Community-acquired pneumonia: physician judgment of need for chest radiograph helped diagnosis

Clinical bottom line (level 1b)

  1. One in fourteen patients with suspected pneumonia had it.
  2. A physician's prediction of the need for a chest X-ray made pneumonia slightly more likely (LR + 2.0) , and the prediction of no need for a chest X-ray slightly less likely (LR - 0.25) .
  3. No clinical diagnosis rule was signficantly better than a physician's prediction.
Emerman et al: Annals of Emergency Medicine 1991; 20 : 1215-1219
Expires December 2004

The study

Setting: emergency department and medical outpatient clinic of a teaching hospital, USA

290 patients (aged mean 39, 69% male) Adults presenting with recent history of acute cough or exacerbation of chronic cough plus either fever, sputum production or haemoptysis with onset within six weeks of presentation.

Excluded if
  • unable to give informed consent
  • pregnancy
  • acutely ill and unable to go to radiography for standard chest radiograph


Independent blinded reference standard, applied in all patients from a consecutive appropriate spectrum.
Reference standard:
  • Standard posteroanterior and lateral chest radiographic examination interpreted independently by two blinded radiologists. Disagreement was resolved by a third radiologist.
Diagnostic test: 39 physicians (15 attending physicians and 24 medical house staff) were asked whether or not they would ordinarily order a chest radiograph for the patient to diagnose pneumonia. Four decision aids were applied retrospectively.
  • Diehr et al aid - point score based on seven clinical variables, including temperature, respiratory rate, absence of rhinorrhea or sore throat, presence of night sweats, myalgia or sputum production throughout the day.
  • Signal et al - based on clinical variables including cough, fever and presence of crackles (threshold probability of 0.26 used to indicate needed for radiograph).
  • Heckerling et al - predictors of pneumonia of decreased breath sounds, absence of asthma, temperature >37.8 C, pulse rate of >100, or presence of rales (threshold of two or more).
  • Gennis et al - ordering chest radiographs only for patients with at least one abnormal vital sign, including pulse, respirations or temperature.

The evidence

pre-test probability of pneumonia: 7%, (95% CI: 4% to 10%)

diagnostic test pneumonia no pneumonia LR+
(95% CI)		 post-test probability LR-
(95% CI)		 post-test probability
physician orders chest X-ray 18 114 2.0
(1.6 to 2.5) 		14% 0.25
(0.087 to 0.71) 		2%
Gennis CDR orders chest X-ray 13 64 2.6
(1.7 to 3.9) 		17% 0.50
(0.29 to 0.87) 		4%
Diehr CDR orders chest X-ray 14 89 2.0
(1.4 to 2.9) 		14% 0.50
(0.27 to 0.92) 		4%
Heckerling CDR orders X-ray 15 88 2.2
(1.6 to 3.0) 		15% 0.42
(0.21 to 0.84) 		3%
Singal CDR orders X-ray 16 121 1.7
(1.3 to 2.2) 		12% 0.43
(0.20 to 0.94) 		3%
total 21 269

Comments

  1. Housestaff were not significantly worse than attending physicians for diagnosing pneumonia.

Citation

  1. Emerman CL, Dawson N, Speroff T, et al: Comparison of physician judgement and decision aids for ordering chest radiographs for pneumonia in outpatients. Annals of Emergency Medicine 1991; 20 : 1215-1219
Contributor: Clare Wotton, December 1999
Reviewer: Chrsi Ball

Clinical Question.
Patient pneumonia
Intervention or Exposure physician judgement
Comparison decision aids
Outcome ordering chest radiographs