Supporters of the Affordable Care Act (ACA / PPACA / ObamaCare) are very interested in targeting uninsured consumers and encouraging them to seek health insurance coverage through their state insurance marketplace or the federal marketplace. Health plans may also wish to identify uninsured consumers in order to maximize their marketing dollars or even direct them to a private exchange that provides more control over how their products are presented.

Predicting who is uninsured

In order to strategically target the uninsured (or avoid them if that is part of your marketing plan), a model is needed that will identify those most likely to be uninsured from among prospect lists and marketing databases. To build our predictive model, we examined over 7,000 survey responses from a national study on insurance status that included adults of all ages and income levels. Using only third-party data that can be purchased from numerous list vendors, we paired survey responses to these third-party variables based on the name and address of each survey respondent. We then created predictive models of their self-reported insurance status using only third-party data.


A logistic regression model predicts insurance status with 70% accuracy using only third-party data. Almost 73% of the uninsured are correctly identified by the model along with 67% of the insured. CHAID and discriminant analysis models produced similar results. The table below shows the percentage of consumers who fall into each of the four possible outcomes from the logistic regression model:

Prediction Actual insurance status
Insured Uninsured
Predicted to be Insured 67% 27%
Predicted to be Uninsured 33% 73%

Roughly 40 variables are used to predict insurance status. Standard demographics such as age, household size and occupation are used along with financial inputs (e.g. online purchases, home ownership, investments, etc.) and attitudinal factors (e.g. political party, voting history, media usage, etc.).

Putting the model to use

The next time you are purchasing a prospect list, consider acquiring variables like these in order to filter out the insured or uninsured households in your list. You can also work directly with the list vendor to target your list purchase by filtering out undesirable prospects prior to purchasing the list.

If you want to know more, give DSS Research a call.