A PICO research question sets out four elements in one sentence: Population, Intervention, Comparison and Outcome (PICO). It says who you will study, what is done, what it is compared with and what you will measure. Define each element until two people read it the same way, then check that a feasible design can answer it.
A clinical idea usually arrives as a hunch: "I wonder whether the new discharge leaflet helps." A hunch cannot be tested, searched or powered. A research question can. This guide covers the step between the two, before you write a protocol or an introduction.
What each element of a PICO research question makes you decide
The four-part structure behind PICO is usually credited to Richardson and colleagues, who described the "well-built clinical question" in 1995. Each letter forces a decision that a vague idea leaves open.
| Element | The decision it forces | Vague | Specific |
|---|---|---|---|
| Population | Who is eligible, in which setting, at which stage of care | Surgical patients | Adults having day-case laparoscopic cholecystectomy at one hospital |
| Intervention or exposure | Exactly what is done or experienced, by whom and when | Better follow-up | A nurse telephone call on the second day after discharge |
| Comparison | What the intervention is measured against, described as fully as the intervention | Standard care | Written discharge information alone, as given before the call was introduced |
| Outcome | One primary measurement, its definition and its time point | Fewer problems | Unplanned attendance at the emergency department within 30 days |
The example in the table is invented to illustrate the structure. It describes no real study.
Choose the variant that fits your type of question
PICO was built for questions about treatment. Other question types need different slots. Forcing a diagnostic or qualitative question into PICO produces an empty comparator or a meaningless outcome. Pick the variant first.
| Question type | Framework | Its elements |
|---|---|---|
| Treatment or intervention | PICO | Population, Intervention, Comparison, Outcome |
| Treatment with a fixed follow-up period | PICOT | PICO plus Time: the point at which the outcome is assessed |
| Exposure, risk factor or cause | PECO | Population, Exposure, Comparator (unexposed or less exposed), Outcome |
| Diagnostic accuracy | A diagnostic form of PICO | Population, index test, reference standard, target condition |
| Qualitative experience or views | SPIDER | Sample, Phenomenon of Interest, Design, Evaluation, Research type |
| Mapping a broad field (scoping review) | PCC | Population, Concept, Context |
SPIDER was proposed by Cooke and colleagues in 2012 as a search tool for reviews of qualitative and mixed-methods research. PCC suits scoping reviews, which rarely have an intervention or a comparator.
Match the question type to a design that can answer it
A well-formed question still fails if the design cannot answer it. The question type points to the design, and the design limits the claim. Our guide to choosing between a case report, case series or cohort study explains why a series without a comparison group cannot show that one treatment beats another.
| Question type | Designs that can answer it | What the answer can claim |
|---|---|---|
| Does the intervention change the outcome? | Randomised controlled trial; a cohort study when a trial is not possible | A causal effect from a trial; an adjusted association from a cohort |
| Is the exposure linked to harm or disease? | Cohort study; case-control study for rare outcomes | An association, not proof of cause |
| How accurate is the test? | Cross-sectional diagnostic accuracy study against a reference standard | Sensitivity and specificity in that setting and spectrum of patients |
| What happens to people with this condition? | Cohort study, ideally from a defined starting point | Incidence and prognosis in that group |
| How common is it now? | Cross-sectional study | Prevalence at one point in time |
A question asked of a retrospective dataset you already hold needs one extra check. The records were collected for care, not research. Write the question around what the records actually contain. "Is the outcome recorded, consistently, for every patient?" decides whether the question survives. A retrospective question is usually framed as an association or an estimate, not as an effect.
Worked example: from a hunch to a PICO research question
The study below is invented to illustrate the steps. It contains no clinical advice and no real data.
- The hunch. "Patients seem to come back less since we started phoning them."
- Name the population. Adults having day-case laparoscopic cholecystectomy at [hospital] between [date] and [date].
- Name the intervention. A structured nurse telephone call on the second day after discharge, following a written script.
- Name the comparison. Patients from the year before the calls began, who received written discharge information only.
- Name one outcome and its time point. Unplanned attendance at the emergency department within 30 days of the operation, taken from the hospital record.
- Check the design. The comparison is historical, so the study is a retrospective before-and-after cohort. It can estimate an association, not prove the call caused the difference.
The finished question reads:
In adults having day-case laparoscopic cholecystectomy at [hospital], is a structured nurse telephone call on the second day after discharge, compared with written discharge information alone, associated with fewer unplanned emergency department attendances within 30 days of the operation?
Note the verb. "Associated with" matches a historical comparison. "Does the call reduce" would promise a causal answer the design cannot give.
Test the question with FINER before you commit
Feasible, Interesting, Novel, Ethical and Relevant (FINER) is a checklist from Hulley and colleagues' textbook Designing Clinical Research. Run it before the protocol, while changing the question is still cheap.
- Feasible: how many eligible patients do you see each year? Is the outcome recorded? Do you have the time, skills and funding?
- Interesting: would the answer matter to you and to the clinicians you work with?
- Novel: has a search shown the question is still open for your population or setting?
- Ethical: can the study gain approval, or a documented waiver for a retrospective design?
- Relevant: which decision, service or future study would change once the question is answered?
A failed feasibility check can end a question outright. Count eligible patients from real records, not memory. If the count is small, narrow the outcome or widen the period before you start.
Where each PICO element goes after the question is set
A clear question saves work later, because each element is reused in a fixed place.
- Eligibility criteria: the population becomes your inclusion and exclusion criteria, written operationally in the research protocol.
- Primary outcome: the outcome element is the single measurement that drives the sample size calculation for a small clinical study.
- Search concept blocks: population and intervention, and sometimes the comparison, each become a block of synonyms for your literature search.
- Title: name the design and population, and the intervention if there is one.
- Objective sentence: the last sentence of the introduction of a research paper restates the question as an aim.
For reviews, chapter 3 of the Cochrane Handbook shows how the PICO elements become eligibility criteria. It also notes that outcomes usually do not decide which studies a review includes.
Six faults that make a research question unanswerable
| Fault | Example | Fix |
|---|---|---|
| Two questions in one | "Does the call reduce attendance and improve satisfaction?" | Pick one primary question; list the other as secondary |
| An undefined outcome | "Better recovery" | State what is measured, how and when |
| A composite outcome nobody can unpick | "Any complication or readmission or attendance" | Define each part and report each one separately as well |
| An undescribed comparator | "Compared with standard care" | Describe standard care as fully as the intervention |
| A population too broad | "Surgical patients" | Name the procedure, age group, setting and period |
| An outcome chosen after seeing the data | Switching to whichever measure looked best | Fix the primary outcome before collecting or analysing data |
The last fault is the hardest to repair. A question written down and dated before data collection is your evidence that it was not shaped by the results.
A fill-in template for your own question
In [population, setting and period], does [intervention or exposure, with who delivers it and when], compared with [comparator, described in full], change / is it associated with [one primary outcome, with its definition] at [time point]? Design: [named design]. Eligible patients per year: [number from records]. Outcome recorded in: [data source].
Keep the verb that fits your design, and delete the other. If you find an error in this guide, please tell us how to report a problem so we can correct it.