Sepsis is a severe medical condition caused by an inordinate immune response to an infection. Early detection of sepsis symptoms is important to prevent the progression into the more severe stages of the disease, which kills one in four it effects. Electronic medical records of 1492 patients containing 233 cases of sepsis were used in a clustering analysis to identify features that are indicative of sepsis and can be further used for training a Bayesian inference network. The Bayesian network was constructed using the systemic inflammatory response syndrome criteria, mean arterial pressure, and lactate levels for sepsis patients. The resulting network reveals a clear correlation between lactate levels and sepsis. Furthermore, it was shown that lactate levels may be predicative of the SIRS criteria. In this light, Bayesian networks of sepsis patients hold the promise of providing a clinical decision support system in the future.
展开▼