Use Geographic Information Systems to Decrease Fraud
Geographic Information Systems can be used to dramatically decrease fraud. It can be a great tool to help you: identify geographic areas to focus on, illustrate that fraud occurred, and identify participants in fraud rings.
While we’ve talked about GIS for years, it is currently underutilized as a fraud and abuse detection tool. I think one reason it is not used more is that we are intimidated. But we don’t need to be. We aren’t preparing academic papers, so we don’t have to be perfect. We just need to be able to use the tools well enough to be able to spot potential abuses.
I’ll describe three examples ranging from very simple to fairly complex: Over-utilization, Transportation Fraud, and Fraud Rings.
Using GIS to Identify a Region with Over-utilization
In August 2003, Redding Medical Center paid $54 million to settle government allegations that they performed medically unnecessary Coronary Artery Bypass Grafts (CABG). The case developed out of a Qui Tam complaint, but you can clearly see the problem in Dartmouth’s geographic analysis of CABG. The dark red areas of the map indicate hospital referral regions with a very high rate of CABG. Redding is the big red area in the northern part of the state.
If you compare Redding to the rest of the country using a distribution graph, you see that Redding was an extreme outlier. This graph also shows that you can use non-mapping tools to do geographic comparisons.
With these two graphs, it is easy to see that something was amiss in Redding cardiology care. Figuring out that two doctors were responsible for most of these surgeries takes further research.
Illustrating Transportation Fraud
GIS systems can calculate mileage for ambulance trips to verify billing accuracy. The calculation of distance is going to work similar to the way MapQuest.com estimates travel distances for you when you ask for directions. There is no guarantee that the transportation provider followed the shortest route. However, when you look at totals across time, you should be able to identify providers who seem to be billing far in excess of what they should have. Being able to show this on a map can be quite compelling if you take them to court.
Some GIS software includes algorithms for calculating distance, but first you’ve got to prepare the data you’ll analyze. The transportation claim does not have the pickup and drop-off addresses, so you’ll have tag this information from other sources.
1) Identify all transportation claims (or the subset you’re most interested in)
2) Tag the patient address to these claims
3) Link the transportation claims to the medical destination claims using patient ID and date of service (if a patient has multiple services on the same day, you may want to exclude their transportation claim from this analysis)
4) Tag the provider address from the medical destination to the transportation claim
Identifying Participants in Fraud Rings
Chiropractors, dentists and outpatient surgery centers sometimes use runners to recruit patients. The runners are paid a fee for bringing in patients. For example, outpatient surgery clinics in Southern California used runners to bring patients from out of state to the clinics. The centers paid runners and patients cash for undergoing surgery. Usually the runners targeted large employers who provided health insurance for low-wage positions.
GIS analysis can identify providers who serve pockets of patients who live near each other, but not near the provider. There are several ways you could approach this. I’ll describe one approach:
1) The first thing you want to do is limit your analysis to one provider type or specialty at a time. We expect different kinds of providers to have different patterns, so we don’t want to confuse things by viewing multiple provider types.
2) However, when you have a large number of providers to analyze, you’ll probably want to limit your map to providers you think are most likely to be serving large numbers of patients who don’t live near them. How do you define that? In rural areas, far away may mean 50 miles, while in urban areas it may mean one mile. Luckily there are tools for defining market areas.
Market areas will be geographically small in urban areas and larger in rural areas. You can use the GIS to automatically calculate market areas or you can use some simple calculations with zipcode level data. See (unknown link #394) for examples of different ways of calculating market/trade areas.
3) You can layer your provider-patient networks on top of the market area layer.
4) Finally, run a query to limit the display of provider-patient networks to providers with a high percent of out-of-area patients.
Overcoming Barriers to Use
Three of the biggest barriers to using GIS to find fraud are: cost of software, expense of geo-coding addresses, and lack of examples. Here are three reasons these need not be barriers:
- GIS doesn’t have to be expensive. GRASS - GIS is an open-source alternative to ArcInfo/ArcView and MapInfo. It is available for Mac, Windows and Linux.
- GIS works best if you have geo-coded addresses, but it can even work if you only have zipcodes.
- I’ve described three examples that can help you get started with using GIS to find fraud.
Many thanks to the following websites for use of their graphics: Dartmouth Atlas Updated and University of Wisconsin Center for Community Economic Development (unknown link #394).