Statistics are essential to how medical writers produce, present, and interpret medical data. However, it is not widely acknowledged that it can be difficult to get a good grounding in statistics. A small study published in the New England Journal of Medicine over 40 years ago found that the majority of doctors overestimated the positive predictive value of a laboratory test result.
This test was repeated in 2014 with a similar result finding that most respondents could not assess the positive predictive value with most errors coming from an overestimation of the positive predictive value.
Medical writers are not the ones designing or carrying out statistical analyses. A critical component of the job is the interpretation and the ability to translate statistics into meaningful results. It is difficult to provide a comprehensive review of medical statistics, so here are some common mistakes to watch out for in various aspects of research:
Design of a study: Although medical writers are not always involved in the design stages of a study, they have a critical role at the writing stage. Key information about study design that can be easily forgotten are the study aims and outcome measures, number of participants, and an a priori statement or description of the null hypothesis. Journal checklists are a great resource to use when writing and reviewing research to ensure this information is not left out.
Data analysis: These errors should be picked up by statisticians involved with the research. One of the biggest errors to be aware of is inflation of a type 1 error. This is when the null hypothesis is true, but it is rejected. The Khan Academy provides some good basic videos on type 1 and type 2 errors.
Presentation: This is perhaps the most significant area where medical writers can have an impact. It is important that data are appropriately described. The mean should always come with a description of variability. Furthermore, standard deviations should be presented in parentheses to make it easier to differentiate between this and a 95% confidence interval. Confidence intervals should be included with the P-value for primary outcomes to provide information about the size of an effect. P-values should also be reported as the exact number obtained, rather than stating a generalised P<0.05 or P>0.05.
Interpretation: Conclusions can be drawn from results that are not supported by the data. A result that is not statistically significant does not mean that there was no effect and should not be interpreted as such. On the contrary, results should not be claimed as significant without the analyses to back it up. Any interpretation of results should also include a discussion on confounding and bias.
This is a very brief overview of some common statistical mistakes. Medical writers are not expected to have a statistician level of knowledge! It is important that statistics can be interpreted correctly as they influence changes that can potentially impact people’s lives.
To find more information on basic statistics, the BMJ provides a good basic principles article.
Elion Medical Communications is an owner-operated, freelance medical communications business offering outsourced medical writing support for researchers, medical communications agencies, healthcare advertising agencies, medical device, and pharmaceutical companies worldwide.