For 2000, the 170 core journals we selected published 60,352 articles. The total number of pass articles was 3059 for original studies and 1073 for reviews. An article could be counted more than once if it passed for multiple categories. Six journals did not publish any pass articles. The complete list of journals and their yield appears in the Additional File 1
The category breakdown of pass articles for original studies and review articles, respectively, was 1639 and 662 for therapy and prevention, 152 and 47 for screening and diagnosis, 195 and 22 for prognosis, 290 and 308 for etiology and harm, 35 and 10 for economics, 358 and 8 for qualitative studies, and 93 and 4 for clinical prediction guides.
The top 20 journals for yield of pass articles are included in Table . The titles varied considerably in both the total number and proportion of clinically relevant articles that they published. For example, 95.0% of the articles (all reviews) in the Cochrane Database of Systematic Reviews passed our criteria, while only 2.8% of the articles in the American Journal of Gastroenterology met standards for clinically applicable studies. (The American Journal of Gastroenterology is a specialty journal and a substantial proportion of its content is preclinical. These preclinical articles, by definition, did not meet the clinical criteria in this study.) Generally, a clinical reader would need to read in the range of 13–14 articles from these top 20 journals to obtain one that is directly clinically important in any healthcare area, although the range is substantial (1.1 to 36.9). We call this number the "number of articles needed to be read" or NNR.
Number of high-quality, clinically relevant articles in the top 20 clinical journals for 2000
The number of pass articles did not correlate with SCI Impact Factors for the top 20 journals (correlation coefficient 0.29, P = 0.24). Analysis of the top 50 titles showed a weak correlation (correlation coefficient 0.41, P = 0.004) for the same analysis.
The breakdown by discipline was done using the total number of articles that were selected for inclusion in each of the four abstract journals–internal medicine, general/family practice, general care nursing, and mental health (Tables ,,,). Both the total number of articles abstracted and the total number of articles in each journal (abstracted and "Other Articles Noted") are included in Tables , , and giving a two-level assessment of "clinical worth". EBM does not publish an "Other Articles Noted" Section.
ACP J Club (internal medicine) journal-specific content of high-quality, clinically relevant articles
EBM (general/family practice) journal-specific content of high-quality, clinically relevant articles†
EBN (general practice nursing) journal-specific content of high-quality, clinically relevant articles
EBMH (mental health) journal-specific content of high-quality, clinically relevant articles
Internal medicine content (ACP J Club)
The journals contributing articles important to the practice of internal medicine (ACP J Club) are shown in Table . Substantial drop-off is seen after the top three titles (New England Journal of Medicine, JAMA, and Lancet). These three journals and the Cochrane Database of Systematic Reviews provided 56.5% of the articles abstracted, with 28 additional journals providing the other 43.5%. Fifteen titles provided only one article each and, overall, 32 journals provided at least one article for abstraction. Another 51 journals provided at least one article in the "Other Articles Noted" section. Thus, 83 journals from our list of 170 published studies important to internal medicine.
The NNR to obtain one high-quality and clinically relevant study or review varied considerably across the titles. For the more stringent definition of clinical relevance (article abstracted in ACP J Club), the range of NNR for internal medicine was from 40.4 for the Cochrane Database of Systematic Reviews to 1334 for Neurology. For the less stringent definition (article abstracted or noted in ACP J Club), the NNR range for internal medicine was from 3.4 for the Cochrane Database of Systematic Reviews to 242 for Acta Obstetrica et Gynaecologica Scandinavica.
Correlating the number of articles published in ACP J Club
with their SCI Impact Factor showed a large and positive correlation for both levels of clinical importance (correlation coefficient 0.786, P
< 0.001 for the more stringent definition; correlation coefficient 0.688, P
< 0.001 for the less stringent definition of clinical importance). These findings support the findings by Lee et al.
] that SCI Impact Factors were correlated with quality articles for general internal medicine.
General/family practice content (EBM)
The most important articles for general/family practice (publication in EBM) were published in BMJ, Lancet, Cochrane Database of Systematic Reviews, Archives of Disease in Childhood, and Annals of Internal Medicine–these journals provided 55.6% of EBM content (Table ). Overall, 45 titles provided abstracts for general/family practice coverage. The "shape" of the data is different for general/family practice than for general internal medicine, with more journals providing articles for abstraction. This is consistent with the discipline because general/family practitioners must use knowledge from a broader range of health conditions (including pediatrics and obstetrics, for example) than general internists and other specialists. Only the most stringent definition for clinical worth could be evaluated for EBM content because the "Other Articles Noted" section of the journal did not exist in the year 2000. The NNR for general/family practice ranged from 55 for the Cochrane Database of Systematic Reviews to 1351 for Circulation.
Correlation analysis showed that the number of qualified articles in each journal title was associated with the journal's SCI Impact Factor (correlation coefficient 0.546, P = < 0.001). This shows substantial agreement between SCI Impact Factors and number of articles but slightly less agreement than that found using the general internal medicine data (correlation coefficients > 0.688).
General care nursing content (EBN)
Nursing content came from many journals, including journals that are considered to be primarily targeted at physicians, and was not concentrated in a small set of journal titles (Table ). To reach 51.0% of the abstracted articles, seven titles were needed (Qualitative Health Research, Cochrane Database of Systematic Reviews, Pediatrics, JAMA, Lancet, BMJ, and Journal of Advanced Nursing). Thirty-two other journals provided articles for abstraction and 33 journals provided studies that were listed only in the "Other Articles Noted" section; 72 journals in total provided content for general care nursing. The NNR for general practice nursing was variable, ranging from 6.0 for Qualitative Health Research to 1530 for New England Journal of Medicine for the more stringent definition of clinical relevance. For the less stringent definition of clinical relevance the NNRs ranged from 4.7 for Qualitative Health Research to 923 for American Journal of Gastroenterology. The low NNR for Qualitative Health Research undoubtedly reflects the fact that only clinical criteria for relevance were applied in the selection of qualitative studies, not explicit methodologic criteria. The reason for lack of methodologic criteria for qualitative studies was that we have been unable to obtain agreement from qualitative researchers of what the quality criteria should be.
No correlation was seen between number of articles published per journal title and SCI Impact Factors for either the stringent definition of clinical relevance (correlation coefficient 0.096, P = 0.57) or the less strict definition (correlation coefficient 0.256, P = 0.038).
Mental health content (EBMH)
Mental health content was also spread over a broader range of journals than was internal medicine (Table ). To reach 53.2% of the articles abstracted, nine titles needed to be read: Archives of General Psychiatry, Cochrane Database of Systematic Reviews, American Journal of Psychiatry, British Journal of Psychiatry, JAMA, Lancet, International Journal of General Psychiatry, Journal of the American Academy of Child and Adolescent Psychiatry, and Journal of Consulting and Clinical Psychology. Forty-one titles provided at least one article for abstraction. The titles in Table show that studies related to mental health are published in many journals and specialties–a reflection of the broad nature of the discipline. The NNR for mental health for the most stringent definition of clinical relevance ranged from 20.1 for Archives of General Psychiatry to 1142.7 for BMJ. Archives of General Psychiatry also has the lowest NNR for the less stringent definition (11.5), with CMAJ having the highest NNR (1007) of those journals with at least one article on mental health. EBMH has a smaller "Other Articles Noted" section. Only eight additional journals provide articles for this section beyond the 61 that provide articles for abstraction.
A weak association was shown between the number of published mental health articles and SCI Impact Factors (correlation coefficient 0.386, P = 0.02 for the more stringent definition; and correlation coefficient 0.381, P = 0.01 for the less stringent definition).
Combining the content across the four discipline areas, we again see the concentration of important clinically relevant articles in a small subset of journals. Eight journals provided at least one article for abstraction to all four abstract journals: Annals of Internal Medicine, Archives of Internal Medicine, BMJ, CMAJ, Cochrane Database of Systematic Reviews, JAMA, Lancet, and New England Journal of Medicine. Another 10 journals provided at least one article to three of the four abstract journals: American Journal of Medicine, Archives of General Psychiatry, British Journal of General Practice, British Journal of Surgery, Health Psychology, Journal of Clinical Epidemiology, Journal of Clinical Psychopharmacology, Journal of Family Practice, Journal of the American Geriatrics Society, and Pediatrics. Twenty-eight journals provided studies to two of the abstract journals, 36 provided articles to at least one abstract journal, and 82 titles provided no articles for abstraction (excluding the six titles that did not publish any pass articles).