Objective: To answer the following questions: Can the digital chest roentgenogram (CXR) be used to differentiate patients' volume status? Do clinical data alter radiologists' accuracy in interpreting the digital CXR?
Design: Prospective cohort study.
Setting: Nine adult intensive care units of a tertiary care medical center.
Patients: One hundred thirty-five consecutive patients with pulmonary artery catheters, of whom 35 were excluded because of unacceptable pulmonary artery occlusion pressure (PAOP) tracings.
Methods: Each patient had a portable, anteroposterior, supine digital CXR. Clinicians evaluated volume status and then measured hemodynamic data within 1 hr of the CXR. Digital CXRs were independently interpreted on two separate occasions (with and without clinical information) by three experienced chest radiologists, and these interpretations were compared with hemodynamic data.
Results: Of the 100 patients, 39 had PAOP >18 mm Hg, whereas 61 had PAOP <18 mm Hg. Radiologists' accuracy in differentiating volume status increased with incorporation of clinical data (56% without vs. 65% with clinical data, p =.009). Using objective receiver operating characteristic-derived cutoffs of 70 mm for vascular pedicle width and 0.55 for cardiothoracic ratio, radiologists' accuracy in differentiating PAOP >18 mm Hg from PAOP <18 mm Hg was 70%. The intrareader and the inter-reader correlation coefficients were very high. The likelihood ratio of the CXR in determining volume status using the objective vascular pedicle width and cardiothoracic ratio measures was 3.1 (95% confidence interval, 1.9-6.0), significantly higher than subjective CXR interpretations with and without clinical data (p <.001).
Conclusions: Differentiating intravascular volume status with portable, supine, digital CXRs may be improved by using objective cutoffs of vascular pedicle width >70 mm and cardiothoracic ratio >0.55 or by incorporating clinical data.