Fever detection from free-text clinical records for biosurveillance

RODS Laboratory, Center for Biomedical Informatics, University of Pittsburgh, Pittsburgh, PA, USA.
Journal of Biomedical Informatics (Impact Factor: 2.19). 05/2004; 37(2):120-7. DOI: 10.1016/j.jbi.2004.03.002
Source: DBLP


Automatic detection of cases of febrile illness may have potential for early detection of outbreaks of infectious disease either by identification of anomalous numbers of febrile illness or in concert with other information in diagnosing specific syndromes, such as febrile respiratory syndrome. At most institutions, febrile information is contained only in free-text clinical records. We compared the sensitivity and specificity of three fever detection algorithms for detecting fever from free-text. Keyword CC and CoCo classified patients based on triage chief complaints; Keyword HP classified patients based on dictated emergency department reports. Keyword HP was the most sensitive (sensitivity 0.98, specificity 0.89), and Keyword CC was the most specific (sensitivity 0.61, specificity 1.0). Because chief complaints are available sooner than emergency department reports, we suggest a combined application that classifies patients based on their chief complaint followed by classification based on their emergency department report, once the report becomes available.

Download full-text


Available from: Wendy Chapman
  • Source
    • "The approaches range from hand-written rule-based systems to fully automated methods using machine learning. For example, Chapman et al. (2004) use heuristical keyword-driven as well as supervised machine learning techniques (Naïve-Bayes classifier) for detecting mentions of fever in free-text clinical records. Similarly, the BioCaster system (Collier et al., 2006; Collier et al., 2008) relies on a carefully constructed medical ontology combined with a Naïve-Bayes classifier as an input filter. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Public health surveillance systems rely on the automated monitoring of large amounts of text. While building a text mining system for veterinary syndromic surveillance, we exploit automatic and semi-automatic methods for terminology construction at different stages. Our approaches include term extraction from free-text, grouping of term variants based on string similarity, and linking to an existing medical ontology.
    Full-text · Conference Paper · Jan 2015
  • Source
    • "In addition, systems use varied definitions for influenza or influenza-like illness (ILI), and depending on the source of information, symptoms, physical findings or laboratory tests included could underestimate or more accurately predict those with confirmed influenza[21], [22], [23], [24], [25], [26]. Although other approaches such as natural language processing free-text extraction or utilization of over-the-counter pharmaceutical sales, number of emergency room visits, absenteeism and triage telephone calls appear to improve the timeliness of detection, no strong conclusions could be made as to the best data indicator[27], [28]. The 2008–09 influenza season was the first flu season in which VA utilized the ESSENCE system for weekly influenza monitoring and surveillance. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The 2008-09 influenza season was the time in which the Department of Veterans Affairs (VA) utilized an electronic biosurveillance system for tracking and monitoring of influenza trends. The system, known as ESSENCE or Electronic Surveillance System for the Early Notification of Community-based Epidemics, was monitored for the influenza season as well as for a rise in influenza cases at the start of the H1N1 2009 influenza pandemic. We also describe trends noted in influenza-like illness (ILI) outpatient encounter data in VA medical centers during the 2008-09 influenza season, before and after the recognition of pandemic H1N1 2009 influenza virus. We determined prevalence of ILI coded visits using VA's ESSENCE for 2008-09 seasonal influenza (Sept. 28, 2008-April 25, 2009 corresponding to CDC 2008-2009 flu season weeks 40-16) and the early period of pandemic H1N1 2009 (April 26, 2009-July 31, 2009 corresponding to CDC 2008-2009 flu season weeks 17-30). Differences in diagnostic ICD-9-CM code frequencies were analyzed using Chi-square and odds ratios. There were 649,574 ILI encounters captured representing 633,893 patients. The prevalence of VA ILI visits mirrored the CDC's Outpatient ILI Surveillance Network (ILINet) data with peaks in late December, early February, and late April/early May, mirroring the ILINet data; however, the peaks seen in the VA were smaller. Of 31 ILI codes, 6 decreased and 11 increased significantly during the early period of pandemic H1N1 2009. The ILI codes that significantly increased were more likely to be symptom codes. Although influenza with respiratory manifestation (487.1) was the most common code used among 150 confirmed pandemic H1N1 2009 cases, overall it significantly decreased since the start of the pandemic. VA ESSENCE effectively detected and tracked changing ILI trends during pandemic H1N1 2009 and represents an important temporal alerting system for monitoring health events in VA facilities.
    Full-text · Article · Mar 2010 · PLoS ONE
  • Source
    • "Symptoms , diseases, mechanisms of injury, and other medical or non-medical concepts are commonly seen in CCs. ED CCs are a popular data source used by many syndromic surveillance systems because of their timeliness and availability [3] [4] [5] [6] [7]. CCs are among the first data elements collected for any ED visit and many hospitals increasingly have free-text CCs available in electronic form. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Emergency department free-text chief complaints (CCs) are a major data source for syndromic surveillance. CCs need to be classified into syndromic categories for subsequent automatic analysis. However, the lack of a standard vocabulary and high-quality encodings of CCs hinder effective classification. This paper presents a new ontology-enhanced automatic CC classification approach. Exploiting semantic relations in a medical ontology, this approach is motivated to address the CC vocabulary variation problem in general and to meet the specific need for a classification approach capable of handling multiple sets of syndromic categories. We report an experimental study comparing our approach with two popular CC classification methods using a real-world dataset. This study indicates that our ontology-enhanced approach performs significantly better than the benchmark methods in terms of sensitivity, F measure, and F2 measure.
    Full-text · Article · May 2008 · Journal of Biomedical Informatics
Show more