We sought to determine which of 3 methods used to evaluate cardiac index (CI) is the most accurate using focused bedside echocardiography (ECHO). We hypothesized that the fractional shortening (FS) method would provide a more accurate estimate of CI than the left ventricular outflow tract/velocity-time integral (LVOT/VTI) or Simpson's methods. This was a prospective observational cohort study conducted in the surgical ICU of an urban level 1 trauma center utilizing all patients with a pulmonary artery catheter (PAC) in place. Three surgical intensive care unit (SICU) faculty and 3 fellows underwent focused cardiac ultrasound training. Focused ECHO exams-bedside echocardiographic assessment in trauma/critical care (BEAT)- were performed using the Sonosite portable ultrasound device (Bothall, Washington). Stroke volume (SV) measurements were prospectively obtained on all trauma/SICU patients, with a PAC in place, using FS, LVOT/VTI, and Simpson's methods. The investigators were blinded to the PAC data. From each measurement, CI was calculated and categorized as low, normal, or high, based on a normal range of 2.4 to 4.0 L/min per m(2). Each CI obtained from the PAC was similarly categorized. The association between the BEAT and PAC estimates of CI was evaluated for each method using chi-square goodness of fit. Eighty five BEAT exams were performed on consecutive SICU patients, 56% were on trauma and 44% on emergency general surgery patients. There was a statistically significant association between the CI estimate using the FS method (P = .012), but not the LVOT/VTI (P = .33) or Simpson's method (P = .74). Our data showed a significant association between the PAC estimate of CI and our estimate using the FS method. The other methods were difficult to obtain, subjective, and inaccurate. Fractional shortening was the method of choice to estimate CI for the BEAT exam performed by intensivists in SICU patients.