Computerinterpretable Case Definitions

An ideal case detection system would also support case finding during outbreak investigations. Case definitions (described in Chapter 3) are the basis for case-control studies and investigations of emerging diseases. A computer-interpretable case definition is a prerequisite for providing computer-support to case finding during investigations. As discussed in Chapter 3, case definitions are Boolean (logical) statement of findings (Figure 13.14).

Case definitions, as currently written, are not well suited for automation. The authors of the SARS case definition intended it for use by physicians and epidemiologists, not computers. The clause "findings of lower respiratory illness (e.g., cough, shortness of breath, difficulty breathing)" does not enumerate all findings of lower respiratory illness. A computer requires a complete enumeration of all findings that it should count as evidence of lower respiratory illness (e.g., cough, shortness of breath, difficulty breathing, wheezing, cyanosis, tachypnea, dullness to percussion, fremitus, whispered pectoriloquy, rales, and rhonchi). The findings would also have to be described more precisely. For example, a physician or an epidemiologist would not count chronic cough or cough associated with asthma as a finding of lower respiratory illness when applying this case definition, but a computer would (unless told otherwise). Note that it is difficult, if not impossible, to enumerate all of the possible exceptions to the counting of a finding as evidence of a disease. This difficulty is the reason that diagnostic expert systems in medicine are probabilistic. They quantify the number of exceptions to a categorical statement about the relationship between findings and disease using probabilities. For example, 70% of patients with cough have an acute lower respiratory illness, but 30% (the exceptions) have some other cause. This observation suggests that computer-interpretable case definitions will employ Bayesian networks, as illustrated by Figure 13.15.

For readers interested in the topic of knowledge representations for computer-interpretable case definitions, there is

Case definition for confirmed case of SARS-CoV disease: (early OR mild-to-moderate OR severe illness) AND laboratory confirmation

Early illness: two or more of the following findings: fever (might be subjective), chilis, rigors, myalgia, headache, diarrhea, sore throat, rhinorrhea.

Mild-to-moderate respiratory illness: Temperature of >100,4° F (>38° C) AND one or more findings of lower respiratory illness (e.g.- cough, shortness of breath,

Laboratory confirmation: serum antibody to SARS-CoV by a test validated by CDC (e.g.. enzyme immunoassay [EI A]), OR isolation in cell culture of SARS-CoV from a clinical specimen, OR SARS-CoV RNA by a reverse-transcription-polymerase chain reaction (RT-PCR) test validated by CDC and with subsequent confirmation in a figure 13.14 Excerpt from the CDC case definition for confirmed SARS-CoV disease (http://www.cdc.gov/ncidod/sars/guidance/b/app1.htm). Only the definitions relevant for classifying a patient as a probable case of SARS CoV are shown. (We omitted the definition of severe respiratory illness for clarity. The illness clause in the case definition for confirmed case of SARS-CoV disease is a disjunction, and all patients that satisfy the definitional criteria for severe disease also satisfy the criteria for mild to moderate respiratory illness.

figure 13.15 A Bayesian network "case definition" for anthrax.The network computes the posterior probability for a patient addmitted to the emergency department. Cxr order, electronic record of an order for a chest radiograph; wide med, chest radiograph finding of wide mediastinum; Gpr, gram-positive rods in blood or cerebrospinal fluid (CSF) smear; micro order, order for a blood or CSF culture. (From Espino J., Tsui, F.-C. (2000). A Bayesian network for detecting inhalational anthrax outbreaks. Pittsburgh, PA: Center for Biomedical Informatics, University of Pittsburgh, with permission.)

figure 13.15 A Bayesian network "case definition" for anthrax.The network computes the posterior probability for a patient addmitted to the emergency department. Cxr order, electronic record of an order for a chest radiograph; wide med, chest radiograph finding of wide mediastinum; Gpr, gram-positive rods in blood or cerebrospinal fluid (CSF) smear; micro order, order for a blood or CSF culture. (From Espino J., Tsui, F.-C. (2000). A Bayesian network for detecting inhalational anthrax outbreaks. Pittsburgh, PA: Center for Biomedical Informatics, University of Pittsburgh, with permission.)

literature from research on computer-interpretable patient care guidelines that is relevant to this topic (Shiffman et al., 2004, Tu and Musen, 2001, Peleg et al., 2003, Boxwala et al., 2004, Wang et al., 2003, Seyfang et al., 2002, Fox et al., 1997, Terenziani et al., 2003, de Clercq et al., 2004, Johnson et al., 2001, Ciccarese et al., 2004, Parker et al., 2004).

Swine Influenza

Swine Influenza

SWINE INFLUENZA frightening you? CONCERNED about the health implications? Coughs and Sneezes Spread Diseases! Stop The Swine Flu from Spreading. Follow the advice to keep your family and friends safe from this virus and not become another victim. These simple cost free guidelines will help you to protect yourself from the swine flu.

Get My Free Ebook


Post a comment