How to Build a Search Strategy for Finding Scientific Literature

Step 1: Define Your Research Question Precisely

A vague question produces a vague search. Before you type anything into a database, you need to know exactly what you are looking for. The most effective way to sharpen your question is to use a structured framework that forces you to specify each component.

The PICO framework is the most widely used structure for clinical and health-related questions. It stands for Population (who are you interested in?), Intervention (what treatment, exposure, or factor are you investigating?), Comparison (what is the alternative?), and Outcome (what result are you measuring?). A well-formed PICO question might be: "In adults with type 2 diabetes (P), does metformin (I) compared to lifestyle modification alone (C) reduce the incidence of cardiovascular events (O)?"

For non-clinical research, variations of PICO work equally well. PEO (Population, Exposure, Outcome) is useful for observational studies. SPIDER (Sample, Phenomenon of Interest, Design, Evaluation, Research type) works well for qualitative research. The specific framework matters less than the discipline of breaking your question into distinct, searchable components.

Define the boundaries of your question before searching. Are you interested in a specific age group, geographic region, or time period? Do you want only randomized controlled trials, or will you include observational studies? Are you looking at human studies only, or do animal studies matter? These boundaries become your inclusion and exclusion criteria, and defining them in advance prevents scope creep during searching.

Write your question down in a single sentence. If you cannot express it clearly in one sentence, it is probably too broad or has multiple components that should be searched separately. A focused question leads to a focused search, which leads to manageable, relevant results.

Step 2: Identify Search Terms and Synonyms

Researchers use different words to describe the same concepts. If you search only for "heart attack," you will miss papers that use "myocardial infarction," "acute coronary syndrome," "cardiac event," or "ST-elevation MI." Comprehensive searching requires identifying all the terms that authors might use for each component of your question.

For each element of your PICO question, create a list of synonyms, abbreviations, and related terms. Start with the obvious terms, then expand by checking how different papers describe the same concept. If you find a relevant paper early in your search, scan its abstract and keywords for additional terminology you had not considered.

Add controlled vocabulary to your keyword lists. PubMed uses Medical Subject Headings (MeSH), which are standardized terms assigned to every paper by trained indexers. When you search a MeSH term, PubMed automatically includes all the specific subtypes within that category. Searching the MeSH term "Neoplasms" captures papers about every type of cancer without listing each one individually. Other databases have their own controlled vocabularies: Emtree in Embase, Subject Headings in CINAHL, and Thesaurus terms in PsycINFO.

Consider spelling variations. British and American English differ (randomised vs randomized, oestrogen vs estrogen, behaviour vs behavior), and missing one spelling means missing papers from an entire geographic region. Truncation (using an asterisk to capture all word endings) helps: "randomis*" captures both randomised and randomized, and also randomisation, randomising, and other variants.

Organize your terms in a concept table with one column for each PICO element. List all synonyms, MeSH terms, and variants in each column. This table becomes the blueprint for building your search query in the next step.

Step 3: Construct Your Search Query

With your concept table complete, you can now build the actual search query using Boolean logic. The fundamental principle is: combine synonyms within each concept using OR (to capture any of the terms), then combine the different concepts using AND (to require all concepts to be present).

For example, if your question involves diabetes, exercise, and blood sugar control, your query structure would be: (diabetes OR "type 2 diabetes" OR T2DM OR "diabetes mellitus") AND (exercise OR "physical activity" OR walking OR running OR "aerobic exercise") AND ("blood glucose" OR "glycemic control" OR HbA1c OR "glycated hemoglobin"). Each parenthetical group captures one concept with all its synonyms, and the AND operators ensure that only papers addressing all three concepts are returned.

Use phrase searching (quotation marks) for multi-word terms. Searching "physical activity" as a phrase finds papers about that specific concept, while searching physical AND activity separately returns papers that contain both words anywhere, including irrelevant contexts. Phrase searching is essential for multi-word concepts, method names, specific outcomes, and proper nouns.

Apply truncation strategically. "Exercis*" captures exercise, exercises, exercising, and exercised in a single term, which is more efficient than listing each variant. However, be careful with short roots: "car*" would match car, cars, cardiac, cardiology, carpet, and career, producing many irrelevant results. The root needs to be specific enough to capture only relevant variations.

Use field restrictions when appropriate. Searching for a term in the title field ([ti] in PubMed) is more precise than searching all fields because a paper with your search term in its title is almost certainly about that topic. Author field searching ([au]) prevents matching author names in the body text. Most databases support field-specific searching through syntax codes or dropdown menus.

Build your query iteratively. Start with a simple version and test it. If you get too many results, add more AND restrictions or use field-specific searching. If you get too few, add more OR synonyms or remove restrictive terms. Most databases let you view the number of results for each line of your search, helping you identify which terms are contributing or restricting your results.

Step 4: Select Databases and Run Your Search

No single database covers all of the scientific literature. The databases you choose should match your topic and the level of comprehensiveness you need.

For a quick overview of a topic, Google Scholar alone may suffice. It has the broadest coverage and the simplest interface, and sorting by relevance or citation count quickly surfaces the most important papers. However, Google Scholar's search syntax is limited, its results are not reproducible in a strict sense (because the algorithm changes), and it does not support the advanced query building that systematic searching requires.

For a comprehensive search, such as a systematic review or formal literature review, you should search at least two or three databases and document your strategy for each. PubMed is essential for biomedical topics. Web of Science or Scopus provides broader multidisciplinary coverage and citation analysis tools. Field-specific databases (Embase for pharmacology, PsycINFO for psychology, CINAHL for nursing) add depth in their respective areas.

Adapt your query syntax for each database. The same search concept requires different syntax in different databases. PubMed uses [ti] for title searching and [MeSH] for controlled vocabulary, while Web of Science uses TI= for titles and has no equivalent of MeSH terms. Copy-pasting a PubMed query into Web of Science will not work correctly. Take the time to translate your query into each database's native syntax.

Run each search on the same date or within a narrow window and record the date. Databases are updated continuously, and running the same search a week apart may produce different results. Recording the search date makes your results reproducible and allows others to evaluate your strategy.

Step 5: Screen Results and Refine Your Strategy

Your initial search results need to be reviewed and filtered. Even a well-designed search returns some irrelevant papers, and this is expected. The screening process identifies which papers are actually relevant to your question.

Start with title and abstract screening. Read the title and abstract of each result and decide whether the paper potentially addresses your question. At this stage, be inclusive rather than exclusive because you can always discard a paper after reading the full text, but you cannot recover a paper you dismissed too hastily. For systematic reviews, two independent screeners should review each paper to minimize the risk of missing relevant studies.

Refine your search based on what you learn during screening. If you notice relevant papers using terminology you did not include, add those terms and re-run your search. If you find that most results from a particular synonym are irrelevant, consider removing that term. The search strategy is not fixed after the first run; refinement is a normal and expected part of the process.

Apply database filters to narrow results if needed. Filter by publication date, study type, language, age group, or species depending on your inclusion criteria. Apply filters after your initial search rather than building them into the query itself, so you can see how each filter affects your result count.

Supplement keyword results with citation tracking. Forward citation tracking (finding papers that cite your key papers) and backward tracking (examining the references of relevant papers) identify studies that keyword searches miss. These methods are particularly valuable for finding papers that approach your question from unexpected angles or use non-standard terminology.

Consider grey literature if your question warrants it. Dissertations, conference proceedings, government reports, and working papers may contain relevant data that never appeared in journal publications. Sources like ProQuest Dissertations, conference websites, and institutional repositories can fill gaps in the journal literature, though the quality of grey literature varies more widely than peer-reviewed publications.

Step 6: Document Your Search Process

Documentation transforms a personal search into a transparent, reproducible process. If you cannot tell someone else exactly what you searched, where, when, and how, your search is not reproducible, and any conclusions drawn from it are harder to evaluate.

Record the following for each database searched: the database name and interface (e.g., PubMed via pubmed.ncbi.nlm.nih.gov), the date the search was run, the complete search query with all Boolean operators and field codes, any filters applied (date range, language, study type), and the number of results returned. This documentation should be detailed enough that another researcher could log into the same database and reproduce your search.

For formal literature reviews and systematic reviews, PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) provides a standardized framework for reporting your search process. The PRISMA flow diagram shows how many records were identified, how many duplicates were removed, how many were screened, how many were excluded and why, and how many were included in the final review. This transparency allows readers to assess whether your search was thorough and your inclusion decisions were reasonable.

Save your search queries in the database itself when possible. PubMed, Web of Science, and Scopus all allow you to save searches to your account and set up alerts for new papers matching your criteria. This makes it easy to update your search periodically without rebuilding it from scratch.

Keep a search log that records not just the final successful queries but also the queries you tried and abandoned, including why they did not work. This log helps you avoid repeating unsuccessful approaches and provides a record of your iterative refinement process. If you are collaborating with others, a shared search log keeps everyone on the same page about what has been searched and what remains to be done.