Article
Crafting effective research questions for AI tools

Editorial illustration.

How to Write a Clinical Research Question with PICO

INRA Logo

INRA.AI Team

AI Research Platform

A useful research question tells the team what it needs to learn. PICO helps specify an intervention question by naming the population, intervention, comparison and outcome. The team still needs to decide whether the question matters and how to answer it.

Start with the reader's problem. Do you need to assess an intervention, explain a topic or map what has been studied? That choice affects the question and the work required to answer it. Use the review-approach guide if the intended output is still unclear.

Use PICO for the decision it helps express

Cochrane chapter 2 describes population, intervention, comparison and outcome as components of intervention-review questions. It also emphasizes that developing a useful question requires time, expertise and engagement with the people who will use the review.

A precise sentence can still ask the wrong question. Before refining the wording, explain who needs the answer, what decision it will inform and what existing reviews already cover.

Work through one clinical planning example

Fictional clinical planning example: a research team wants to examine follow-up after discharge for adults with heart failure. The question below is a planning illustration, not evidence that telephone follow-up works.

In adults with heart failure discharged home, how does nurse-led telephone follow-up compared with usual follow-up affect unplanned hospital readmissions within 30 days?
  • Population: adults with heart failure discharged home. The team still needs to define eligible settings and relevant exclusions.
  • Intervention: nurse-led telephone follow-up. Specify which forms, timing and intensity the question intends to cover.
  • Comparison: usual follow-up. Record how usual care is defined in each included study rather than assuming it is identical.
  • Outcome and time: unplanned readmission within 30 days. Agree on the event definition and handling of different reporting periods.

The remaining decisions matter more than making the sentence sound polished. Treat them as planning questions for the team; do not invent eligibility rules or study results to fill the gaps.

Keep the question separate from the search string

The question explains what you want to learn. A database query translates selected concepts into the terms and syntax that database supports. Copying every PICO component into every query is not automatically the right search strategy. Work with appropriate search expertise and inspect the search-planning example.

When PICO is not the useful starting point

A narrative review might ask how explanations of post-discharge follow-up differ across settings. A scoping review might map which forms of support have been studied. Neither question needs an invented intervention-versus-comparator structure merely to satisfy a template.

Use a framework that serves the review purpose. For intervention reviews, also distinguish the review's overall question from the particular comparisons in each synthesis and from what the included studies actually investigated; Cochrane chapter 2 makes these distinctions.

Use the builder as a draft to check

The free PICO question builder can propose a structure for an intervention question. Check every added detail, revise the input when it changes your intent and keep unresolved decisions visible. The builder's result is not a validated protocol or proof that enough evidence exists.

You can use the question without generating an INRA review. If you later consider NLR drafting, inspect its scope, source checks and current paid-plan requirement. Check the resulting draft against its sources before relying on it.

Updated October 19, 2026 to clarify product limits and source checks. Qualified methods review of this update is pending.