Biomedical informatics--a confluence of disciplines?

BACKGROUND: Biomedical informatics is a broad discipline that borrows many methods and techniques from other disciplines. OBJECTIVE: To reflect a) on the character of biomedical informatics and to determine whether it is multi-disciplinary or inter-disciplinary; b) on the question whether biomedical informatics is more than the sum of its supporting disciplines and c) on the position of biomedical informatics with respect to related disciplines. METHOD: Inviting an international group of experts in biomedical informatics and related disciplines on the occasion of the 50th anniversary of Methods of Information in Medicine to present their viewpoints. RESULTS AND CONCLUSIONS: This paper contains the reflections of a number of the invited experts on the character of biomedical informatics. Most of the authors agree that biomedical informatics is an interdisciplinary field of study where researchers with different scientific backgrounds alone or in combination carry out research. Biomedical informatics is a very broad scientific field and still expanding, yet comprised of a constructive aspect (designing and building systems). One author expressed that the essence of biomedical informatics, as opposed to related disciplines, lies in the modelling of the biomedical content. Interdisciplinarity also has consequences for education. Maintaining rigid disciplinary structures does not allow for sufficient adaptability to capitalize on important trends nor to leverage the influences these trends may have on biomedical informatics. It is therefore important for students to become aware of research findings in related disciplines. In this respect, it was also noted that the fact that many scientific fields use different languages and that the research findings are stored in separate bibliographic databases makes it possible that potentially connected findings will never be linked, despite the fact that these findings were published. Bridges between the sciences are needed for the success of biomedical informatics.

Mapping the domain of medical informatics.

OBJECTIVES: The domain of medical informatics (MI) is not well defined. It covers a wide range of research topics. Our objective is to characterize the field of MI by means of the scientific literature in this domain. METHODS: We used titles and abstracts from MEDLINE records of papers published between July 1993 and July 2008, and extracted uni-, bi- and trigrams as features. Starting with the ISI category of medical informatics, we applied a semi-automated procedure to identify the set of journals and proceedings pertaining to MI. A clustering algorithm was subsequently applied to the articles from this set of publications. RESULTS: MI literature can be divided into three subdomains: 1) the organization, application, and evaluation of health information systems, 2) medical knowledge representation, and 3) signal and data analysis. Over the last fifteen years, the field has remained relatively stable, although most journals have shifted their focus somewhat. CONCLUSIONS: We identified the scientific literature pertaining to the field of MI, and the main areas of research. We were able to show trends in the field, and the positioning of different journals within this field.

Triangulation applied to Jan H. van Bemmel.

OBJECTIVE: To describe the person of Jan H. van Bemmel from different points of view. METHOD: Triangulation. RESULTS AND CONCLUSIONS: Jan H. van Bemmel successfully contributed to research and education in medical informatics. He inspired a lot of people in The Netherlands and internationally.

Evaluation and implementation: a call for action.

OBJECTIVES: To raise awareness for actions that are urgently needed to accompany the large scale implementations of ICT in Health Care that are currently taking place in many countries around the world. METHODS: An analysis of a few studies that have recently been described in the literature guided by recent suggestions for research and development of evaluation of health ICT. RESULTS AND CONCLUSION: Six specific recommendations for action are specified: Development of good implementation practice, Development of an experience base of implementation of ICT in health care, Setting up a surveillance system for unintended effects, Build an evidence base of best evaluation practice, Developing guidelines for proper reporting of evaluation studies, Education of clinicians and decision makers.

Diagnostic reference frames for seizures: A validation study.

INTRODUCTION: We developed structured descriptions of signs and symptoms for specific seizure types (called Diagnostic Reference Frames-DRFs-by us) that can serve as a frame of reference in the process of classifying patients with epileptic seizures. In this study the validity of the DRFs for clinical use is evaluated and described. MATERIAL AND METHODS: In this study we use a decision support system based on the DRFs and using Bayes's rule for the validation of the DRFs. Patient's manifestations are entered in the decision support system and by successively applying Bayes's rule posterior probabilities are calculated. The DRFs with the highest posterior probability gives an indication of the classification of the seizure. The validation of the DRFs was performed by comparing the seizure type with the highest posterior probability with the classification of experienced epileptologists on a series of test cases with known epileptic seizures. In this way we assessed the accuracy of the DRFs in classifying patients with epileptic seizures. RESULTS: We included sixty-six patients in this efficacy study. The patients and/or their relatives described the manifestations occurring during a seizure. Sixty cases (91%) were correctly classified using the decision support system. DISCUSSION: The accuracy of 91 % indicates that the knowledge encoded in the DRFs for the included seizure types is valid. The next step is to test the DRFs in a clinical setting to evaluate the applicability in daily practice.

Three-Layer Model for the design of a Protocol Support System.

OBJECTIVE: The aim of the PropeR project is to investigate the impact of Active Computerized Protocol Support (ACPS) on daily care processes in different settings (home care and hospital care). ACPS consists of an active Protocol Support System (PSS) that is linked to an Electronic Patient Record system. The aim of this paper is to describe how we have taken the organizational and social aspects into account in the hospital setting and the consequences of this approach for the design of the PSS. METHODS: Socio-technical approaches have been applied. Observations and interviews with various health care providers were performed at the hematology and oncology department of the University Hospital Maastricht. Ten extensive sessions with a specialist physician and research nurse took place to further elaborate a study protocol and to discuss how it is integrated in daily practice. The knowledge editor component of Gaston was used to build a computer interpretable version of the selected protocol. RESULTS AND CONCLUSIONS: To support the representation of a study protocol integrated in routine clinical care, a Three-Layer Model was developed. This model distinguishes the protocol description, local adaptations to the protocol and communication as three separate layers. These layers have been incorporated into the knowledge acquisition tool Gaston. The Three-Layer Model makes easy updating possible, and also supports transferability of computerized (study) protocols to other organizations.

An approach to knowledge base construction based on expert opinions.

OBJECTIVES: To describe, validate and demonstrate an approach for knowledge base construction based on expert opinions. METHODS: A knowledge base containing the frequency of occurrence of manifestations in epileptic seizures is constructed based on information provided by neurologists/epileptologists. The reliability of the responses is determined with the inter-rater intraclass correlation coefficient (ICC). If the ICC is not large enough the Spearman-Brown prophecy formula can be used to predict the number of additional experts. We propose a method to assess whether an additional expert provides information consistent with the already acquired data as well as a method to detect experts with deviating opinions. The power of the first method was determined. RESULTS: Data were collected for five seizure types. The ICCs determined from the responses for the various seizure types after inclusion of the additional experts was in all cases almost equal to 0.9, the target value. Yet one expert with diverging opinions concerning the frequency of occurrence of manifestations for different seizure types could be identified. Excluding this participant improved the reliability of the data. The power of the methods was good (> or =0.75). CONCLUSIONS: It is shown that human experts can provide reliable information about the frequency of occurrence of manifestations in epileptic seizures. In addition, the described approach correctly identified neurologists/epileptologists with both consistent and diverging opinions about the frequency of occurrence of manifestations in a number of seizure types.

The PropeR way to support medical doctors in daily practice. Developing the protocol based DSS.

This paper describes the first phase of the development of a Protocol based Decision Support System (PDSS) that will be linked to an Electronic Patient Record system (EPR system). The protocol system will be pro-active: the physician will be automatically prompted from the EPR of a particular patient if the protocol that applies for that patient defines it necessary. The PropeR project studies the impact of a PDSS that is linked to an EPR on daily care processes. There are two areas of research: hospital and home care. This paper describes the application in the hospital. The protocol that is being computerized is a treatment protocol for Acute Myelogenous Leukaemia (AML) that also studies treatment alternatives (conventional versus experimental treatment). This paper based AML protocol has been translated into a formal representation. The KA-tool Gaston is used to make this representation. Twenty-eight subprotocols have been organized in a hierarchical structure with three levels. One of the aims of the project is to make a representation of the AML protocol that can be used in other organizations as well. The main problem we encountered is that the representation not only contains the content of the protocol, but also aspects of application of that protocol in daily care of the hospital and aspects of support. The solution to this problem is the creation of two layers of representation: the first layer is an exact copy of the protocol and thus sharable and the second layer focuses on the support of the protocol in the daily working processes and is mainly domain specific: for the University Hospital Maastricht. At the moment, this division into two layers is being discussed.

How many neurologists/epileptologists are needed to provide reliable descriptions of seizure types?

We are developing seizure descriptions as a basis for decision support. Based on an existing dataset we used the Spearman-Brown prophecy formula to estimate how many neurologist/epileptologists are needed to obtain reliable seizure descriptions (rho = 0.9). By extending the number of participants to the required level we found that the number of participants needed to obtain a reliability coefficient of 0.9 were in accordance with the number of participants determined from the Spearman-Brown prophecy formula. Systematic differences between the participants were minor and not statistically significant.

Development of diagnostic reference frames for seizures. Part 1: inter-participant agreement in the selection of symptoms.

OBJECTIVE: Our aim is to develop reliable descriptions of various seizure types, which will be used as a basis for decision support. We use expert opinions in this process. In this contribution we evaluate the inter-participant agreement in the selection of frequently occurring symptoms for the description of seizure types. METHOD: We compared the actual agreement among participants with the agreement that would result from random symptom selection as well as with the maximal agreement attainable. For each seizure type we calculated the reliability coefficients of the responses. RESULTS: For all seizure types we found that the agreement in symptom selection among the participants is significantly higher than expected by chance, but not reaching the maximum agreement attainable. The reliability coefficients varied between 0.56 and 0.74 for the various seizure types. CONCLUSION: Although the participants do not reach the maximum agreement attainable in the selection of symptoms, the majority agreement on characteristic frequently occurring symptoms for the different seizure types does approach the maximum agreement attainable. Therefore, we conclude that expert opinions can be used for building descriptions of seizure types. However, to derive a reliable set of symptoms for the construction of the diagnostic reference frames (DRFs) more participants are needed.

Development of diagnostic reference frames for seizures. Part 2: are seizure descriptions discriminative?

OBJECTIVE: To determine whether the seizure descriptions given by a group of neurologists/epileptologists are discriminative. METHOD: We constructed templates for various seizure types describing how often symptoms were selected by the participants. We defined a matching score to indicate the match between such a template and the symptoms selected by each neurologist/epileptologist individually and computed the scores for each of the sets of selected symptoms with all templates. Correlation coefficients were calculated between the templates. RESULTS: Data were collected from 24 participants. The matching scores and the correlation coefficients both show that participants provide discriminative descriptions of the seizure types. Descriptions of aggregated seizure types, such as primary generalized seizures, are less discriminatory than the descriptions of more specific seizure types. CONCLUSION: We concluded that the participants in our study selected symptoms that result in discriminative descriptions of the seizure types. This indicates that knowledge elicitation by using the opinions of a group of clinical experts is possible. The study also indicates that the design of the study could be ameliorated in several ways. These findings will be taken into account when designing the final study.

Medical informatics as a discipline at the beginning of the 21st century.

OBJECTIVES: To analyse the present situation of the discipline medical informatics and to propose actions for change. METHODS: Evaluation of the current situation mainly based on anecdotal evidence. RESULTS: The difference between the scientific and the engineering aspects of medical informatics get blurred. Because of the requirements of European funding medical informatics focuses more on engineering than on science. Too many manuscripts are submitted that describe engineered artefacts without a scientific purpose. Some of the subjects (like security issues) that are studied in medical informatics are not considered important by medical faculties thus impeding support. CONCLUSIONS: The methodological underpinnings of our research should be strengthened, impact studies should be more frequently performed; the quality of results reporting should be increased.

Inter-participant agreement in the selection of symptoms for the description of seizures.

PURPOSE: To evaluate the inter-participant agreement in the selection of symptoms for the description of seizure types. METHODS: We evaluated the inter-participant agreement for a number of seizure types by comparing the actual agreement to the agreement that would result from random symptom selection as well as the maximal attainable agreement among participants that selected a different number of symptoms. RESULTS: For all seizure types the agreement in the symptom selection among the majority of the participants is significantly higher than expected by chance, but not reaching the maximum agreement attainable. CONCLUSION: Although the maximum agreement possible is not obtained, the symptoms selected by a majority of the participants seem to be adequate for the description of seizure types.

Robustness metrics for measuring the influence of additive noise on the performance of statistical classifiers.

This paper presents a novel quality measure called robustness. The robustness measure quantifies the influence of measurement noise in the attribute values on the credibility of the classification of a case. It is assumed that the type of distribution of the noise-generating process is known. It is not simple to measure the robustness in the general situation where the noise-free distribution of the attributes is unknown. Therefore, two approximations are suggested and compared with the robustness measure based on the noise-free distribution of the attributes. The usefulness of the suggested robustness measure is explored in a simulation experiment.

Evaluating the MACRO solution. Multimedia Application for Clinical Research in Oncology.

This paper describes the evaluation approach of the MACRO project. This approach is based on earlier work in the European 3rd Framework Programme (AIM). We describe how user requirements and expectations play a role in the design of the various assessment studies in MACRO.

The discriminative value of patient characteristics and dyspeptic symptoms for upper gastrointestinal endoscopic findings: a study on the clinical presentation of 1,147 patients.

The discriminative value of patient characteristics and dyspeptic symptoms for upper gastrointestinal endoscopic findings was prospectively assessed in 1,147 patients attending for their first diagnostic endoscopy and who answered paper (n = 431) or computerized (n = 716) questionnaires. The questionnaires provided detailed information concerning present dyspeptic symptoms, with special attention to provoking and/or relieving factors, and smoking and/or drinking habits. In logistic regression models each of a number of 'specific endoscopic diagnoses' was contrasted with normal endoscopy (n = 390), and 'relevant endoscopic disease' (oesophagitis, peptic ulcers, cancers; n = 269) was contrasted with 'irrelevant' and normal endoscopic findings (n = 878). From the regression model a receiver operating characteristic (ROC) curve could be constructed, and the area under the ROC curve (AUC) was calculated to summarize the discriminative power of the regression model. The best discrimination from patients with a normal endoscopy was achieved for patients with gastric (AUC = 0.86) or duodenal (AUC = 0.85) ulcers, followed by patients with hiatus hernia (AUC = 0.78 or oesophagitis (AUC = 0.77). The discriminative performance of the regression models was somewhat less for duodenitis/bulbitis (AUC = 0.75) and endoscopic gastritis (AUC = 0.73). In an open-access endoscopy unit setting, the value of preinvestigation history-taking for the prediction of clinically relevant endoscopic disease was very limited (AUC = 0.63).

Appropriateness of indications for diagnostic upper gastrointestinal endoscopy: association with relevant endoscopic disease.

BACKGROUND: Since the institution of open access endoscopy units there has been a considerable increase of referrals for UGI examinations. Therefore, guidelines for the appropriate use of UGI endoscopy are needed. METHODS: The outcome of first diagnostic UGI endoscopy was prospectively assessed for several referral indications in a consecutive series of 2900 patients. Indications were judged "appropriate" when significantly (p < 0.01) associated with clinically "relevant" endoscopic findings. RESULTS: The proportion of relevant disease for various indications was as follows: signs of UGI bleeding (42.2%); history of peptic ulcer (40.5%); dysphagia (31.9%), short-term (24.4%), and without therapy (20.9%). Relevant endoscopic findings were observed in 21.0% of dyspeptic patients aged 45 years or less, and in 25.3% of those older than 45 years of age. CONCLUSIONS: The generally approved alarm symptoms should be a reason to perform endoscopy without hesitation. Dyspeptic symptoms, despite adequate empiric treatment, as well as first dyspeptic symptoms in patients older than 45 years should also be a reason for endoscopic investigation. Our results support the strategy to treat patients younger than 45 years who have isolated dyspepsia by a limited course of antipeptic agents, provided that they are seen for re-evaluation within 4 to 6 weeks.

Evaluation of new self-learning techniques for the generation of criteria for differentiation of wide-QRS tachycardia in supraventricular tachycardia and ventricular tachycardia.

This study presents a comparison of three different methods for differentiating between supraventricular and ventricular tachycardias with wide-QRS complex. One set of criteria, derived using classical statistical techniques, was compared with two new self-learning computer techniques: the artificial neural networks and the induction algorithm approach. By analyzing the results obtained in an independent test set, using these new techniques, the criteria defined by the classical method could be improved.

KAVAS-2: Knowledge Acquisition, Visualization and Assessment System.

The objective of KAVAS-2 is the development of a tool, named KAVIAR, with which domain experts can make their knowledge explicit. It contains components for (computer assisted) knowledge elicitation and for machine learning. A key issue in KAVAS is the assessment of the quality of the classification and domain models built. Various quality measures are available and implemented in KAVIAR to assess the quality of models, specifically those developed from data bases by machine learning techniques.