Healthcare Fraud Prevention - How to Represent Studies in Software

The Healthcare Fraud Prevention Partnership (HFPP) is an initiative to combat fraud in healthcare reimbursements. TopCoder is building a data exchange network for this initiative. The partners are healthcare insurers who will share data through the network to help one another identify fraudulent claims. An effort to identify a particular kind of fraud is called a study. We are challenging you to come up with ideas for representing a study in software. The software representation of a study might be textual or graphical, old or new, simple or complicated. Take a look at the study examples and tell us what you think would work best.

Background

We are building a system that enables healthcare insurers to detect fraudulent claims with shared data. In the American system of health insurance, a patient who receives medical treatment or a provider of medical treatment submits a claim for reimbursement to the healthcare insurer. Providers and patients can carry out scams with several insurers, making it difficult to identify the fraud and prevent it from occurring at yet another insurer. The HFPP initiative seeks to detect fraud by analyzing data provided by a network of healthcare insurers. An effort to identify a particular type of fraud is called a study.

We will be using the following terminology throughout this contest.

• insurer: an entity that pays for a patient's medical services
• provider: an entity that provides medical services to a patient
• claim: a provider's request for payment from an insurer
• HFPP: Healthcare Fraud Prevention Partnership, a government initiative to detect and prevent fraudulent claims
• partner: a healthcare insurer that has joined the HFPP
• study: a plan to search for fraudulent claims
• data request: a unit of execution for a study; the study is broken down into one or more requests for data from partners
• data response: a partner's reply to a data request

 

Study examples

A study will seek to identify fraud by looking at claims provided by partners in the data exchange network. Each claim includes the following information.

• the patient who was treated
• the provider of the treatment
• the medical condition that was treated
• the type of treatment that was performed
• the date and duration of the medical treatment
• the cost of the treatment

Now let's consider some typical studies that partners would want to carry out. Bear in mind that each study has many possible variations. The date range may vary, or the study may focus on a certain set of providers or patients.

The impossible day: If insurers don't share data, they are unable to identify providers who are billing several insurers for a total of more than 24 hours in a single day. The data exchange network will allow partners to detect this fraud with a study that asks the following question: Is a provider billing insurers for more than 24 hours of service in a single day?

Duplicate claims: An unscrupulous provider may provide a single treatment and submit multiple claims for it. To make the claims look different, the provider changes one detail such as the date, keeping everything else the same. A study might ask the following question: Is there a pair of claims that are identical in every way except for one value?

Unbundled services: Like a combo meal at a fast-food restaurant, some medical services are performed together for cost savings. This is often done with blood tests, for example. A common type of fraud involves billing separately for the components of the bundled treatment in order to receive a larger reimbursement. A study might ask: Did a patient receive individual services X, Y, and Z instead of the bundled service XYZ?

Excessive services: A provider can seek to increase its income by providing services in greater quantity or more frequently than required. For example, a study might ask: Is a provider charging for 30 wound care kits a week for a patient who only requires a new dressing once a day? Or: Is a patient receiving daily visits for a condition that only requires weekly visits?

Compromised numbers: A partner may circulate a list of patient numbers or provider numbers that are known to have been involved in fraud. The study seeks to measure the extent of fraudulence for these numbers throughout the network.

 

Requirements

We are interested in hearing your ideas for representing a study in software. Should users express a study in some kind of declarative language? Should they fill out a multiple-choice form? Is there a drag-and-drop interface that would work well? We are open to all kinds of possibilities. All you have to do is explain how users would formulate a study with your proposed software representation. You can propose an existing solution or outline something new for us to develop. You can use text, diagrams, sketches, or any combination of these to explain your idea.

 

Optional checkpoint

If you upload your submission within 72 hours of the contest launch, you will receive an individual review. Up to five checkpoint submissions will receive a checkpoint prize of $50 each.

 

Target audience

You are targeting IT professionals who are in charge of building the new data exchange network.

 

Judging criteria

Reviewers will be most interested in the expressive power and computational convenience of your proposed software representation.

Expressive power: Does this software representation make it easy to express a wide range of studies?

Computational convenience: In this software representation, do we see a clear pathway to breaking down the study into data requests?

 

File format

You may submit your completed submission template in Word or OpenDocument format according to your preference. Please choose one of the attached template files and complete a separate copy for each submission.

 

What to submit

Submission zip file: One completed copy of the submission template.

Source zip file: Identical to the submission zip file.

Preview image: Make a 1024 x 1024 screenshot of your submission and save it in JPEG or PNG format.

 

Final fixes

You may be asked to complete one round of minor changes to ensure that your submission meets the stated requirements of this contest. More information about final fixes.