A tale of two citations: counting on numeracy in the digital divide

The BMJ (British Medical Journal) and JAMA (the Journal of the American Medical Association) are vanguard clarions of their respective medical communities. In 1999 the BMJ published an analysis of JAMA cover art that was critical of the choices made.¹ This criticism inflamed many post-colonial physicians, who rose to defend their homeland. With pride, those west of the Atlantic pointed to the technological advances, the rise of Silicon Valley, and the importance of numbers in the daily lives of Americans.

To demonstrate this last item we evaluated whether numeracy in titles of articles in medical journals is more prevalent, and by inference more important to physicians, in the United States.

Methods and results

We reviewed issues of JAMA and BMJ for 2001. To maintain equivalence—that is, not comparing apples and oranges²—we examined only original contributions in JAMA and research papers in BMJ. A number in a title was identified as such if it was cardinal or ordinal, in digits or spelt out. Ethical discussions and case reports were not considered as numeric.

We carried out a Medline search for 1990-2000 using the same criteria, but this could not be restricted to original contributions or research papers.

Our null hypothesis was that the BMJ had equal numeric citations to JAMA. We computed relative risk ratios and used standard error for the log ratio with comparison of proportions to determine 95% confidence intervals. The Mann-Whitney U test was used for comparison of citation numbers over the years 1990-2000. Trends were examined with the Spearman rank correlation.³

We found that in 2001 JAMA had significantly fewer research titles with numbers (16% of 184 articles v 26% of 234 for the BMJ; relative risk 0.63, 95% confidence interval 0.44 to 0.89). Additionally, Medline had fewer JAMA citations with numbers in the title for each year we searched except one (figure; P=0.011). Over the last 11 years of the millennium, the BMJ actually increased numeric citations (P<0.001 for trend). In 2001 not only were numbers less prevalent in titles in JAMA but research articles with numeric based conclusions were also less densely presented (0.13 conclusions per page for 1359 pages v 0.34 per page for 642 pages in the BMJ, relative risk 0.39, 0.32 to 0.46).

Comments

We found that the BMJ surpassed JAMA in its pursuit of numeracy and therefore claim that the much lamented decline of British medical numeracy⁴ is inaccurate.

The upward trend over the past decade in the quantity of BMJ citations with numbers may reflect an increasing awareness of the importance and power of numeracy in medicine.⁵ Medical authors often feel the need to be more exact. Although a number is not worth a thousand words, it conveys a concept that may otherwise need more than one word. Finding the ideal ratio of words to numbers may prove a fruitful area for further research.

We cannot exclude the possibility that BMJ editors, already experts at delivering a high density of information, have begun increasing their insistence on numbers in titles to save space. We must hope they will moderate this stance before BMJ titles become unbroken strands of digits.

British medical education has never slighted the number. The series Statistics at Square One is often used as a reference text in medical schools on both sides of the Atlantic. Clinical Evidence, published twice a year by the BMJ Publishing Group, is a standard of medical meta-analyses.

Our findings show that the numeracy of the BMJ exceeds that of its colonial cousin. The land of Newton, Hardy, Turing, and Wiles can remain proud of its mathematical abilities.

Figure.

Number of Medline citations with numbers in their titles, 1990-2000