Studying a Study and Testing a Test: How to Read the Medical Evidence

5 Responses to “Studying a Study and Testing a Test: How to Read the Medical Evidence”

1. Birger Hjørland says:

   June 1, 2010 at 2:32 am

   This book assumes that the reading of medical texts is not just something that is self-evident, but something that can be learned by reading the book. “The habitual course in reading the medical literature consists of “Here’s the New England Journal of Medicine. Read it!”. Unlike this “sink or swim” method, this book provides a step-by step, active-partaking approach to a clinical review of the medical literature.

   In a way, this book is a habitual book on research methods turned upside down. Instead of telling the researcher how to do and report a study, this book tells the reader how to apply such knowledge in determining the quality of a medical paper. As such it is well written, clear and relevant.

   I want to suggest, but, that research methods are everlastingly depending on a views in the philosophy of science. (Courses in the Philosophy of Medicine are apt more and more vital in the education of Doctors). As an example of a well received book I can mention “Philosophy of Medicine. An Introduction” by Henrik R. Wulff, Stig Andur Pedersen & Raben Rosenberg (Oxford, UK: Blackwell Scientific Publications, 1986). Knowledge of this kind should enable the reader to read and interpret the medical literature at a still higher level. (It is of course more trying to write simple “how to read books” based on such a more theoretical and philosophical level compared to the more statistical and methodical level).

   There exists a broader literature on “Clinical reasoning in the health Professions” (Higgs et al), “Medical semiotics” (Baer et al.), diverse philosophical studies of medicine, “Quality in science” and much more. It would be fascinating if anyone would try to expose how such knowledge may possibly be turned upside down as guide on how to read the medical literature on a still deeper level.

   It is my claim that a general background in philosophy and science studies should provide readers with even better qualifications to read the scientific literature. This is not an extraordinary position. In Denmark courses in the philosophy of science are very well loved, and just now are we discussing to make such courses compulsory in all university studies.
   Rating: 4 / 5
2. Daniel H. Freeman, Jr. says:

   June 1, 2010 at 4:23 am

   I ongoing to review the book, but may possibly not stand it after reading a few pages. On p 56 interaction is incorrectly defined and is referred as being theme to low power. This is neither right nor helpful. Even as tests of interaction may have less power then main effect tests, they can still be performed and are often critical to an analysis.
   
   P 323 Estimation is incorrectly limited to effect size. Inference is incorrectly equated with hypothesis testing. Inference is both hypothesis testing and estimation. The latter includes both top and interval estimates. Later on the page the book states data in graphic form are hard to communicate. This is flat incorrect. “One picture is worth one thousand words” and at least as many statistics. Students and researchers today are routinely urged to plot their data before undertaking an analysis.
   
   P 324 Asserts that mean is a unique case of average. The author incorrectly asserts that averages are weighted and means use weights of one. This is news to most statisticians who routinely use weighted means. In EXCEL the average function is unweighted.
   
   P 329 uses the neologisms univariable and bivariable rather than the English words univariate and bivariate. The jargon multivariable is also used, but should not be encouraged. If the analyst means multiple regression or several variables, then that should be said.
   
   P 329 asserts the choice of statistical technique depends on scale of measurement. This is far too narrow, the choice depends on the research design. The author gives an incomplete definition of continuous and numerical discrete scales. A critical top is that for either the distance from, for example, 1 to 3 is twice the distance from 1 to 2.
   
   P 331 Ordinal is required to have at least three values and “… at most a limited number of values, such as stages of cancer.” This is flat incorrect. A binary variable can be treated as ordinal and an ordinal variable can have an unlimited number of values. For example non parametric statistics can be based on ranks with one rank for each value of the original discrete data. Moreover, not all cancer stages are ordinal, for example TNM production is not ordinal. Colon cancer stages 3A and 3B are not ordered.

   At this top I resolute not to read on as I had had enough of a cardiovascular work out. This book should be avoided.
   
   Studying a Study and Testing a Test: How to Read the Health Science Literature
   Rating: 1 / 5
3. K. Pelch says:

   June 1, 2010 at 4:49 am

   Fantastic book that was really helpful as I was reviewing the literature and preparing to write a grant proposal.
   Rating: 5 / 5
4. peakibu says:

   June 1, 2010 at 7:25 am

   Very nicely written… for non-math oriented people. Very “user” friendly. Provides examples and explanations clearly.
   Rating: 5 / 5
5. Peds Resident says:

   June 1, 2010 at 7:39 am

   A wonderful book to help out with all those journal clubs and intro epi courses we have to take. I used it in medical school and continue to use it during residency. I refer to it before journal clubs and before discussing studies with medical students. Simple to read, concise, fantastic overview of the theme! Highly reccomended!
   Rating: 5 / 5