Objective: To test the hypothesis that heart rate variability (HRV) can provide an early indication of illness severity among patients presenting to the emergency department (ED) with sepsis. Methods: The authors enrolled a convenience sample of 15 ED patients meeting the American College of Chest Physicians/Society of Critical Care Medicine criteria for sepsis. Each patient had continuous Holter monitoring performed in the ED. Acute Physiology and Chronic Health II (APACHE II) and Sequential Organ Failure (SOFA) scores were calculated for the day of presentation. Holter tapes obtained in the ED were analyzed off-line to calculate HRV variables for the 5-minute segment with the least artifact and non-sinus beats. These variables were correlated with APACHE II and SOFA scores. Results: LFnu (normalized low-frequency power), an assessment of the relative sympathetic contribution to overall HRV, was correlated with increased illness severity as calculated using APACHE II (r = Ϫ0.67, r 2 = 0.43) and SOFA (r = Ϫ0.80, r 2 = 0.64) scores. LF/HF ratio (low-frequency/high-frequency ratio), a measure of sympathovagal balance, was correlated with the SOFA score [r = Ϫ0.54 (95% CI = Ϫ0.83 to Ϫ0.01), r 2 = 0.29]. All five patients who required critical care monitoring or ventilatory support or who died during the first 5 days of their hospitalization had LFnu values below 0.5 and LF/HF ratios less than 1.0. None of the patients with measurements greater than these threshold values died or required these interventions during the five days following admission. Conclusions: A single variable, LFnu, which reflects sympathetic modulation of heart rate, accounted for 40-60% of the variance in illness severity scores among patients presenting to the ED with sepsis. HRV, as reflected in LFnu and the LF/HF ratio and measured with a single brief (5-minute) period of monitoring while in the ED, may provide the emergency physician with a readily available, noninvasive, early marker of illness severity. The threshold effect of LFnu and LF/HF in the prediction of early clinical deterioration was an unexpected finding and should be regarded as hypothesis-generating, pending further study.