Predictive Risk Models for Underwriting
Technological advancements have steepened the evolutionary curve of underwriting. Two things are happening simultaneously:
- New data sources are available from wearable health monitoring devices to electronic health records (EHRs).
- Advancements in data science have fueled a set of predictive risk models of medical and behavioral data to generate mortality scores. In turn, these mortality scores can emulate, and in a few cases, outperform underwriting decisions.
Historical analysis of underwriting data has enabled a myriad of predictive models, each targeted at a different aspect of underwriting. These include applying behavioral science to elicit instances of non-disclosure; leveraging a combination of diagnosis, prescriptions, and lab results to flag potential risks and/or follow-up investigations; or triaging applications for straight-through process or underwriter reviews.
In this whitepaper, we will be discussing each of the three different categories of models – non-disclosure models, mortality risk models, and triage models.