重新分类死亡风险（英）2025

How wearable-driven mortalitymodeling reshapes risk stratificationin life insurance underwriting C O N T E N TS Section 1:Executive summary3 Section 2:Background5 Section 3:The Klarity Model7 Mortality scoring9 How does the model compare with currentpreferred guideline risk assignment?9 Benefit gained from activity data15 Section 4:Use cases and next steps18 Section 5:About Klarity20 Section 6:About WTW21 Section 1:Executive summary This has added to the challenges for risk classificationand difficulty in truly differentiating the risk profiles ofthe preferred risks as well as the healthier impaired. “Sitting is the new smoking” is a catchphraseoften used to encourage people to get somelevel of physical activity. The need to rethink the risk stratification process in thelife insurance industry has become increasingly evidentover the past decade. With the proliferation of new datasources and advancements in technology, there is asignificant opportunity to enhance the accuracy andefficiency of underwriting processes. The Klarity modelaims to address this need by leveraging nontraditionaldata to produce individual-level mortality scores thatcan predict and classify risks more effectively thantraditional methods. Medical personnel, underwriters, actuaries and mortalityresearchers understand activity level is an importantmeasure to assess one’s health and expected longevity.Unfortunately, activity level information is often overlookedor has been measured only through self-reporting orcorrelation to other measures such as body mass index(BMI) in the life insurance risk selection process. Since the proliferation of multiple risk classes, companieshave used traditional measures such as cholesterollevel, blood pressure, BMI, tobacco usage, and personaland family history, to name a few, for stratifying anddetermining risk class criterion and placement forapplicants. While each is an important health metric,these traditional approaches and metrics often misclassapplicants because the measures only provide part ofan individual’s health profile and often miss importantindividualized measures such as resting heart rate, heartrecovery rate, sleep and activity versus inactivity levels. Over the past year, WTW's Insurance Consulting andTechnology (ICT) team has analyzed a new risk scoringtool developed by Klarity, which incorporates dataobtained from a wearable device such as a fitnesswatch, a smartphone or other device that capturesactivity levels, sleep patterns, heart rate and pulsedata. The model, originally trained on U.K. nationalhealth data, was built on data spanning over 12 years,covering 6.1 million life years, focused on ages 40 to 70.Within this data set, there are over 37,000 deaths.To analyze the model and its applicability for lifeinsurance, WTW partnered with Klarity to test theefficacy of the model’s mortality score predictions ondata from the U.S., leveraging the National Health andNutrition Examination Survey (NHANES) data set. Over the past 10-plus years, the industry has beenmoving toward changing the underwriting process.For life insurance, this has meant rethinking the datasources used, improving the customer’s experienceand shortening the time from application to policy issue. Key observations of WTW’s analysis show: The Klarity model demonstrates a promising approach toimproving risk stratification in the life insurance industry.One of our key findings is that while the Klarity modelvalidates the risk ranking of traditional underwritingclasses used to assess mortality risk by insurers,the Klarity model can improve and further differentiaterisks significantly even within a single risk class.We found the risk assignments by the Klarity modelto be highly correlated with actual mortality resultsrelative to mortality demographic baselines. •The Klarity model has the ability to betterclassifyrisksand reduce the overlap inherent in today’s riskclassification systems. In our analysis, when activitylevel and the Klarity model risk score are considered: –34% of the second-best nonsmoker risks and 16%of the residual standard class risks are identified ashaving a better risk score and exhibit similar mortalityrisk profiles to the best nonsmoker risks –6% of risks currently classified in the best and 13%classified in the second-best preferred nonsmokerclasses are identified as having a lower risk scoreand exhibit actual-to-expected (A/E) ratios, moreakin to a residual (not preferred) risk By utilizing a broader range of data sources andadvanced predictive techniques, the model canprovide more accurate and individualized riskassessments, ultimately leading to better pricing,improved customer engagement and enhancedin-force management. •The level of activity, including Step Count and ActivityDuration,provide high correlation to mortalityoutcomes, even more than some traditional markerssuch as cholesterol, BMI, and family history of heartdisease and diabetes •Though trained mostly on U.K.