Clinical Question Formulation and Literature Search
What this domain tests
This domain tests whether you can convert clinical uncertainty into an answerable question and build a reproducible search. Items often ask you to identify PICO elements, choose the best question type, or spot a search strategy that will miss evidence.
Core rules
- Background questions ask “what is…” and are answered by textbooks; foreground questions compare interventions, exposures, or diagnoses and require primary studies or syntheses.
- PICO: Population, Intervention, Comparison, Outcome. Add T for timing and S for study design when useful (PICOT/PICOTS). Qualitative questions use PICo or SPIDER; diagnostic accuracy questions use PIRD.
- FINER criteria: feasible, interesting, novel, ethical, relevant.
- Search strategy: combine keywords with Boolean operators (AND narrows, OR broadens), use subject headings (MeSH in PubMed, CINAHL Headings), truncation, phrase searching, and filters cautiously. Search at least two relevant databases plus grey literature and reference lists.
- Document the search: databases, dates, full strings, limits, and yield. PRISMA-style flow diagrams make the process auditable.
- Sensitivity versus precision: a sensitive search finds more relevant studies but returns more noise; a precise search returns fewer irrelevant records but may miss studies. Systematic reviews favor sensitivity.
Worked example
Question: “In hospitalized adults aged 65 and older, does a nurse-led multifactorial fall prevention program compared with usual care reduce injurious falls within 12 months?” P: hospitalized adults aged 65 and older; I: nurse-led multifactorial program; C: usual care; O: injurious falls; T: 12 months. A search would use “accidental falls,” “fall prevention,” “hospital,” and “aged” as MeSH terms, with OR synonyms, then combine with AND. Add CINAHL and Cochrane because PubMed alone may miss nursing journals.
Traps that cost marks
- Writing a question with no comparison or an unmeasurable outcome.
- Using only keywords and no controlled vocabulary, or searching only one database.
- Applying filters that exclude the population or study design you need.
- Failing to record the search, making it irreproducible.