OBJECTIVE To correlate patient condition and reasons for obtaining chest radiographs (CXRs) with the utility of CXRs in critical illness and to determine the potential impact of stricter criteria for obtaining a CXR in a surgical intensive care unit (ICU). METHODS Inception cohort study of 1003 CXRs examined prospectively. METHODS A total of 157 consecutive patients admitted to the general surgical ICU of a 780-bed, urban, university-affiliated, tertiary care hospital. METHODS Nothing was done to influence the ordering of CXRs. METHODS Influence of CXR findings on clinical management. RESULTS The likelihood of a clinically important finding was 17% for CXRs obtained for no clear clinical indication (routine), 26% for those obtained to verify the position of a medical device, and 30% for those obtained for suspected clinical conditions. By univariate analysis, suspected pathophysiologic condition, admission APACHE II (Acute Physiology and Chronic Health Evaluation II) score, presence of a central venous or Swan-Ganz catheter, and length of ICU stay were all predictors of a significant finding. By multivariate analysis, the only independent predictor of a finding was a suspected clinical condition, and the only indwelling medical device that was an independent predictor of a finding was a Swan-Ganz catheter. If the criterion that routine CXRs should only be obtained in patients with Swan-Ganz catheters had been used, 200 CXRs would have been avoided during the 3-month study period. The only findings missed by not obtaining those CXRs would have been two malpositioned nasogastric tubes and one malpositioned central venous catheter. CONCLUSIONS Chest radiographs should only be obtained on surgical ICU patients for specific indications. Routine CXRs for ICU patients are justified only for patients with indwelling Swan-Ganz catheters.