Predicting pneumonia in primary care: The 3C cohort study of lower respiratory tract infection in primary care

Talk Code: 
Michael Moore
Stuart B, Smith S, Thompson MJ, Knox K, van den Bruel A, Lown, M, Little P, Mant D
Author institutions: 
University of Southampton, University of Oxford, University of Washington


Lower respiratory tract infection (LRTI) is one of the commonest acute infections presenting in primary care. Antibiotics are frequently prescribed although there is little evidence to support symptomatic benefit. One of the drivers of continued prescribing is concerns over missed pneumonia and evidence from the GRACE trial suggests that primary care physicians miss two thirds of radiographic pneumonia.


Adult patients presenting in UK general practice with lower respiratory tract infections had symptoms, signs and treatment recorded. Participants were followed-up for 30 days to determine clinical outcome. In those receiving chest x-ray within the 30 day window reports were reviewed and assigned a pneumonia/non-pneumonia diagnosis. The predictive value of patient characteristics, presenting symptoms, and clinical findings for the diagnosis of pneumonia was assessed in the entire cohort using reported findings from those x-rayed within one week of the index consultation.


A cohort of 28867 adult patients with acute cough was recruited with informed consent by 522 practices between October 2009 and April 2013. 1485/28867 (5%) patients were x-rayed within 30 days of the initial consultation and 718 (2.5.%) within one week of the index consultation and of these 106 (14.8%) were assigned a pneumonia diagnosis. Patient characteristics were not helpful in predicting the probability of pneumonia. In contrast clinical symptoms and signs provided predictive information. The significant independent predictors of X ray confirmed pneumonia within one week of consultation (in order of strength of association) are temperature >37.8 degrees, crackles on auscultation, pulse >100/min, and oxygen saturation <95%. Using the group x-rayed within 30 days rather than within one week gave the same set of variables with the addition of chills and absence of coryza. A prediction model based on patients x-rayed within one week and including the four significant variables had an AUC of 0.67 (0.62,0.73)In the whole cohort using the four factor model 40% had one of the four clinical factors sensitivity 84% specificity 49% PPV 0.6% NPV 99.9% and 9% had any two clinical factors present with sensitivity 42% specificity 89% PPV 1.4% NPV 99.8%.


In routine practice x-ray confirmed pneumonia as a short term complication of LRTI is unusual (0.37%). The absence of key clinical signs has a high negative predictive value for pneumonia. The utility of the model alone may be limited by its positive predictive value.

Submitted by: 
Michael Moore
Funding acknowledgement: 
NIHR Programme Grant