Reduce drug spend using medical claims analytics

Question

I have access to medical claims data from a US health insurance company.

I believe there's an opportunity to find some cost savings by switching the site of service from high cost outpatient facilities to possibly home infusion or physician for some procedures.

An example would be that Botox costs(in dollars) the company 10000 per claim at the outpatient facilities compared to 3000 at the physicians office.

I am interested in learning about how a data scientist would approach this problem of reducing costs from medical claims.

Could I run a cluster analysis on the population of claims to see who is most similar to the existing Botox patients who go to their physician and target these to move them to the lower cost sites of service?

Answer 1

Things depend upon the dataset we have on hand here. See what observations you can grab from them using combination of different features.

If i were to approach this problem, first i would find what kind of patients go to their physicians (Age, area of living, insurance plan, previous diseases, previous admissions, income zone etc). Then would analyse which outpatients facilities serves in which area, their average charges & other features you can find.

After the above analysis i would go with prediction algorithm so to make a theory that an particular set of groups will visit that facility. And using that deduction along with clustering (what i have in mind is 2 clusters 1 going to physicians & other with outpatient facilities) i will select the overlapping clients (I mean the ones who have features similar to lower cost going site but prefer higher cost sites) of insurance company to move towards lower cost site.

Reason behind using predictive model is because it will help you in predicting the future customers, just normal clustering will just work with the current data. So to make it effective for future i would go with predictive model -> clustering model.