With the increasing popularity of accelerated underwriting (AUW), fewer life insurance applicants are being underwritten with paramedical labs and exams. This allows some applicants who would have previously been declined coverage to now be placed into insurance pools. To monitor AUW programs, life insurers often estimate the pricing impact by studying samples of newly issued policies. These samples are frequently limited in size due to operational challenges and expenses associated with larger samples. However, using studies from small sample sets to make pricing decisions may come with significant estimation risk.
Our research indicates that foregoing exam and labs introduces low frequency, high severity events, which may create more sample volatility. In this article, we explore the challenges of appropriately quantifying AUW mortality risk using small audit samples through illustrative simulation based on declinable lab research findings. This work builds on ongoing research into the mortality impact of declinable lab results using data provided by ExamOne, which will be detailed in an upcoming white paper.
This simulation utilizes mortality study research that links lab and exam results to observed mortality outcomes for approximately 11.5 million insurance applicants who applied from 2001 to 2024. It is designed to capture the possible audit outcomes and to demonstrate the difficulty of creating accurate best estimates using small sample sets.
The general parameters of the simulation are:
The parameters for this simulation were selected for illustrative purposes, with 1,000 policies being a common sample size for an audit cohort. Labs provide most value above age 40 and many AUW programs limit applicants to below age 60, so age 50 splits the difference. Additionally, select lab measurements in this research apply only to males.
Our declinable lab mortality study safely covers 10 years of experience, drawing all mortality impacts directly from our research. As this is meant to simulate a cohort of AUW business, the face amounts of the policies are simulated as a distribution, with the average of the distribution being $1,000,000. The distribution used is shown below and based on industry AUW composition data.
| Death benefit | Count | ||
| $250,000 | 200 | ||
| $500,000 | 250 | ||
| $1,000,000 | 325 | ||
| $2,000,000 | 175 | ||
| $3,000,000 | 50 | ||
| Total | 1,000 |
Table 1: Distribution of face amounts used for profit simulation.
We selected 6 lab measurements that could generate a decline in isolation. We elaborate on the selection process and our underwriting team’s role in that process in our 2025 article Cracking the Code on Declinable Labs: How Cross-Team Collaboration Elevates Risk Assessment.
Using our mortality study dataset, we can quantify both the frequency and severity of a declinable lab outcome. The severity is given in terms of relative mortality by duration, allowing nominal comparison against the net single premium paid for a standard policy.
The raw frequency is estimated simply from incidence in the dataset. This estimate requires adjustment, however, considering a declinable policy may be caught by other elements of the AUW program before it is approved. Therefore, we assigned each lab a detection rate assumption based on underwriting expertise. This parameter is referred to as “AUW information overlap” and represents the crossover between digital health data and prior lab and exam outcomes.
With these components, the mortality cost of each declinable lab outcome can be modeled as a binomial random variable. The sum of all mortality cost variables represents the cost to the group. Using a Monte Carlo simulation with 100,000 simulations, we develop a distribution of group mortality cost under each of the following AUW information overlap scenarios:
| AUW Information Overlap | Description | Detection Rate Range | |
| Very High | Aspirational | 90% to 95% | |
| High | Ideal | 75% to 90% | |
| Medium | Current average | 50% to 80% |
Table 2: AUW information overlap scenarios based on industry trends; ranges include detection rates for all lab measurements in that scenario. Detection rate ranges are multiplicatively scaled between scenarios and are not meant to line up with other scenarios at each endpoint. Simulation includes very low frequency events, therefore simulation count was increased accordingly.
These scenarios represent AUW programs with varying levels of digital health data coverage, as digital health data is the main driver of detecting these severe conditions in the absence of labs and exams. The scenarios range from a representative current program with medium data overlap to an aspirational, most effective program with very high data overlap. As data overlap increases, the frequency of the labs entering the simulated pool decreases. Simulating these scenarios will allow us to assess the mortality projection’s sensitivity to AUW program effectiveness.
The simulated counts of issued declinable applicants are shown in Table 3 below, with AUW information overlap scenarios in rows and simulated outcome percentiles for each scenario in columns:
Outcome – Count
| AUW Information Overlap | 10th percentile | Median | 90th percentile | |
| Very High | 0 | 2 | 4 | |
| High | 2 | 4 | 7 | |
| Medium | 5 | 9 | 13 |
Table 3: Simulation outcome by count.
As expected, the number of adverse lab events display an increasing pattern of instances for each scenario, starting at 0 events in the best random scenario with the best data overlap to 13 for the worst random scenario with the worst data overlap. As the information overlap declines, both the counts at each percentile and the spread between the 10% and 90% outcome increase. This indicates that the volatility of results increases as the program’s information overlap decreases.
The pattern of increased spread is also evident on an amount basis.
Outcome – Amount ($)
| AUW Information Overlap | 10th percentile | Median | 90th percentile | |
| Very High | 0 | 46,550 | 149,078 | |
| High | 34,750 | 123,000 | 263,300 | |
| Medium | 123,925 | 261,750 | 452,153 |
Table 4: Simulation outcome by amount.
The mortality cost is impacted by the distribution of face amounts in the cohort, creating another layer of volatility where an adverse lab result can find its way not just into an issued cohort, but also be associated with a smaller or larger policy.
These cost figures are interesting; however, they are only meaningful in the context of the specified distribution and product mix being simulated. We can analyze this on a profit margin basis to ground these figures in a relative context. Assuming this block of business targets a profit margin of 5% of the net single premium, the following table shows the remaining profit margin after deducting the mortality cost of declinable lab cases.
Outcome – Remaining Profit Margin
| AUW Information Overlap | 10th percentile | Median | 90th percentile | |
| Very High | 5% | 5% | 4% | |
| High | 5% | 4% | 3% | |
| Medium | 4% | 3% | 1% |
Table 5: Simulation outcome by remaining profit margin.
The profit margin range in this small sample indicates the potential for significant error when estimating the true mortality slippage for a cohort.
Most carriers with AUW offerings in the market have established monitoring programs to address inherent mortality slippage, however AUW monitoring and decisions can often be based on small audit sample sizes like the 1,000-life cohort simulated in this article. Our analysis indicates that the low-frequency, high-severity nature of declinable labs can materially shift AUW monitoring best estimates away from the true underlying mortality slippage.
AUW programs naturally introduce uncertainty, and this uncertainty can be amplified by the randomness present in small audit samples. When this volatility is thoughtfully anticipated and managed, carriers are better positioned to ensure pricing assumptions remain consistent with emerging mortality results.
Given the inherent volatility of AUW audit samples, life insurance carriers can take a prudent approach to managing potential sampling error. The following considerations can help support more informed and resilient audit outcomes:
PartnerRe continues to develop our PRe EMPT™ suite of AUW monitoring and pricing tools that explore the mortality consequences associated with accelerated underwriting. PartnerRe works with our carrier partners to support their AUW risk assessments. The PRe EMPT™ research that formed the basis of this simulation can create new guidance for quantification of declinable mortality risk. Contact our team to learn more.
Contributors
Ehren Nagel, Head of Actuarial Innovation
Joshua Herzog, Actuary – Accelerated Underwriting & Analytics