Table 1

Author, date and country	Patient group	Study type (level of evidence)	Outcomes	Key results	Study weaknesses
PPV, positive predictive value; NPV, negative predictive value; LR+, positive likelihood ratio; LR-, negative likelihood ratio
Hopstaken et al, 2003, Netherlands	246 adult patients presenting to 25 GP’s with symptoms and signs consistent with lower respiratory tract infection.	Prospective diagnostic cohort study	Auscultation abnormality for diagnosis of pneumonia	Present in 84% of patients; OR 2.0 (95% CI 0.6–6.9). PPV 14.2%; NPV 92.3%	Interobserver reliability not assessed.
	Signs and symptoms recorded by GP on standard form; PA and lateral chest radiographs blindly interpreted by radiologist on day 3 formed the gold standard.		Accuracy of crackles for diagnosis of pneumonia	Present in 20.6% of patients; Odds ratio 1.5 (0.7–3.7); PPV 18.0%; NPV 87.6%	No sample size calculation. Wide confidence intervals suggest the study may be underpowered.
Osmer and Cole, 1966, USA	200 ‘random’ cases of young men admitted to hospital with radiographic evidence of acute pneumonia between September 1963 and August 1964.	Retrospective diagnostic cohort	Absence of auscultatory findings	Occurred in 50 (25%) of cases.	Retrospective analysis.
	Auscultatory findings recorded by “internists skilled in chest diseases” compared to chest radiograph results, as reported by three radiologists.		Absence of rales	Occurred in 98 (49%) of cases.	No sample size calculation.
			Auscultatory abnormality in same location as chest radiograph abnormality	Occurred in 52 (26%) of cases.	“Random” method for identification of cases not described.
					No mention of blinding.
					Only cases of pneumonia included (not an undifferentiated group).
Metlay et al, 1997, USA	Published studies of patients suspected of having pneumonia, which evaluated clinical signs for diagnostic accuracy (identified in Medline).	Systematic review	Interobserver reliability of chest signs	Kappa scores: crackles 0.41; wheezes 0.51; bronchial breath sounds 0.32. (Indicates only fair-moderate agreement for each).	Only Medline was searched.
			Any chest finding for diagnosis of pneumonia	LR+ 1.3, LR- 0.57 (95% CI 0.39–0.83) (p<0.05). This is insufficient to safely confirm or exclude pneumonia in practice.	No attempts to retrieve unpublished data.
	Gold standard was chest radiography. All studies were reviewed for quality.		Crackles for diagnosis of pneumonia	LR+ ranged from 1.6–2.7; LR- 0.78–0.87 (p<0.05)	No attempt to meta-analyse the data.
			Bronchial breath sounds for diagnosis of pneumonia	LR+ 3.5; LR- 0.90 (p<0.05)	Confidence intervals not always stated.
			Rhonchi for diagnosis of pneumonia	LR+ ranged from “not significant” - 1.5; LR–“not significant”–0.76	Exact p values not stated.
Wipf JE, et al, 1999 USA	54 patients with respiratory symptoms (cough and change in sputum) who presented to the Emergency Department.	Prospective, blinded diagnostic study	Sensitivity of overall clinical diagnosis	Range 47% to 69%	1. Small patient group.
		Chest exam performed by 2 (of 3) physicians (blinded to patient vital signs and history). Chest radiograph read by radiologist used as gold standard.	Specificity of overall clinical diagnosis	58% to 75%	2. Group characteristics: Entirely male, all late-middle aged, many with underlying respiratory/cardiac pathology.
			Multivariate analysis of rales (crackles) in presence of pneumonia	Odds Ratio 3.73	3. Long study period affecting standardisation between patients.
			Interrobserver reliability for clinical diagnosis of pneumonia	Paired kappa ranged from 0.18 to 0.32, indicating only fair agreement	4. Chest radiograph used as gold standard which can be insensitive and nonspecific.
Leuppi JD, et al, 2005, Switzerland	243 consecutive patients attending the Emergency Department with chest symptoms.	Prospective, double blinded study	Contribution of lung auscultation	Unchanged diagnosis after auscultation in 96.4% of cases	1. Chest symptoms were not described for initial recruitment.
		Diagnosis proposed before and after auscultation by a physician (with initial access to referring letter and patient history) and compared with the (seperate) diagnosis made on discharge letter.	Predictive value of normal lung auscultation for absence of lung or heart disease	OR 0.12 (95% CI 0.053–0.29)	2. Auscultatory findings were not described.
					3.Influence of patient’s history to the diagnosis proposed would contribute to findings despite attempts to control this.
					The study was not specific for pneumonia and so a cardiac history (eg chest pain with orthopnea) with crackles will give a very different impression to sputum with crackles.