AEHRC & QUT at TREC 2011 Medical Track : a concept-based information retrieval approach

The Australian e-Health Research Centre and Queensland University of Technology recently participated in the TREC 2011 Medical Records Track. This paper reports on our methods, results and experience using a concept-based information retrieval approach. Our concept-based approach is intended to overcome specific challenges we identify in searching medical records. Queries and documents are transformed from their term-based originals into medical concepts as de ned by the SNOMED-CT ontology. Results show our concept-based approach performed above the median in all three performance metrics: bref (+12%), R-prec (+18%) and Prec@10 (+6%).

Agreement between self-reported use of in vitro fertilization or ovulation induction, and medical insurance claims in Australian women aged 28–36 years

STUDY QUESTION: What is the self-reported use of in vitro fertilization (IVF) and ovulation induction (OI) in comparison with insurance claims by Australian women aged 28–36 years? SUMMARY ANSWER: The self-reported use of IVF is quite likely to be valid; however, the use of OI is less well reported. WHAT IS KNOWN AND WHAT THIS PAPER ADDS: Population-based research often relies on the self-reported use of IVF and OI because access to medical records can be difficult and the data need to include sufficient personal identifying information for linkage to other data sources. There have been few attempts to explore the reliability of the self-reported use of IVF and OI using the linkage to medical insurance claims for either treatment. STUDY DESIGN: This prospective, population-based, longitudinal study included the cohort of women born during 1973–1978 and participating in the Australian Longitudinal Study on Women's Health (ALSWH) (n = 14247). From 1996 to 2009, participants were surveyed up to five times. PARTICIPANTS AND SETTING: Participants self-reported their use of IVF or OI in two mailed surveys when aged 28–33 and 31–36 years (n = 7280), respectively. This study links self-report survey responses and claims for treatment or medication from the universal national health insurance scheme (i.e. Medicare Australia). MAIN RESULTS AND THE ROLE OF CHANCE: Comparisons between self-reports and claims data were undertaken for all women consenting to the linkage (n = 3375). The self-reported use of IVF was compared with claims for OI for IVF (Kappa, K = 0.83), oocyte collection (K = 0.82), sperm preparation (K = 0.83), intracytoplasmic sperm injection (K = 0.40), fresh embryo transfers (K = 0.82), frozen embryo transfers (K = 0.64) and OI for IVF medication (K = 0.17). The self-reported use of OI was compared with ovulation monitoring (K = 0.52) and OI medication (K = 0.71). BIAS, CONFOUNDING AND OTHER REASONS FOR CAUTION: There is a possibility of selection bias due to the inclusion criteria for participants in this study: (1) completion of the last two surveys in a series of five and (2) consent to the linkage of their responses with Medicare data. GENERALIZABILITY TO OTHER POPULATIONS: The results are relevant to questionnaire-based research studies with infertile women in developed countries. STUDY FUNDING/COMPETING INTEREST(S): ALSWH is funded by the Australian Government Department of Health and Ageing. This research is funded by a National Health and Medical Research Council Centre of Research Excellence grant.

Exploiting medical hierarchies for concept-based information retrieval

Search technologies are critical to enable clinical sta to rapidly and e ectively access patient information contained in free-text medical records. Medical search is challenging as terms in the query are often general but those in rel- evant documents are very speci c, leading to granularity mismatch. In this paper we propose to tackle granularity mismatch by exploiting subsumption relationships de ned in formal medical domain knowledge resources. In symbolic reasoning, a subsumption (or `is-a') relationship is a parent-child rela- tionship where one concept is a subset of another concept. Subsumed concepts are included in the retrieval function. In addition, we investigate a number of initial methods for combining weights of query concepts and those of subsumed concepts. Subsumption relationships were found to provide strong indication of relevant information; their inclusion in retrieval functions yields performance improvements. This result motivates the development of formal models of rela- tionships between medical concepts for retrieval purposes.

Exploiting SNOMED CT concepts and relationships for clinical information retrieval : Australian e-Health Research Centre and Queensland University of Technology at the TREC 2012 Medical Track

The Australian e-Health Research Centre and Queensland University of Technology recently participated in the TREC 2012 Medical Records Track. This paper reports on our methods, results and experience using an approach that exploits the concept and inter-concept relationships defined in the SNOMED CT medical ontology. Our concept-based approach is intended to overcome specific challenges in searching medical records, namely vocabulary mismatch and granularity mismatch. Queries and documents are transformed from their term-based originals into medical concepts as defined by the SNOMED CT ontology, this is done to tackle vocabulary mismatch. In addition, we make use of the SNOMED CT parent-child `is-a' relationships between concepts to weight documents that contained concept subsumed by the query concepts; this is done to tackle the problem of granularity mismatch. Finally, we experiment with other SNOMED CT relationships besides the is-a relationship to weight concepts related to query concepts. Results show our concept-based approach performed significantly above the median in all four performance metrics. Further improvements are achieved by the incorporation of weighting subsumed concepts, overall leading to improvement above the median of 28% infAP, 10% infNDCG, 12% R-prec and 7% Prec@10. The incorporation of other relations besides is-a demonstrated mixed results, more research is required to determined which SNOMED CT relationships are best employed when weighting related concepts.

Dying in hospital: medical failure or natural outcomes?

The purpose of this study was to describe patterns of medical and nursing practice in the care of patients dying of oncological and hematological malignancies in the acute care setting in Australia. A tool validated in a similar American study was used to study the medical records of 100 consecutive patients who died of oncological or hematological malignancies before August 1999 at The Canberra Hospital in the Australian Capital Territory. The three major indicators of patterns of end-of-life care were documentation of Do Not Resuscitate (DNR) orders, evidence that the patient was considered dying, and the presence of a palliative care intention. Findings were that 88 patients were documented DNR, 63 patients' records suggested that the patient was dying, and 74 patients had evidence of a palliative care plan. Forty-six patients were documented DNR 2 days or less prior to death and, of these, 12 were documented the day of death. Similar patterns emerged for days between considered dying and death, and between palliative care goals and death. Sixty patients had active treatment in progress at the time of death. The late implementation of end-of-life management plans and the lack of consistency within these plans suggested that patients were subjected to medical interventions and investigations up to the time of death. Implications for palliative care teams include the need to educate health care staff and to plan and implement policy regarding the management of dying patients in the acute care setting. Although the health care system in Australia has cultural differences when compared to the American context, this research suggests that the treatment imperative to prolong life is similar to that found in American-based studies.

Graph-based concept weighting for medical information retrieval

This paper presents a graph-based method to weight medical concepts in documents for the purposes of information retrieval. Medical concepts are extracted from free-text documents using a state-of-the-art technique that maps n-grams to concepts from the SNOMED CT medical ontology. In our graph-based concept representation, concepts are vertices in a graph built from a document, edges represent associations between concepts. This representation naturally captures dependencies between concepts, an important requirement for interpreting medical text, and a feature lacking in bag-of-words representations. We apply existing graph-based term weighting methods to weight medical concepts. Using concepts rather than terms addresses vocabulary mismatch as well as encapsulates terms belonging to a single medical entity into a single concept. In addition, we further extend previous graph-based approaches by injecting domain knowledge that estimates the importance of a concept within the global medical domain. Retrieval experiments on the TREC Medical Records collection show our method outperforms both term and concept baselines. More generally, this work provides a means of integrating background knowledge contained in medical ontologies into data-driven information retrieval approaches.

Representation of assertions in clinical free-­text using SNOMED CT

Information retrieval (IR) by clinicians in the healthcare setting is critical for informing clinical decision-making. However, a large part of this information is in the form of free-text and inhibits clinical decision support and effective healthcare services. This makes meaningful use of clinical free-­text in electronic health records (EHRs) for patient care a difficult task. Within the context of IR, given a repository of free-­text clinical reports, one might want to retrieve and analyse data for patients who have a known clinical finding.

The role of perceived usefulness and attitude on electronic health record acceptance : an empirical investigation using response surface analysis

Information and communications technologies are a significant component of the healthcare domain and electronic health records play a major role within it. As a result, it is important that they are accepted en masse by healthcare professionals. How healthcare professionals perceive the usefulness of electronic health records and their attitudes towards them have been shown to have significant effects on their overall acceptance. This paper investigates the role of perceived usefulness and attitude on the intention to use electronic health records by future healthcare professionals using polynomial regression with response surface analysis. Results show that the relationship is more complex than predicted in prior research. The paper concludes that the predicting properties of the above determinants must be further investigated to clearly understand their role in predicting the intention to use electronic health records and in designing systems that are better adopted by healthcare professionals of the future.

Identify disorders in health records using Conditional Random Fields and Metamap: AEHRC at ShARe/CLEF 2013 eHealth Evaluation Lab Task 1

The Australian e-Health Research Centre (AEHRC) recently participated in the ShARe/CLEF eHealth Evaluation Lab Task 1. The goal of this task is to individuate mentions of disorders in free-text electronic health records and map disorders to SNOMED CT concepts in the UMLS metathesaurus. This paper details our participation to this ShARe/CLEF task. Our approaches are based on using the clinical natural language processing tool Metamap and Conditional Random Fields (CRF) to individuate mentions of disorders and then to map those to SNOMED CT concepts. Empirical results obtained on the 2013 ShARe/CLEF task highlight that our instance of Metamap (after ltering irrelevant semantic types), although achieving a high level of precision, is only able to identify a small amount of disorders (about 21% to 28%) from free-text health records. On the other hand, the addition of the CRF models allows for a much higher recall (57% to 79%) of disorders from free-text, without sensible detriment in precision. When evaluating the accuracy of the mapping of disorders to SNOMED CT concepts in the UMLS, we observe that the mapping obtained by our ltered instance of Metamap delivers state-of-the-art e ectiveness if only spans individuated by our system are considered (`relaxed' accuracy).

Principles of information accountability : an eHealth perspective

Information accountability is seen as a mode of usage control on the Web. Due to its many dimensions, information accountability has been expressed in various ways by computer scientists to address security and privacy in recent times. Information accountability is focused on how users participate in a system and the underlying policies that govern the participation. Healthcare is a domain in which the principles of information accountability can be utilised well. Modern health information systems are Internet based and the discipline is called eHealth. In this paper, we identify and discuss the goals of accountability systems and present the principles of information accountability. We characterise those principles in eHealth and discuss them contextually. We identify the current impediments to eHealth in terms of information privacy issues of eHealth consumers together with information usage requirements of healthcare providers and show how information accountability can be used in a healthcare context to address these needs. The challenges of implementing information accountability in eHealth are also discussed in terms of our efforts thus far.

De-Identification of health records using Anonym: Effectiveness and robustness across datasets

Objective Evaluate the effectiveness and robustness of Anonym, a tool for de-identifying free-text health records based on conditional random fields classifiers informed by linguistic and lexical features, as well as features extracted by pattern matching techniques. De-identification of personal health information in electronic health records is essential for the sharing and secondary usage of clinical data. De-identification tools that adapt to different sources of clinical data are attractive as they would require minimal intervention to guarantee high effectiveness. Methods and Materials The effectiveness and robustness of Anonym are evaluated across multiple datasets, including the widely adopted Integrating Biology and the Bedside (i2b2) dataset, used for evaluation in a de-identification challenge. The datasets used here vary in type of health records, source of data, and their quality, with one of the datasets containing optical character recognition errors. Results Anonym identifies and removes up to 96.6% of personal health identifiers (recall) with a precision of up to 98.2% on the i2b2 dataset, outperforming the best system proposed in the i2b2 challenge. The effectiveness of Anonym across datasets is found to depend on the amount of information available for training. Conclusion Findings show that Anonym compares to the best approach from the 2006 i2b2 shared task. It is easy to retrain Anonym with new datasets; if retrained, the system is robust to variations of training size, data type and quality in presence of sufficient training data.

eHealth-as-a-Service (eHaaS) : a data-driven decision making approach in Australian context

A commitment in 2010 by the Australian Federal Government to spend $466.7 million dollars on the implementation of personally controlled electronic health records (PCEHR) heralded a shift to a more effective and safer patient centric eHealth system. However, deployment of the PCEHR has met with much criticism, emphasised by poor adoption rates over the first 12 months of operation. An indifferent response by the public and healthcare providers largely sceptical of its utility and safety speaks to the complex sociotechnical drivers and obstacles inherent in the embedding of large (national) scale eHealth projects. With government efforts to inflate consumer and practitioner engagement numbers giving rise to further consumer disillusionment, broader utilitarian opportunities available with the PCEHR are at risk. This paper discusses the implications of establishing the PCEHR as the cornerstone of a holistic eHealth strategy for the aggregation of longitudinal patient information. A viewpoint is offered that the real value in patient data lies not just in the collection of data but in the integration of this information into clinical processes within the framework of a commoditised data-driven approach. Consideration is given to the eHealth-as-a-Service (eHaaS) construct as a disruptive next step for co-ordinated individualised healthcare in the Australian context.

eHealth and the modern society

The integration of Information and Communication Technologies (ICT) into healthcare processes “eHealth” is driving enormous change in healthcare delivery and productivity. The transformations empower patients and present opportunities for new synergies between healthcare professionals, clinical decision makers, policy makers and educators. Technologies that are directly driving changes include Tele-medicine, Electronic health records (EHR), Standards to ensure computer systems inter-operate, Decision support systems, Data mining and easy access to medical information. This workshop provides an introduction to key informatics initiatives in eHealth using real examples and suggests how applications can be applied to modern society.

Information and communication technologies transform the practice of medicine

Information and Communication Technologies are dramatically transforming Allopathic medicine. Technological developments including Tele-medicine, Electronic health records, Standards to ensure computer systems inter-operate, Data mining, Simulation, Decision Support and easy access to medical information each contribute to empowering patients in new ways and change the practice of medicine. To date, informatics has had little impact on Ayurvedic medicine. This tutorial provides an introduction to key informatics initiatives in Allopothic medicine using real examples and suggests how applications can be applied to Ayurvedic medicine.

Prognostic validity of 3-Minute Nutrition Screening (3-MinNS) in predicting length of hospital stay, readmission, cost of hospitalisation and mortality : a cohort study

Background: It is important to identify patients who are at risk of malnutrition upon hospital admission as malnutrition results in poor outcomes such as longer length of hospital stay, readmission, hospitalisation cost and mortality. The aim of this study was to determine the prognostic validity of 3-Minute Nutrition Screening (3-MinNS) in predicting hospital outcomes in patients admitted to an acute tertiary hospital through a list of diagnosis-related groups (DRG). Methods: In this study, 818 adult patients were screened for risk of malnutrition using 3-MinNS within 24 hours of admission. Mortality data was collected from the National Registry with other hospitalisation outcomes retrieved from electronic hospital records. The results were adjusted for age, gender and ethnicity, and matched for DRG. Results: Patients identified to be at risk of malnutrition (37%) using 3-MinNS had significant positive association with longer length of hospital stay (6.6 ± 7.1 days vs. 4.5 ± 5.5 days, p<0.001), higher hospitalisation cost (S$4540 ± 7190 vs. S$3630 ± 4961, p<0.001) and increased mortality rate at 1 year (27.8% vs. 3.9%), 2 years (33.8% vs. 7.2%) and 3 years (39.1% vs. 10.5%); p<0.001 for all. Conclusions: The 3-MinNS is able to predict clinical outcomes and can be used to screen newly admitted patients for nutrition risk so that appropriate nutrition assessment and early nutritional intervention can be initiated.