Next-gen data sources in life underwriting: Novel attributes to segment mortality risk

Next-gen data sources in life underwriting

Looking beyond BMI to assess mortality risk

8 minutes read

Published 03/25/2025

Portrait of a female doctor measuring a patient's height to calculate BMI

Share
LinkedIn

Facebook

Twitter

Whatsapp

E-Mail

Copy Link

This paper is the third in our series examining the potential of third-party data sources to enhance life insurance underwriting.

Background

This publication is the final in a series of analyses of the UK Biobank data, conducted between Klarity, a UK-based health data analytics firm, and Munich Re Life US.¹ This study considers alternative attributes to body mass index (BMI) for assessing mortality risk: waist circumference, waist-to-height ratio, and waist-to-hip ratio. Although BMI is an attribute that has been used in life insurance underwriting since the early 20^th century to assess obesity and its mortality impact, BMI alone may be an imprecise indicator for metabolic health.² BMI does not consider the distribution of fat tissue or differences in body type. While BMI is predictive of mortality, we consider other physical measurements that can be layered onto BMI to help refine risk assessment. In particular, waist circumference, waist-to-height ratio, and waist-to-hip ratio can better capture different distributions of body fat than BMI.

Objective

This study examines the UK Biobank data to accomplish the following:

Analyze the effectiveness of alternative build-related attributes in stratifying mortality risk of a simulated insurance population.
Consider how these alternative build-related attributes, such as waist-to-height ratio, can be used in addition to the traditional life insurance BMI factor to refine risk assessment.

Data and extrapolation to life insurance 

Similar to the earlier publications in this series, the analysis in this study focuses on the UK Biobank data that Klarity extracted. Variables relevant to the study include: 

Demographic information: age, gender, income, index of multiple deprivation (IMD, a UK measurement of poverty and access to services within a small geographic area)

Physical measurements/vitals: weight, height, waist circumference, BMI, blood pressure, hip circumference

Note that the waist and hip measurements were recorded in a well-controlled setting that may not be easily replicable in a life insurance underwriting setting. Also, the Biobank study participants were not underwritten for life insurance, so any anti-selection associated with applying for life insurance is not relevant in this setting. 

Like the previous studies, the analysis and results discussed onwards are based on the filtered “insurable” block of individuals, unless stated otherwise. 

Classic actuarial methodology 

We performed a classical actuarial actual-to-expected (A/E) mortality analysis on both the overall and “insurable”  datasets. The expected mortality basis shown in the graphs below uses the UK National Life Tables, which are split by age, gender, and calendar year, without any mortality improvement. The overall A/E using the UK National Life Tables basis is 48%. For the “insurable” dataset the A/E is 23%. 

Results

Research suggests that measures of abdominal obesity are more predictive of cardiovascular disease and diabetes than BMI.³ Height, weight, waist circumference, and hip circumference measurements are available for over 99% of the insurable group of individuals. Table 1a summarizes these physical measurement attributes across different subgroups of individuals and lays out the differences in these physical measurements between men and women. The difference across genders is especially pronounced for waist-to-hip ratio and grip strength, but more similar for waist-to-height ratio. 

Table showing physical measurement statistics like BMI and waist-to-hip ratio for different populations.

BMI

In Figure 1, we see that BMI stratifies mortality risk; there is higher mortality at low and high BMI groupings, resulting in a J-shaped curve. Most individuals fall into BMI 18.5-30 groupings, where BMI 18.5-25 indicates ”normal” BMI individuals, BMI 25-30 overweight individuals, and BMI 30+ obese individuals. From Table 1 and the below relative risk curves by gender, we see that BMI for females tends to be lower than for males, with a median BMI of 25.5 for females compared to 26.8 for males. We use Figure 1 and the demonstrated mortality segmentation by BMI to compare against the subsequent analyses of the alternative build attributes.

Graph showing relative actual-to-expected mortality by BMI with lines for males, females, and all insurable lives, and bars for insurable lives by BMI.

Waist-to-height ratio

We focus on waist-to-height ratio as it shows promise in mortality segmentation from Table 1, and the median waist-to-height ratio is roughly similar between males and females. Although waist-to-hip and waist circumference show evidence of mortality segmentation, their distribution for males and females differ and would likely require different thresholds for meaningful analysis. In addition, waist-to-hip ratios involve obtaining an extra measurement, hip circumference, which may be a potential source of error. Thus, waist-to-height ratio is easier to analyze and generalize across genders. This is confirmed in Figure 2, where we see that the relative risk curves are roughly similar between males and females. We remove the under 0.4 waist-to-height ratio for males due to lack of credibility in this group. Therefore, waist-to-height ratio was our first choice for analysis as a supplementary measure to BMI.  

Around half of females and a quarter of males have a waist-to-height ratio between 0.4 and 0.49, defined as the healthy range.⁴ A ratio between 0.5 and 0.59 is elevated risk, and above 0.6 is highest risk, with an increased chance of cardiovascular disease. We see that waist-to-height ratio stratifies mortality risk, and there is an associated higher mortality risk for the high ratio group. 

Graph showing relative actual-to-expected mortality by waist-height ratio with lines for males and females, and bars for male and female proportion of insurable lives.

Considering BMI and waist-to-height ratio together

Relative risk and death counts for different waist-to-height ratio and BMI combinations are shown in Figure 3. Cells with fewer than five deaths are left blank. In addition, we exclude the BMI under 18.5 category, as these individuals are concentrated in the under 0.5 waist-to-height category. Note that these BMI and waist-to-height ratio groupings are selected such that the ranges go from healthy to increasing risk. The 18.5-25 BMI grouping is the “normal” BMI range, and the <0.5 waist-to-height ratio is the healthy waist-to-height ratio category. 

Color-coded chart with red and green squares showing relative actual-to-expected mortality across BMI and waist-height ratio, with values in each cell.

We note from Figure 3 that the combination of waist-to-height ratio and BMI allows us to better segment mortality risk than BMI alone, and segmentation can be seen across all BMI groups In fact, individuals with an overweight BMI but healthy waist-to-height ratio have 15% lower mortality risk than those with a healthy BMI but unhealthy waist-to-height ratio. Within the 25-30 BMI category, we see that the healthy waist-to-height ratio of under 0.5 has a mortality risk 40% less than the unhealthiest waist-to-height ratio of 0.6+. Thus, waist-to-height ratio can offer additional mortality risk segmentation even after controlling for BMI.  

We can use waist-to-height ratio to identify low mortality risk individuals in the overweight BMI category. In fact, individuals with an overweight BMI but healthy waist-to-height ratio have 15% lower mortality risk than those with a healthy BMI but unhealthy waist-to-height ratio.

As in our previous study, Klarity trained its mortality models on different subsets of variables to proxy various underwriting scenarios. The variables used in the models are listed below in Table 2. We focus on Model E, which includes all attributes intended to proxy the fully underwritten (FUW) scenario and adds in the novel attributes that are not currently considered in life insurance underwriting, including step count. This allows us to directly compare the significance of these attributes in predicting mortality. 

Table of predictor variables across five model types, grouped by demographics and diagnoses, lab tests, step count, and novel attributes.

The SHAP plot listing the most significant variables for Model E is shown in Figure 4a. Note that waist-to-height ratio is a significant predictor of mortality and ranks as more important than BMI in predicting mortality.

Vertical bar chart showing 30 health-related variables used for Model E plotted against their SHAP values.

Separately, Klarity also trained an alternative version of its Model A, called Model A2, that replaces BMI with waist-to-height ratio while keeping all other predictors the same. Models A and A2 represent different versions of an accelerated underwriting (AUW) program. When comparing A and A2’s SHAP plots (Figures 4b, 4c) with the most significant predictors, waist-to-height ratio is more predictive in Model A2 than BMI is in Model A. 

Vertical bar chart showing 12 health-related variables used for Model A plotted against their SHAP values.

Vertical bar chart showing 12 health-related variables used for Model A2 plotted against their SHAP values.

Conclusion 

Based on our analyses of the UK Biobank dataset, we conclude that there is strong evidence that waist-to-height ratio can effectively segment mortality risk on top of BMI and can be used to identify high-risk individuals across all BMI categories. This is also true for other measurements such as waist-to-hip and waist circumference. Note that these conclusions were drawn on a UK population, which has a lower prevalence of overweight and obese individuals than the US; in 2022, 26.2% of UK adults are estimated to be obese⁵ compared to 40.3% of US adults.⁶ Given the US has a higher proportion of obese lives than the UK, adding a metric like waist-to-height ratio has an even greater value add in stratifying out healthier lives in typically substandard or declined BMI categories. The higher percentages of overweight and obese individuals in the US imply a greater need to effectively classify mortality risk in these BMI categories.

However, it is important to note that these physical measurements require controlled conditions for accurate measurement, making their widespread use in life insurance underwriting more challenging. Carriers looking to adopt these measures into their underwriting processes can start by piloting programs to collect data, assess customer participation rates, and gain insights into how these attributes are distributed across their insured population. Such pilots serve as a practical step toward evaluating the feasibility and value of integrating these novel measures. Munich Re Life US has experience in developing and monitoring such programs and can work with carriers through incorporating these newer variables into their existing underwriting.  

References

¹This research has been conducted using the UK Biobank Resource under Application Number 88308. ²Is BMI a Reliable Predictor of Mortality Risk?, UNC School of Medicine, July 2023.   ³Abdominal obesity: the most prevalent cause of the metabolic syndrome and related cardiometabolic risk, European Heart Journal Supplements, May 2006.   ⁴Why your waist size matters, British Heart Foundation, October 2024. ⁵Obesity Profile: short statistical commentary May 2024, UK Office of Health Improvement and Disparities, May 2024. ⁶Obesity and Severe Obesity Prevalence in Adults: United States, August 2021–August 2023, CDC National Center for Health Statistics, September 2024.

Back to all articles

Contact

Angela Gong

Staff Data Scientist, Integrated Analytics

Munich Re North America Life

Email agong@munichre.com

Murali Niverthi, PhD

Director & Actuary, Underwriting Risk Assessment & Research

Munich Re Life US

Email mniverthi@munichre.com

Newsletter

Subscribe to our newsletter for the latest industry research, insights on life and health-related topics, and relevant Munich Re Life US updates. Join today and stay connected with our business.

Subscribe now

properties.trackTitle