Publication Languages of Impact Factor Journals and of Medical Bibliographic Databanks

Inhaltsübersicht

Background and objective: A preference for English-language sources during determination of Journal Impact Factors (IF) was discussed, IF being published in the annual Journal Citation Reports (JCR). The JCR are derived from data in Science Citation Index (SCI). The aim of this study was, therefore, (i) to review publication countries and languages in JCR, (ii) publication languages in SCI in comparison to further recognised medical bibliographic databanks.

Methods: Searching (i) countries and languages in JCR Science-Editions 1997 and 1998, (ii) language distributions in publication years 1995-2000 in bibliographic databanks SCI, MEDLINE (ME) and EMBASE (EM).

Results: (i) Almost 70 % journals in JCR 1997 and 1998 were published in USA, United Kingdom, or The Netherlands. Of two language options present, a number of English-classified journals contained > 90 % articles in other languages, whereas > 90 % publications in English could occur in Multi-Language (ML) journals, thereby complicating statistical comparisons. 83,9 % JCR-periodicals in 1997 and 85,6 % in 1998 were classified English. English/ ML ratios increased exponentially with increasing IF. (ii) 95,5 % of the articles documented 1995- 2000 in whole SCI and in our constructed SCI segment ”Medicine and related areas” were written in English, compared to 88,5 % in ME and 89,8 % in EM. The SCI Medicine segment was 15 % more comprehensive than either MEDLINE or EMBASE. Highly significant differences of language distributions in SCI vs. MEDLINE and especially SCI vs. EMBASE were observed. Retrieval rates in SCI of German-, French-, Japanese- and Chinese-language medical papers published in 2000 were impressively augmented by EMBASE and MEDLINE.

Conclusions: (i) Anglo-American publishers’ countries and English- language journals prevail in JCR with respect to numbers and IF levels. Publication language English favours citation frequency. (ii) Of databanks studied, SCI shows a maximum preference for English-language sources, thereby causing an English Language Bias during IF derivation.

Glossary

Bias: statistic-epidemiological skewing

Difference test for two shared values resulting from two very large independent samples: These statistics hold for the predictions of language distributions that, on the one hand, the tested medical databases may be viewed as random samples from the (same) universe of international medical articles, that secondly, subjects areas are comparable, and that no disparate qualitative standards between these systems occur.

EMBASE: Excerpta Medica Database/ Elsevier Science B.V. Amsterdam

ISI: Institute for Scientific Information - Thomson Scientific, Philadelphia/ USA

JCR: annual Journal Citation Reports, by ISI; there are ‘Science Editions’ (Medicine, Natural and Applied Sciences) as well as ‘Social Science Editions’

MEDLINE: Index Medicus On-Line/ National Library of Medicine, created in the USA (NLM)

SCI: Science Citation Index (also: ”Current Contents”)/ ISI

SCISEARCH: Database version of SCI at hosts DIMDI, STN Karlsruhe, DIALOG

SCI Medicine segment: a constructed part of SCI, comprising the referral documents relating to medicine and related areas (constructed in this work on the basis of the pertinent 65 journal classification codes in SCI)

IF: Journal Impact Factor (11). The formula for the year 2000 is presented in Fig. [1]:

Fig. 1 Journal Impact Factor (11): formula for the year 2000.
*) as documented in the bibliographic databases of ISI

So, an Impact Factor (IF) in the annual Journal Citation Reports (JCR) greater than, or equal to, unity is reported for a journal which published in 1998 - 1999, for example, 100 citable articles according to the Science Citation Index - SCI (original articles, short communications, reports, reviews - so-called meaty articles) and whose articles are cited at least 100 times in the subsequent year (2000) in the literature databases produced by the Institute for Scientific Information - Thomson Scientific (ISI).

However, the entirety of cited document types enters the numerator of the IF formula ([11]), i. e., cited editorials, letters and conference reports are added to the meaty articles’ citations.

Background and objective

Critical analyses [(13], [14], [18], [20]) of the annual Journal Citation Reports speak of a language bias, i. e., a skewing of results in the calculation of the JCR impact factors. This is reportedly due to the preferential selection of English-language sources relative to those in other languages. The IF is supposed to be a measure of the citation rates of scientific journals ([11]) and is internationally applied for academic evaluations ([5], [12]). Repercussions of the JCR on German medical publishing have been extensively discussed ([3], [4], [6], [8], [19]); the Zurich gynecologist Haller ([15]) even raised the question, ”Is the IF killing the German language?”

The JCR are based on evaluations of cited references in publication records stored in ISI databases. E. Garfield, founder of ISI, designated the JCR as a by-product of the Science Citation Index and the Social Science Citation Index ([10]). So the annual JCR Science Editions (medicine, natural and applied sciences) are derived from SCI data. In the literature there is a lack of concrete numbers on the distribution of publication languages in the JCR as well as in the underlying SCI database. Hence, the following questions have been raised:

1. What is the representation of publication countries and languages for journals in the JCR Science Editions in their entirety and in the area of medicine?

2. Is the distribution of publication languages in the SCI similar to that in other pertinent medical literature databases?

Methodology

JCR

Data on publication countries and languages were researched and evaluated for the JCR 1997 and 1998 Science Editions (CD ROMs, November 1998 and 1999). Country data in the JCR were derived from publisher addresses. Data on publication languages referred to the entire volume of a journal. Only the designations English (E) and Multi-Language (ML) were used.

ML periodicals, however, often contained articles predominantly in English. SERLINE/NLM and MEDIZET/Deutsche Zentralbibliothek der Medizin (ZBMED) are journal cataloguing databases using mostly different classifications in these cases. For example, we found:

• Tropical Medicine & International Health (JCR: ML; SERLINE/MEDIZET classification: English; 1990 - 2001 proportion of English-language articles in the SCI: 98.1 %)

• Experimental Brain Research (JCR: ML; SERLINE/MEDIZET: English; SCI: 100 % English articles)

• Ultrasound in Medicine and Biology (JCR: ML; SERLINE/MEDIZET: English; SCI: 100 % English articles)

Conversely, a majority of the papers in English-designated JCR journals could also be written in one other language. Four relevant examples were compared with data reported in both SERLINE and MEDIZET. The English-language share of individual publications reported for the 1990 - 2001 SCI is given in parentheses:

• Annales d’Urologie: Language French in SERLINE and MEDIZET (SCI: 7.3 % articles in English)

• Deutsche Medizinische Wochenschrift: German; English editions 1956 - 1970; 1971 - 1973 (SCI: 0.1 % in English)

• Semaines des Hopitaux: French (SCI: 0.2 % in English)

• Zeitschrift fuer Kardiologie: German (SCI: 9.1 % in English)

Therefore, statistical findings based on language classifications in JCR should be interpreted with caution.

In order to demonstrate a possible connection between IF magnitude and English-language share, the percentages classified in the JCR 1997 and 1998 as English or ML periodicals were analyzed as a function of IF. Therefore, IF intervals of width 0.5 were formed: 0.000 - 0.500, 0.501 - 1.000,..., 5.501 - 6.000, 6.000+.

Using the JCR 1998 journal subject classifications (analogous codes also exist in the SCI) a journal group M (Medicine and Related Fields) was selected from all JCR periodicals. This JCR segment should correspond somewhat to the coverage of the MEDLINE/EMBASE databases. So language shares specifically in medicine as a function of IF magnitudes could also be considered. Group M journals were listed among journals with 65 classification codes (list available upon request) from medicine and related fields. In the case of multiple categories in the JCR 1998, the first code took precedence.

Literature Databases

In bibliographic databases, language designations always refer to individual documents; however, in the JCR, they refer to entire volumes. Each database record refers to one publication and contains bibliographic data, key words and often abstracts. In addition, cited references are stored in the Science Citation Index.

The SCI (database version SCISEARCH) covers the fields of medicine and the natural and applied sciences. Comparable in size and specializing in medicine and related fields, MEDLINE and EMBASE cover, respectively, 45 % and 44 % of the scope of the SCI. With respect to its source material, MEDLINE is generally oriented to the United States. EMBASE focuses on European and on pharmacological literature and significantly complements MEDLINE search results ([1], [17], [22], [23], [24]).

The three databases mentioned were searched at DIMDI (German Institute for Medical Documentation and Information) (7) using grips commands. Determined was the number of records for the data field Language (LA), which specifies the language of the original text. The publication years 1995 - 2000 were chosen. This time span covers the evaluation period 1997 and 1998 for the IF calculation along with 1995 - 1997 (Fig. [1]) and includes the three following years to provide a sense of the possible trends of language distribution in the database.

To provide thematic comparisons, the database category ’’medicine and related areas’’ was formed using the 65 SCI classification codes. This category provides content similar to that of MEDLINE and EMBASE. In the area of medicine, the two SCI subfiles SCI-CLIN and SCI-BIOM were defined by the publisher up to and including the year 2000.

The SCI medicine section covered twice as many documents as SCI-CLIN (47.77 million versus 23.71 million) for the period 1990-XII.3.2001. However, SCI-CLIN was hardly a proper subset of the SCI medicine section, but covered 1,6 % other documents for the same period.

The scope and the scientific relevance of the constructed SCI medicine section were examined using test searches whose results were compared with those from SCI-CLIN, SCI-BIOM and SCI-CLIN+SCI-BIOM. Using medical terms, for example, such as myocardial infarct, blood pressure, gene therapy, SCI medicine searches yielded more results than a combined search of both SCI subfiles. The scope and scientific relevance of the constructed SCI Medicine section were on the whole quite acceptable. (Note: SCI-CLIN, SCI-BIOM and further SCI subfiles edited by the publishers were discontinued, e. g., at DIMDI, on I.01.2001).

The difference test for two ratios from two large independent samples was used to examine SCI medicine versus MEDLINE and SCI Medicine versus EMBASE. It was assumed that each database constituted a large independent sample (with over 2 million records each for 1995 - 2000) of all the internationally published medical articles. Such a statistical test is valid only if the differences in language distribution are not attributable to differences in subject concentrations or in qualitative criteria.

Differences in scientific coverage were controlled by the composition of the SCI Medicine segment. For the databases evaluated here, similar qualitative parameters can be assumed (and should further be imputed) in view of the decades long worldwide testing and usage of MEDLINE, SCI and EMBASE by the scientific public. Detailed database descriptions are available through DIMDI ([7]) and other hosts, DIALOG, STN Karlsruhe, etc.)

Furthermore, the content overlap of non-English-language records for combined databases (SCI Medicine+MEDLINE+EMBASE for the year 2000) was determined. To do this, duplications of all German, French, Japanese and Chinese documents retrievable in this combined database (”superbase”) were removed. During this procedure, the results of the SCI medicine section remained unaffected. Next, additional results from MEDLINE were obtained and finally the unique records present in EMBASE, but absent from SCI Medicine and MEDLINE, were added Since this extremely technical ’’duplicate checking’’ and removing method ([7]) is quite (cpu) time consuming, documents published in the year 2000 were evaluated as examples.

Results

Country and Language Distribution in the JCR 1997 and 1998 (Science Editions)

Table [1] shows the results on the origin of journals in the JCR 1998, arranged according to frequency of publication country and classified according to language (English or Multi-Language). The corresponding list for the JCR 1997 showed - for a somewhat smaller number of periodicals - similar ratios and is available upon request. An average of 68 % of the JCR journals for both volumes were published in three countries: the United States, the United Kingdom and The Netherlands. If we include German and Swiss publications, the percentage increases to almost 80 %. Forty percent of the periodicals with an impact factor were published in the United States alone.

Table 1 Periodicals in Journal Citation Report 1998 ”Science Edition”: Publishers’ Countries and Publication Languages.

Publishers’ Country	Number of Journals	% of the total of Journals in JCR	Journals classified English	ML1 * - classified periodicals	English : ML ratio	Corresponds to % Englisch
USA	2.110	38,6	2.033	77	26,4	94,4
UK	1.068	19,5	953	115	8,3	89,2
The Netherlands	535	9,8	422	113	3,7	78,9
Germany	408	7,5	255	153	1,7	62,5
Switzerland	173	3,2	123	50	2,5	71,1
France	165	3,0	88	77	1,1	53,3
Japan	141	2,6	115	26	4,4	81,6
Russia	112	2,0	112	0	112:0	100
Canada	77	1,4	30	47	0,6	39,0
Denmark	58	1,1	47	11	4,3	81,0
Australia	49	0,9	48	1	48,0	98,0
Italy	69	1,3	57	12	4,8	82,6
Sweden	38	0,7	30	8	3,8	79,0
...
China PR	31	0,6	29	2	14,5	93,5
Other countries	433	7,9	338	95	3,6	78,1
JCR, in toto	5.467	100	4.680	787	5,9	85,6
* ML = Multi-Language

In particular, the contributions of The Netherlands (about 10 %) and Switzerland (over 3 %) proved the international role of successful domestic publishing efforts in so-called smaller countries.

Of the journals with impact factors of one or greater, i. e., journals in the top 35 % of the JCR, 51 % were based in the United States, 21 % in the United Kingdom and about 10 % in The Netherlands. Germany was home to 6 % of this periodicals. Switzerland followed with about 3 %; Denmark and France were next. Where IF > = 1, it was found that 89 % of the journals were classified as English. The top journals, whose IF > = 3, had the following distribution for the years 1997 and 1998: United States at 70.3 % and at 71 %, United Kingdom at 19 % and 19.7 %, Netherlands at 5.4 % and 6.3 % With respect to this set of journals, 3.5 % and 3.4 % were published in Germany for the same years. Following these countries in the same IF range were Denmark, Switzerland and New Zealand. The share of English-classified journals with impact factors of 3 or more was 97 - 98 % for each year (detailed tables available upon request).

In the JCR, the ratio of English to Multi-Language was 5.2 to 1 (83.9 % in English) in 1997 and 5.9 to 1 (85.6 % in English) in 1998. Table [1] shows that in the publishing countries English-classified journals prevailed. For France, the English/Multi-Language ratio was close to one, likely attributable to policies promoting the French language (1997: 75 to 78; 1998: 88 to 77). For the sequence Germany (about 3 to 2), Switzerland, Netherlands, Japan and China, this ratio increases quite markedly in favor of English. For journals based in the Russian Federation, those classified exclusively as English obtained an impact factor.

Among the EU countries mentioned here, including Switzerland, there were 3.2 English-designated journals for every Multi-Language-designated journal (1766 to 546) in the year 1997.

Canada was noteworthy in both years with an English/Multi-Language ratio of about 1 to 1.5 (30 to 47 in the year 1998). This may have been due mostly to the mixed-language publications (English and French).

JCR Languages and Impact Factor Intervals

Fig. [2] shows the number of English and Multi-Language journals in the JCR 1998 as a function of IF at intervals of 0.5. The central value of the interval was chosen in each interval; for example, the interval 0.00 - 1.00 is represented in Fig. [2] by the value 0.75. Both groups diminish exponentially as the IF increases (logarithmic transformation of these numbers yields straight lines). Hence, the number of Multi-Language journals approaches zero for impact factors above three. This means that the most highly cited journals, IF > = 3, were almost exclusively classified as English.

Fig. 2 English-language and Multi-Language Journals in JCR 1998 as a Function of Impact Factor Levels. X axis: Impact Factor (Citation Frequency); Y axis: Number of Journals. Blue dots: English; black dots: multi-language.

If the log of the English/Multi-Language ratio is plotted against the impact factor, a nearly straight line with a positive slope develops. This indicates that the citation frequency, as calculated by ISI, grows exponentially with English-language representation.

In the JCR 1997, the 4,963 journals (4166 English and 797 Multi-Language) listed exhibit similar results (graphs available upon request).

Languages of Medical Journals in JCR

Table [2] illustrates that the level of English classifications for JCR 1998 periodicals depended on the discipline. In our constructed group of medical and related journals, it was 93.4 %, well above the average of 85.6 %.

Table 2 Periodicals ”Medicine and Related Areas” included in Journal Citation Report/ Science Edition 1998 (JCR): Ratios of Publication Languages English: Multi-Language

	Number of Journals	English-classified Periodicals (%)	ML-classified Periodicals (%)	Ratios English : ML
Medicine and Related Areas	2.191	2.001 (91,3 %)	190 (8,7 %)	10,5 : 1
Entire JCR for comparison	5.467	4.680 (85,6 %)	787 (14,4 %)	5,9 : 1

Publication Languages in Literature Databases

For the period 1995 - 2000, the SCI indicated the greatest coverage with around 5.57 million documents, 95.5 % of which were indicated as English. MEDLINE indexed approximately 2.52 million publications during the same period; 88.5 % of the documents were in English. In EMBASE, 89.8 % of the 2.44 million articles were written in English. If one examined the two periods 1995 - 1997 and 1998 - 2000 separately, similar values would be discovered in the three databases, with a slight tendency in favour of the English language. When compared to MEDLINE and EMBASE, the SCI showed the highest preference English-language sources in the period 1995 - 2000; other languages had representation below 5 %.

Medicine and Related Areas

MEDLINE and EMBASE database records (Table [3]) were compared with the entirety of those SCI documents for the publication years 1995 - 2000 bearing subject codes (65) for Medicine and Related Areas. In addition, the corresponding data from the SCI subfile SCI-CLIN were included in the Table.

Table 3 Publication Languages in Medicine and Relates Areas in the Period 1995 - 2000 in pertinent bibibliographic databases - with special emphasis on the relevant parts of the Science Citation Index.
Number of documented Publications¹ (percentages in parentheses) in the individual files

	MEDLINE		EMBASE		SCI - Medicine2		SCI - CLIN3
English	2.226.324	(88.5 %)	2.177.085	(89.8 %)	2741.246	(95.3 %)	1.394.388	(94.5 %)
German	51.747	(2.1 %)	72.853	(3.0 %)	48.930	(1.7 %)	37.843	(2.6 %)
French	48.145	(1.9 %)	59.704	(2.5 %)	44.178	(1.5 %)	29.677	(2.0 %)
Japanese	42.894	(1.7 %)	38.161	(1.6 %)	3.185	(0.1 %)	1	(0 %)
Russian	38.981	(1.6 %)	1.640	(0.1 %)	15.019	(0.5 %)	4.280	(0.3 %)
Spanish	31.347	(1.2 %)	23.688	(1.0 %)	13.929	(0.5 %)	8.685	(0.6 %)
Chinese	11.388	(0.5 %)	8.012	(0.3 %)	2.211	(0.1 %)	2	(0 %)
Italian	13.960	(0.6 %)	13.647	(0.6 %)	1.810	(0.1 %)	0.6	(0 %)
Other	49.708	(2.0 %)	29.631	(1.2 %)	6.016	(0.2 %)	124	(0 %)
1= Number of records retrieved by grips (7) commands in the combined data fields Publication Year (PY) und Language (LA). 2=selected Journal Subject Codes (JSC (7)) ‚Medicine and Related Fields‘ 3= SCI segment Clinical Medicine compiled by ISI (Subfile SF=CLIN in SCISEARCH at DIMDI (7) - in Jan 2001 no longer continued )

For the period 1995 - 2000, MEDLINE and EMBASE had about equal coverage, whereas the constructed SCI Medicine segment contained about 15 % more documents. Based on the more specific area definition, the ISI subfile SCI-CLIN had less coverage than our SCI medicine segment.

The table further shows that both SCI Medicine and SCI-CLIN held the largest share of English-language records with an average of 95 %. For the same period (1995 - 2000), MEDLINE indexed some 3,000 more German-language publications than the SCI Medicine segment; EMBASE indexed 24,000 more. MEDLINE included 4,000 more French-language articles than the SCI Medicine segment; EMBASE had 15,500 more. Chinese and Japanese had negligeable shares of 0.1 % in the SCI segments - compared to, e. g., 1,7 % Japanese and 0.5 % Chinese in MEDLINE.

The comparative evaluation of language shares from Table [3] according to the difference test for two ratios from two large independent samples yielded highly significant differences for language distribution in SCI Medicine versus MEDLINE and SCI Medicine versus EMBASE tests. For example, the German language had a highly significant greater representation in EMBASE than in the SCI Medicine segment. This was also true in the EMBASE versus MEDLINE case.

For publications in German, French, Japanese and Chinese, which were retrievable from a combined 2000 database incorporating SCI Medicine, MEDLINE and EMBASE, there emerges - as Fig. [3]a illustrates - a similar picture as in Table [3]. MEDLINE and, in particular, EMBASE turned out to provide significantly more results than the SCI Medicine segment for all four non-English publication languages. Determination of the content overlap based on SCI Medicine documents indicated that these records were clearly enhanced by MEDLINE and especially EMBASE. Fig. [3]b shows 100 % of the results for the three databases. The results from SCI Medicine underlies the additional records from MEDLINE and EMBASE with their stepwise organization. (Similar results related to the gross and overlap contents biomedical databases we described in (23) from detailed comparative searching in about 30 pertinent files).

Fig. 3 a. Gross Database Yields
German-, French-, Japanese- and Chinese Language Publication Referrals of Publication Year 2000 in the Science Citation Index (SCI) Medicine segment (blue), in MEDLINE (ME) (grey) and in EMBASE (EM) (white). X axis: German/ French/ Japanese/ Chinese; Y axis: Number of publication referrrals in publication year 2000.
b. Content Overlap of the individual publication refererrals: supplementing the documents in SCI-Medicine first by MEDLINE and then [SCI+MEDLINE] by EMBASE.

X axis: German/ French/ Japanese/ Chinese; Y axis: Yield of publication referrals in 2000 (in Percent), composed of 1. all pertinent SCI Medicine documents (blue), 2. additional documents from MEDLINE (grey), and 3. unique EMBASE documents in addition to the SCI+MEDLINE result (white)

[Composite result from SCI-Medicine + MEDLINE + EMBASE taken as 100 %]

Discussion

The results indicate that in the late nineties Anglo-American (58 %) and Dutch (10 %) publishers dominated among journals with impact factors. >From 7.5 % to 2.5 % of the JCR journals were published, by order of importance, in Germany, Switzerland, France and Japan.

According to the JCR language classifications - which must be considered with some reservations - publishing in English favored a high IF level and, therefore, a high international citation frequency.

Periodicals published solely in non-Roman alphabets have apparently the lowest likelihood of obtaining an impact factor. According to Table [1], all journals with an impact factor published in Russia were classified as English by ISI. Table [2] would imply that the medical Multi-Language journals tend to be cited less than the JCR average.

The data on publication languages in the JCR and in databases are not comparable for two reasons: first, the language specifications in the JCR refer to complete volumes of journals, whereas those of literature databases refer to individual articles; secondly, it was already pointed out that JCR journals classified as English may include a large portion of original articles written in other languages, such as the Deutsche Medizinische Wochenschrift. Conversely, a series of journals with predominantly English-language articles, such as Experimental Brain Research, was indexed in the JCR as Multi-Language. By way of contrast, the literature databases classified each of the articles from Exp Brain Res as English-language.

Approximately 85 % of the journals in the JCR are classified as English. (If the set is restricted to medical JCR journals, the share rises to 93.4 %.) At first glance, the English-language records in the JCR command a significantly lower share than in MEDLINE (89 %), EMBASE (90 %) and especially the SCI (95 %) and thus suggest that non-English JCR journals could have a greater representation. Based on partially unexplained language classifications in the JCR, a comparison of the proportion of English-language representation is not possible. However, on the databank level, one may compare the SCI - representing the data basis of the JCR Science Editions - and its Medicine segment with MEDLINE and EMBASE.

It becomes apparent then that the SCI Medicine segment examined here has a 15 %-larger scope than MEDLINE or EMBASE and reflects the largest number of English-language sources.

Table [3] shows major differences in the language distribution of individual databases. English-language sources in the SCI Medicine segment have the highest representation according to the difference test with a high level of significance. On the other hand, all other publication languages examined have a significantly higher frequency in MEDLINE and EMBASE than in SCI Medicine. These statistics hold for the predictions that, on the one hand, the tested medical databases may be viewed as random samples from the (same) universe of international medical articles, that secondly, subjects areas are comparable, and that no disparate qualitative standards between these systems occur.

In the constructed SCI Medicine segment, coverage is quite identical to that of MEDLINE/EMBASE. After decades long usage by the scientific community, these products of ISI, NLM and Elsevier appear qualitatively comparable.

The aforementioned high under-representation of non-English-language records in the SCI and its Medicine segment is significant for the impact factor, because the IF is calculated using cited references from the 5,700 SCI source journals ([16]). These are the active citing journals selected by ISI for IF determinations.

The preferred selection of English-language sources in medicine and the natural sciences for the SCI shows up, therefore, in the citation analysis of the JCR as an English-language bias. Total cited references in, for example, non-SCI journals publishing in German or French are not considered for the IF calculation. Thus, the following speculative question seems justified: How did the JCR change with the evaluation of additional cited references from other relevant databases? For example, MEDLINE, EMBASE and BIOSIS Previews (Biological Abstracts) include no citation data in their bibliographic information ([21], [23]). In the field of chemistry/ biochemistry, the Chemical Abstracts Service at Columbus/ Ohio has begun in 2000 to store cited references in CA/CA-plus databases.

In the area of emergency medicine, Barnaby and Gallagher ([2], [9] ) reported that the SCI-based JCR does not reflect the relevant advancements in their field, since this discipline is not represented by any citing journal for the IF calculation. One must therefore suspect the presence of a bias during the IF derivation. Any distinction between ’’actively citing journals’’ and ’’only cited journals’’ is technically without grounds. The trauma surgeon Meenen ([18]) made similar claims and mentioned the insufficient international visibility of important German-language articles, for example on the method of osteosynthesis. This may lead to omissions in applied therapy and, consequently, to disadvantages for patients.

In light of the international significance of the Journal Impact Factors, these findings overall suggest that it would be desirable if additional relevant sources excluded by the SCI could be evaluated in calculating the IF. In the area of medicine, such sources could be the MEDLINE and EMBASE journals which the SCI does not cover.

Translation in English by M. Smith and E. Garfield, Chairman Emeritus of ISI, for purposes of ASIS&T, SIG-Metrics.

Authorized by G. Winkmann ([e-mail]), with kind assistance of Georg Thieme Verlag Stuttgart ([e-mail] , [e-mail]).

References

• 1 Barillot M J, Sarrut B, Doreau C G. Evaluation of drug interaction document citation in nine on-line bibliographic databases.  Ann Pharmacother. 1997;  31 45-49
• 2 Barnaby D P, Gallagher E J. Alternative to the Science Citation Index impact factor as an assessment of emergency medicine’s scientific contributions.  Ann Emerg Med. 1998;  31 78-82
• 3 Beller F K. Der Zusammenhang zwischen Index Medicus, dem Impact Factor und der deutschen Sprache.  Dtsch Med Wochenschr. 1999;  124 A18
• 4 Beller F K. Die Zukunft der deutschen Sprache in der Wissenschaft.  Gynakol geburtshilfliche Rundsch. 2000;  40 50-54 (1)
• 5 Benitez-Bribiesca L. The impact factor of medical journals: its use and misuse.  Arch Med Res. 1999;  30 161-162
• 6 Dietrich G V, Hempelmann G. Welchen Stellenwert hat eine Publikation in deutscher Sprache?.  Anasthesiol Intensivmed Notfallmed Schmerzther. 2000;  35 543-544
• 7 DIMDI. Deutsches Institut für Medizinische Dokumentation und Information, Einstiegspunkt Database Searching (SCISEARCH-, MEDLINE-, EMBASE-Memocards; grips command language).  Internet URL http://www.dimdi.de/.
• 8 Finzen A, Hoffmann-Richter U, Dittmann V, Haug H J. Deutsch lesen - Englisch schreiben. Fachzeitschriften zwischen Science Citation Index und Nulltarif.  Psychiat Prax. 1996;  23 1-3
• 9 Gallagher E J, Barnaby D B. Evidence of methodological bias in the derivation of the Science Citation Index impact factor.  Ann Emerg Med. 1998;  31 83-86
• 10 Garfield E. How can impact factors be improved?.  Br Med J. 1996;  313 411-413
• 11 Garfield E. The Impact Factor.  Internet URL www.isinet.com/isi/hot/essays/journalcitationreports/7.html (as of XXI-02 - 2002).
• 12 Gieler A. Zur Evaluierung der Forschungsleistung eines Universitätsklinikums.  Dtsch Med Wochenschr. 2000;  125 979-983
• 13 Golder W. Der Impact Factor: eine kritische Analyse.  ROFO Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr. 1998;  169 220-226 (3)
• 14 Golder W. Wer kontrolliert die Kontrolleure? Zehn Thesen zum sogenannten Impact Factor.  Onkologie. 2000;  23 73-75
• 15 Haller U, Hepp H, Reinold E. Tötet der ”Impact Factor” die deutsche Sprache?.  Gynakol geburtshilfliche Rundsch. 1997;  37 117-118
• 16 ISI. Science Citation Index; Science Citation Index Expanded; SciSearch.  Internet URL http://www.isinet.com/products/citation/citsci.html.
• 17 Kleijnen J, Knipschild P. The comprehensiveness of MEDLINE and EMBASE computer searches. Searches for controlled trials of homoeopathy, ascorbic acid acid for common cold and gingko biloba for cerebral insufficiency and intermittent claudication.  Pharm Weekbl Sci. 1994;  14 316-320
• 18 Meenen N M. Der Impact-Faktor - ein zuverlässiger scientometrischer Parameter?.  Unfallchirurgie. 1997;  23 128-134 (4)
• 19 Middeke M. Zukunft braucht Herkunft - Die DMW im Jahr 2000.  Dtsch Med Wochenschr. 2000;  125 1099-1102 (38)
• 20 Seglen P O. Citations and journal impact factors: questionable indicators of research quality.  Allergy. 1997;  52 1050-1056
• 21 Stegmann J. Building a list of journals with constructed impact factors.  Journal of Documentation. 1999;  55 310-324
• 22 Stegmann J. Zitierstatus von MEDLINE- und Nicht-MEDLINE-Artikeln. Deutsche Gesellschaft für Informationswissenschaft und Informationspraxis (DGI) 1999. 51. Jahrestagung, Hamburg; 21 - 23. September: S. 147-158
• 23 Winkmann G, Schweim H G. Medizinisch-biowissenschaftliche Datenbanken und der Impact-Faktor.  Dtsch Med Wochenschr. 2000;  125 1133-1141 (38)
• 24 Woods D, Trewheellar K. MEDLINE and EMBASE complement each other in literature searches.  Br Med J. 1998;  316 1166

correspondence