Write an abstract that stands alone
An abstract is read far more often than the paper it summarises, and it has to make sense without the paper: it carries the paper's key components in the paper's own order and says what the paper says. Search engines also read it to find and rank the article. It is drafted from the paper, yet many published abstracts do not match their papers, and the errors are caught by checking every detail of the abstract against the full text after each revision.
How widely the abstract is read
The title and the abstract are the two most frequently read parts of a manuscript. Of the two, the title is the more widely read. Most readers never read more of an article than its abstract. Abstracts are widely reproduced in electronic databases and form the basis of most electronic searches, so they are how articles are discovered, interpreted and judged.
Abstracts are perhaps the most important part of a manuscript, because they are often the only part that is read and used as an information source. In most cases an editor commits to reading a whole manuscript only after reading its title and abstract. Readers use the abstract, consciously or not, to decide whether to read the full article, and editors often use it to decide whether to send the manuscript out for peer review. So the abstract is an advertisement for the manuscript and its authors as well as a source of information.
It is worth the effort to make the abstract catch the audience's attention with the paper's key components. The time and effort writers and editors put into abstracts are well spent; someone's life may depend on them.
What the abstract contains
The abstract is a synopsis of the paper written by its authors. The abstract should stand alone: a summary from which a reader can understand the key points of the study without checking the full paper. It should clearly describe the nature of the study and any key findings and their implications or applications, and it should entice the reader to read the complete paper.
An abstract should be concise and easy to read, summarising the essential information without repetition and with the reader in mind. The abstract should present the paper's sections in the same order: the objective or scientific question, how the study was done, what was found, and the answer to the question. It should focus on the question, the main methods, the results and the conclusion, and being specific and selective can achieve that focus. For example, it may give the primary objective, its result and the conclusion drawn from it, and add secondary objectives and their results if space allows. The International Committee of Medical Journal Editors (ICMJE) recommends that the abstract give the context or background for the study and state its purpose, its basic procedures (the selection of participants, settings, measurements and analytical methods), its main findings, with specific effect sizes and their statistical and clinical significance where possible, and its principal conclusions.
The methods and the results are probably the most important parts of an abstract, because a reader decides whether a study is worth reading from its design and methods and from whether it found anything.
One author who teaches medical writing has always taught that, in general, the abstract should mirror the paper's IMRAD structure, presenting its introduction, methods, results and discussion or conclusions each in a sentence or two, and many good abstracts have exactly four short paragraphs. Many abstracts set the stage by starting with an unassailable general statement, such as Sudden cardiac death among children and young adults is a devastating event. The methods and results describe work that has been done and should be written in the past tense; the discussion explains what the authors think and is written in the present tense, often with a phrase such as We conclude.
The abstract must never contain a conclusion that the body of the paper does not support. The purpose statements and the conclusions in the abstract must reflect those written in the main text. The abstract should end its conclusions by stating the results of the research; a conclusion that says only that a study provides data on adverse events and rates of resolution, for example, leaves the reader asking what the authors concluded.
Structured abstracts and the journal's format
An abstract should follow the length, headings and format that the journal's instructions for authors set. The trend is for journals to require structured abstracts, which present their information under specific headings that differ a little with each journal. All structured abstracts include the four key components of the IMRAD model, though synonyms for these headings may be used, with variations such as Context and Main Outcome Measures. To tempt readers on to further reading and download, a structured abstract is worth the time it takes to write.
A structured abstract is a microcosm of the whole paper and may include an introduction or background, objectives, methods, results, discussion, study limitations and conclusions. One experienced writer always tries to include at least one comparison, P value or both.
For reports of original data, JAMA asks for an abstract of no more than 350 words under the headings Importance, Objective, Design, Setting, Participants, Intervention(s) or Exposure(s), Main Outcome(s) and Measure(s), Results, Conclusions and Relevance, and Trial Registration, and its instructions for authors go on to say what each section should present. The tight word limit and the many items to cover get the important data into the abstract, but at the expense of some very complicated, number-laden and almost incomprehensible sentences in the results.
Some journals prefer abstracts written in full sentences; others encourage phrases. For brevity, most journals allow incomplete sentences in the abstract. In one well-written structured abstract, the Participants section is the only one of the first five (Objective, Design, Setting, Participants and Main Outcome Measure) that contains a complete sentence; later, under Results and Conclusions, the paragraphs become longer and the style changes from phrases to complete sentences.
An abstract that search engines find
Search engines read the abstract page, which anyone can view free online. The abstract both tempts researchers to read the article and gives a search engine the data it needs to find and rank it. Optimising the abstract for search engines can help the article be discovered, read, used and cited, which in turn helps the impact factor of the journal it appears in and further raises the article's visibility. The better the abstract is written, the better the article's chance of appearing high in the search rankings.
The chief requirement of a structured abstract is to include the key words and other terms that search engines will locate. The abstract should focus on three or four phrases that a researcher interested in the paper might search for, repeated in a natural way. Repeating keywords and phrases used in the title boosts search-engine ranking. Search engines show the first two sentences of the abstract in their results, so those sentences should count. Ideally the keywords appear in those sentences and are then reused three to six times through the abstract. Synonyms that the audience might search for can raise the chance of the article being found.
Drafting the abstract
The methods used in journalism can be adapted to drafting an abstract. The model is the lead, the first sentence of a news story, which in its standard form carries all the essential information: who, what, where, when and why. The lead is the news story's equivalent of the abstract. After the title, the lead is the first part of a news story the reader sees, and the abstract is the first part of a paper that gives context about the study. Each is the most-read part of its piece, and each should make sense on its own. The abstract's parts correspond to the questions: the objective to why, the methods to how, the results to what, and the discussion to so what. A key difference between the two is why the reader should take an interest in the topic: an abstract's discussion or conclusion explains why the results should be of interest and how they contribute to the scientific literature, which a news story does not commonly do.
The method is to read the paper carefully, marking key words and sentences to bring out its key components; to think in terms of the lead, drafting the objective, methods, results and discussion as the why, how, what and so what; to list all the marked material; and then to edit the list to condense and refine the content.
Most of an abstract's information is in its results, and part is probably in its background; the abstract is almost a very small outline of the paper's story. Writing the abstract first can be a way to start building that story; some people prefer to write it at the end, and the order is a personal choice. Many advise making the final version of the abstract the last item written, since only then does the writer know exactly what is in the paper being summarised.
Gaps and errors in published abstracts
Abstracts are quick to read and give instant answers, but they are often incomplete or wrong, and readers have been warned many times to be sceptical of them. Research has shown that many published abstracts have serious gaps or inaccuracies. The issue was first highlighted by a 1999 study that found discrepancies between full text and abstract in 18% to 68% of articles in six major medical journals, which led the publishing community to renew its efforts to improve abstracts. Three studies published in the eight years up to 2012 found that published abstracts in various medical disciplines still lacked key data or carried inaccuracies.
Cases have been documented in which readers made clinical decisions from abstracts because the full text was unavailable, although they should not have. Lack of time, lack of access to the full text, especially in the developing world, and unwillingness to pay for it have all led clinicians to reach conclusions solely from abstracts. Concern has been expressed that incomplete and inaccurate abstracts are not only misleading but also harmful to patients' health.
One experienced editor holds that investigators do not set out to write poor abstracts: often they do not understand the importance of the abstract and treat it as an afterthought. They may cut and paste it from an earlier conference presentation without considering the changes made to their data and conclusions when they prepared the paper for submission, and they may forget to revise it after revising the manuscript. All of these can lead to discrepancies or missing data.
Checking the abstract against the paper
Peer reviewers, in one editor's experience, are inconsistent about commenting on problems in the abstract, so it falls to the manuscript editor to check every detail in the abstract against the full text and point out gaps, discrepancies and missing data. Statements in the abstract should also match the full text in tone and in strength of claim. The CONSORT checklist is a guideline that helps ensure an abstract of a randomised trial has all the data needed. The checklist is limited to randomised trials, but Annals of Internal Medicine takes a broader approach and sets out headings for abstracts of original research, cost-effectiveness studies and systematic reviews and meta-analyses; writers and editors should use these headings as an additional checklist to make sure all important points are covered. Writers and editors should recheck the abstract after any revision to the full text.
Adding a systematic consistency check between the abstract and the rest of the manuscript to one journal's copyediting procedure generated additional author queries. A study of that peer-reviewed medical journal examined how often queries noting inconsistencies in abstracts led to revisions in its articles. It covered three issues, classed each query involving an inconsistency between the abstract and another part of the manuscript as proactive, revising the text and asking whether the revision was acceptable, or passive, adding a query without a revision, and flagged particularly important inconsistencies, such as a statement from which the word not had been omitted.
The three issues held 49 articles with full abstracts, of which 41 (83.7%) carried at least one consistency query involving the abstract. The 129 queries split evenly between information present in the abstract and nowhere else in the manuscript (63, or 48.8%) and information that differed between the abstract and another part (66, or 51.2%). Overall, 92 of the 129 queries (71.3%) led to revisions in the published articles, and all seven particularly important inconsistencies were corrected. Authors accepted all the revisions attached to the 27 proactive queries, while 65 of the 102 passive queries (63.8%) led to revisions.
Copyeditors should perform a systematic consistency check between the abstract and the rest of a manuscript, because the majority of the resulting queries lead to revisions. Where possible, they should proactively revise the text to make it consistent and ask whether the change is acceptable, rather than simply querying the author.
The abstract is a key point where a writer or editor can add value and prevent a dangerous misinterpretation. Readers cannot be prevented from using or misusing abstracts, but the completeness and accuracy of abstracts can be ensured. Abstracts can be given special attention at each step to bring them up to standard.