Currently, the sources of medical data are mainly medical institutions (e.g., medical pharmacy laboratories, medical rehabilitation centers, etc.) and the Internet. The data collected is wide-ranging, high-dimensional, diverse, and not specific to a particular problem.

2. Uncertainty Measurement Issues

Most of the mature big data models that have entered the practical stage are for pharmaceutical companies and insurance companies. In the U.S., it is often difficult to find a suitable entry point for the business of doctors and patients in the application of big data in healthcare. Enterprise-oriented business is relatively easy, especially for insurance companies and pharmaceutical companies, while it is relatively difficult. Because of the limited accuracy of Big Data models, they are of limited practical value in safety-critical settings and for physicians. For example, a model that is 95% accurate may still not be accurate enough for a physician, who makes decisions on an individual patient basis, rather than on a statistical basis.

Also, statistical learning models are poorly interpretable, and often only statisticians and computer scientists can accurately and completely explain the model, while there are significant barriers to real users of the model, such as physicians and government officials.