Evidence-Based Medicine · Course for Techies

Lesson 2: Asking Answerable Clinical Questions (PICO) and Finding the Evidence

Part 2 — turning a vague worry into a searchable question, then finding the paper that answers it.

In Lesson 1 we put the patient at the centre: history, examination, and a treatment plan built with SOAP notes and PaJR. We ended with homework — bring a patient's story, and bring a clinical question. This lesson is about that second piece: how do you turn a real, messy clinical worry into a question that a database can actually answer?

Why this matters for techies: this is essentially a search-query optimization problem. "Is this drug good?" returns garbage. A well-structured question returns exactly the paper you need. If you've ever rewritten a bad SQL query or a vague Google search into something precise, you already have the instinct for this.

From Vague Worry to Structured Question

Real clinical questions start messy: "Should this patient be on aspirin?" "Is this new inhaler better than the old one?" "Does this test actually help?" None of these are searchable as written — they're missing the pieces that let you filter a database of millions of papers down to a handful of relevant ones.

The PICO Framework

PICO breaks any clinical question into four parts. Think of it as the schema for a well-formed query:

Letter	Stands for	What you fill in
P	Population / Patient / Problem	Who is this about? Age, condition, setting — e.g. "adults with type 2 diabetes"
I	Intervention	What are you considering doing? A drug, test, procedure, lifestyle change
C	Comparison	Compared to what? Placebo, standard care, an alternative drug, "no treatment"
O	Outcome	What are you actually trying to change or measure? Mortality, symptom relief, side effects, cost

Some versions add a fifth letter — T for Time frame (PICOT) — when duration matters, e.g. "risk of stroke over 5 years."

Worked Example

Vague version: "Does metformin help with weight loss?"

PICO version:

P — Adults with type 2 diabetes and obesity
I — Metformin
C — Placebo or lifestyle intervention alone
O — Change in body weight at 6 and 12 months

Answerable question: "In adults with type 2 diabetes and obesity, does metformin, compared to lifestyle intervention alone, lead to greater weight loss at 6–12 months?"

Two Kinds of Questions

Not every clinical question is about treatment. It helps to know which bucket you're in before you search, because it changes what kind of study you're looking for:

Background questions — general knowledge. "What is atrial fibrillation?" Answered by textbooks, review articles, UpToDate-style resources.
Foreground questions — specific, patient-facing, PICO-structured. "In a 68-year-old with AF, does anticoagulation reduce stroke risk more than antiplatelet therapy?" Answered by original research, systematic reviews, guidelines.

As a rule of thumb: the more experienced you get, the more your questions shift from background to foreground. Beginners often ask background questions dressed up as foreground ones — part of this course is learning to tell the difference.

Matching the Question Type to the Study Type

Once your question is PICO-structured, its type tells you what kind of study design to look for. This is the fastest way to filter your search:

Question type	Best study design to search for
Therapy / treatment effect	Randomized controlled trial (RCT), systematic review of RCTs
Diagnosis / test accuracy	Cross-sectional study comparing test vs. gold standard
Prognosis	Cohort study (following patients over time)
Harm / side effects	Cohort or case-control study (RCTs are often too short/small to catch rare harms)
Cost-effectiveness	Economic evaluation / cost-effectiveness analysis

This is why Lesson 1's evidence pyramid matters here: knowing the shape of your question tells you where on the pyramid to start looking.

Turning PICO into a Search Query

PubMed and similar databases reward structured searching. A rough workflow:

Pull out the key concepts from P, I, C, O.
Search each concept separately first, then combine with AND / OR.
Use quotation marks for exact phrases, and consider MeSH terms (PubMed's controlled vocabulary) once you're comfortable.
Apply filters: publication type (e.g. "Randomized Controlled Trial", "Systematic Review"), date range, human studies.

Example search string (from the worked PICO above):

("metformin"[Title/Abstract]) AND ("weight loss"[Title/Abstract] OR "body weight"[MeSH]) AND ("type 2 diabetes"[MeSH]) AND (randomized controlled trial[Publication Type])

Pro tip: if a plain search returns thousands of results, your question probably isn't specific enough yet — go back and tighten the P, I, C, or O. If it returns zero, you've probably over-specified — loosen one term at a time.

Where to Search

PubMed — the largest free index of biomedical literature. Best general-purpose starting point.
Cochrane Library — systematic reviews specifically; good when you want the "already appraised for you" answer.
Google Scholar — broader net, less structured filtering, useful as a supplement.
CDSS tools (mentioned in Lesson 1) — UpToDate, DynaMed, BMJ Best Practice — good for background questions and quick pre-appraised summaries.

Homework for Lesson 2

Take the clinical question you brought from Lesson 1 (or your patient's question) and rewrite it in full PICO form.
Identify which question type it is (therapy, diagnosis, prognosis, harm) and which study design you should be hunting for.
Run your PICO question through PubMed. Bring back the single best-matching paper's title, and one sentence on why it fits your P, I, C, and O.

Resources for This Lesson

PubMed — search medical research papers Cochrane Library — systematic reviews CEBM Oxford — EBM tools, including PICO worksheets Google Scholar