Knowledge-based systems: contrasting realities and expectations and possible ways forward.

Correspondence to: Casimir A. Kulikowski Department of Computer Science Rutgers – The State University of New Jersey New Brunswick NJ 08901 USA E-mail: [email protected] The paper by Spreckelsen et al. [1] in the present issue of Methods of Information in Medicine discusses the results of an online survey of researchers and developers of medical knowledge-based systems (KBSs) in Germany, Austria and Switzerland, and reports on insights into the current applications and their clinical roles, as well as the obstacles, challenges and expectations faced by developers of KBSs in trying to broaden the impact of these systems in clinical and research settings. The online survey had a response rate of about 33% which is reasonable. It is noteworthy that almost half of the participants stated that they had experience in medical informatics for 10 years or longer, and over 72% indicated they were currently working on a research project. Questions were designed to elicit both quantitative and qualitative responses. Quantitative responses included the type of application fields of KBS developed by the institutions of the responding researchers, which were predominantly (70%) in decision support, data interpretation, pattern recognition/ classification, information retrieval or text mining, with only 9% in ontology development, 8% in control or simulation, 6% in modeling or process modeling, 5% in image processing/interpretation and 1% in language processing. This illustrates a strong bias towards KBSs designed for decision-support and related interpretation tasks which may explain some of the contrasting or apparently contradictory results of the survey. The most striking results of the survey come from the Likert scale responses to the questions about the demand, acceptance, routine use and future application of KBSs. In specifying whether they agreed with the statement “There is high demand for knowledge based systems in medicine” 67.8% of respondents agreed at the two highest levels of the Likert scale. However, in saying whether they agreed with the statement “There is a high acceptance for KBS in the clinical context”, only 14.6% agreed at the two highest levels, with 33.4% strongly or somewhat strongly disagreeing, and 34.4% indicating neutrality. This con trast between perceived demand and acceptance is even more pronounced on the issue of routine use. In specifying agreement with the statement “KBSs are well established in the clinical practice” no respondents agreed strongly, only 6.3% agreed somewhat strongly, but 31.2% disagreed either most strongly or somewhat strongly with the statement. Yet, on future application the numbers swing again to the optimistic side with 68.8% of respondents agreeing that KBSs will grow strongly in the future, and only 4.2% disagreeing somewhat (and none disagreeing strongly). The qualitative questions from the survey produced answers that included: 1. listing the main fields of application of KBSs in medicine, reported by the developer-respondents in German-speaking countries. These cited Arden syntax, ontologies, pharmaco-vigilance/prescribing as the top three principal fields, with CPOE, guidelines, laboratory information systems, ambient assisted living scoring systems, semantic web frameworks, and search engines as the others; 2. citing specific exemplary applica tions, for which a drug information system, an