Diagnosis support system based on clinical guidelines: comparison between case-based fuzzy cognitive maps and Bayesian networks

Several studies have described the prevalence and severity of diagnostic errors. Diagnostic errors can arise from cognitive, training, educational and other issues. Examples of cognitive issues include flawed reasoning, incomplete knowledge, faulty information gathering or interpretation, and inappropriate use of decision-making heuristics. We describe a new approach, case-based fuzzy cognitive maps, for medical diagnosis and evaluate it by comparison with Bayesian belief networks. We created a semantic web framework that supports the two reasoning methods. We used database of 174 anonymous patients from several European hospitals: 80 of the patients were female and 94 male with an average age 45±16 (average±stdev). Thirty of the 80 female patients were pregnant. For each patient, signs/symptoms/observables/age/sex were taken into account by the system. We used a statistical approach to compare the two methods.

Building a time-saving and adaptable tool to report adverse drug events

The difficult task of detecting adverse drug events (ADEs) and the tedious process of building manual reports of ADE occurrences out of patient profiles result in a majority of adverse reactions not being reported to health regulatory authorities. The SALUS individual case safety report (ICSR) reporting tool, a component currently developed within the SALUS project, aims to support semi-automatic reporting of ADEs to regulatory authorities. In this paper, we present an initial design and current state of of our ICSR reporting tool that features: (i) automatic pre-population of reporting forms through extraction of the patient data contained in an Electronic Health Record (EHR); (ii) generation and electronic submission of the completed ICSRs by the physician to regulatory authorities; and (iii) integration of the reporting process into the physician's work-flow to limit the disturbance. The objective is to increase the rates of ADE reporting and the quality of the reported data. The SALUS interoperability platform supports patient data extraction independently of the EHR data model in use and allows generation of reports using the format expected by regulatory authorities.

An ontological approach for the exploitation of clinical data

Clinical data captured in hospital information systems may be unusable in their original format due to missing information or knowledge. The use of external resources (e.g. domain ontology) could be a way of dealing with this lack of knowledge. Our study thus aimed to develop a framework allowing a user to perform medical queries in the context of infectious diseases. By creating an interaction between a knowledge source and clinical data, using semantic and semantic web tools and methods, the users are able to perform queries on a database to obtain results about antibiotic resistance. This work has been performed in the context of the DebugIT European project that aims to control and monitor the antibioresistance growth via a semantic interoperability platform. The results obtained by the use of different semantic web tools were quantitatively evaluated by comparison of the number of results and the query execution time. We have compared our approach with classic business intelligence approaches in terms of usability and functionality.

Improving information retrieval using Medical Subject Headings Concepts: a test case on rare and chronic diseases

BACKGROUND

As more scientific work is published, it is important to improve access to the biomedical literature. Since 2000, when Medical Subject Headings (MeSH) Concepts were introduced, the MeSH Thesaurus has been concept based. Nevertheless, information retrieval is still performed at the MeSH Descriptor or Supplementary Concept level.

OBJECTIVE

The study assesses the benefit of using MeSH Concepts for indexing and information retrieval.

METHODS

Three sets of queries were built for thirty-two rare diseases and twenty-two chronic diseases: (1) using PubMed Automatic Term Mapping (ATM), (2) using Catalog and Index of French-language Health Internet (CISMeF) ATM, and (3) extrapolating the MEDLINE citations that should be indexed with a MeSH Concept.

RESULTS

Type 3 queries retrieve significantly fewer results than type 1 or type 2 queries (about 18,000 citations versus 200,000 for rare diseases; about 300,000 citations versus 2,000,000 for chronic diseases). CISMeF ATM also provides better precision than PubMed ATM for both disease categories.

DISCUSSION

Using MeSH Concept indexing instead of ATM is theoretically possible to improve retrieval performance with the current indexing policy. However, using MeSH Concept information retrieval and indexing rules would be a fundamentally better approach. These modifications have already been implemented in the CISMeF search engine.

Evaluation of automated term groupings for detecting anaphylactic shock signals for drugs

Signal detection in pharmacovigilance should take into account all terms related to a medical concept rather than a single term. We built an OWL-DL file with formal definitions of MedDRA and SNOMED-CT concepts and performed two queries, Query 1 and 2, to retrieve narrow and broad terms within the Standard MedDRA Query (SMQ) related to 'anaphylactic shock' and the terms from the High Level Term (HLT) grouping related to 'anaphylaxis'. We compared values of the EB05 (EBGM) statistical test for disproportionality with 50 active ingredients randomly selected in the public version of the FDA pharmacovigilance database. Coefficient of correlation was R(2) = 1.00 between Query 1 and HLT; R(2) = 0.98 between Query 1 and SMQ narrow; R(2) = 0.89 between Query 2 and SMQ Narrow+Broad. Generating automated groupings of terms for signal detection is feasible but requires additional efforts in modeling MedDRA terms in order to improve precision and recall of these groupings.

Automatic generation of MedDRA terms groupings using an ontology

In the context of PROTECT European project, we have developed an ontology of adverse drug reactions (OntoADR) based on the original MedDRA hierarchy and a query-based method to achieve automatic MedDRA terms groupings for improving pharmacovigilance signal detection. Those groupings were evaluated against standard handmade MedDRA groupings corresponding to first priority pharmacovigilance safety topics. Our results demonstrate that this automatic method allows catching most of the terms present in the reference groupings, and suggest that it could offer an important saving of time for the achievement of pharmacovigilance groupings. This paper describes the theoretical context of this work, the evaluation methodology, and presents the principal results.

The Electronic Healthcare Record for Clinical Research (EHR4CR) information model and terminology

A major barrier to repurposing routinely collected data for clinical research is the heterogeneity of healthcare information systems. Electronic Healthcare Record for Clinical Research (EHR4CR) is a European platform designed to improve the efficiency of conducting clinical trials. In this paper, we propose an initial architecture of the EHR4CR Semantic Interoperability Framework. We used a model-driven engineering approach to build a reference HL7-based multidimensional model bound to a set of reference clinical terminologies acting as a global as view model. We then conducted an evaluation of its expressiveness for patient eligibility. The EHR4CR information model consists in one fact table dedicated to clinical statement and 4 dimensions. The EHR4CR terminology integrates reference terminologies used in patient care (e.g LOINC, ICD-10, SNOMED CT, etc). We used the Object Constraint Language (OCL) to represent patterns of eligibility criteria as constraints on the EHR4CR model to be further transformed in SQL statements executed on different clinical data warehouses.

Data Definition Ontology for clinical data integration and querying

This paper describes an approach to build a Data Definition Ontology (DDO) in the context of full domain ontology integration with datasets in order to share and query clinical heterogeneous data repositories. We have adapted an existing semantic web tool (D2RQ) to implement a process that automatically generates the DDO from a database information model, thanks to reverse engineering and schema mapping approaches. This study has been performed in the context of the DebugIT European project (Detecting and Eliminating Bacteria UsinG Information Technology) that aims to control and monitor the bacterial growth via a semantic interoperability platform (IP). The evaluation of the process is based, first, on the accuracy of the produced DDO for different samples of database storage and second, by checking the congruency between the DDO and the D2RQ database mapping file.

Clinical diagnosis support system based on case based fuzzy cognitive maps and semantic web

Incorrect or improper diagnostic tests uses have important implications for health outcomes and costs. Clinical Decision Support Systems purports to optimize the use of diagnostic tests in clinical practice. The computerized medical reasoning should not only focus on existing medical knowledge but also on physician's previous experiences and new knowledge. Such medical knowledge is vague and defines uncertain relationships between facts and diagnosis, in this paper, Case Based Fuzzy Cognitive Maps (CBFCM) are proposed as an evolution of Fuzzy Cognitive Maps. They allow more complete representation of knowledge since case-based fuzzy rules are introduced to improve diagnosis decision. We have developed a framework for interacting with patient's data and formalizing knowledge from Guidelines in the domain of Urinary Tract Infection. The conducted study allowed us to test cognitive approaches for implementing Guidelines with Semantic Web tools. The advantage of this approach is to enable the sharing and reuse of knowledge from Guidelines, physicians experiences and simplify maintenance.

Decision support system based semantic web for personalized patient care

Personalized medicine may be considered an extension of traditional approaches to understanding and treating diseases, but with greater precision. A profile of a patient's genetic variation can guide the selection of drugs or treatment protocols that minimize harmful side effects or ensure a more successful outcome. In this paper we describe a decision support system designed to assist physicians for personalized care, and methodology for integration in the clinical workflow. A reasoning method for interacting heterogeneous knowledge and data is a necessity in the context of personalized medicine. Development of clinical decision support based semantic web for personalized patient care is to achieve its potential and improve the quality, safety and efficiency of healthcare.

Integrating clinical research with the Healthcare Enterprise: from the RE-USE project to the EHR4CR platform

BACKGROUND

There are different approaches for repurposing clinical data collected in the Electronic Healthcare Record (EHR) for use in clinical research. Semantic integration of "siloed" applications across domain boundaries is the raison d'être of the standards-based profiles developed by the Integrating the Healthcare Enterprise (IHE) initiative - an initiative by healthcare professionals and industry promoting the coordinated use of established standards such as DICOM and HL7 to address specific clinical needs in support of optimal patient care. In particular, the combination of two IHE profiles - the integration profile "Retrieve Form for Data Capture" (RFD), and the IHE content profile "Clinical Research Document" (CRD) - offers a straightforward approach to repurposing EHR data by enabling the pre-population of the case report forms (eCRF) used for clinical research data capture by Clinical Data Management Systems (CDMS) with previously collected EHR data.

OBJECTIVE

Implement an alternative solution of the RFD-CRD integration profile centered around two approaches: (i) Use of the EHR as the single-source data-entry and persistence point in order to ensure that all the clinical data for a given patient could be found in a single source irrespective of the data collection context, i.e. patient care or clinical research; and (ii) Maximize the automatic pre-population process through the use of a semantic interoperability services that identify duplicate or semantically-equivalent eCRF/EHR data elements as they were collected in the EHR context.

METHODS

The RE-USE architecture and associated profiles are focused on defining a set of scalable, standards-based, IHE-compliant profiles that can enable single-source data collection/entry and cross-system data reuse through semantic integration. Specifically, data reuse is realized through the semantic mapping of data collection fields in electronic Case Report Forms (eCRFs) to data elements previously defined as part of patient care-centric templates in the EHR context. The approach was evaluated in the context of a multi-center clinical trial conducted in a large, multi-disciplinary hospital with an installed EHR.

RESULTS

Data elements of seven eCRFs used in a multi-center clinical trial were mapped to data elements of patient care-centric templates in use in the EHR at the George Pompidou hospital. 13.4% of the data elements of the eCRFs were found to be represented in EHR templates and were therefore candidate for pre-population. During the execution phase of the clinical study, the semantic mapping architecture enabled data persisted in the EHR context as part of clinical care to be used to pre-populate eCRFS for use without secondary data entry. To ensure that the pre-populated data is viable for use in the clinical research context, all pre-populated eCRF data needs to be first approved by a trial investigator prior to being persisted in a research data store within a CDMS.

CONCLUSION

Single-source data entry in the clinical care context for use in the clinical research context - a process enabled through the use of the EHR as single point of data entry, can - if demonstrated to be a viable strategy - not only significantly reduce data collection efforts while simultaneously increasing data collection accuracy secondary to elimination of transcription or double-entry errors between the two contexts but also ensure that all the clinical data for a given patient, irrespective of the data collection context, are available in the EHR for decision support and treatment planning. The RE-USE approach used mapping algorithms to identify semantic coherence between clinical care and clinical research data elements and pre-populate eCRFs. The RE-USE project utilized SNOMED International v.3.5 as its "pivot reference terminology" to support EHR-to-eCRF mapping, a decision that likely enhanced the "recall" of the mapping algorithms. The RE-USE results demonstrate the difficult challenges involved in semantic integration between the clinical care and clinical research contexts.

Grouping pharmacovigilance terms with semantic distance

Pharmacovigilance is the activity related to the collection, analysis and prevention of adverse drug reactions (ADRs) induced by drugs or biologics. Besides other methods, statistical algorithms are used to detect previously unknown ADRs, and it was noted that groupings of ADR terms can further improve safety signal detection. Standardised MedDRA Queries are developed to assist retrieval and evaluation of MedDRA-coded ADR reports. Dependent on the context of their application, different SMQs show varying degrees of specificity and sensitivity; some appear to be over-inclusive, some might miss relevant terms. Moreover, several important safety topics are not yet fully covered by SMQs. The objective of this work is to propose an automatic method for the creation of groupings of terms. This method is based on the application of the semantic distance between MedDRA terms. Several experiments are performed, showing a promising precision and an acceptable recall.

The Information Quality Triangle: a methodology to assess clinical information quality

Building qualitative clinical decision support or monitoring based on information stored in clinical information (or EHR) systems cannot be done without assessing and controlling information quality. Numerous works have introduced methods and measures to qualify and enhance data, information models and terminologies quality. This paper introduces an approach based on an Information Quality Triangle that aims at providing a generic framework to help in characterizing quality measures and methods in the context of the integration of EHR data in a clinical datawarehouse. We have successfully experimented the proposed approach at the HEGP hospital in France, as part of the DebugIT EU FP7 project.

Bridging the semantics gap between terminologies, ontologies, and information models

SNOMED CT and other biomedical vocabularies provide semantic identifiers for all kinds of linguistic expressions, many of which cannot be considered terms in a strict sense. We analyzed such "non-terms" in SNOMED CT and concluded that many of them cannot be interpreted as directly referring to objects or processes, but rather to information entities. Discussing two approaches to represent information entities, viz. the OBO Information artifact ontology (IAO) and the HL7 v3 Reference Information Model (RIM), we propose an integrative solution for representing information entities in SNOMED CT, in a way that is still compatible with RIM and the IAO and uses moderately enhanced description logics.

Documentation in pharmacovigilance: using an ontology to extend and normalize Pubmed queries

OBJECTIVES

To assess and understand adverse drug reactions (ADRs), a systematic review of reference databases like Pubmed is a necessary and mandatory step in Pharmacovigilance. In order to assist pharmacovigilance team with a computerized tool, we performed a comparative study of 4 different approaches to query Pubmed through ADR-drug terms. The aim of this study is to assess how an ontology of adverse effects, used to normalize and extend queries, could improve this search.

MATERIAL AND METHOD

The ontological resource OntoEIM contains 58,000 classes and integrates MedDRA terminology. The entry point is a ADR-Drug term and the four methods are (i) a direct search on Pubmed (ii) a search with a normalized query enhanced with domain-specific Mesh Heading criteria, (iii) a search with the same elaborated query extended to the MeSH sub-hierarchy of the adverse effect entry and (iv) a search with a set of MedDRA terms grouped by subsomption in the OntoEIM ontology. For each of the 16 queries performed and analysed, relevant publications are selected "manually" by two pharmacovigilant experts.

RESULTS

The recall is respectively of 63%, 50%, 67% and 74%, the precision of 13%, 26%, 29% and 4%. The best recall is provided by the ontology-based method, for 4 cases out of 16 this method returns relevant publications when the others return no results.

CONCLUSION

Results show that an ontology-based search tool improves the recall performance, but other tools and methods are needed to raise the precision.

Ensuring HL7-based information model requirements within an ontology framework

This paper describes the building of an HL7-based Information Model Ontology (IMO) that can be exploited by a domain ontology in order to distribute querying over different clinical data repositories. We employed the Open Medical Development Framework (OMDF) based on a model driven development methodology. OMDF provides model transformation features to build an HL7-based information model that covers the conceptual scope of a target project. The resulting IMO is used to mediate between ontologically queries and information retrieval from semantically less defined Hospital Information Systems (HIS). In the context of the DebugIT project - which scope corresponds to the control of infectious diseases and antimicrobial resistances - Information Model Ontology is integrated to the DebugIT domain ontology in order to express queries.

Evaluation of an ontological resource for pharmacovigilance

In this work, we present a methodology for evaluating an ontology designed in a previous study to describe adverse drug reactions. We evaluate it in term of its fitness for grouping cases in pharmacovigilance. We define as gold standard the Standardized MedDRA Queries (SMQs) developed manually to group terms representing similar medical conditions. We perform an automatic search in the ontology in order to retrieve concepts related to the medical conditions. An optimal query is built for each medical condition. The evaluation relies on the comparison between the terms in the SMQ and the terms subsumed by the query. The result is quantified by sensitivity and specificity. We applied this methodology for 24 SMQs and we obtain a mean sensitivity of 0.82. This work allows validating the semantic resource and provides, in perspective, tools to maintain the ontology while the knowledge is evolving.

Combination of endogenous clues for profiling inferred semantic relations: experiments with Gene Ontology

Acquisition and enrichment of lexical resources is acknowledged as an important research in the area of computational linguistics. While such resources are often missing, specialized domains, ie biomedicine, propose several structured terminologies. In this paper, we propose a high-quality method for exploiting a structured terminology and inferring elementary synonym lexicon. The method is based on the analysis of syntactic structure of complex terms. The inferred synonym pairs are then profiled according to different clues endogenously computed within the same terminology. We apply and evaluate the approach on the Gene Ontology biomedical terminology.

A case report: using SNOMED CT for grouping Adverse Drug Reactions Terms

BACKGROUND

WHO-ART and MedDRA are medical terminologies used for the coding of adverse drug reactions in pharmacovigilance databases. MedDRA proposes 13 Special Search Categories (SSC) grouping terms associated to specific medical conditions. For instance, the SSC "Haemorrhage" includes 346 MedDRA terms among which 55 are also WHO-ART terms. WHO-ART itself does not provide such groupings. Our main contention is the possibility of classifying WHO-ART terms in semantic categories by using knowledge extracted from SNOMED CT. A previous paper presents the way WHO-ART term definitions have been automatically generated in a description logics formalism by using their corresponding SNOMED CT synonyms. Based on synonymy and relative position of WHO-ART terms in SNOMED CT, specialization or generalization relationships could be inferred. This strategy is successful for grouping the WHO-ART terms present in most MedDRA SSCs. However the strategy failed when SSC were organized on other basis than taxonomy.

METHODS

We propose a new method that improves the previous WHO-ART structure by integrating the associative relationships included in SNOMED CT.

RESULTS

The new method improves the groupings. For example, none of the 55 WHO-ART terms in the Haemorrhage SSC were matched using the previous method. With the new method, we improve the groupings and obtain 87% coverage of the Haemorrhage SSC.

CONCLUSION

SNOMED CT's terminological structure can be used to perform automated groupings in WHO-ART. This work proves that groupings already present in the MedDRA SSCs (e.g. the haemorrhage SSC) may be retrieved using classification in SNOMED CT.

Analysis of overridden alerts in a drug-drug interaction detection system

OBJECTIVE

The aim of this study was to evaluate the relevance of the signals generated by a computerized drug-drug interaction detection system and to design a classification of overridden drug-drug interaction alerts.

STUDY DESIGN

Prospective study over two months.

SETTING

Five hundred and ten-bed university paediatric hospital.

MAIN OUTCOME MEASURES

In Robert Debré Hospital physicians generate drug orders online using a computerized physician order entry system that also detects drug-drug interactions in real time. We analysed the relevance of a sample of alerts overridden by physicians.

RESULTS

We analysed a sample of 613 overridden alerts. We defined three categories of overridden alerts: informational errors (35); system errors (244) and accurate alerts (334). Two reasons account for 40% of false-positive alerts: an inability of the system to recognize real conflicts between drug treatments and guidelines stating that the two drugs can be used together, because the benefit outweighs the risk of side effects due to the drug-drug interaction.

CONCLUSIONS

We created a classification of overridden alerts, in the context of computerized physician order entry system coupled with a drug-drug interaction detection system. There is clearly room for improvement in the development of drug-drug interaction software. This classification should make it possible to break this work down into smaller tasks, making it possible to decrease the sensitivity to background noise of drug-drug interaction detection systems.

Assessment of biomedical knowledge according to confidence criteria

The characterisation of biomedical knowledge taking into account the degree of confidence expressed in texts, is an important issue in the biomedical domain. The authors of scientific texts use grammatical and lexical devices to qualify their assertions. We named these markers of qualification "confidence markers". We present here the results of our efforts to collect confidence markers from full texts and abstracts, to classify them on the basis of semantics, and their use within our knowledge extraction system. We propose in this study, an implementation of these confidence markers for functional annotation of the human gene Apolipoprotein (APOE) thought to be involved in Alzheimer's disease. As a result, we obtain, through the extraction system, triplets: (G, F, PMID), in which G is the gene APOE, F is its function found in texts and the PMID of the article from which this knowledge was extracted. Moreover, a spatial 3D of the triplets, relative to each other, is proposed depending on their respective confidence degree.

Maintenance of a computerized medical record form

Structured entry forms for clinical records should be updated to take into account the physicians' needs during consultation and advances in medical knowledge and practice. We updated the computerized medical record form of a hypertension clinic, based on its previous use and clinical guidelines. A statistical analysis of previously completed forms identified several unnecessary items rarely used by clinicians. A terminological analysis of guidelines and of free-text answers on completed forms identified several new topics relevant to current clinical practice. We therefore added new items to the form and some topics previously recorded as free text were itemized. We collaborated with clinicians in interpretation of the results of the statistical and terminological analyses used as the starting point and guide for this updating process.

Archetypes as interface between patient data and a decision support system

We propose an experiment to validate the hypothesis that archetypes enable better access and reliable use of patient data by a decision support system, mainly because they are designed to consistently link patient data with terminological systems and metadata.

Classification of health webpages as expert and non expert with a reduced set of cross-language features

Making the distinction between expert and non expert health documents can help users to select the information which is more suitable for them, according to whether they are familiar or not with medical terminology. This issue is particularly important for the information retrieval area. In our work we address this purpose through stylistic corpus analysis and the application of machine learning algorithms. Our hypothesis is that this distinction can be performed on the basis of a small number of features and that such features can be language and domain independent. The used features were acquired in source corpus (Russian language, diabetes topic) and then tested on target (French language, pneumology topic) and source corpora. These cross-language features show 90% precision and 93% recall with non expert documents in source language; and 85% precision and 74% recall with expert documents in target language.

Building an ontology of pulmonary diseases with natural language processing tools using textual corpora

Pathologies and acts are classified in thesauri to help physicians to code their activity. In practice, the use of thesauri is not sufficient to reduce variability in coding and thesauri are not suitable for computer processing. We think the automation of the coding task requires a conceptual modeling of medical items: an ontology. Our task is to help lung specialists code acts and diagnoses with software that represents medical knowledge of this concerned specialty by an ontology. The objective of the reported work was to build an ontology of pulmonary diseases dedicated to the coding process. To carry out this objective, we develop a precise methodological process for the knowledge engineer in order to build various types of medical ontologies. This process is based on the need to express precisely in natural language the meaning of each concept using differential semantics principles. A differential ontology is a hierarchy of concepts and relationships organized according to their similarities and differences. Our main research hypothesis is to apply natural language processing tools to corpora to develop the resources needed to build the ontology. We consider two corpora, one composed of patient discharge summaries and the other being a teaching book. We propose to combine two approaches to enrich the ontology building: (i) a method which consists of building terminological resources through distributional analysis and (ii) a method based on the observation of corpus sequences in order to reveal semantic relationships. Our ontology currently includes 1550 concepts and the software implementing the coding process is still under development. Results show that the proposed approach is operational and indicates that the combination of these methods and the comparison of the resulting terminological structures give interesting clues to a knowledge engineer for the building of an ontology.