Medical Ethics: Research Themes and Intellectual Base

The basic principle of science is to publicize its research results. An emphasis on a field's publications and patterns of formal communication has the potential ability to reflect the cognitive structures of a scientific field. Oncea discovery or research result is made public and put on permanent record, it could be said to constitute an entity of primary scientific communication and together such entities constitute an archive of public knowledge. On a personal level, such a work put on record is seen as the scientist's intellectual property. When works are read by other scientists and cited in their own works. the cited authors are recognized for their contribution to knowledge. These Citations made and received by authors/scientists constitute a formal linking mechanism between the research published by individuals and other members of the scientific community. These previous research results to which a new research publication is linked backward in time through its citations are thus formally cited as authoritative sources(Ziman, 1984). As the number of publications increase, so does the number of citations constructing an expanding citation network. Any scientist's publication is thus embedded in this network constituted by the citations to and from many other scientists and depending on their work.Analyzing this network might contribute to our understanding of a field's intellectual structure and content.In this paper the network constituted by the formalcommunica-tion of researchers from the field of medical ethics or bioethics is analyzed using bibliometric methods.The term bibliometric has been used since the beginning of the 1970st and refers to the mathematical and statistical analysis of patterns from publications and the use of publications. Medical ethics refers to the study of moral issues in the fields of medical treatment and research. The Term is also sometimes used more generally to describe ethical issues in the life sciences and the distribution of scarce medical resources. The professional fields that deal with ethical issues in medicine include medicine, nursing,law, sociology, philosophy, and theology, though today medical ethics is also recognized as its own discipline. For This analysis the Journal of Medical Ethics was chosen as a good representative of this field. First Published in 1975, it has become a leading international journal, reflecting the whole field of medical ethics. Thus, the objective of this paper is to focus on the intellectual structure of medical ethics by analyzing the citation network constituted by references in this journal, depicting discernable research themes and the structure of the intellectual base2. The Objective of this paper is purely descriptive

A bibliographic data file was downloaded from the Institute for Scientific Information's index Science Citation Index Expanded. This index, accessed by the interface of Web of Science, contains complete bibliographic data, including cited references, citations received, and author abstracts for each bibliographic item. In all, 477 articles on Medical Ethics from the period 1993–2001 were downloaded, formatted, and appended to one file. Most of the computing was accomplished using the software Bibexcel. 2.1 Citation Analysis Citation analysis encompasses a wide-ranging area of bibliometric research methods. The use of citation analysis may focus on the documents themselves, their authors, or the journals in which the documents appear. There is also a great variety as to how citation analysis is applied and for what purposes. To mention a few, citations can be considered as indicators of scientific communication patterns, where the assumption is that there is a certain amount of congruence between documental and social structures. The study of the historical, scientific process using citation analysis is based on a literary model of the scientific process where scientific work is represented by papers, and citations can be used to trace the chronology of events, relationships between them, and their importance (Smith, 1981). The study of the cognitive structure of science, where the linkages among key papers establish a structure map for the specialty. Through the study of changing structures, the development of disciplines and their interrelationship could be monitored (Small, 1973). The techniques applied for this citation analysis, bibliographic coupling and co-citation coupling, are well established. Kessler (1963) first presented bibliographic coupling as a new method for grouping technical and scientific papers on the basis of bibliographic coupling units. A coupling unit is a single reference used by two papers, and the strength of the coupling is measured by the number of coupling units between them. When two citing items have a strong bibliographic coupling, they are presumably dealing with the same subject matter. Vladutz and Cook (1984) performed a large-scale study of the concept of bibliographic coupling, where it was found that the use of bibliographic coupling yielded valid results in terms of subject relatedness in a large-scale database. The co-citation technique was initially presented by Small (ibid.) as a new form of document coupling defined as the frequency with which two documents are cited together. The number of identical citing items decides the strength of co-citation between two documents. Therefore, this relationship between documents is established by the citing authors and the way they choose to cite works pertinent to their own work. Like bibliographic coupling, co-citation is a measure of similarity as to subject area, and co-citation patterns can be used to model scientific specialties. Co-citation and bibliographic coupling can be illustrated as follows: Document I is bibliographically coupled with Document II, as they both cite Document C and Document F. Documents A, C, F, and H are all co-cited by Document I. Documents B, C, F, and L are all co-cited by Document II. Documents C and F are co-cited by Document I and II. Two complementary statistical techniques are applied in this paper: multidimensional scaling (MDS) and cluster analysis. Both techniques have their starting point in a matrix of proximity values. MDS refers to a class of techniques that use the proximity values of objects as input. The chief output is a geometric configuration of points, each point representing an object. The display of this geometric configuration is the "map" of (co-occurring) objects. The geometric configuration could be said to reflect the "hidden structure" of the matrix and often makes the data easier to comprehend (Kruskal & Wish, 1978). The operations used to obtain the geometric configuration are highly mathematical, and due to their complexity, all calculations are performed with the aid of a computer. However, the principle of MDS can be explained: Let A, B, C, and D be the representations of four objects in the matrix. Let X be the value of similarity for A and B, and Y for C and D. Ideally, the following conditions should be the rule: If X = Y, then the distance between both points in the configuration representing A and B should be the same as the distance between the points representing C and D. If X < Y, then the distance between both points representing A and B in the configuration should be greater than the distance between the points representing C and D. Clustering is a kind of classification of objects into meaningful sets (clusters). This classification might discover systems of organizing observations, usually people, into groups where members share properties in common (Stockburger, 1998). In this paper, a cluster could be defined as a group or a set consisting of a number of documents sharing some sort of property. The clusters should also be mutually exclusive, meaning no document may exist in more than one cluster. The routine is explained by a metaphorical example taken from the help files of Bibexcel: Imagine you have the following list of pairs: css Copy Edit 10 A-B 9 D-F 8 B-C 7 A-C 6 F-G 5 H-I 4 A-H Note that the pairs are sorted by a co-occurrence frequency in column 1 and that the pair list also must have tabs separating the two units of a pair. The clustering routine will have the following sequence of events (the pairs are invited to a party): A-B comes first, has to wait in the hall. D-F comes next, has to wait in the hall. B-C comes next, forms a cluster table with A-B in the living room. A-C comes next, will be deleted since A-C is already in the living room. F-G comes next, will not find a friend in the room. Goes to the hall and finds D-F. Then D-F-G will form a cluster in the room. H-I comes next, has to wait in the hall. A-H comes next, H will cluster with A-B-C, then searches the hall and finds H-I. H-I will cluster with A-B-C-H. Cluster 1 will hold: A-B-C-H-I Cluster 2 will hold: D-F-G In order to estimate interrelations (based on co-citations) between clusters, the average co-citation strengths between clusters were calculated. The average co-citation frequency for C1 and C2, AvgCC(C1, C2), is defined as: 𝐴 𝑣 𝑔 𝐶 𝐶 ( 𝐶 1 , 𝐶 2 ) = ∑ 𝐶 𝐶 ( 𝑑 𝑖 , 𝑑 𝑗 ) 𝑛 1 × 𝑛 2 AvgCC(C1,C2)= n 1 ​ ×n 2 ​ ∑CC(d i ​ ,d j ​ ) ​ where 𝑛 1 n 1 ​ is the number of documents in 𝐶 1 C1, 𝑛 2 n 2 ​ is the number of documents in 𝐶 2 C2, and 𝐶 𝐶 ( 𝑑 𝑖 , 𝑑 𝑗 ) CC(d i ​ ,d j ​ ) is the co-citation frequency for 𝑑 𝑖 d i ​ (∈ C1) and 𝑑 𝑗 d j ​ (∈ C2). Based on these normalized values (proximities) of co-citation strength, an MDS map was computed showing the relative similarity between clustered groups of documents. 2.2 The Internal Citation Network The network constituted by citations given and received exclusively by documents of a sample of selected articles from Medical Ethics will also be investigated. This approach demands a technique that enables the extraction of documents that cite or are being cited by other documents in the selected sample. This technique is implemented in Bibexcel, where a search key is constructed that contains the surname of the first author, publication year, volume number, and starting page for every record. 2.3 Word Analysis To investigate and comprehend the subject content of a set of source articles from a field of investigation, an additional method uses title words or keywords as input data. The assumption is that the frequency and co-occurrence of terms can reflect the content and structure of research. In this paper, descriptor terms are used, and after computing a pair list of co-occurring index terms, MDS is used to create a spatial configuration in two dimensions showing how concepts of medical ethics relate to one another during the time period chosen for this investigation. On a more detailed level, the co-occurrence of descriptors are analyzed in order to apprehend the context of some descriptor terms selected on the basis of their frequency ofuse and oftheir share of articles' total numberof descriptors

3.1 Research themes through bibliographic coupling The research themes derived from clusters constituted by bibliographically coupled, highly cited source documents reveal recent research themes from the Journal of Medical Ethics and are labeled according to the content indicated by titles and abstracts. The citation threshold of the citing documents was set to 4 citations, and exactly 100 documents were cited at least 5 times. The distribution of citations over source documents shows that 35% of all source documents have not, as yet, been cited at all and that a few are heavily cited. The distribution of citations—eight documents were excluded in the clustering. In all, 49 documents of 100 had at least one common reference, constituting 6 clusters. The clusters are presented with the first author, title, publication year, volume, issue, times cited, and, at the end, the number of times they appear as one of the parts of the links forming the clusters. Typically, the more connected documents appear at the beginning of each cluster, and the less relevant (as to the common theme of the cluster) documents are usually located at the end. As for the relationships between clusters, cluster 2 and 3 were found to have 21 references in common. Cluster 2 and 1 had one common reference, and the remaining clusters had no common references at all. This indicates that there probably is a greater resemblance in subject content between cluster 2 and 3 than between the other clusters. The probability that cluster 2 and 3 contain common references is, of course, enhanced by the fact that they also contain most documents. In all, the clusters have a clear emphasis on medical treatment and care. Cluster 1, Nutrition at the End of Life, is concerned with issues about terminal care and medical-moral issues concerning the alimentation of the patient near the end of life. Cluster 2, Principles of Resuscitation, has the second most documents, which indicates that the ethical question of resuscitation is a central theme in medical ethics research. In this context, questions concerning euthanasia, patient autonomy, and the age of the patient are also debated. Cluster 3, Ethics of Care and Education of Ethics, is the largest cluster and has a more general content such as the basic issues of medical ethics. In focus are educational aspects of medical ethics in the context of medical care. A more specialized theme is visible in cluster 4, Clinical Trials, where the role of uncertainty and scientific evidence as well as issues of patient and public benefit are discussed. In cluster 5, Medical Research, the emphasis is on genetic issues, but questions about screening, prevention, and obstetrics are also debated. Finally, cluster 6, Patient Autonomy, is focused on patient autonomy in the context of the refusal of blood transfusion. Cluster Breakdown Cluster 1: Nutrition at the End of Life (n=4) Stevens CA / Management of death, dying, and euthanasia—attitudes and practices of medical practitioners in South Australia / 1994 / 20 / 291 Dunlop RJ / On withholding nutrition and hydration in the terminally ill—has palliative medicine gone too far? A reply / 1995 / 21 / 8 Jochemsen H / Euthanasia in Holland—an ethical critique of the new law / 1994 / 20 / 13 Craig GIW / On withholding artificial hydration and nutrition from terminally ill sedated patients. The debate continues / 1996 / 22 / 7 Cluster 2: Principles of Resuscitation (n=15) Bruce-Jones P / Resuscitating the elderly: What do the patients want? / 1996 / 22 / 12 Robertson GS / Resuscitation and senility—a study of patients' opinions / 1993 / 19 / 14 Hilberman M / Marginally effective medical care: Ethical analysis of issues in cardiopulmonary resuscitation (CPR) / 1997 / 23 / 8 Orr RD / Requests for "inappropriate" treatment based on religious beliefs / 1997 / 23 / 10 Van Delden JJM / Deciding not to resuscitate in Dutch hospitals / 1993 / 19 / 15 Mead GE / Cardiopulmonary resuscitation in the elderly—patients and relatives' views / 1995 / 21 / 39 Cluster 3: Ethics of Care and Education of Ethics (n=16) Mitchell KR / Assessing the clinical ethical competence of undergraduate medical students / 1993 / 19 / 16 Robertson DW / Ethical theory, ethnography, and differences between doctors and nurses in approaches to patient care / 1996 / 22 / 14 Smith LFP / Ethical dilemmas for general practitioners under the UK new contract / 1994 / 20 / 7 McHaffie HE / Withholding/withdrawing treatment from neonates: Legislation and official guidelines across Europe / 1999 / 25 / 10 Cluster 4: Patient Autonomy (n=5) Savulescu J / Rational non-interventional paternalism: Why doctors ought to make judgments of what is best for their patients / 1995 / 21 / 327 Malyon A / Transfusion-free treatment of Jehovah's Witnesses: Respecting the autonomous patient's rights / 1998 / 24 / 302 Cluster 5: Using the Dead (n=3) Iserson KV / Postmortem procedures in the emergency department—using the recently dead to practice and teach / 1993 / 19 / 2 Ardagh MW / May we practice endotracheal intubation on the newly dead? / 1997 / 23 / 289 Cluster 6: Prioritization in Medicine (n=3) Nord E / The relevance of health state after treatment in prioritizing between different patients / 1993 / 19 / 37 Ryynanen OP / Random paired scenarios—a method for investigating attitudes to prioritization in medicine / 1996 / 22 / 238 3.3 Concepts of Research To make full use of the information contained in the bibliographic descriptions, the keywords assigned to the documents by the authors were analyzed. The fallout of documents was acceptable as 95% of the citing documents contained descriptors. Compared to the techniques used in 3.1 and 3.2, there is no discrimination of documents, and all documents containing descriptors are included. In this way, most of the articles of the Journal of Medical Ethics initially have the same chance to make a contribution. However, at the descriptor level, there is a choice of which ones to include. In this case, 1911 descriptors were gathered in one file, and the total number of unique descriptors was 1126, which is too many to be meaningfully reflected. Some selection had to be made. The distribution of occurrences of unique descriptors is highly skewed (Figure 2), and excluding descriptors only on the basis of low occurrence might exclude important concepts. Instead, each descriptor term was assigned a fraction according to its share of all descriptors in a particular document, and all fractions of this term were then summed up (Table 3). The reason for this approach is that if a document is assigned only a single or a few descriptors, each descriptor should have a greater weight compared to cases where many descriptors are used. The mode, median, and mean are all 4-10, suggesting that mostly 3-6 descriptors are needed to reflect the content of a document. In rare cases, one or two descriptors are sufficient, while in a few cases, more than four descriptors are used (Figure 3). Since a large portion of all descriptors occur only once (46%), it is challenging to conceptualize research foci, or it may indicate that the research field is fragmented. This suggests a problem in reflecting research structure through descriptor terms. Therefore, this analysis is limited to descriptors with relatively high scores, keeping in mind that a broad range of concepts may not be visualized. To identify changes over time, the set of descriptors was divided into three periods (Table 3). 3.1 Research themes through bibliographic coupling The research themes derived from clusters constituted by bibliographically coupled, highly cited source documents reveal recent research themes from the Journal of Medical Ethics and are labeled according to the content indicated by titles and abstracts. The citation threshold of the citing documents was set to 4 citations, and exactly 100 documents were cited at least 5 times. The distribution of citations over source documents shows that 35% of all source documents have not, as yet, been cited at all and that a few are heavily cited. The distribution of citations—eight documents were excluded in the clustering. In all, 49 documents of 100 had at least one common reference, constituting 6 clusters. The clusters are presented with the first author, title, publication year, volume, issue, times cited, and, at the end, the number of times they appear as one of the parts of the links forming the clusters. Typically, the more connected documents appear at the beginning of each cluster, and the less relevant (as to the common theme of the cluster) documents are usually located at the end. As for the relationships between clusters, cluster 2 and 3 were found to have 21 references in common. Cluster 2 and 1 had one common reference, and the remaining clusters had no common references at all. This indicates that there probably is a greater resemblance in subject content between cluster 2 and 3 than between the other clusters. The probability that cluster 2 and 3 contain common references is, of course, enhanced by the fact that they also contain most documents. In all, the clusters have a clear emphasis on medical treatment and care. Cluster 1, Nutrition at the End of Life, is concerned with issues about terminal care and medical-moral issues concerning the alimentation of the patient near the end of life. Cluster 2, Principles of Resuscitation, has the second most documents, which indicates that the ethical question of resuscitation is a central theme in medical ethics research. In this context, questions concerning euthanasia, patient autonomy, and the age of the patient are also debated. Cluster 3, Ethics of Care and Education of Ethics, is the largest cluster and has a more general content such as the basic issues of medical ethics. In focus are educational aspects of medical ethics in the context of medical care. A more specialized theme is visible in cluster 4, Clinical Trials, where the role of uncertainty and scientific evidence as well as issues of patient and public benefit are discussed. In cluster 5, Medical Research, the emphasis is on genetic issues, but questions about screening, prevention, and obstetrics are also debated. Finally, cluster 6, Patient Autonomy, is focused on patient autonomy in the context of the refusal of blood transfusion. Cluster Breakdown Cluster 1: Nutrition at the End of Life (n=4) Stevens CA / Management of death, dying, and euthanasia—attitudes and practices of medical practitioners in South Australia / 1994 / 20 / 291 Dunlop RJ / On withholding nutrition and hydration in the terminally ill—has palliative medicine gone too far? A reply / 1995 / 21 / 8 Jochemsen H / Euthanasia in Holland—an ethical critique of the new law / 1994 / 20 / 13 Craig GIW / On withholding artificial hydration and nutrition from terminally ill sedated patients. The debate continues / 1996 / 22 / 7 Cluster 2: Principles of Resuscitation (n=15) Bruce-Jones P / Resuscitating the elderly: What do the patients want? / 1996 / 22 / 12 Robertson GS / Resuscitation and senility—a study of patients' opinions / 1993 / 19 / 14 Hilberman M / Marginally effective medical care: Ethical analysis of issues in cardiopulmonary resuscitation (CPR) / 1997 / 23 / 8 Orr RD / Requests for "inappropriate" treatment based on religious beliefs / 1997 / 23 / 10 Van Delden JJM / Deciding not to resuscitate in Dutch hospitals / 1993 / 19 / 15 Mead GE / Cardiopulmonary resuscitation in the elderly—patients and relatives' views / 1995 / 21 / 39 Cluster 3: Ethics of Care and Education of Ethics (n=16) Mitchell KR / Assessing the clinical ethical competence of undergraduate medical students / 1993 / 19 / 16 Robertson DW / Ethical theory, ethnography, and differences between doctors and nurses in approaches to patient care / 1996 / 22 / 14 Smith LFP / Ethical dilemmas for general practitioners under the UK new contract / 1994 / 20 / 7 McHaffie HE / Withholding/withdrawing treatment from neonates: Legislation and official guidelines across Europe / 1999 / 25 / 10 Cluster 4: Patient Autonomy (n=5) Savulescu J / Rational non-interventional paternalism: Why doctors ought to make judgments of what is best for their patients / 1995 / 21 / 327 Malyon A / Transfusion-free treatment of Jehovah's Witnesses: Respecting the autonomous patient's rights / 1998 / 24 / 302 Cluster 5: Using the Dead (n=3) Iserson KV / Postmortem procedures in the emergency department—using the recently dead to practice and teach / 1993 / 19 / 2 Ardagh MW / May we practice endotracheal intubation on the newly dead? / 1997 / 23 / 289 Cluster 6: Prioritization in Medicine (n=3) Nord E / The relevance of health state after treatment in prioritizing between different patients / 1993 / 19 / 37 Ryynanen OP / Random paired scenarios—a method for investigating attitudes to prioritization in medicine / 1996 / 22 / 238 3.3 Concepts of Research To make full use of the information contained in the bibliographic descriptions, the keywords assigned to the documents by the authors were analyzed. The fallout of documents was acceptable as 95% of the citing documents contained descriptors. Compared to the techniques used in 3.1 and 3.2, there is no discrimination of documents, and all documents containing descriptors are included. In this way, most of the articles of the Journal of Medical Ethics initially have the same chance to make a contribution. However, at the descriptor level, there is a choice of which ones to include. In this case, 1911 descriptors were gathered in one file, and the total number of unique descriptors was 1126, which is too many to be meaningfully reflected. Some selection had to be made. The distribution of occurrences of unique descriptors is highly skewed (Figure 2), and excluding descriptors only on the basis of low occurrence might exclude important concepts. Instead, each descriptor term was assigned a fraction according to its share of all descriptors in a particular document, and all fractions of this term were then summed up (Table 3). The reason for this approach is that if a document is assigned only a single or a few descriptors, each descriptor should have a greater weight compared to cases where many descriptors are used. The mode, median, and mean are all 4-10, suggesting that mostly 3-6 descriptors are needed to reflect the content of a document. In rare cases, one or two descriptors are sufficient, while in a few cases, more than four descriptors are used (Figure 3). Since a large portion of all descriptors occur only once (46%), it is challenging to conceptualize research foci, or it may indicate that the research field is fragmented. This suggests a problem in reflecting research structure through descriptor terms. Therefore, this analysis is limited to descriptors with relatively high scores, keeping in mind that a broad range of concepts may not be visualized. To identify changes over time, the set of descriptors was divided into three periods (Table 3). 3.4 The Intellectual Base of Journal of Medical Ethics, 1993–2001 3.4.1 Document Types and Journals The intellectual base of the Journal of Medical Ethics is constituted of a total of 1,412 unique references. In order to reflect the distribution of citations to the more cited items, citations to non-journal items as well as to journal articles were counted, and a rank-ordered listing was produced to show the top journals as well as the top non-journal items cited (Table 4). Not surprisingly, most of the journal articles cited at least four times are from the Journal of Medical Ethics, and 37% of the citations to articles are self-citations at the journal level. In all, 28% of the citations to items cited at least four times are directed to non-journal items, and 72% to journal articles. As for the total number of citations, 2,875 or 39% are to non-journals. This means that the intellectual base of the Journal of Medical Ethics, to a greater extent, can be traced to journal articles and, to a lesser extent, to books. In order to get a general view of the citations to journals, the total distribution of citations to journals was calculated (Table 5). Comparing Table 4 and Table 5 as to the 10 top positions, one can see that two new journals have entered in Table 5: Hastings Center Report, which is a genuine medical ethics journal, and Social Science & Medicine. The two journals that lost their more prominent positions were the Journal of Clinical Ethics and the Journal of the Royal College of Physicians of London. In all, the dominance of journals not exclusively focusing on medical ethics is obvious. Table 4 The rank order of journal articles & non-journal items Note: The rank order is based on the distribution of citations to items cited at least four times by documents in the Journal of Medical Ethics, 1993–2001. Journal articles are merged with the journal in which they are published. Table 5 The distribution of citations to journals cited by documents published in Journal of Medical Ethics, 1993–2001 Note: Only journals cited at least 30 times are shown. Figure 1 Journal Co-Citation Map – Journals cited by Journal of Medical Ethics, 1993–2001 Note: Only journals cited at least 10 times, a total of 48 journals, were included. The circle size is proportional to the number of citations received by a journal, and the width of connecting lines indicates the co-citation strength. Journal titles are abbreviated. It can be of interest not only to know about the degree of visibility or use reflected by the number of citations received by journals but also to reflect the structure of the intellectual base on a journal level and appreciate the whole pattern of relations between journals cited by Journal of Medical Ethics. In order to accomplish this, a journal co-citation map, based on the number of times that journals in the reference lists of documents from the Journal of Medical Ethics co-occur, was produced by means of MDS (Figure 9). The relations between journals and the spatial configuration of the map could be viewed in a center-periphery manner: In the center, a core is constituted by the most cited journals and corresponds well with the listing of highly cited journals in Table 5, while less cited journals are located in the periphery. Peripheral journals in this field represent a wide spectrum of journals focused on different aspects of medical research: cardiovascular & respiratory diseases, critical care, emergency medicine, genetics, geriatrics, internal medicine, obstetrics, and pediatrics. Other fields of research present in the periphery are nursing science and philosophy, medical education, and law medicine. To conclude, it is obvious that a few highly cited journals from the field of medical ethics, as well as a few journals with a general medical focus, construct the core. An additional number of journals on medical ethics, like Bulletin of Medical Ethics, Bioethics, Cambridge Quarterly of Healthcare Ethics, and Kennedy Institute of Ethics Journal, are quite distant from the core and thus more seldom co-cited with journals of the same field. The circle size is proportional to the number of citations received by a journal, and the width of connecting lines indicates the co-citation strength. Journal titles are abbreviated. 3.4.2 The Cluster Structure of the Intellectual Base The distribution of citations by cited documents is extremely skewed, and 93% of all references are cited only once (Figure 10). As all documents are not of the same importance, or at least, not used or made visible to the same extent, some citation threshold for inclusion in the analysis has to be decided on. In this case, all documents cited at least four times were included, constituting a set of 55 documents. The resulting clustering produced six clusters, containing 53 documents, which are represented with the number of objects per cluster, the median publication year, and the number of citing documents. The objects in the clusters are presented with the author's name, publication year, volume number, starting page, abbreviated journal title, number of citations, and, at the end, the number of times they appear as one of the parts of the links forming the clusters. In order to interpret and label the clusters, all documents citing a particular cluster are collected, and titles, abstracts, identifiers, and descriptors are studied. Thus, the labeling of clusters is based on the content of the citing documents. However, it seems valid, to a certain extent, that there should be a resemblance between the subject content of the cited documents and the subject content of the citing documents. Looking at the literature that authors in the Journal of Medical Ethics refer to, the publication period of the analyzed items has no limitation. Certainly, an author can refer to any document he or she sees as pertinent regardless of its age. Nevertheless, a publication seems to lose interest for readers the older it is, and more recent publications are usually more visible through their use, as reflected by citations received. In accordance with this notion, the mean median publication year is 1994, which means that relatively recent publications are among the more cited documents selected. Scanning the labels of these clusters, it is obvious that the subject content of the intellectual base does not diverge radically from the content of clusters based on the citing documents in sections 3.1 and 3.2. As far as the authors of the Journal of Medical Ethics base their articles on previous knowledge and results contained in their referenced publications, this should, to some extent, be reflected in the subject content of their own articles as well as in their titles and assigned keywords. Nevertheless, new themes can be seen to emerge, where referenced items are seen in a new light and linked together in new combinations by the way they are co-cited. The subject-content dimension is rather clear: Clinical aspects of medical ethics are reflected by clusters 1, 2, 4, and 5. A more general theme of ethics in connection with education is reflected by cluster 3, which is also the largest cluster. Using MDS, it is possible to reflect the interrelations between clusters by counting the number of times that documents in clusters are co-cited with documents of other clusters (Figure 11). As there are differences in the number of objects in each cluster, the mean co-citation strength between clusters is calculated as presented in section 2.1. The most related clusters are cluster 5 and cluster 6. Apparently, as indicated by similar labels, they focus on similar research topics, and it is, in fact, not unlikely that they are part of the same "specialty" of medical ethics research. Proportionally, they seem to be equally cited, and the median publication year is about the same. Thus, the concept of autonomy is connected to the concept of life-death decisions and, through cluster 3, connected to aspects of education and ethical principles. The latter cluster has the most central position, indicating that aspects of ethics and education are connected to several themes of medical ethics. Clusters 1, 2, and 4 reflect more specialized aspects of medical ethics research and have fewer interconnections with other clusters. Obviously, the time factor has some importance as to how documents cohere in terms of being co-cited. A certain aspect of research may increase in size and connections to other clusters over time, especially for a "young" cluster reflecting a relatively new research theme. Cluster 1 Clinical and Research Ethics Committees (N=9, median publication year=1995, number of citing documents=22) Thornton JG, 1995, V311, P667, Brit Med J/4/4; Gilbert C, 1989, V299, P1437, Brit Med J/8/3; Garfield R, 1995, V311, P660, Brit Med J/4/2; While AE, 1995, V311, P661, Brit Med J/4/2; Gillon R, 1997, V23, P203, J Med Ethics/5/1; Harries UI, 1994, V28, P150, J Roy Coll Phys Lond/4/1; Royal Coll Phys, 1990, Guid Pract Eth Comm/4/1; Brit Med Ass, 1999, Withh Withdr Life Pro/6/1; Craig GM, 1994, V20, P139, J Med Ethics/4/1 Cluster 2 Patient Autonomy - Transfusion-Free Treatment (N=3, median publication year=1998, number of citing documents=6) Muramoto O, 1998, V24, P223, J Med Ethics/6/2; Malyon D, 1998, V24, P376, J Med Ethics/4/1; Muramoto O, 1998, V24, P295, J Med Ethics/4/1 Cluster 3 Ethical Principles in Health Care and Teaching Ethics (N=15, median publication year=1990, number of citing documents=49) Hebert P, 1990, V16, P141, J Med Ethics/6/5; Culver CM, 1985, V312, P253, New Engl J Med/8/4; Beauchamp TL, 1994, Principles Biomedical/9/3; Parfit D, 1984, Reasons Persons/5/2; Sulmasy DP, 1993, V19, P157, J Med Ethics/4/2; Buchanan AE, 1989, Deciding Others Ethi/4/2; Mitchell KR, 1993, V19, P230, J Med Ethics/5/2; Gillon R, 1994, V309, P184, Brit Med J/5/1; Miles SH, 1989, V64, P705, Acad Med/5/1; Hebert PC, 1992, V18, P142, J Med Ethics/4/1; Duster T, 1990, Backdoor Eugenics/4/1; Pellegrino ED, 1990, V1, P175, J Clin Ethics/4/1; Pellegrino ED, 1992, V268, P1734, Jama-J Am Med Assoc/4/1; Sulmasy DP, 1990, V150, P2509, Arch Intern Med/4/1; Gillon R, 1994, Principles Hlth Care/4/1 Cluster 4 Covert Video Surveillance (N=3, median publication year=1995, number of citing documents=6) Evans D, 1995, V21, P9, J Med Ethics/4/2; Southall DP, 1987, V294, P1637, Brit Med J/4/1; Southall DP, 1995, V21, P104, J Med Ethics/4/1 Cluster 5 Autonomy and Life-Death Decisions (N=14, median publication year=1991, number of citing documents=51) SE, 1983, V309, P569, N Engl J Med/5/1; Tomlinson T, 1990, V264, P1276, Jama-J Am Med Assoc/5/1; Emanuel LL, 1991, V324, P889, New Engl J Med/5/1; Vanderwal G, 1996, V335, P1706, New Engl J Med/4/1; Bakhurst D, 1992, V18, P63, J Med Ethics/4/1 Cluster 6 Life-Death Decisions (N=8, median publication year=1990, number of citing documents=25) Morgan R, 1994, V308, P1677, Brit Med J/5/4; Jonsson Figure 11 Interrelations between cocitation clusters Note: Circle sizes are proportional to the number of objects in clusters and width of lines to the average cocitation strength between clusters. Schneiderman LJ, 1990, V112, P949, Ann Intern Med/8/5; Beauchamp TL, 1989, Principles Biomedical/7/4; Wreen MJ, 1991, V11, P124, J Med Ethics/6/3; Blackhall LJ, 1987, V317, P1281, New Engl J Med/4/3; Truog RD, 1992, V326, P1560, New Engl J Med/4/3; Jackson J, 1991, V17, P5, J Med Ethics/5/2; Gilligan C, 1982, Different Voice/4/1; On RD, 1997, V23, P142, J Med Ethics/4/1; Savulescu J, 1995, V21, P327, J Med Ethics/4/1; Bedell PV, 1988, V148, P2373, Arch Intern Med/5/3; Murphy DJ, 1994, V330, P545, New Engl J Med/4/2; Duff RS, 1973, V289, P890, New Engl J Med/4/1; Rawls J, 1971, Theory Justice/5/1; Seckler AB, 1991, V115, P92, Ann Intern Med/5/1; Vandermaas PJ, 1991, V338, P669, Lancet/4/1; Wanzer SH, 1989, V320, P844, New Engl J Med/5/1 CluA(1), Clinical and Research Ethics Committees; (N=9, median publication year=1995, number of citing documents=22) CluB(2), Patient Autonomy - Transfusion-Free Treatment; (N=3, median publication year=1998, number of citing documents=6) CluC(3), Ethical Principles in Health Care and Teaching Ethics; (N=15, median publication year=1990, number of citing documents=49) [Cluster 4, Covert Video Surveillance; (N=3, median publication year=1995, number of citing documents=6). No documents of Cluster 4 are cocited with any documents of the other clusters.] CluE(5), Autonomy and Life-Death Decisions; (N=14, median publication year=1991, number of citing documents=51) CluF(6), Life-Death Decisions; (N=8, median publication year=1990, number of citing documents=25)

Based on the assumption that the Journal of Medical Ethics is a prominent representative for the field of medical ethics, the object of this paper has been to examine if its content can reflect the intellectual structure of the field. It Was found that core journals in the journal citation structure comprised several journals not explicitly focused on medical ethics. It is not clear to what extent cited publications in these journals focus on mere medical ethics issues or on medical-clinical issues. Furthermore, a number of journals with a clear emphasis on medical ethics were found to be recited to a lesser extent by the Journal ofMedical Ethics. Therefore, the result is indicative and not an exhaustive reflection of all possible research foci in the field.Concerning the description of subject content, number of research foci or research themes have been identified. Educational aspects and ethical principles is a pronounced theme of the intellectual base, represented by the largest and most central cluster. Autonomy and ethical aspects of life-death decisions is another pronounced theme and so is medical research ethics in a clinical context. Other Mclre peripheral and delimited foci concern blood transfusion in a religious context, and covert surveillance.The analysis of documents linked by bibliographic coupling visualized clinical aspects of medical ethics through clusters dealing with ethical problems of resuscitation and nutrition at the terminal stage. The largest cluster is concerned with issues about the ethical principles of 'care and ethics education'. The ethics of medical research and patient autonomy were two additional themes.Clustering documents that cite or are cited by other documents in the Journal of Medical Ethics during the period 1993-200I, the objective was to see how research themes are formed around the top-cited documents of this particular journal and thus reflect this journals' internal,formal communication. The research themes were in large found to be similar to those found in clusters based on bibliographic coupling, though a few new research themes emerged.The analysis of words gave insights into which major concepts that have a central meaning to the field, reflectingthe use and combination of descriptor terms. Autonomy, ethics, euthanasia and medical ethics were found to be used frequently over 3 consecutive periods. The context of these terms was analyzed as co-occurrence with other terms and provided a more detailed understanding of each term. A general view of term relations by means of MDS based on co-occurrences further ascertained the cognitive context of more frequent descriptor terms. Over the time period of this investigation a few terms have remained stable and a large share of low frequency descriptors indicate difficulties in describing the research structure on this basis.