Insight

The Might of Modeling ACCURATE MODELS OF POTENTIAL PATIENTS HELP HEALTHCARE MARKETERS REFINE STRATEGY AND SAVE MONEY BY DEVELOPING LEAN CAMPAIGNS THAT ACCURATELY TARGET THOSE MOST LIKELY TO USE CARE. S A M B U S S M A N N , M A , data scientist with True North Custom, makes a case for data-driven modeling over traditional audience selection. INSIGHT: FOR BACKGROUND, WHICH TARGETING METHODOLOGIES DO HEALTHCARE MARKETERS MOST COMMONLY USE? Bussmann: Many look something like this: all households within certain ZIP codes that have members older than 45, income greater than $50,000 per year, and a female present. While this is one way of narrowing your audience, there's no way to ensure that all the individuals in this group are more valuable than all individuals in the excluded group. INSIGHT: LET'S TAKE THE FIRST DIFFERENTIATOR—AGE. WHAT'S WRONG WITH USING 45 AS A CUT-OFF? Bussmann: Narrowing by age is an arbitrary way of looking at what is probably the most important predictor of healthcare use. A 45-year-old likely uses health care in a different way than a 75-year- old. By putting them in the same rough group, those differences are lost. A data-driven model gives value to each age group commensurate with their actual healthcare use. INSIGHT: WHAT ABOUT INCOME? ISN'T IT IMPORTANT TO TARGET PATIENTS WHO HAVE THE ABILITY TO PAY WELL FOR THEIR CARE? Bussmann: Of course, the ability to afford health care is important. However, our research has shown income is not sufficient to determine return-on-investment potential. For instance, household income often drops in retirement, when many individuals are living off of savings. Those in retirement are often on Medicare—i.e., they have insurance—and are the biggest consumers of health care. They are also the biggest revenue drivers. In other examples, entrepreneurs often have high incomes but poor insurance coverage. Government employees often have low incomes but great insurance coverage. INSIGHT: AS LONG AS WE'RE QUESTIONING OUR USUAL TARGETING METHODOLOGIES, WHAT ABOUT ZIP CODE? IS THAT A PROBLEMATIC DIFFERENTIATOR? Bussmann: A narrowly defined area, such as a five-mile radius of the hospital, can be too restrictive, whereas a data-driven model can help us predict how far people are willing to travel. There are likely some individuals living 10 miles away that a predictive model would identify as strong prospects. A ZIP code selection method omits those individuals. INSIGHT: NOW THAT YOU'VE HELPED US QUESTION THE ASSUMPTIONS BEHIND OUR USUAL TARGETING STRATEGIES, TELL US HOW A DATA-DRIVEN MODEL IMPROVES ON THEM. Bussmann: A sound model understands that factors influencing value are not rigid or defined by just a few variables. By looking at the relationship between actual healthcare usage and hundreds of characteristics about each individual, a model makes predictions about how that individual will use health care in the future. And when it comes to the bottom line, the model can synthesize each individual's demand for care with their ability to pay for that care. Let True North Custom help you make the most of your data. Contact Chief Healthcare Strategist Tim Hanners at 423-490-9962 to get started. Sam Bussmann, MA data scientist T R U E N O R T H C U S T O M TRUENORTHCUSTOM.COM INSIGHT V11N3 11 insight DATA SCIENTIST Q&A

• Cover