The purpose of this study was to find out differences between septicemic and healthy premature newborns from the respiratory signal. An statistical analysis of breathing trace's temporal features showed significant differences between the inspiration/expiration ratio, being its average value shorter in healthy patients. All patient's total respiratory cycle time exceeding the apnea threshold were classified, and two different distributions had been found: log-normal for non septicemic and GEV (Generalized Extreme Value) distribution for septicemic. Finally, assessing fractal signal properties, we found an increased Hurst exponent (H) in non infected population. Results suggest that respiratory parameters could be helpful in sepsis diagnosis together with heart rate variability (HRV), demonstrated a relevant marker by our research team.