A lot of rewards and more → and also have a medical degree → so good for the traditional educational system.
Good presentation skills.
Some of the algorithms are going to damage the healthcare industry → and we need to know this → and prevent this.
We need to have good data → a good kind of data.
Some of the methods → we are going to be classifying the images. (this is just one of the algorithm that is going to be coming in the industry).
But we are going to talk about some other methods → these methods are not a medical algorithm rather → it is more of a policy algorithm. (population health management). (health systems are incentivized to manage population → and this can damage patient).
If you have complex diseases → you are not going to do well → and bad outcomes will happen. (cuz both medical and policy complexity).
How can we improve this? → care coordination might be one of the answers. (someone can just call you and the problem will be solved → this is personalized medicine and more).
And these type of phone calls → really does work! → this is super cool. (these programs are good for the patient and more).
But these programs are very expensive → these are a lot of stuff → targeting is critical. (so we need some method of scale) → and it seems like there is an algorithm that targets patient → but who gets flagged may be biased.
These algorithms are going to be used → to generate predictions.
There are going to be a difference in racial ratio. (it is hard to get into data → this is not an easy task).
There is going to be some bias in any algorithm. (in this case who gets into the special program). (there is a risk score generated → for each patient).
Again, there is a lot of white people in the data → this will going to affect the overall problems. (black patients have the worst health than → white people → and this might be due to many reasons).
So → basically, black people who did not go into the program.
But we can use biomarkers.
Black people have more unstable blood pressure. (these types of bias is going to happen if we are not careful with the data collection stage).
The price difference between white and black.
There might be some economic reasons why this is happening → so who knows why this is happening. (this is not a simple problem → rather a very complex problem → that needs to be solved by multiple people).
Healthier whites → are taking the spot for the sick black people.
So the original algorithm is going to change → first we rank by health and other stuff are considered.
Now → more black patients are added! → less bias → this was an algorithm problem → that has been created due to data collection problem → and this is fundamentally an economic problem.
There is a cost for creating an algorithm → that optimizes accuracy. (these are a very complex problem → some data scientist → might → just optimize for accuracy) → but this might have some consequence.
Social policy is hard to solve.
The problem formulation → might cause a lot of problems in the long run → so it is a good idea to be very careful with this.
Even how the loss function is created → will cause a lot of different problems → and this can be related to a mathematical problem. (overfitting is a general problem → and this is a traditional machine learning problem).
So the company has started to work with this team → and they have got the team. (very good collaboration). (there are a couple of ways to remove racial bias → without collecting more data).