A pandemic presents a perfect moment to rethink old premises, to unlearn established practices, and to ask unexpected questions. For example, why not perform an additional risk assessment upfront – before ordering evidence – to determine if the outcome is likely to justify the expense? By using statistical probabilities and predictive models, a more dynamic risk assessment process has the potential to deliver substantial cost savings.
Try this thought experiment: Ask an American to name the color of a “Yield” sign, a common sight on U.S. roads.
New arrivals to the U.S. are likely know the correct answer, but the native-born often insist that the yield sign is yellow; it has been red since 1971.
Why do so many get this answer so wrong? The answer lies in human nature. We accumulate a store of generally-accepted truths over time, but just because information was once accurate does not mean it always remains so. We have to adapt – to learn to unlearn.
Consider the standard “AA” or age and amounts table. Lockdowns and stay-at-home measures are making collecting fluids and conducting parameds less feasible, yet underwriters still base the decision to order evidence on an age and policy face amounts table that has not changed substantially in decades. The result? Insurers often order evidence unnecessarily or waive needed requirements.
A pandemic presents a perfect moment to rethink old premises, to unlearn established practices, and to ask unexpected questions. For example, why not perform an additional risk assessment upfront – before ordering evidence – to determine if the outcome is likely to justify the expense? By using statistical probabilities and predictive models, a more dynamic risk assessment process has the potential to deliver substantial cost savings, but only if insurers can learn to unlearn what we think we know.
To understand how the idea of dynamic risk assessment works, we first need to unlearn certain assumptions. Ordering any form of evidence, from lab tests to attending physician statement (APS), can be expensive. Yet all too often carriers can disregard other less tangible costs. In-person exams can seem intrusive to the applicant and can lengthen the time for a decision. Such high-friction purchasing processes diminish the customer experience and can prompt some to abandon an insurance purchase. What’s more, due to COVID-19, it may not be logistically possible to safely conduct traditional evidence gathering under certain circumstances. All of these costs can add up, and yet often carriers do not assess the full expense of ordering evidence, nor do insurers often look at the other side of the ledger to weigh whether new information is likely to alter the underwriting decision.
Dynamic risk assessment challenges what all underwriters think they know about how to use an AA table. It asks insurers to estimate the full costs of a given piece of evidence and weigh these expenses against possible benefits – or protective value – and consider this equally alongside applicant age and policy face amount. Age and face amount, of course, are well understood. Underwriters use both as the primary basis for the decision to either order or waive an additional evidence requirement and send applications through full underwriting or through an accelerated process. You could say the life industry is built upon this sturdy age-and-amount foundation.
But AA requirements alone fail to tell the full story of an applicant, and that’s where the third element – this risk assessment – makes all the difference. Predicting risk is a probabilities game. Even knowing nothing more than the age, gender, and lifestyle from a basic insurance application, the underwriter can determine certain conditional probabilities based on mortality and morbidity trends in certain demographic populations. For example, our risk of heart attack rises dramatically with age – and incidence is even greater in men. So the protective value of certain cardiac tests on an older-age male may be far greater, when compared to the cost, than the same tests performed on a younger man or woman. Insurers can build predictive models to apply these probabilities and help the underwriter calculate protective value and assess the likelihood that additional fluid testing or scoring will be able to detect an impairment that will result in full underwriting outcome in worst class.
This process also is dynamic, evolving with each new test or score. Every item of evidence deepens our understanding of the applicant and either raises or reduces the risk and the resulting return on investment for yet more evidence. The result is an increasingly accurate and personalized analysis that becomes more precise over time. In a study using internal experience data, RGA demonstrated that a dynamic risk assessment model could deliver substantial cost savings.
But wait… isn’t the purpose of a predictive model to replace certain forms of evidence, such as fluids? Not entirely: carriers could benefit from unlearning settled beliefs about modeling. Consider the smoking propensity predictive model: it was intended to evaluate whether an applicant would be urine-cotinine positive – a smoker – without having to conduct this expensive test. Insurers instead found that the model generated too many false positives and false negatives. The problem may lie, not entirely with the model itself, but what underwriters were asking it to do. Rather than asking the model to diagnose smokers – to essentially replicate the accuracy of the urine-cotinine test – underwriters should be asking the model to assess the probability that the cotinine test is necessary at all.
Elements of dynamic risk assessment modeling are nothing new. Underwriters have been conducting similar “reflex testing” assessments for years. Insurers identify cases in which a positive test result automatically triggers a “reflex” requirement for additional evidence gathering. For example, a high glucose reading suggests an elevated likelihood the applicant may have pre-diabetes or diabetes, and underwriters commonly seek an HbA1c test to confirm that the disease is present. Insurers require hepatitis testing if a liver function lab result is elevated, and an attending physician statement is often required if underwriters suspect an impairment as a result of an exam. In each circumstance, the insurer is determining that new data on a particular applicant elevates the protective value and, thus, justifies the expense of additional evidence.
Accelerated underwriting itself draws on the basic principles of dynamic risk assessment. If an applicant meets eligibility requirements for acceleration based on age or face amounts, insurers perform risk assessment as a second step, using application questions or other data sources, such MIB, motor vehicle, pharmacy and other electronically available records. If all of these factors, collectively, fall below certain risk threshold, the underwriter typically sends the application through an accelerated process that does not require fluid testing.
If the concepts behind this form of risk assessment are so well known, why isn’t it in widespread practice today? Dynamic risk assessment is different than a product – it’s a whole new way of operating, and it requires insurers to dislodge fixed beliefs and try something new. For too many carriers, the Yield sign is still glowing yellow. Leaders are still focused on evidence evaluation practices that worked well in the past rather than systems that are a better fit for the future. Dynamic risk assessment may not work for all insurers, and any predictive models would likely need to be tailored to reflect unique underwriting rules and actuarial assumptions. Still, early proof-of-concept studies by RGA have been encouraging, and the savings potential could be significant. Insurers no longer need to slow down and yield; it’s time to go.
At RGA, we are eager to engage with clients to better understand and tackle the industry’s most pressing challenges together. Contact us to discuss and learn more about the RGA capabilities, resources, and solutions available to you.
