One of the major problems that faces predictive medicine is the huge amounts of data it needs to be accurate. AI's involvement in medicine is still in its early stages and there are many cases where there is simply not enough data to create a reliable predictive algorithm.
The typical significance value for most statistical tests is 0.05. But for statistical tests conducted for medical purposes, a much lower signifance value is used - 0.000005. But is this enough? This is a question we will keep having to ask ourselves untill we can be fully sure that predictive and precision medicine algorithm provide reliable results.
Another challenge is the inconsistency of most data. There are many different companies making internet-connected medical device for hospitals around the world. Most of them format the data they collect in completely different ways. We will have to tackle this problem if we hope to build universal predictive medicine algorithms.
Another thing that concerns many patients is the fact that almost all of their medical data is being used by an algorithm on a computer that could be hacked. While this does present some security issues, most patients' medical data is already being stored on some computer. And unlike predictive medicine algortihms which don't require data about specific individuals, precision medicine does require personal health-related information. This makes data security an even more concerning problem when it is used by precision medicine algorithms. Much has been done to try to stop hackers from gaining confidential information but almost a hundred cybersecurity attacks were targeted at healthcare companies last year sparking even more concern.
One thing that concerns many physicians is the fact that predictive medicine cannot establish scientific reasons for its conclusions. This is a problem that will always exist when machine learning or AI is applied to anything.