At the 2015 WCI Conference this session addressed how companies use predictive analytics and what they measure to understand if it is successful.
The speakers included Scott Rodgers, Executive Vice President Casualty Operations, Sedgwick, George Furlong, Senior Vice President Managed Care Outcomes, Sedgwick, Dr. Marcos Iglesias, M.D., Vice President, Medical Director, The Hartford, and Mark Sidney, Vice President of Claims, Midwest Employers Casualty Company.
When we think about Big Data, we think about gathering all the information we can that is available to use for predictive modeling. One of the panelist stated that data in general has three components: volume, velocity, and variety. Velocity refers to the sheer volume of information a company receives. Velocity refers to how fast the information is coming at you which is constantly. Variety refers to the unstructured data and the amount of variance there is in the data received.
Most companies try to capture as much data as they can. This includes claimant info, nurse data, HR, payroll, age, gender, and other demographics. Then the company will take a look at this data and scan for trends. There is no such thing as too much data. Not only should companies collect data but they should compare that data with outside sources. The downfall to outside source information is that it is usually a couple years behind. So the company cannot really use this data for predictive analytics but they can as baseline information to make sure they are seeing the same trends in their data.
A trending topic discussed during this session is pharmacogenomic testing. This is genetic testing that looks at genetics then physicians look into to see how their bodies would react to certain treatments. This could mean the right treatment at the right time. There have been patients whose genetic composition metabolizes prescriptions at a different rate than others. Is this useful for workers’ compensation? No. In workers’ compensation, there could be use for this regarding opioids and narcotics. These tests are very expensive and have the potential to not be as beneficial as originally thought. Regarding opioids and narcotics, the first question that should be asked is if the injured worker even needs this strong of a prescription. It is important to remember not to forget the simple things.
When companies take a look at the data, look for outlayers and items things that just don’t seem to fit in with the rest of the data. These outlayers might be great predictors of a trend. Not only do you want to look at the outlayers but also look at the data that seems to be perfect. Nothing is every perfect so this could be a red flag.