PICO and related question frameworks

Why a structured question

A vague question gives a vague review. If the question does not say who is being studied, what is being done, what it is compared with and what is being measured, the review cannot decide which studies to include, cannot design a search and cannot say what a result means. Different team members will read the question in different ways and select different studies. A structured question forces precision at the start, when it is cheapest to settle, and gives the later stages a fixed reference. The structure also makes the review easier to read, register and appraise.

The most widely used structure comes from evidence-based medicine, where a well-built clinical question was described as having four parts. It is named PICO after them. Variants have been proposed for other kinds of questions, and none is universally better. The right choice is the one that captures the elements of the question that matter for deciding eligibility. The resulting criteria are then written into the protocol, as described under protocol development and registration.

The elements of PICO

Population

Who is being studied: the condition or problem, and the relevant characteristics such as age group, severity and setting. It should be specific enough to select studies and broad enough to find them.

Intervention

What is being done: a treatment, a test, a program or an exposure of interest, described by type, dose, duration and delivery where these matter.

Comparator

What the intervention is compared with: placebo, no treatment, usual care or an alternative treatment. The choice defines the contrast being estimated.

Outcome

What is measured to judge the effect: the outcome domain, the measure, the time point and, where relevant, the way it is summarized.

Each element becomes part of the eligibility criteria. For a systematic review of the effect of exercise on depressive symptoms in adolescents, for instance, the population is adolescents with depressive symptoms or a diagnosis of depression, the intervention is structured exercise programs, the comparator is no exercise or usual care, and the outcome is depressive symptoms measured by a validated scale. Each of these is a statement that can be applied to a study to decide whether it counts.

Variants and when to use them

The names vary, and there are more of them. Authors sometimes add elements such as context, setting or stakeholder. The aim in each case is the same: to state the components that decide eligibility. See the pages on scoping review, diagnostic accuracy meta-analysis and prognostic meta-analysis.

Defining the outcome properly

Of the four elements, the outcome is the one most often left too loose. An outcome such as quality of life or pain is a domain, and studies measure it with dozens of different instruments at different times. A review needs to say which domains are of interest, which measures are acceptable, at what time points and in what form, for example as a change from baseline or a final value, or as a proportion who improved. Several outcomes are usually of interest, and the question should say which are primary and which secondary, and why. Outcomes that matter to patients are preferred to surrogate markers, and the list should include harms as well as benefits.

Setting the primary outcome before looking at studies is one of the most important protections against selective reporting. If the review reports outcomes after the fact, it can emphasize those that favor a conclusion. A helpful way to specify an outcome is to state its domain, the specific measurement, the metric, the method of aggregation and the time point. The Cochrane Handbook describes the approach. See GRADE for the selection of critical and important outcomes.

PICO and the search

The elements of the question are not all used in the search. The population and the intervention are usually searched, as they are most often named in titles and abstracts and indexed consistently. The comparator and the outcome are generally not searched, because they are reported inconsistently in abstracts, use many different terms, and adding them would drop relevant studies. The study design may be added through a validated filter when the review is limited to a design and a filter exists. The decision about which elements to search is part of building the strategy, described in building a search strategy. Research on the use of PICO as a search tool has found that it helps to structure searches, and that the benefit is greatest when it is combined with an understanding of the databases' indexing, and not used mechanically.

Worked examples

Intervention question (PICO)

Population: adults with type 2 diabetes. Intervention: structured education programs. Comparator: usual care. Outcome: change in glycated hemoglobin at 6 to 12 months. Design limited to randomized trials (PICOS).

Exposure question (PECO)

Population: pregnant women. Exposure: high levels of fine particulate air pollution. Comparator: lower levels. Outcome: low birth weight. Designs: cohort and case-control studies.

Diagnostic question (PIRD)

Population: adults with suspected pulmonary embolism. Index test: a rapid blood test. Reference standard: imaging. Target condition: pulmonary embolism.

Scoping question (PCC)

Population: older adults. Concept: community programs to reduce loneliness. Context: high-income countries.

Qualitative question (SPIDER)

Sample: parents of children with a chronic illness. Phenomenon of interest: experiences of managing care at home. Design: interviews and focus groups. Evaluation: perceived burden and coping. Research type: qualitative.

The examples are illustrative. In each, the question is turned into criteria that can be applied to a study.

Getting the breadth right

The most difficult judgment is how broad to make each element. A broad population, such as all adults with depression, gives more studies and wider applicability but more heterogeneity. A narrow one, such as women over 65 with treatment-resistant depression, gives a coherent set but may find too few studies. The same applies to the intervention: lumping different types of psychotherapy together may answer a question about psychotherapy in general and hide differences between types. A common approach is to define a broad question and plan subgroup analyses by the most important differences, or to define a narrower question and say that a broader one is outside the scope. A scan of the literature, before the protocol is finalized, shows how many studies exist and how much they vary, and helps to set the breadth. If a recent review exists on a very similar question, the new review should say what it adds.

Refining the question before the protocol

The first draft of a question is rarely the final one. A practical sequence helps. Write the question in a sentence, then split it into its elements. Run a quick scoping search to see how many studies and existing reviews there are, and which terms they use. Ask whether the population is homogeneous enough, whether the intervention is one thing or several, whether the comparator is clear, and which outcomes the studies actually report. Talk to the people who will use the review, since their needs define which outcomes matter and which contrasts are useful. Revise, and repeat until the question is both answerable and worth answering. At the end the question should be stated in a form that a stranger could use to decide whether a given study is eligible. That test, applied to a few real studies, usually exposes remaining vagueness. The result then goes into the protocol and, where relevant, the registry record, described in the guide to PROSPERO registration.

Beyond PICO: other things a question may need

Some questions need more than the four elements. The setting, such as hospital or community, may change the effect. The timing of the intervention or of the follow-up may be critical. Subgroups of interest, such as age, sex or severity, may be named as planned analyses, along with the direction in which a difference is expected. For questions about harms, the type of adverse event and how it is ascertained matter. For complex interventions, the components, the context and the mechanism may need to be spelled out, and frameworks from implementation science are sometimes used alongside PICO. For questions about equity, the characteristics along which disadvantage operates are specified. These additions are best treated as part of the eligibility criteria and the analysis plan, written down before the work begins, and not as afterthoughts when the results are in.

Recording the question in the protocol and registry

Once the question has been framed, its elements are written into the protocol in the form of explicit inclusion and exclusion criteria, and into any registry entry, so that the question is fixed in a public record. A good entry states each element specifically enough to be applied, names the primary outcome, and records any planned subgroups. If the question changes later, the change is recorded as an amendment with the reason, so that readers can see whether it could have been influenced by the results. This is the simplest and strongest protection that a structured question provides.

Common mistakes

• Leaving the outcome as a domain with no measures or time points.
• Defining the comparator loosely, so that studies with different comparators are combined.
• Using PICO for a question it does not fit, such as a mapping question or a qualitative one.
• Changing the elements after screening has begun without recording and justifying the change.
• Putting every element into the search and losing relevant studies.
• Making the population so broad that the studies cannot sensibly be compared, or so narrow that none can be found.
• Not stating which outcomes are primary.

Support

Frequently asked questions