All papers in EPIDEMIOLOGY—except for letters, research letters, and commentaries—require an abstract. After the title, the abstract is the most important public-facing aspect of a paper; it is always freely available, even when the rest of the paper is not. As such, the abstract should convey key methods and results against the background and importance of the study. This goal can be challenging given the strict limit of 250 words, so we have come up with some guidelines.
The abstract serves several purposes. It primarily summarizes your work, both in the print and electronic versions, as well as in indices such as PubMed. It also entices readers to invest time in reading the whole paper. Literature summaries start with reviews of abstracts to select papers that meet inclusion criteria. For these reasons, the abstract might be the most important writing you will do. Unfortunately, the abstract often receives the least attention. Authors often simply cut and paste key phrases from each section of their main text to stitch together an abstract. To write better abstracts, write them, don't cut and paste them. Write a first draft before you have written anything else. Then write a second independent draft after you have drafted the other sections. Compare these drafts and keep the best from both. Then start rewriting. As with all writing, revisions lead to the best text.
The usual structure for an EPIDEMIOLOGY research paper has four sections: Introduction, Methods, Results, and Conclusions. A structured abstract should follow this outline. In the abstract, the Introduction section use a sentence or two to focus on the rationale for the study: what is still unknown in a given area and what the study aims to accomplish. The Methods should briefly describe the study design and population investigated, and main methods used. The Results section should include a key numerical result or two, as this adds interest; however, beware of the temptation to make sentences unwieldy with a lot of risk estimates and confidence intervals. (Have a look at our strategies for presenting numbers in text here.) Finally, the Discussion should provide the take-home message of the study, avoiding (per journal policy) undue causal language or strictly avoiding public health or policy recommendations.
Not all abstracts need be structured; methodology papers, particularly those with less internal structure than a typical outcomes paper, can have an abstract free of the constraint of the sections above.
Although we do not have a goal for a formal level of readability, the abstract should be as readable as practicable for most readers. Our audience mostly has a background in biomedical sciences, but includes non-epidemiologists. We are also a general interest epidemiology journal, so jargon specific to a topic area should be avoided, especially in the abstract. For these reasons, we are somewhat more strict regarding abbreviations (about which you can read more here—in brief, use only the most familiar ones, e.g. BMI and CI, but they should be defined on first use and, if possible, avoided entirely, particularly when there is potential ambiguity. More than once, for example, a paper using the expression 'men who have sex with men' has appeared in the same issue as one that used 'marginal structural models,' and both were abbreviated MSM (which has various other meanings as well), so we made sure that this abbreviation was not used at all in either abstract.
As mentioned above, our limit of 250 words is strict, primarily to avoid PubMed truncating any excess. If the abstract is too long and/or if your editor spells out an abbreviation, she or he will also adjust elsewhere or ask you to do so. Remember that it is fine to save some details of methods or results for the body of the paper.