Dataset Information

Supervised methods to extract clinical events from cardiology reports in Italian.

ABSTRACT: Clinical narratives are a valuable source of information for both patient care and biomedical research. Given the unstructured nature of medical reports, specific automatic techniques are required to extract relevant entities from such texts. In the natural language processing (NLP) community, this task is often addressed by using supervised methods. To develop such methods, both reliably-annotated corpora and elaborately designed features are needed. Despite the recent advances on corpora collection and annotation, research on multiple domains and languages is still limited. In addition, to compute the features required for supervised classification, suitable language- and domain-specific tools are needed. In this work, we propose a novel application of recurrent neural networks (RNNs) for event extraction from medical reports written in Italian. To train and evaluate the proposed approach, we annotated a corpus of 75 cardiology reports for a total of 4365 mentions of relevant events and their attributes (e.g., the polarity). For the annotation task, we developed specific annotation guidelines, which are provided together with this paper. The RNN-based classifier was trained on a training set including 3335 events (60 documents). The resulting model was integrated into an NLP pipeline that uses a dictionary lookup approach to search for relevant concepts inside the text. A test set of 1030 events (15 documents) was used to evaluate and compare different pipeline configurations. As a main result, using the RNN-based classifier instead of the dictionary lookup approach allowed increasing recall from 52.4% to 88.9%, and precision from 81.1% to 88.2%. Further, using the two methods in combination, we obtained final recall, precision, and F1 score of 91.7%, 88.6%, and 90.1%, respectively. These experiments indicate that integrating a well-performing RNN-based classifier with a standard knowledge-based approach can be a good strategy to extract information from clinical text in non-English languages.

SUBMITTER: Viani N

PROVIDER: S-EPMC6948016 | biostudies-literature | 2019 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Supervised methods to extract clinical events from cardiology reports in Italian.

Viani Natalia N Miller Timothy A TA Napolitano Carlo C Priori Silvia G SG Savova Guergana K GK Bellazzi Riccardo R Sacchi Lucia L

Journal of biomedical informatics 20190528

Clinical narratives are a valuable source of information for both patient care and biomedical research. Given the unstructured nature of medical reports, specific automatic techniques are required to extract relevant entities from such texts. In the natural language processing (NLP) community, this task is often addressed by using supervised methods. To develop such methods, both reliably-annotated corpora and elaborately designed features are needed. Despite the recent advances on corpora colle ...[more]

PMID: 31150777

Similar Datasets

Project description:Background: Identifying individuals at heightened cardiovascular risk is a priority for reducing the global burden of cardiovascular disease. Aspirin is widely used to prevent cardiovascular events, though with variable results. Therefore, we hypothesized that aspirin exposure would reveal novel biological pathways relevant to the development of cardiovascular events. Methods: We administered aspirin, followed by peripheral blood RNA microarray profiling, in a discovery cohort of healthy volunteers (n = 50, HV1), followed by two validation cohorts of healthy volunteers (n = 53, HV2) or outpatient cardiology (OPC, n = 25) patients, in conjunction with platelet function testing with the platelet functions score (PFS, HV1 and HV2) or the VerifyNow Asprin (VN, OPC) test. Sets of coexpressed genes, or “Factors” were identified via Bayesian sparse factor analysis and associated with platelet function in HV1 and validated in HV2 and OPC. Validated factors were associated with death/MI in observational (n = 191) and case:control (n = 447) patient cohorts with available RNA data collected at the time of cardiac catheterization. Results: Factor analysis yielded 20 Factors, of which one, Factor 14, contained 60 genes and was associated with PFS in HV1 (r = -0.31, p-value = 0.03). Factor 14 was associated with platelet function with the same strength and direction in HV2 (r = -0.34, p-value = 0.02) and OPC (one-sided p-value for aspirin resistant vs. aspirin sensitive = 0.046), thus validating the association. Factor 14 was associated with death/MI in the two patient cohorts, odds ratio (OR) = 1.2, 95% CI [1.02-1.4], p-value = 0.01 and hazard ratio = 1.5, [1.2-1.9], p = 0.001, respectively, independent of known cardiovascular risk factors (combined OR = 1.2, CI = [1.02, 1.4], p = 0.03). Factor 14 and the expression of the Factor 14 transcript most highly correlative of PFS, ITGA2B, improved reclassification compared to traditional risk factors (category-free net reclassification index = 31% and 37%, p ≤ 0.0002 for both). Conclusions: By challenging humans subjects with aspirin, a medication used for cardiovascular risk reduction, we elucidated genes and pathways that may underlie platelet function and mechanisms responsible for cardiovascular death/MI. This accession represents the OPC cohort microarray data

Project description:BackgroundThere is growing interest in the impact of national clinical audit programmes on the quality of healthcare. There is also an evolving evidence-base for enhancing the design and delivery of audit and feedback. We assessed the extent to which a sample of UK national clinical audit feedback reports met a set of good practice criteria over three time points.MethodsWe undertook three cross-sectional content analyses. We developed good practice criteria for the content and delivery of feedback based upon evidence, behavioural theory and expert opinion. We applied these to a feedback reports from 23 national audits listed on the Healthcare Quality Improvement Partnership (HQIP) website in November 2015. We repeated our assessments in January 2017 for 20 repeat feedback reports, after HQIP had published reporting guidance for national audits, and in August 2019 for a further 14 repeat feedback reports. We verified our assessments, where possible, with audit leads.ResultsFeedback reports consistently included strengths at baseline, including past or planned repeated audit cycles (21; 91%), stating the importance of the topic in relation to patient care (22; 93%), using multi-modal data presentation (23; 100%), and summarising key findings (23; 100%). We observed improvements over subsequent assessments, so that by 2019, at least 13 out of 14 (93%) feedback reports presented easily identifiable key findings and recommendations, linked recommendations to audit standards, and proposed easily identifiable action plans. Whilst the use of regional comparators did not improve, audit leads highlighted that programmes now provide local data via additional means. The main shortcoming was the time lag between data collection and feedback; none of the 14 reports assessed in 2019 presented performance data less than 6 months old. Audit leads highlighted that some of these data might be available via programme websites.ConclusionWe identified increased adherence to good practice in feedback by national clinical audit programmes that may enhance their impact on service delivery and outcomes. There is scope for improvement, especially in the recency of performance data. With further refinements, a criterion-based assessment offers an efficient means of monitoring the quality of national clinical audit feedback reports.

Dataset Information

Supervised methods to extract clinical events from cardiology reports in Italian.

Publications

Supervised methods to extract clinical events from cardiology reports in Italian.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets