Framing literature review questions with PICO, PCC, SPIDER and other frameworks https://yila.ai
|
|
3 週間 前 | |
|---|---|---|
| README.md | 3 週間 前 |
Every literature review rests on its question. A clear question determines the search terms, the eligibility criteria, the data to extract and the shape of the synthesis. A vague question produces a vague review, no matter how much effort goes into searching and screening. Question frameworks are simple tools that help researchers break a question into components and make sure none is missing. This guide explains the most common frameworks, shows how to choose among them, and offers tips for turning a framed question into a workable protocol. It also notes where an AI research assistant like Yila can help with early exploration, while the question itself should be settled by the team and its stakeholders.
A framework does three things. It forces precision, because each component must be stated. It links the question to the search, because each component usually becomes a concept block in the search strategy. And it links the question to eligibility, because each component becomes a criterion for inclusion. When a reviewer later asks why a study was excluded, the answer can usually be traced to one component of the framework.
Frameworks are aids, not rules. The best framework is the one that fits the question, and some questions need a hybrid or adapted version.
PICO stands for Population, Intervention, Comparison and Outcome. It is the standard framework for questions about the effects of interventions, and it is built into much of the guidance for systematic reviews.
Example: In adults with chronic low back pain (P), does a structured exercise programme (I), compared with usual care (C), reduce pain intensity at twelve months (O)?
Common variants add elements. PICOS adds Study design, useful when the review will include only certain designs. PICOT adds Time, the period over which outcomes are measured. PECO replaces Intervention with Exposure for questions about environmental or occupational exposures, where the "intervention" is not chosen by anyone.
PCC stands for Population, Concept and Context. It is recommended by JBI for scoping reviews, which ask broader questions than effectiveness reviews.
Example: What is known about peer support (Concept) for young people with long-term health conditions (Population) in school settings (Context)?
PCC deliberately leaves out comparison and outcome, because scoping reviews typically describe what has been studied rather than estimating effects.
Qualitative evidence syntheses ask about experiences, perceptions and meanings, which PICO does not capture well.
SPIDER stands for Sample, Phenomenon of Interest, Design, Evaluation and Research type. It was designed to improve the specificity of searches for qualitative studies, though evaluations suggest it may miss relevant studies if applied too narrowly, so many searchers use it to structure thinking rather than as a strict search template.
SPICE stands for Setting, Perspective, Intervention or phenomenon of Interest, Comparison and Evaluation. It is often used in social care, library and information science and service evaluation questions.
Example using SPIDER: How do nurses (S) experience the introduction of electronic health records (PI), as explored in interview or focus group studies (D), in terms of workload and professional identity (E), using qualitative research (R)?
Several other frameworks address particular question types. ECLIPSE (Expectation, Client group, Location, Impact, Professionals, Service) suits health policy and management questions. CoCoPop (Condition, Context, Population) is used for prevalence and incidence reviews. PIRD (Population, Index test, Reference standard, Diagnosis of interest) suits diagnostic accuracy reviews. PEO (Population, Exposure, Outcome) is a simpler option for some observational questions.
Ask what kind of answer the review must provide:
If none fits well, adapt one and explain the adaptation in the protocol.
Once the components are stated, each should be translated into three things:
Writing these three translations side by side often exposes ambiguities. If the team cannot agree whether a study of adolescents aged sixteen to nineteen fits the population, the population definition needs work before screening begins.
Before finalizing the question, it helps to explore the literature briefly. How much research exists? What terms do authors use? Are there existing reviews? An AI research assistant such as Yila can help with this kind of quick orientation, surfacing related work and terminology. The findings should then inform a deliberate decision about the question, ideally with input from people who will use the review, and the final question should be recorded in a registered protocol.
Too many components joined with AND. Each additional required component reduces the number of eligible studies and makes the search more fragile.
Outcomes defined too narrowly. Studies often measure the same construct with different instruments. Define the outcome conceptually and list acceptable measures.
Comparison left implicit. If the comparison matters, state it. If any comparison is acceptable, say so.
Changing the question silently. Questions sometimes need refinement after initial searching. That is acceptable if the change is documented and justified.
Question frameworks turn an interest into a precise, answerable question and connect it to every later stage of a review. Choose PICO and its variants for effects, PCC for scoping, SPIDER or SPICE for qualitative questions, and specialized frameworks for tests, prevalence or policy. Translate each component into search concepts, criteria and extraction fields, explore the literature before committing, and document every refinement.