Predictive Models and Underwriting: What’s the Potential?

Not long ago I went on a tear about American General’s use of the CRL Lab’s proprietary laboratory analyzing methodology called “Smart Score”. The blasting cap for that little tiff was a case that American General declined, obviously using the Smart Score numbers to pile on top of information they already knew. I did my best to get to the facts surrounding the Smart Score system to no avail. CRL refused to discuss it because it is a proprietary methodology, sending me back to American General where, I kid you not, it does not appear as though anyone actually knows how the system works.

In response to this revelation I wrote a blog post. I was hoping to elicit some action from AIG where they might actually find out how the system works and be able to explain it to the boots on the ground, me. They did respond by reviewing the case and deciding that it should have been approved, saying that the Smart Score was used improperly in this instance. I asked, sensing a breakthrough, how it was actually supposed to be used and ran into the same knowledge block. “Yes we’re using it in underwriting and yes, in most cases it should help clients get lower rates and no, we’re not sure how it does that”.

You might remember in response to one of my tirade’s I was contacted by Ammon Dixon the ExamOne Director of Informatics and Analytics, who asked if he could meet with me concerning the Risk IQ system and methodology created by ExamOne and Quest Diagnostics. I accepted and invited Rich Fuller with Special Risk Services to attend, honestly not sure what to expect but on one level hoping I would at least gain a better understanding of their competition in lab analytics, the Smart Score.

That meeting happened this Monday and was an extremely eye opening experience. Both Rich and I have done battle with and had clients be victims of the underwriting mindset that one lab result, in isolation can be a knockout blow for a client. At some point or another over the last 15-20 years we have experienced almost every lab value being the cause of a decline or postpone and what we found out in this meeting was how potentially unfair that can be in certain cases.  Rich and I wondered if underwriters who declined based on, for instance, an elevated ALT (liver function), elevated Glucose/A1c (diabetic markers) or elevated creatinine (kidney function), understand or are willing to entertain the notion that while those results are notable, in the context of the full blood and urine panel, taking into account age, sex and BMI, they may not be the best determinants of an applicant’s mortality.

ExamOne is the largest insurance laboratory in the country and they have compiled lab results for years, to the tune of 80 million or so. They have cross referenced these results against the social security death master file and have created a comprehensive, super cross referencing tool for going beyond the one result knockout and measuring what all of the lab results as a whole mean to mortality. What it shows is that there are cases where one bad result has no  major impact on applicant mortality and consequently, a set of lab results with all values in the normal range could give good reason to question the mortality. While this would be rare, it can occur with lab results in the normal range but scraping either the top or bottom of the range.

Mr. Dixon confirmed some of my thoughts about current underwriting, especially that knockout criteria can blow one or two lab results out of proportion in comparison to an applicant’s actual mortality risk. The truth is that far too many underwriters are depending on these knockout criteria and in at least some cases denying applicants a policy or a policy premium that accurately reflects the individual’s mortality risk. 

Keep in mind that Risk IQ is just about lab results, age, gender and BMI. Risk IQ is not influenced by known impaired risks or medical records and yet, in several comparisons with actual life insurance company underwriting, Risk IQ alone was as accurate as or more accurate than fully underwritten approvals. I don’t throw that out to say that companies should quit all of the expensive acquisition of medical records. In fact, Mr. Dixon pointed out that Risk IQ should not be used as a sole indicator of mortality risk, but rather in conjunction with other underwriting criteria in order to fully assess applicant mortality risk.  However, at a relatively small cost per applicant, why would a company not implement this type of methodology to essentially double their chances of underwriting correctly? ExamOne is going above and beyond in trying to improve the life insurance and company underwriting experience.

Bottom line: I’ve rattled on over the years about, sometimes how quickly and most times how slowly, the life insurance industry embraces new medical science. I don’t want underwriting to become a computer function but, especially as part of the world of impaired risk life insurance, I do constantly look for ways to make the process more fair. Risk IQ seems to have that potential. I will be writing more about this soon and hopefully sharing some actual examples where Risk IQ could have brought important value to the table. If you have any questions or would like a fresh look at a decline or rated approval, call or email me directly. If you’re an agent and haven’t looked into some of these new methodologies for underwriting, educate yourselves. Several companies are holding back from seriously considering this because they aren’t sure agents will understand and be able to explain it. We’re better than that. My name is Ed Hinerman. Let’s talk.