Rapid Responses to:

LETTERS: Lucas M Bachmann, Johann Steurer, and Gerben ter Riet Simple presentation of test accuracy may lead to inflated disease probabilities BMJ 2003; 326: 393a [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Use a simple spreadsheet for accuracy calculations

GH Hall   (15 February 2003)

Improving statistical interpretation using natural frequency format

Amit Ghosh, Karthik Ghosh   (17 February 2003)

Use a simple spreadsheet for accuracy calculations 15 February 2003
GH Hall, Retired physician EX1 2HW Send response to journal: Re: Use a simple spreadsheet for accuracy calculations	Mistakes are always being made- even by experts- when questions of prior odds, probability, sensitivity, specificity, likelihood ratios etc are raised in diagnostic accuracy estimations. Nomograms are available but most people nowadays would prefer to use a simple spreadsheet, which is easy to construct. Readings of prior probability, sensitivity and specificity can be entered and the odds version of the Bayes formula used to calculate posterior odds and hence posterior probability. My own version is available on request. Competing interests:   None declared
Improving statistical interpretation using natural frequency format 17 February 2003
Amit Ghosh, Assistant Professor Medicine, Associate Program Director General Internal Medicine Fellowship Mayo Clinic, 200 First Street, SW, Rochester, Minnesota, USA 55905, Karthik Ghosh Send response to journal: Re: Improving statistical interpretation using natural frequency format	Dear Sir, The recent letter by Dr. Bachmann and colleagues, highlights the ongoing struggle with statistical numeracy among most physicians. While it is easy to blame such an error among GPs, a perusal of medical students, senior physicians, lawyers reveal a similar lack of ability to decipher statictical information(1). The problem lies with the fact that we process the same information differently with the difference in presentation. It is for the same reason that multiplication using arabic numericals became more popular than one using Roman numericals. Dr. Gigerenzer and his colleagues from Max Plank Institute, in Berlin have popularized the use of natural frequency and have found it to improve the understanding of statistical information.(2) Using the information presented by Dr. Bachmann ( Pretest probaility of endometrial cancer of 10%, sensitivity 80% and specificity of 60%), one could describe the results using natural frequency as follows, " 100 out of every 1000 women over 65 years have endometrial cancer. Of these 100 women with endometrial cancer, 80 women are detected as having an abnormal tests using transvaginal ultrasound. Out of 900 women without endometrial cancer, 360 patients will also test positive by tranvaginal ultrasound. How many women who actually test positive using the transvaginal ultrasound actually have endometrial cancer?" An increasingly effort need to be made by authors to present data in more than one format to improve understanding of scientific information. References: 1) Hoffrage U, Lindsey S, Hertwig R, Gigerenzer G. Communicating statistical information. Science 2000;290:2261-2261. 2) Hoffrage U, Gigerenzer G. Using natural frequencies to improve diagnostic inferences. Acad Med 1998;73:538-540 Competing interests:   None declared