Essential Skills: BJOG Launches The Research Methods Guides: Editorial

DOI: 10.1111/1471-0528.
15235 Editorial
www.bjog.org
Essential skills: BJOG launches the Research

Methods Guides
This series includes an interview with series Editor Dr Natalie AM Cooper in an
audio podcast available at https://soundcloud.com/bjog/rmg-interview
A certain elementary training in sta- giving the wrong information and advice increased complexity and sophistication
tistical method is becoming as neces- to our patients. Furthermore, if we cannot in the statistical method makes it difficult
sary for everyone living in this world put the findings of a recent paper or meta- to keep up, BJOG launches its Research
of today as reading and writing. analysis or the recommendations of a Methods Guides, with short understand-
recent guideline into perspective (appreci- able articles that should help us all. &
H.G. Wells (1938)1 ating its flaws and strengths, and the flaws
and strengths of other publications con- References
Unfamiliarity with statistics can make us cerning the subject) then the advice we
poor doctors, poor researchers and poor give to our patients and our management 1 Wells HG. World Brain. London: Methuen &
editors of journals. The statistical method of them is likely to veer backwards and for- Co., Ltd.; Garden City, NY: Doubleday, Doran &
influences the diagnostic tools we use, the wards according to the latest flavour of Co., Inc.
treatments we prescribe, what we say to the month. Deliverers of health care policy 2 Gill CJ, Sabin L, Schmid CH. Why clinicians are
our patients and what we write when we may fall foul of such statistical naivety, the natural bayesians. BMJ 2005;330:1080–3.
generate clinical guidelines and local hos- ‘pill scare’ of the mid 1990’s being one 3 Goodman SN. Toward evidence-based
pital protocols. It will shape the research example. The little research that we have medical statistics. II. The Bayes factor. Ann
we may choose to undertake and how we concerning the statistical abilities of clini- Intern Med 1999;130:1005–13.
do it, and whether such research is or cians is not reassuring. For example, about 4 Sir Peter Rubin, chair of the GMC. [https://
should be published. a half of clinicians, including obstetricians www.gmc-uk.org/guidance/10058.asp]
Good diagnosticians may be good and gynaecologists, are unable to correctly Accessed 5 March 2018.
statisticians in some respects, even if they interpret the statistical evidence about 5 Wegwarth O, Schwartz LM, Woloshin S,
do not know it. It has been argued that screening and diagnostic tests.5–8 Gaissmaier W, Gigerenzer G. Do physicians
the clinical diagnostic process is a clear Statistical errors in papers submitted to understand cancer screening statistics? A
example of the Bayesean statistical journals are not uncommon. Errors include national survey of primary care physicians in the
method.2 Bayes’ theorem states that the incorrect use or presentation of descriptive United States Ann Intern Med 2012;156:340–9.
pre-test odds of a hypothesis being true analysis, inappropriate choice of the statis- 6 Wegwarth O, Gigerenzer G. “There is
multiplied by the weight of new evidence tical test, unsuitable use of a statistical test nothing to worry about”: gynecologists’
(likelihood ratio) generates post-test odds for comparing three or more groups for counseling on mammography. Patient Educ
of the hypothesis being true.3 In simpler differences, incorrect presentation or Couns 2011;84:251–6.
terms, the probability of the previous interpretation of the P-value, faulty inter- 7 Bramwell R, West H, Salmon P. Health
hypothesis being correct is altered by the pretation of correlation analysis, and fail- professionals’ and service users’ interpreta-
new data. During the diagnostic process ure to provide a power analysis. Such tion of screening test results: experimental
we interpret the sequence of data that we errors have been reported to be found in study. BMJ 2006;333:284.
gather with history, examination and around three quarters of submitted manu- 8 Steurer J, Fischer JE, Bachmann LM, Koller
investigation not as a series of evaluations scripts.9 These errors often remain unde- M, ter Riet G. Communicating accuracy of
of the probability of a false positive, but tected, with about a third of published tests to general practitioners: a controlled
as an ongoing evaluation of the degree to papers containing statistical errors10, even study. BMJ 2002;324:824–6.
which a positive or negative result adjusts in the most renowned scientific journals.11 9 Simundic A-M, Nikolac N. Statistical errors in
the probability of a given disease. Some of the errors identified may be trivial, manuscripts submitted to Biochemia Medical
The process of successful diagnosis in but may indicate that that there are more journal. Biomedica Chemica 2009;19:294–300.
the clinical setting is Bayesean, but, being a serious statistical flaws elsewhere in a 10 Altman DG. Statistical reviewing for medical
good diagnostician is not all there is to paper.10 Alas, many published studies are journals. Stat Med 1998;17:2661–74.
being a good doctor. In the current climate of low quality. They may be susceptible to 11 Garcıa-Berthou E, Alcaraz C. Incongruence
of respect for patient autonomy a crucial bias, perhaps because the data are from a between test statistics and p values in medical
skill is the ability to explain the risks of dif- small single centre or because of confound- papers. BMC Med Res Methodol 2004;4:13–7.
ferent management strategies to their ing, which remains important no matter
patients and to share this information what the size of the trial. Perhaps a study is
‘openly and honestly with our patients, flawed because there is a lack of allocation
always respecting that the final decision is concealment. These are the crucial things Michael S Marsh
theirs’.4 If physicians cannot interpret the that doctors need to be able to spot. Interim Editor-in-Chief,
data they have in this data-rich world then So, knowledge of statistics is impor- on behalf of BJOG Editorial Board
we run the risk of ‘honestly and openly’ tant for many reasons. At a time when
1056 ª 2018 Royal College of Obstetricians and Gynaecologists

Essential Skills: BJOG Launches The Research Methods Guides: Editorial

Uploaded by

Copyright:

Available Formats

Essential Skills: BJOG Launches The Research Methods Guides: Editorial

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Essential Skills: BJOG Launches The Research Methods Guides: Editorial

Uploaded by

Copyright:

Available Formats

DOI: 10.1111/1471-0528.

Essential skills: BJOG launches the Research

1056 ª 2018 Royal College of Obstetricians and Gynaecologists

You might also like