Skip to content

2. Ask a question

Asking the question a review answers is one of four ways to find its papers, beside searching the databases, starting from one paper and importing a file. Ask it as you would ask a colleague: Does … improve …?, What are the effects of … on …?, or simply the topic. The question becomes a search you see and edit before anything is sent.

  1. In the review's Searches step, open Ask a question, beside Search the databases, Start from a paper and Import a file; or use its Question step. Tick the databases under the box, type the question and press Enter.
  2. The planned search appears:
    • the title of the review, which you can change;
    • each concept of the question, named as the PICOS element it fills: Population or setting, Intervention, Comparator, Outcome, or Topic where the question asks no relation. Change it with the menu beside the concept;
    • its words: the words as you wrote them, and the synonyms found for them in MeSH (the U.S. National Library of Medicine's vocabulary), the STW Thesaurus for Economics and a curated list in English and Portuguese. Some are ticked and searched; the others are offered for you to tick. Click a word to tick or untick it, and point at it to see where it came from. Add a word adds one of your own to the concept;
    • the search, made from the ticked words: a concept's words joined by OR, the concepts by AND, phrases in quotes. Edit it as you like; once you do, ticking words no longer changes it, and Make it from the chips again makes it from the ticked words once more. A search is at most 1,000 characters.
  3. The planner reads which concept is the intervention and which the outcome from the question's grammar, and under the search it says what is asked:
    • an effect: whether the intervention improves or increases, reduces or prevents, worsens or harms, or affects, either way, the outcome, read from the question's verb. The direction is yours to set;
    • a comparison (Is metformin better than insulin for gestational diabetes?): whether the intervention is better than or worse than the comparator;
    • a property (Is coeliac disease hereditary?): whether the papers say it of the intervention.

Where the question you asked is not a yes/no question (What are the effects of exercise on depression?, or a topic such as Hypophysitis and longevity), a note says so and shows the yes/no question its verdict answers instead, such as Does hypophysitis improve longevity?: the first concept as the intervention and the next as the outcome, asked whether it improves or increases it, or Is … effective? where the question names one concept. Your question and its search stay exactly as you asked them; only the verdict uses the paraphrase. Change a concept to Intervention or Outcome, or change the direction, and the verdict answers that instead. A question that asks what something is (What factors predict burnout in nurses?) is searched with no verdict, and a note says so. 4. Enter asks: the question becomes the review's, and its search is run at once in each database ticked, as Search the databases runs one, which Searches then shows. It waits for you instead where a word may be misspelt one of several ways (tick the one you meant), and where the question was read further than it says outright (a paraphrase, a comparison or a property): check how it was read, then press Ask and search. The planned search stays open: edit it and press Ask and search to run it again; if you have changed the question in the box, it is planned again first. With no database ticked, the question is kept without searching. 5. A question's search brings in each database's 100 most relevant papers, then 100 more at a time until 100 of them name both the intervention and the outcome, or the database has no more (1,000 papers at most). A paper the review already holds, or one the database sends twice, is not added or counted again.

How the question is read. The question's grammar is read by spaCy's English and Portuguese pipelines, which read the same question the same way every time, and its words are matched to the vocabularies ignoring case, accents, hyphens and plural endings. No large language model is used, nothing is sent anywhere to plan the search, and the planned search is shown to you before any database is asked.

The Question step

In a review, the Question step, first under Setup & Discovery, shows the question it asks and its planned search. Correct what a concept is, or which way the question asks, and Save: that makes a new version of the question. Where the question was paraphrased to a yes/no question, the step shows the paraphrase too. Change the question, beside the question here and in the bar above Papers, opens it in the question box to reword: it is planned the same way, becomes the review's question and runs its search, which Searches then shows. Earlier versions are listed below it; a verdict a person gave stays with the version it was given under.

Next step, at the top right of every step of a review, takes you to the step after it, in the order a systematic review takes them: from the question and its searches, through screening and extraction, to the report.

The question and the protocol

The question's concepts fill the protocol's PICOS elements: a concept marked Intervention offers its ticked words to Intervention, Outcome to Outcomes, Comparator to Comparator and Population or setting to Population. In Protocol (PICOS), they are offered under each element as From the question: click one, or Add all, then save the protocol. Nothing reaches the protocol, or screening, until you add it.

What the papers say

A review that asks a question has a What the papers say step, first under Reporting: what its papers' conclusions say of it, or of the yes/no question it was paraphrased as, which the step then shows, Yes, Possibly, Mixed, Possibly not or No, as a bar with the number of papers (N) and a legend with each share. Papers shows the same bar above the list. The verdicts are worked out each time the step opens, paper by paper, and a bar shows how far the reading is; a paper whose conclusion cannot be read is left out and counted, so it never holds up the others.

Each verdict is worked out from the paper's own conclusion, the same way every time:

  • only papers with an abstract that screening has not excluded are read, and only those whose title or abstract names both the intervention and the outcome (the intervention alone where the question names one concept): a paper on the intervention alone says nothing of what it does to the outcome;
  • the conclusion is the text after the abstract's conclusion heading, or else its last two sentences;
  • phrases that report a finding, a finding of no effect, a hedge or qualifier, or mixed results are found in it, in English and Portuguese. A negation just before a finding turns it against the question, and so does a finding the other way from the direction you set. Only sentences that name the intervention or the outcome are read: a hedge or a finding in any other sentence is about something else. Within a sentence, only findings in the part that names them count, and a word that says how something else went (a lower mortality rate, where the question asks of longevity) is not a finding on the question. Leads to and results in read as an increase, unless the words after them name the change (led to a reduction in pain). A paper whose conclusion has no such sentence is left out;
  • Mixed: findings both for and against, or results said to be mixed. Possibly: a hedge or qualifier in those sentences, with any finding for the question, since authors who hedge a finding do not stand behind it. Possibly not: a hedged finding against the question. No: findings against only, unhedged. Yes: findings for only, unhedged.

Explain verdict, in this step and under the bar above Papers, lists each paper with its verdict and the exact sentences of its abstract that decided it, the phrases read marked. How was this decided? opens the steps for that paper: why it counts, which sentences were read as its conclusion, each phrase found and what was made of it, and the rule that gave the verdict. How the verdicts are worked out explains the method for every paper.

Every verdict the rule gives is a suggestion. Change sets your own verdict, which always stands over the rule's and is recorded in the paper's History; Back to the suggestion gives the paper back to the rule.