case studies April 23

Journal of Clinical Epidemiology 66 (2013) 945

WRITING TIPS SERIES

Effective writing and publishing scientific papers, part V: results

1. What you should know

The results section of an article presents a clear, concise, and objective description of the findings from a particular study and is mostly written in the past tense. The findings are presented without interpretation, as this should occur in the discussion section only. You may think of the results sec- tion as mirroring the methods section: For every method (what you did), there should be a corresponding result (what you found) and vice versa. A common order of elements is: recruitment/response, characteristics of the sample, findings from the primary analyses, secondary analyses, and any addi- tional (unexpected) findings. Ideally, the results section is a dynamic interplay between text and figures/tables; themost important data will be shown in both. Tables and figures are particularly useful to present larger quantities of data (see part 7 of this series on ‘‘Tables and figures’’).

The word ‘‘significant’’ is often used in everyday language to stress something that is important or substantial, but in a scholarly article, it is probably better to use thewords ‘‘statis- tically significant’’ if you want to report a difference proven by a statistical test. Although the reporting of P-values is very common in the medical literature, interpretation of findings based solely onP-values canbemisleading and is therefore dis- couraged. The 95% confidence interval not only contains the information from P-values but also additionally shows the di- rection of the treatment effect (whether toward harm or bene- fit), the size of the effect estimate, and its degree of precision.

Checklist for the results section

� Write the results section in the past tense. � Structure roughly into: recruitment/response, sample characteristics, primary analyses, secondary analyses, and ancillary analyses.

� Match the results section with the methods section. � Present findings without interpretation. � Highlight findings from tables and figures in the text. � Present estimates with 95% confidence intervals. � Consider providing additional results in tables and figures as web- only supplementary material.

2. What you should do

Keep the story line of your paper in mind: Findings in the results section should match and answer the research questions from the introduction, using the procedures ex- plained in the methods section. Retaining this focus will help you to be more concise, that is, to decide which find- ings to present and which to leave out.

Start the results section with a description of the recruit- ment/response of participants, or rather the yield of other procedures by which you obtained the data for your analy- ses. In prospective research, such as randomized controlled trials, it is particularly useful to present a flow chart of the recruitment procedure and the response of participants to treatment or measurement events (this typically becomes Figure 1 of your paper). The next step is to describe the characteristics of the study sample. Data on the sample can be presented very efficiently in a table (typically

0895-4356/$ - see front matter � 2013 Elsevier Inc. All rights reserved.

http://dx.doi.org/10.1016/j.jclinepi.2013.04.003

Table 1) and should include basic demographic characteris- tics as well as the major clinical and lifestyle variables.

Use more tables and figures to support the main text of the results section. As with all information from tables and figures, you should not repeat this information in its en- tirety in the text but only highlight the findings that support your hypothesis and those which are unexpected.

Begin a new paragraph for the results from the primary analyses. These should be presented early in the results sec- tion to stress their importance. Also use a new paragraph for results from secondary analyses. End the results section, if applicable, with a short paragraph on any additional (un- expected) findings. Make it clear that these findings result from ancillary (post hoc) analyses and are intended to gen- erate new hypotheses. Avoid words such as ‘‘remarkably’’ or ‘‘strikingly,’’ which imply an interpretation of the find- ings. Use similar sentences and words to present similar re- sults and do not try to find new ways to write the same (i.e., synonyms), as this will only confuse the reader.

Always use the same order when presenting data. For ex- ample, always report findings from the experimental group be- fore those from the control group. Provide effect sizes, such as odds ratios or relative risks, together with their 95% confi- dence intervals. Never report results withP-values only.Make consistent use of meaningful decimals for reported figures. So unless you have a very large sample size (let us say NO 1,000), present numerical values with one decimal place. Fur- thermore, present measures of central tendency together with their appropriatemeasures of variability:mean (standard devi- ation) or median (interquartile range). Always present the ab- solute number of cases in addition to relative measures (e.g., ‘‘The percentage was 22% (33/150) in the intervention group compared with 15% (23/150) in the control group’’).

Daniel Kotz Jochen W.L. Cals

E-mail address: [email protected] (D. Kotz)