Original Research
Performance Characteristics of Clinical Diagnosis, a Clinical Decision Rule, and a Rapid Influenza Test in the Detection of Influenza Infection in a Community Sample of Adults

https://doi.org/10.1016/j.annemergmed.2005.05.020Get rights and content

Study objective

The accurate diagnosis of influenza remains a diagnostic dilemma. We examine the performance of various strategies for diagnosing influenza infection in an unselected sample of adults during influenza season.

Methods

Consecutive adults presenting to a university emergency department or urgent care clinic between January and March 2002 with acute respiratory complaints were eligible for this prospective observational study. The performance of clinician judgment, a rapid influenza test, and a clinical prediction rule in predicting influenza infection was evaluated using referent standard of reverse transcriptase polymerase chain reaction. Statistical significance was assessed using McNemar's test of proportions.

Results

Fifty-three of 258 (21%) patients had a positive influenza reverse transcriptase polymerase chain reaction test. Overall, clinician judgment showed sensitivity of 29% (95% confidence interval [CI] 18% to 43%) and specificity of 92% (95% CI 87% to 95%). The rapid influenza test showed a sensitivity of 33% (95% CI 22% to 47%) and specificity of 98% (95% CI 96% to 99%). The clinical prediction rule showed a sensitivity of 40% (95% CI 27% to 54%) and specificity of 92% (95% CI 87% to 95%). Clinician judgment when patients presented within 48 hours showed a sensitivity of 67% (95% CI 39% to 86%) and specificity of 96% (95% CI 81% to 99%). Neither the rapid influenza test (P=.10) nor the clinical prediction rule (P=.42) was superior to clinician judgment alone in the diagnosis of influenza.

Conclusion

The suggestion that a clinical decision rule or a rapid influenza test is better than clinical judgment alone for the diagnosis of influenza in an unselected patient population is not supported by this study.

Introduction

The annual influenza epidemic and the continued threat of an impending pandemic constitute major infectious disease concerns worldwide.1, 2, 3 Ten percent to 20% of US residents contract influenza annually, accounting for an average of 36,000 deaths and more than 120,000 hospitalizations per year.2 Approval of neuraminidase inhibitors prompted a renewed interest in the accurate diagnosis of influenza so that patients who might benefit from such therapy could be treated promptly and appropriately.4, 5 In addition, with antigenic shifts, including the threat of avian flu, the recent shortages of influenza vaccine, and an aging patient population, physicians can expect more patients with suspected influenza in acute care settings.

Historically, influenza has been a clinical diagnosis. The clinical identification of influenza is improved when physicians are aware that influenza virus is circulating in their geographic area.6 Several trials have attempted to analyze specific symptoms or signs, but no single clinical finding has consistently shown a high enough positive likelihood ratio to clinically diagnose influenza or a low enough negative likelihood ratio to eliminate it.7 Few studies have quantified the performance characteristics of various combined diagnostic strategies. Monto et al8 reported that the presence of cough and objective fever had the best positive predictive value (79%) for culture-proven influenza during influenza seasons. However, this study was conducted in a population that had a baseline prevalence of influenza of 69%, higher than one would expect in a general outpatient population. In their conclusion, the authors suggest that when influenza is circulating, patients with an influenza-like illness who have both cough and fever within 48 hours of symptom onset are likely to have influenza and that the administration of antiviral therapy may be appropriate. However, selection bias limits this study's results because the study population included only patients who fulfilled prespecified clinical criteria for influenza, including fever or symptom of fever, symptom onset within 48 hours, and those who agreed to be randomized to zanamivir treatment. Boivin et al,9 in a study unrelated to the neuraminidase trials, also found that cough and fever were the only clinical features significantly associated with influenza infection, although, again, the study population was also restricted to patients meeting entry criteria similar to those in the study by Monto et al.8

Because selection and spectrum bias threaten the results of the previous studies, the present study was conducted to assess the test characteristics of the Monto et al8 “cough and fever” clinical prediction rule compared with clinician judgment and rapid influenza testing, using a more unselected community-based patient population seeking care for acute respiratory illness. We sought to assess the performance characteristics of these diagnostic methods in a population more representative of the patient spectrum in everyday practice.

Section snippets

Setting and Selection of Participants

Consecutive adults (age >18 years) seeking care at the University of California, San Francisco emergency department (ED) or urgent care ambulatory clinic with symptoms of an acute respiratory tract infection were eligible for participation in this study. The University of California, San Francisco is a large tertiary care university hospital. The ED treats approximately 40,000 patients per year, with a 24% admission rate, and is staffed by 23 attending physicians who are all board certified in

Results

Between January and March 2002, a total of 405 patients were eligible for the study. Seventy-seven patients refused to participate, 12 were excluded for inability to consent, 25 persons were excluded for taking antibiotics for concurrent infections, and 12 patients were eligible but left without being treated by a physician. Because this study had a dual purpose of examining c-reactive protein levels in adults with acute respiratory illness, 13 patients were excluded because of conditions that

Limitations

This study's results should be interpreted in light of its limitations. First, it is possible that our c-reactive protein exclusion criteria caused our population to be biased. However, we excluded only 13 of 405 (3%) of our patient population for this conflict. Nonetheless, we may still have had selection bias in our patient population because of enrolling patients only during daytime hours, although we have no reason to suspect this.

Unfortunately, we did not reliably collect information on

Discussion

Influenza is a highly morbid illness affecting more than 10% of the population annually and accounting for more than 100,000 hospitalizations per year in the United States.12 The societal impact during outbreaks in terms of medical costs, as well as associated costs from missed workdays and reduced productivity, is substantial.13 Several recent events have made the accurate diagnosis of influenza particularly relevant. First, the enthusiasm generated by the publication of several large

References (17)

  • S. Flanders et al.

    Performance of a bedside c-reactive protein test in the diagnosis of community-acquired pneumonia in adults with acute cough illness

    Am J Med

    (2004)
  • N. Cox et al.

    Influenza

    Lancet

    (1999)
  • R. Couch

    Prevention and treatment of influenza

    N Engl J Med

    (2000)
  • W.W. Thompson et al.

    Mortality associated with influenza and respiratory syncytial virus in the united states

    JAMA

    (2003)
  • T. Hampton

    Government drafts flu preparedness plan: concerns about serious pandemic spur effort

    JAMA

    (2004)
  • A.S. Monto et al.

    Zanamivir in the prevention of influenza among healthy adults: a randomized controlled trial

    JAMA

    (1999)
  • F.G. Hayden et al.

    Use of the selective oral neuraminidase inhibitor oseltamivir to prevent influenza

    N Engl J Med

    (1999)
  • N.J. Montalto

    An office-based approach to influenza: clinical diagnosis and laboratory testing

    Am Fam Physician

    (2003)
There are more references available in the full text version of this article.

Cited by (77)

  • Influenza and Viral Pneumonia

    2024, Infectious Disease Clinics of North America
  • Influenza and Viral Pneumonia

    2018, Clinics in Chest Medicine
  • Point-of-Care Testing for Group A Streptococcus Infection and Influenza

    2017, Clinical Microbiology Newsletter
    Citation Excerpt :

    Instead, clinical judgment has been recommended in these scenarios for diagnosing influenza virus infection. One study performed in a large urban emergency department and urgent care ambulatory clinic demonstrated that a physician's clinical judgment (sensitivity, 29%; specificity, 93%) was no better than the performance of an RIDT lateral-flow immunoassay (sensitivity, 33%; specificity, 98%), suggesting that the use of clinical judgment to make decisions to treat patients for influenza is still a poor replacement for an in-laboratory molecular assay [22]. In a computer simulation model study evaluating the potential economic value of several diagnostic strategies for influenza, it was found that clinical judgment, followed by PCR and POC testing, was most cost-effective, given high influenza probability [24].

View all citing articles on Scopus

Supervising editor: David A. Talan, MD

Michael L. Callaham, MD, recused himself from the decisionmaking process.

Author contributions: JS, SF, and RG conceived the study and designed the trial. RG obtained research funding. JS, SF, and RG supervised the conduct of the trial and data collection. JM and RG provided statistical advice on study design and analyzed the data. WLD provided laboratory services. JL and JKH provided RT-PCR data and reviewed the aspects of the manuscript related to these issues. JS drafted the manuscript, and all authors contributed substantially to its revision. JS and RG take responsibility for the paper as a whole.

Funding and support: This study was supported by the Robert Wood Johnson Minority Medical Faculty Development Program (RG). The primary study sponsor, The Robert Wood Johnson Foundation, played no role in any aspects of the conduct of the study or manuscript preparation and submission. Quidel Corporation, the manufacturer of the rapid influenza test used in this study, provided free test kits. Quidel played no role in any aspects of the conduct of the study or manuscript development.

Presented at the Society for Academic Emergency Medicine annual meeting, May 2004, Orlando, FL.

Reprints not available from the authors.

View full text