes against it.
Module 17 (Semantic Qualifiers) is the earliest possible Ask — compressing raw findings into an answerable question before any diagnosis is attempted. Module 1 is the next-stage Ask, forcing commitment to an answer. Modules 15, 16, 18, 19, 20 are all heavily Apply — different flavours of deploying knowledge against a specific case (script retrieval, mechanism linking, network reasoning, social context, time-pressured action). This wasn’t designed as an EBM progression, but it functions as one: the early-numbered cognitive modules (17 → 1 → 15 → 16 → 18 → 20) trace a path from raw case to committed action that tracks Ask through Apply reasonably well.
Until Module 21, almost nothing in the stack trains active search for new external evidence. Modules 2, 3, 6, 7, 9 all involve acquiring data, but it’s data about this patient or this registry — not searching the wider evidence base for what’s currently known or newly discovered. Module 9’s Stage 2 (comparator identification) is the closest thing to genuine literature acquisition before Module 21, and even that is scoped to one case’s published comparators rather than open-ended frontier search. Module 21 is, structurally, the module that completes the cycle — it’s the only one whose entire premise is Acquire, with the EBM steps run in explicit, named order.
There are two different things being appraised across this stack, and they shouldn’t be confused:
Nearly every module’s Phase 3 (Closure/Review) does some form of Assess — a debrief, a checkpoint, a critical-awareness audit. This is structurally guaranteed by the repo’s own three-phase lifecycle (Initiation → Execution → Closure), which independently produces an Assess step almost every time, whether or not the module was designed with EBM in mind. The one place Assess is doing genuinely EBM-specific work — assessing whether applied evidence actually changed an outcome, not just assessing the learner’s reasoning quality — is Module 3’s Step 3.6–3.7 (longitudinal trajectory vs. baseline) and Module 21’s Step 21.7 (appraising whether a frontier finding holds up). Most other “Assess” steps in the stack are closer to metacognitive assessment (how did I reason) than outcome assessment (did the applied evidence work) — a distinction worth being deliberate about if the repo is extended further.
Module 13 (Medication Reconciliation) runs Ask (Step 13.0 framing) → Acquire (Step 13.1 building the medication table) → Appraise (Steps 13.2–13.6 interaction and cascade hunting) → Apply (Step 13.9 the advocate brief) → Assess (Step 13.11 self-assessment checklist) — entirely at the single-patient, single-problem level, without external literature search. Module 21 runs the same five steps but at the evidence-base level, with external literature search as the Acquire mechanism. Between them, they’re the two modules someone could point to as “this is what the full EBM cycle looks like, applied at two different scales” — one scoped to a patient’s drug list, one scoped to the frontier of what’s known.
Not everything in the stack is, or should be, mapped onto EBM — some components serve a different function entirely (persona design, learning-theory scaffolding, longitudinal infrastructure) and forcing them into the table above would be a worse account of what they actually do.
| Item | Why it sits outside Sackett’s cycle |
|---|---|
| Framework A — Humanistic Persona | Governs how the AI communicates throughout every step of every module — affect and confidence-building, not evidence handling. Orthogonal to EBM by design. |
| Framework B — Fink’s FLINK Taxonomy | A reflection-and-meaning framework (human dimension, caring, learning-how-to-learn) layered onto sessions after evidence has already been applied — closer to professional identity formation than evidence methodology. |
| Framework C — Bloom’s Taxonomy | A cognitive-complexity ladder (Remember → Create) that can sit inside any EBM step (you can Appraise at a Remember level or a Create level) — it’s a depth dial across the cycle, not a stage within it. |
| Lifecycle Coverage Summary | A meta-document about the repository’s own structure, not a clinical-reasoning tool. |
Worth naming explicitly: Modules 8 and 12 are structurally the same move applied to two different objects — Module 8 adversarially appraises a teaching prompt, Module 12 adversarially appraises a diagnosis. Both are doing something Sackett’s cycle never anticipated: turning the appraisal lens back onto the reasoning process itself (whether AI-authored or learner-authored) rather than onto an external publication. If this stack is formalised further, that shared pattern — call it reflexive appraisal — is arguably a sixth, unnamed step sitting alongside Sackett’s five, and a stronger conceptual organising principle than trying to force Modules 8 and 12 into the “Appraise” row alongside Module 10’s very different, literature-facing appraisal work.