Monday, May 10, 2010

Reporting Statistics in Healthcare I

Tom Lang is seen by many as the dean of statistical reporting in healthcare. His book on statistical reporting (1) is a leading text to the guidelines used for this purpose, and for novice writers following his recommendations is de riguer. His book opens with a chapter that reviews the reporting of statistics in each part of a scientific research report. This bears your attention.


1. State the purpose of the study. Identify the relationships that were studied and the reasons for studying them. All too often the purpose for the research is left unsaid. In clinical trials, it helps to know if the study is explanatory or pragmatic. Explanatory studies, also known as efficacy studies, are designed to help understand a disease or therapeutic process; as a result, they are best conducted under optimal conditions with extremely tight controls over what occurs during the course of the study. Pragmatic studies, also known as effectiveness studies, are designed to help answer or provide information about clinical decision making. Thus, these are conducted under what might be considered “real world” or normal conditions, representing the reality of working in a clinical environment. Thus, it helps to know the purpose of the study, and in general a study should be one or the other, but not both. Lang states that you will more typically see heterogenous samples in pragmatic studies, but more homogenous samples in explanatory studies, and this makes sense given the above definitions.

2. If the study was designed to test one or more a priori hypotheses, state the hypotheses. This is because good science requires a “clear statement of a testable question.” When we know this in advance, it is called an a priori hypothesis, and the results can then be interpreted in light of that question. When we develop the question after we collect the data, that is termed a post hoc analysis, and this is a more speculative form of analysis because the data was not originally collected to answer the questions being asked “after the fact.”

3. State how the original data may be obtained for reanalysis and the format in which the data are stored. It should be an industry standard that the data collected in scientific studies should be made available upon request. Indeed, several journals require authors to sign a statement that they will provide the data when the editor asks for it. This can help to provide a check on data analysis, as well as scientific misconduct. Authors should report the format in which the data are stored, so as to ensure that those interested will be able to work with that data.

This is an introduction to how we can use report data properly. I will later look at other parts of scientific papers to provide an overview of reporting research designs and activities. I note and recommend that for those interested in conducting research Lang’s book is a must have. Highly recommended.

1. Lang T, Sekic M. How to report statistics in medicine: annotated guidelines for authors, editors and reviewers. 2nd edition. Philadelphia, PA; American College of Physicians, 2006

No comments: