Questions

Below are questions that the organizers found interesting. Please use these samples as inspiration for the question(s) you want to explore!

In parenthesis, we've provided a link to a suggested dataset that you can use to answer the question. You may also wish to consider other datasets for another perspective or complementary information. For example, you may be able to join two datasets about doctors on their National Provider Identifier (NPI).

Describe the Disorder

1. Which companies made the most payments? (CMS Open Payments)
2. Which physicians received the most payments and why? Are there payment outliers? What drives the payments? Note: one physician can bill under another and outliers may be misleading; take a look at middle of distribution in addition to outliers. (CMS Open Payments)
3. Which states received the most payments? (CMS Open Payments)
4. What is the total amount of payments by type of physician. (CMS Open Payments)
5. How much ownership do physicians have in Manu/GPOs? (CMS Open Payments)
6. What dollar amount was disputed? Are any physicians/hospitals more likely to dispute than others? (CMS Open Payments)
7. What is the dollar amount of payments made through third parties? Are any physicians/hospitals more likely to pay through 3rd parties than others? (CMS Open Payments)

Which are the most costly chronic conditions, in terms of total dollar amount (avg payment * bene count) and aggregated by chronic condition? Are there any conditions and age group combinations that are expensive to treat? How do Part A and B differ? How do Part A and B interact? How much did payments increase from 2008 to 2010?

Model the Mess

1. What diseases have good media coverage and why? Find important words from the abstract and use the MeSH terms as features to build a prediction model of newsworthiness. Try to link diseases that have a good coverage in the media (e.g., in the top 50-100 most predictive features) to the healthcare cost. The article considers articles from September 2012 to present, so you could look at the results from a different time period(s) and compare. Are there trends with time? (Elhadad)
2. What features are predictive of the average medicare payment? (Chronic Conditions)

Clean the Clutter

1. Separate the CMS Open Payments data into two tables: one for recipients that are physicians and one for recipients that are teaching hospitals. (CMS Open Payments)
2. Properly normalize the data, create separate tables for physician info, Manu/GPO info, hospital info and pure transaction information. (CMS Open Payments)