Jeroen S. de Bruin

The Arden Syntax standard for clinical decision support

Arden Syntax is a widely recognized standard for representing clinical and scientific knowledge in an executable format. It has a history that reaches back until 1989 and is currently maintained by the Health Level 7 (HL7) organization. We created a production-ready development environment, compiler, rule engine and application server for Arden Syntax. Over the course of several years, we have applied this Arden - Syntax - based CDS system in a wide variety of clinical problem domains, such as hepatitis serology interpretation, monitoring of nosocomial infections or the prediction of metastatic events in melanoma patients. We found the Arden Syntax standard to be very suitable for the practical implementation of CDS systems. Among the advantages of Arden Syntax are its status as an actively developed HL7 standard, the readability of the syntax, and various syntactic features such as flexible list handling. A major challenge we encountered was the technical integration of our CDS systems in existing, heterogeneous health information systems. To address this issue, we are currently working on incorporating the HL7 standard GELLO, which provides a standardized interface and query language for accessing data in health information systems. We hope that these planned extensions of the Arden Syntax might eventually help in realizing the vision of a global, interoperable and shared library of clinical decision support knowledge.

Effectiveness of an automated surveillance system for intensive care unit-acquired infections

This study assessed the effectiveness of a fully automated surveillance system for the detection of healthcare-associated infections (HCAIs) in intensive care units. Manual ward surveillance (MS) and electronic surveillance (ES) were performed for two intensive care units of the Vienna General Hospital. All patients admitted for a period longer than 48 h between 13 November 2006 and 7 February 2007 were evaluated according to HELICS-defined rules for HCAI. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and personnel time spent per surveillance type were calculated. Ninety-three patient admissions were observed, whereby 30 HCAI episodes were taken as a reference standard. Results with MS versus ES were: sensitivity 40% versus 87%, specificity 94% versus 99%, PPV 71% versus 96%, NPV 80% versus 95%, and time spent per surveillance type 82.5 h versus 12.5 h. In conclusion, ES was found to be more effective than MS while consuming fewer personnel resources.

Scientific Workflow Management in Proteomics

Data processing in proteomics can be a challenging endeavor, requiring extensive knowledge of many different software packages, all with different algorithms, data format requirements, and user interfaces. In this article we describe the integration of a number of existing programs and tools in Taverna Workbench, a scientific workflow manager currently being developed in the bioinformatics community. We demonstrate how a workflow manager provides a single, visually clear and intuitive interface to complex data analysis tasks in proteomics, from raw mass spectrometry data to protein identifications and beyond.

Detecting borderline infection in an automated monitoring system for healthcare-associated infection using fuzzy logic

BACKGROUND Many electronic infection detection systems employ dichotomous classification methods, classifying patient data as pathological or normal with respect to one or several types of infection. An electronic monitoring and surveillance system for healthcare-associated infections (HAIs) known as Moni-ICU is being operated at the intensive care units (ICUs) of the Vienna General Hospital (VGH) in Austria. Instead of classifying patient data as pathological or normal, Moni-ICU introduces a third borderline class. Patient data classified as borderline with respect to an infection-related clinical concept or HAI surveillance definition signify that the data nearly or partly fulfill the definition for the respective concept or HAI, and are therefore neither fully pathological nor fully normal. OBJECTIVE Using fuzzy sets and propositional fuzzy rules, we calculated how frequently patient data are classified as normal, borderline, or pathological with respect to infection-related clinical concepts and HAI definitions. In dichotomous classification methods, borderline classification results would be confounded by normal. Therefore, we also assessed whether the constructed fuzzy sets and rules employed by Moni-ICU classified patient data too often or too infrequently as borderline instead of normal. PARTICIPANTS AND METHODS Electronic surveillance data were collected from adult patients (aged 18 years or older) at ten ICUs of the VGH. All adult patients admitted to these ICUs over a two-year period were reviewed. In all 5099 patient stays (4120 patients) comprising 49,394 patient days were evaluated. For classification, a part of Moni-ICUs knowledge base comprising fuzzy sets and rules for ten infection-related clinical concepts and four top-level HAI definitions was employed. Fuzzy sets were used for the classification of concepts directly related to patient data; fuzzy rules were employed for the classification of more abstract clinical concepts, and for top-level HAI surveillance definitions. Data for each clinical concept and HAI definition were classified as either normal, borderline, or pathological. For the assessment of fuzzy sets and rules, we compared how often a borderline value for a fuzzy set or rule would result in a borderline value versus a normal value for its associated HAI definition(s). The statistical significance of these comparisons was expressed in p-values calculated with Fishers exact test. RESULTS The results showed that, for clinical concepts represented by fuzzy sets, 1-17% of the data were classified as borderline. The number was substantially higher (20-81%) for fuzzy rules representing more abstract clinical concepts. A small body of data were found to be in the borderline range for the four top-level HAI definitions (0.02-2.35%). Seven of ten fuzzy sets and rules were associated significantly more often with borderline values than with normal values for their respective HAI definition(s) (p<0.001). CONCLUSION The study showed that Moni-ICU was effective in classifying patient data as borderline for infection-related concepts and top-level HAI surveillance definitions.

Medical Fuzzy Control Systems with Fuzzy Arden Syntax

Arden Syntax is a formal language for representing and processing medical knowledge that is employed by knowledge-based medical systems. In HL7 International’s Arden Syntax version 2.9 (Fuzzy Arden Syntax), the syntax was extended by formal constructs based on fuzzy set theory and fuzzy logic, including fuzzy control. These concepts are used to model linguistic and propositional uncertainty – which is inherent to medical knowledge – in a variety of clinical situations. Using these fuzzy methods, we can create medical fuzzy control systems (MFCSs), in which linguistic control rules are used and evaluated in parallel. Their results are aggregated so that gradual transitions between otherwise discrete control states are enabled. In this paper, we discuss the implementation of MFCSs in Fuzzy Arden Syntax. Through code examples from FuzzyArdenKBWean, an MFCS for weaning support in mechanically ventilated patients after cardiac surgery, we illustrate the implementation of fuzzy control.