Primary objective: To assess and compare the functional state of the autonomic nervous system in healthy individuals with patients in coma using measures of heart rate variability (HRV), and to evaluate its efficiency in predicting mortality. Research design: Retrospective group comparison study of patients in coma, classified into two subgroups, according to their Glasgow coma score, with a healthy control group. Methods and procedures: HRV indices were calculated from 7 minutes of free of artifacts electrocardiograms using the Hilbert-Huang method in the spectral range 0.02-0.6 Hz. A special procedure was applied to avoid the effect of confounding factors. Stepwise multiple regression logistic analysis (SMLRA), followed by ROC curve analysis were applied to evaluate predictions. Results: Progressive reduction of HRV associated with deepening of coma was confirmed and a mortality score model including three spectral HRV indices: absolute power values of the very low, low frequency, and the power in normalized units of the very high frequency bands (0.4-0.6 Hz). The SMLRA model showed sensitivity of 95.65%, specificity of 95.83%, positive predictive value of 95.65%, and overall efficiency of 95.74%. Conclusions: HRV is a reliable method to assess the integrity of the neural control of the caudal brainstem centres upon the heart, of patients in coma, and predict mortality.