A Clinical Prediction Rule for the Severe Acute Respiratory Syndrome

Abstract

Letters1 February 2005A Clinical Prediction Rule for the Severe Acute Respiratory SyndromeFREEMatthew H.M. Ma, MD, PhD, Shey-Ying Chen, MD, Wen-Chu Chiang, MD, Chan-Ping Su, MD, and Wen-Jone Chen, MD, PhDMatthew H.M. Ma, MD, PhDFrom National Taiwan University Hospital, Taipei, Taiwan.Search for more papers by this author, Shey-Ying Chen, MDFrom National Taiwan University Hospital, Taipei, Taiwan.Search for more papers by this author, Wen-Chu Chiang, MDFrom National Taiwan University Hospital, Taipei, Taiwan.Search for more papers by this author, Chan-Ping Su, MDFrom National Taiwan University Hospital, Taipei, Taiwan.Search for more papers by this author, and Wen-Jone Chen, MD, PhDFrom National Taiwan University Hospital, Taipei, Taiwan.Search for more papers by this authorAuthor, Article, and Disclosure Informationhttps://doi.org/10.7326/0003-4819-142-3-200502010-00019 SectionsAboutVisual AbstractPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinkedInRedditEmail TO THE EDITOR:We read with great interest the article by Leung and colleagues (1) on their clinical prediction rule for emergency department diagnosis of the severe acute respiratory syndrome (SARS). During the early period of the 2003 SARS epidemics in Taiwan, we prospectively developed a clinical decision rule from a cohort of febrile patients (2, 3), consisting of a 4-item symptom score and a 6-item clinical score (2). The prediction rule was adopted in our institution and became part of an integrated decision-making process for sorting incoming febrile patients during the epidemic. When validated by a second cohort, our rule showed a sensitivity of 90.2%, a specificity of 80.1%, and an area under the receiver-operating characteristic curve of 0.89 (3). We applied Leung and colleagues' rule to our cohort of 299 febrile patients, including 79 with laboratory-confirmed SARS and 220 without SARS. The sensitivity, specificity, positive predictive value, and negative predictive value were 98.8%, 52.0%, 43.6%, and 99.1%, respectively.The authors should be congratulated for a well-conducted study. However, some aspects of their prediction rule warrant further elaboration.First, the study cohort had a higher proportion of nonfebrile patients with SARS than has been reported elsewhere (4, 5). Because of the study's retrospective design, it is possible that some of these patients might not have had SARS to begin with but were instead cross-infected in hospital settings. This might introduce some misclassification bias and threaten the study's validity. Second, the clinical utility of the rule, if applied as suggested, may be limited in a large outbreak. More than 80% of patients still need hospitalization after triage by the rule. Housing patients with a 21% risk for SARS, quartile 1 in the high-risk group, in a communal isolation ward can be a dangerous practice.Triage of SARS during epidemics depends on the prevalence of the disease in the community and on local policies. Three pieces of information—contact history, fever, and pulmonary infiltrate—need to be considered. Patients with contact history and pulmonary infiltrates should be admitted regardless of body temperature. Afebrile patients without pulmonary infiltrates could be discharged home and receive daily follow-up of body temperature. Clinical decision rules are not needed for these patients. Febrile patients without pulmonary infiltrates are the most challenging for emergency department staff and should be the focus of a prediction rule. Whether to admit patients with pulmonary infiltrates but no contact history, however, is a policy issue. Well-thought-out clinical prediction rules, along with sound policies, would help communities tackle future SARS epidemics or similar outbreaks.Matthew H.M. Ma, MD, PhDShey-Ying Chen, MDWen-Chu Chiang, MDChan-Ping Su, MDWen-Jone Chen, MD, PhDNational Taiwan University Hospital; Taipei, Taiwan