Big data model can improve the prognosis for patients hospitalized

In the United States more than half of hospital deaths are directly related to serious infection or sepsis, recently researchers from Yale University and other institutions have developed a predictive model that can take advantage of big data to local patients, and the use of machine learning methods to help identify those with disease risk, this new method than the current method of clinical practice is better, the study published in the international Journal of Academic Emergency Medicine.

The current emergency room doctors can use a simple calculator or scoring system for clinical decision criteria to help determine which patients are more likely due to nosocomial sepsis and death, but these methods are often not successfully identify high-risk patients, because it is only based on limited information, which is not able to calculate the complexity of the data, only using different patient populations and developed.

In this paper, this new model developed by scientists on the use of large amounts of data from the local electronic health records of patients, this method called "random forest model" the data can be processed from a patient and analyzed to make certain prediction; this new big data analysis method is far better than the current model, and can potentially be classified in the 200-300 patients per 5,000 patients with severe sepsis.

Dr. R. Andrew Taylor pointed out that the use of machine learning techniques and incorporates a large number of mutations (more than 500 mutations), we can develop a new model to help better predict potential hospital mortality in patients with sepsis. The researchers hope this time late in the promotion of the use of large data analysis model, but can also help the patient in real-time monitoring, the researchers aim is to get the patient's data, and to develop new learning health system, namely the development of prediction of models to be applied to improve the health of patients being.