Validity of a clinical model to predict influenza in patients presenting with symptoms of lower respiratory tract infection in primary care

GRACE Consortium

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

Background. Valid clinical predictors of influenza in patients presenting with lower respiratory tract infection (LRTI) symptoms would provide adequate patient information and reassurance. Aim. Assessing the validity of an existing diagnostic model (Flu Score) to detect influenza in LRTI patients. Design and Setting. A European diagnostic study recruited 1801 adult primary care patients with LRTI-like symptoms existing ≤7 days between October and April 2007-2010. Method. History and physical examination findings were recorded and nasopharyngeal swabs taken. Polymerase chain reaction (PCR) for influenza A/B was performed as reference test. Diagnostic accuracy of the Flu Score (1× onset <48 hours + 2× myalgia + 1× chills or sweats + 2× fever and cough) was expressed as area under the curve (AUC), calibration slopes and likelihood ratios (LRs). Results. A total of 273 patients (15%) had influenza on PCR. The AUC of the Flu Score during winter months was 0.66 [95% CI (95% confidence internal) 0.63-0.70]. During peak influenza season, both influenza prevalence (24%) and AUC were higher [0.71 (95% CI 0.66-0.76], but calibration remained poor. The Flu Score assigned 64% of the patients as 'low-risk' (10% had influenza, LR - 0.6). About 12% were classified as 'high risk' of whom 32% had influenza (LR + 2.7). During peak influenza season, 60% and 14% of patients were classified as low and high risk, respectively, with influenza prevalences being 14% (LR - 0.5) and 50% (LR + 3.2). Conclusion. The Flu-Score attributes a small subgroup of patients with a high influenza risk (prevalence 32%). However, clinical usefulness is limited because this group is small and the association between predicted and observed risks is poor. Considerable diagnostic imprecision remains when it comes to differentiating those with influenza on clinical grounds from the many other causes of LRTI in primary care. New point of care tests are required that accurately, rapidly and cost effectively detect influenza in patients with respiratory tract symptoms in primary care.

Original languageEnglish (US)
Pages (from-to)408-414
Number of pages7
JournalFamily Practice
Volume32
Issue number4
DOIs
StatePublished - Aug 1 2015

Fingerprint

Respiratory Tract Infections
Human Influenza
Primary Health Care
Area Under Curve
Calibration
Point-of-Care Systems
Polymerase Chain Reaction
Chills
Sweat
Myalgia
Cough
Respiratory System
Physical Examination
Fever
History
Costs and Cost Analysis

Keywords

  • Cough
  • Diagnostic accuracy
  • Europe
  • Influenza
  • Lower respiratory tract infection
  • Primary health care

ASJC Scopus subject areas

  • Family Practice

Cite this

Validity of a clinical model to predict influenza in patients presenting with symptoms of lower respiratory tract infection in primary care. / GRACE Consortium.

In: Family Practice, Vol. 32, No. 4, 01.08.2015, p. 408-414.

Research output: Contribution to journalArticle

@article{2b41ff038508454295fc569a89b286a1,
title = "Validity of a clinical model to predict influenza in patients presenting with symptoms of lower respiratory tract infection in primary care",
abstract = "Background. Valid clinical predictors of influenza in patients presenting with lower respiratory tract infection (LRTI) symptoms would provide adequate patient information and reassurance. Aim. Assessing the validity of an existing diagnostic model (Flu Score) to detect influenza in LRTI patients. Design and Setting. A European diagnostic study recruited 1801 adult primary care patients with LRTI-like symptoms existing ≤7 days between October and April 2007-2010. Method. History and physical examination findings were recorded and nasopharyngeal swabs taken. Polymerase chain reaction (PCR) for influenza A/B was performed as reference test. Diagnostic accuracy of the Flu Score (1× onset <48 hours + 2× myalgia + 1× chills or sweats + 2× fever and cough) was expressed as area under the curve (AUC), calibration slopes and likelihood ratios (LRs). Results. A total of 273 patients (15{\%}) had influenza on PCR. The AUC of the Flu Score during winter months was 0.66 [95{\%} CI (95{\%} confidence internal) 0.63-0.70]. During peak influenza season, both influenza prevalence (24{\%}) and AUC were higher [0.71 (95{\%} CI 0.66-0.76], but calibration remained poor. The Flu Score assigned 64{\%} of the patients as 'low-risk' (10{\%} had influenza, LR - 0.6). About 12{\%} were classified as 'high risk' of whom 32{\%} had influenza (LR + 2.7). During peak influenza season, 60{\%} and 14{\%} of patients were classified as low and high risk, respectively, with influenza prevalences being 14{\%} (LR - 0.5) and 50{\%} (LR + 3.2). Conclusion. The Flu-Score attributes a small subgroup of patients with a high influenza risk (prevalence 32{\%}). However, clinical usefulness is limited because this group is small and the association between predicted and observed risks is poor. Considerable diagnostic imprecision remains when it comes to differentiating those with influenza on clinical grounds from the many other causes of LRTI in primary care. New point of care tests are required that accurately, rapidly and cost effectively detect influenza in patients with respiratory tract symptoms in primary care.",
keywords = "Cough, Diagnostic accuracy, Europe, Influenza, Lower respiratory tract infection, Primary health care",
author = "{GRACE Consortium} and {van Vugt}, {Saskia F.} and Broekhuizen, {Berna D.L.} and Zuithoff, {Nicolaas P.A.} and {van Essen}, {Gerrit A.} and Ebell, {Mark H.} and Samuel Coenen and Margareta Ieven and Christine Lammens and Herman Goossens and Butler, {Chris C.} and Kerenza Hood and Paul Littleg and Verheija, {Theo J.M.} and Peter Zuithoff and {van Essen}, Ted and Jordi Almirall and Francesco Blasi and Slawomir Chlabicz and Mel Davies and Maciek Godycki-Cwirko and Helena Hupkova and Janko Kersnik and Artur Mierzecki and Sigvard M{\"o}lstad and Michael Moore and Tom Schaberg and {de Sutter}, An and Antoni Torres and Pia Touboul and Paul Little and Theo Verheij",
year = "2015",
month = "8",
day = "1",
doi = "10.1093/fampra/cmv039",
language = "English (US)",
volume = "32",
pages = "408--414",
journal = "Family Practice",
issn = "0263-2136",
publisher = "Oxford University Press",
number = "4",

}

TY - JOUR

T1 - Validity of a clinical model to predict influenza in patients presenting with symptoms of lower respiratory tract infection in primary care

AU - GRACE Consortium

AU - van Vugt, Saskia F.

AU - Broekhuizen, Berna D.L.

AU - Zuithoff, Nicolaas P.A.

AU - van Essen, Gerrit A.

AU - Ebell, Mark H.

AU - Coenen, Samuel

AU - Ieven, Margareta

AU - Lammens, Christine

AU - Goossens, Herman

AU - Butler, Chris C.

AU - Hood, Kerenza

AU - Littleg, Paul

AU - Verheija, Theo J.M.

AU - Zuithoff, Peter

AU - van Essen, Ted

AU - Almirall, Jordi

AU - Blasi, Francesco

AU - Chlabicz, Slawomir

AU - Davies, Mel

AU - Godycki-Cwirko, Maciek

AU - Hupkova, Helena

AU - Kersnik, Janko

AU - Mierzecki, Artur

AU - Mölstad, Sigvard

AU - Moore, Michael

AU - Schaberg, Tom

AU - de Sutter, An

AU - Torres, Antoni

AU - Touboul, Pia

AU - Little, Paul

AU - Verheij, Theo

PY - 2015/8/1

Y1 - 2015/8/1

N2 - Background. Valid clinical predictors of influenza in patients presenting with lower respiratory tract infection (LRTI) symptoms would provide adequate patient information and reassurance. Aim. Assessing the validity of an existing diagnostic model (Flu Score) to detect influenza in LRTI patients. Design and Setting. A European diagnostic study recruited 1801 adult primary care patients with LRTI-like symptoms existing ≤7 days between October and April 2007-2010. Method. History and physical examination findings were recorded and nasopharyngeal swabs taken. Polymerase chain reaction (PCR) for influenza A/B was performed as reference test. Diagnostic accuracy of the Flu Score (1× onset <48 hours + 2× myalgia + 1× chills or sweats + 2× fever and cough) was expressed as area under the curve (AUC), calibration slopes and likelihood ratios (LRs). Results. A total of 273 patients (15%) had influenza on PCR. The AUC of the Flu Score during winter months was 0.66 [95% CI (95% confidence internal) 0.63-0.70]. During peak influenza season, both influenza prevalence (24%) and AUC were higher [0.71 (95% CI 0.66-0.76], but calibration remained poor. The Flu Score assigned 64% of the patients as 'low-risk' (10% had influenza, LR - 0.6). About 12% were classified as 'high risk' of whom 32% had influenza (LR + 2.7). During peak influenza season, 60% and 14% of patients were classified as low and high risk, respectively, with influenza prevalences being 14% (LR - 0.5) and 50% (LR + 3.2). Conclusion. The Flu-Score attributes a small subgroup of patients with a high influenza risk (prevalence 32%). However, clinical usefulness is limited because this group is small and the association between predicted and observed risks is poor. Considerable diagnostic imprecision remains when it comes to differentiating those with influenza on clinical grounds from the many other causes of LRTI in primary care. New point of care tests are required that accurately, rapidly and cost effectively detect influenza in patients with respiratory tract symptoms in primary care.

AB - Background. Valid clinical predictors of influenza in patients presenting with lower respiratory tract infection (LRTI) symptoms would provide adequate patient information and reassurance. Aim. Assessing the validity of an existing diagnostic model (Flu Score) to detect influenza in LRTI patients. Design and Setting. A European diagnostic study recruited 1801 adult primary care patients with LRTI-like symptoms existing ≤7 days between October and April 2007-2010. Method. History and physical examination findings were recorded and nasopharyngeal swabs taken. Polymerase chain reaction (PCR) for influenza A/B was performed as reference test. Diagnostic accuracy of the Flu Score (1× onset <48 hours + 2× myalgia + 1× chills or sweats + 2× fever and cough) was expressed as area under the curve (AUC), calibration slopes and likelihood ratios (LRs). Results. A total of 273 patients (15%) had influenza on PCR. The AUC of the Flu Score during winter months was 0.66 [95% CI (95% confidence internal) 0.63-0.70]. During peak influenza season, both influenza prevalence (24%) and AUC were higher [0.71 (95% CI 0.66-0.76], but calibration remained poor. The Flu Score assigned 64% of the patients as 'low-risk' (10% had influenza, LR - 0.6). About 12% were classified as 'high risk' of whom 32% had influenza (LR + 2.7). During peak influenza season, 60% and 14% of patients were classified as low and high risk, respectively, with influenza prevalences being 14% (LR - 0.5) and 50% (LR + 3.2). Conclusion. The Flu-Score attributes a small subgroup of patients with a high influenza risk (prevalence 32%). However, clinical usefulness is limited because this group is small and the association between predicted and observed risks is poor. Considerable diagnostic imprecision remains when it comes to differentiating those with influenza on clinical grounds from the many other causes of LRTI in primary care. New point of care tests are required that accurately, rapidly and cost effectively detect influenza in patients with respiratory tract symptoms in primary care.

KW - Cough

KW - Diagnostic accuracy

KW - Europe

KW - Influenza

KW - Lower respiratory tract infection

KW - Primary health care

UR - http://www.scopus.com/inward/record.url?scp=84939557472&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84939557472&partnerID=8YFLogxK

U2 - 10.1093/fampra/cmv039

DO - 10.1093/fampra/cmv039

M3 - Article

C2 - 26045544

AN - SCOPUS:84939557472

VL - 32

SP - 408

EP - 414

JO - Family Practice

JF - Family Practice

SN - 0263-2136

IS - 4

ER -