Predicting patients' health and future costs with claims data and socio-economic data
The basic structure is:
- predict outcome only with claims diagnosis data - get error measure A - predict outcome with claims data and socio-economic data - get error measure B
See if error measure B < error measure A.
- Need to check if we require some sort of normalisation because model B uses more parameters.
We want the improvement to be at least a few percentage points. This is still vague.
Outcome variable is open at this stage. It will be easy to swap around.
- Next visit's health status - Time to next visit - Costs of next visit, next twelve months' cost
This is not a technical challenge, it's really depends on how well the data works.
Model is a time series model, as we have visits over time.
|✅||Get base data|
|✅||Get ACS data|
|✅||Decide on model structure|
|✅||Run model A|
|Run model B|