Showing papers presented at "Knowledge Representation for Health-Care in 2012"

Process mining in healthcare: data challenges when answering frequently posed questions

Abstract: In hospitals, huge amounts of data are recorded concerning the diagnosis and treatments of patients. Process mining can exploit such data and provide an accurate view on healthcare processes and show how they are really executed. In this paper, we describe the different types of event data found in current Hospital Information Systems (HISs). Based on this classification of available data, open problems and challenges are discussed that need to be solved in order to increase the uptake of process mining in healthcare.

Embedding conformance checking in a process intelligence system in hospital environments

Abstract: Process intelligence is an effective means to analyze and improve business processes in companies with high degree of automation. Hospitals are also facing high pressure to be profitable with ever decreasing available funds in a stressed healthcare sector, which calls for methods to enable process management and intelligent methods in their daily work. However, traditional process intelligence systems work with logs of execution data that is generated by workflow engines controlling the execution of a process. But the nature of the treatment processes requires the doctors to work with a high freedom of action, rendering workflow engines unusable in this context. In this paper, we introduce a novel method to conformance checking that computes fitness of individual activities in the setting of sparse process execution information, i.e., not all activities of a patient's treatment are logged. We embed this method into a process intelligence approach for hospitals without workflow engines, enabling process monitoring and analysis.

Rule-Based combination of comorbid treatments for chronic diseases applied to hypertension, diabetes mellitus and heart failure

Abstract: The treatment of patients with several chronic diseases (comorbidities) has become a frequent actuation of health-care professionals in their daily practice. As different treatments are needed for each disease, there is a risk of undesired drug interactions that must be detected and solved using evidence-based medical knowledge. In this paper we have extracted part of this knowledge for the comorbidities of hypertension, diabetes mellitus and heart failure, and we have represented it by means of combination rules. A rule execution system has been developed which is able to combine treatments of different diseases into a unique comorbid treatment avoiding undesired drug interactions. The system has been checked by health-care professionals of the SAGESSA Health-care group in 20 medical cases.

Data integration for clinical decision support based on openEHR archetypes and HL7 virtual medical record

Abstract: Clinical Decision Support Systems (CDSS) have gained relevance due to their potential to support patient-centric care, but their deployment still has to overcome barriers to become successful. One of these barriers is the integration of patient data with the CDSS engine, a tough challenge given the need to address interoperability with many different existing systems and medical devices. The MobiGuide project aims to build such a CDSS, providing guideline- based clinical decision support through a Personal Health Record (PHR). This PHR is the main component through which the CDSS could access patient data originating from hospital EMRs and wearable sensors, but it also contains the log of the recommendations provided by the CDSS. Using a case study, we compare data-representation standards through which the PHR could be developed, while considering expressiveness and usability requirements. We propose to develop the PHR by combining openEHR archetypes and the HL7 Virtual Medical Record standard, supported by a service oriented framework for data exchange. This proposal aims to close the gap between the HL7 and the ISO/CEN 13606 by using an openEHR-based approach.

Towards a realistic clinical-guidelines application framework: desiderata, applications, and lessons learned

Abstract: Clinicians can benefit from automated support to guideline (GL) application at the point of care However, several conceptual dimensions should be considered for a realistic application: 1) The representation of the knowledge might be through structured text (semi-formal), or specified in a machine-comprehensible language (formal); 2) The availability of electronic patient data might be partial or full; 3) GL-based recommendations might be triggered by a human-initiated (synchronous) session, or data---driven (asynchronous) In addition, several requirements must be fulfilled, such as an evaluation of the GL application engine by a GL simulation engine Finally, to apply multiple GLs, by multiple users, in multiple settings, the GL-application engine should be designed as an enterprise architecture that can plug into any Electronic Medical Record (EMR) We present an architecture fulfilling these desiderata, describe application examples with different conceptual dimensions and requirements, using our new GL-application engine, PICARD, discuss lessons learned, and briefly describe a clinical evaluation of the current framework in the domain of pre-eclampsia/toxemia of pregnancy

Semantic integration of patient data and quality indicators based on open EHR archetypes

Abstract: Electronic Health Records (EHRs) contain a wealth of information, but accessing and (re)using it is often difficult. Archetypes have been shown to facilitate the (re)use of EHR data, and may be useful with regard to clinical quality indicators. These indicators are often released centrally, but computed locally in several hospitals. They are typically expressed in natural language, which due to its inherent ambiguity does not guarantee comparable results. Thus, their information requirements should be formalised and expressed via standard terminologies such as SNOMED CT to represent concepts, and information models such as archetypes to represent their agreed-upon structure, and the relations between the concepts. The two-level methodology of the archetype paradigm allows domain experts to intuitively define indicators at the knowledge level, and the resulting queries are computable across institutions that employ the required archetypes. We tested whether openEHR archetypes can represent both elements of patient data required by indicators and EHR data for automated indicator computation. The relevant elements of the indicators and our hospital's database schema were mapped to (elements of) publicly available archetypes. The coverage of the public repository was high, and editing an archetype to fit our requirements was straightforward. Based on this mapping, a set of three indicators from the domain of gastrointestinal cancer surgery was formalised into archetyped SPARQL queries and run against archetyped patient data in OWL from our hospital's data warehouse to compute the indicators. The computed indicator results were comparable to centrally computed and publicly reported results, with differences likely to be due to differing indicator definitions and interpretations, insufficient data quality and insufficient and imprecise encoding. This paper shows that openEHR archetypes facilitate the semantic integration of quality indicators and routine patient data to automatically compute indicators.

An approach for representing and managing medical exceptions in care pathways based on temporal hierarchical planning techniques

Abstract: This work presents an approach for representing and managing medical exceptions that may occur during the execution of a patient-centered care pathway. Personalized care pathways are generated automatically by means of a knowledge-driven planning process over a temporal hierarchical task network (HTN), which encodes an evidence-based clinical guideline. The exceptional situations specified in this guideline as well as the recommendations for their management are represented by knowledge-based rules in the task network model. However these rules, which encode the exceptional flow of the guideline, are represented separately from the normal flow in order to not obscure the modelling. Moreover, we propose the use of medical concepts from a standard terminology (UMLS) for the formal representation of these rules. This fact promotes interoperability, knowledge sharing and precision aspects. Finally, a therapy planning system with capabilities for exception detection, analysis and adaptation has been developed. As a result, the proposal, which is evaluated with oncology care plans, seems to be an adequate exception recovery mechanism maintaining guideline adherence.

Knowledge-Driven adaptive execution of care pathways based on continuous planning techniques

Abstract: Care pathways are excellent tools for the standardization of care delivery and the improvement of clinical efficiency. The high dynamism and unpredictability of the clinical environment require pathways to be adaptable to the deviations that may arise during their execution. In this work we present a methodology for the identification, monitoring, detection and managing of these deviations. Such deviations include evolving patient conditions, arbitrary medical modifications and unpredicted clinical settings. The care pathways are dynamically generated based on a knowledge-driven planning process that personalizes treatments according to up-to-date patient conditions and guarantees adherence to clinical guidelines recommendations. The implementation of the proposed methodology in a domain-independent continuous planning architecture is presented.

The elicitation, representation, application, and automated discovery of time-oriented declarative clinical knowledge

Abstract: Monitoring, interpretation, and analysis of large amounts of time-stamped clinical data are tasks that are at the core of tasks such as the management of chronic patients using clinical guidelines, the retrospective assessment of the quality of that application, or the related task of clinical research by learning new knowledge from the accumulating data. I briefly describe several conceptual and computational architectures developed over the past 20 years, mostly by my research teams at Stanford and Ben Gurion universities, for knowledge-based performance of these tasks, and highlight the complex and interesting relationships amongst them. Examples of such architectures include the IDAN goal-directed temporal-mediation and the Momentum data-driven monitoring architectures, both of which are based on the knowledge-based temporal-abstraction method; the KNAVE-II and VISITORS knowledge-based interactive-exploration frameworks for single and multiple longitudinal records; and the KarmaLego interval-based temporal data mining methodology. I point out the progression from individual-subject data-interpretation, monitoring, and therapy, to multiple-patient aggregate analysis and research, and finally to the discovery and learning of new knowledge. This progression can be viewed as a positive-feedback loop, in which new knowledge is brought back to bear upon both individual-patient management and on the learning of new and meaningful (temporal) associations.

A study of semantic proximity between archetype terms based on SNOMED CT relationships

Abstract: The OpenEHR archetypes have been suggested as a standard for detailing data models of electronic healthcare records, as a means of achieving interoperability between clinical systems. But, mapping terms of these clinical data models to a terminology system, such as SNOMED CT, is a crucial step to provide the required interoperability. Through this study, we aim to understand better how archetype clinical information is semantically related using SNOMED CT relationships as a reference. For this purpose, we developed an automated approach to bind archetype terms to the SNOMED CT terminology. Our method revealed a high degree of semantic similarity between the terms modeled in the archetypes and the hierarchical and logical relationships covered by SNOMED CT. It has been detected that more than 75% of the archetype terms are semantically related to other terms of the same archetype. Taking this into account, our approach applies a combination of terminological relationships-based techniques with lexical and linguistic resources. A set of 25 clinical archetypes with 477 bound terms was used to test the method. Of these, 378 terms (79%) were linked with 96% precision, 76% recall. Our approach has proven to take advantage of the SNOMED CT relationship structure, increasing the total recall by 10%. Therefore, this work shows that it is possible to automatically map archetype terms to a standard terminology with a high precision and recall, with the help of appropriate contextual and semantic information of both models.

Distributed ad hoc cooperation in healthcare

Abstract: Inter-institutional cooperation among physicians becomes increasingly important. Yet, it is unrealistic to assume that cooperation can be supported via a homogeneous system that is pre-installed in every organization. Instead, physicians will typically have their own autonomous systems that support internal processes. Traditional activity-oriented workflow platforms do not resolve inter-institutional integration challenges. This paper presents an approach for distributed process management, which enables ad hoc cooperation via active electronic documents without the need to integrate local systems. A distributed case file is used to coordinate cooperating parties. Using this case file does not require any preinstalled system components, so genuine ad hoc information exchange is enabled.

Flexible and efficient retrieval of haemodialysis time series

Abstract: The problem of retrieving time series similar to a specified query pattern has been recently addressed within the Case Based Reasoning (CBR) literature. Providing a flexible and efficient way of dealing with such an issue would be of paramount importance in medical domains, where many patient parameters are often collected in the form of time series. In this paper, we describe a novel framework for retrieving cases with time series features, relying on Temporal Abstractions. With respect to more classical (mathematical) approaches, our framework provides significant advantages. In particular, multi-level abstraction mechanisms and proper indexing techniques allow for flexible query issuing, and for efficient and interactive query answering. The framework is currently being applied to the hemodialysis domain. In this field, experimental results have shown the superiority of our approach with respect to the use of a classical mathematical technique in flexibility, user friendliness, and also quality of results. Tests in other application domains, as well as further enhancements, are foreseen in our future work.