Digital Health Data for Mortality Risk: What Actuaries Should Know

Digital health data for mortality risk assessment is quietly rewriting the actuarial playbook in life insurance. For decades, actuaries relied on a relatively stable set of inputs: age, gender, smoking status, build tables, and whatever a paramedical exam turned up. That worked, mostly, because the data sources themselves didn't change much. But over the past five years, a flood of new digital health data, from electronic health records to prescription drug databases to camera-based vital sign readings, has given actuaries access to risk signals that simply didn't exist before. The question isn't whether this data is useful. It's how much of the mortality picture actuaries have been missing without it.

"All three digital underwriting evidence sources demonstrated significant value, both individually and in combination. EHR exhibited the largest single evidence mortality impact." — RGA, "Assessing Mortality Impact of Digital Underwriting Evidence" (2025)

What counts as digital health data in underwriting

The term gets thrown around loosely, so it helps to be specific. Digital health data in the underwriting context generally falls into a few categories, each with different strengths and actuarial implications.

Electronic health records pull structured clinical data directly from provider systems. This includes diagnosed conditions, lab results, vital signs, medications prescribed, and visit histories. LexisNexis Health Intelligence, which acquired Human API's health data platform in early 2025, now provides formatted EHR summaries designed specifically for underwriting consumption. Their Medical Insights product extracts targeted health attributes like vitals, labs, and material conditions from raw EHR data, making it easier for underwriters to find relevant risk factors without reading through hundreds of pages of medical records.

Prescription drug histories, accessed through pharmacy benefit manager databases, reveal medication usage patterns that correlate with specific conditions. An applicant taking metformin suggests diabetes management. Statins suggest cardiovascular risk factors. The data is structured, consistently available, and relatively easy to incorporate into automated decisioning.

Medical claims data captures billing records from healthcare encounters, including procedure codes, diagnosis codes, and provider types. It offers a broader view of healthcare utilization than EHRs alone, though it lacks clinical detail.

Then there are newer entrants: contactless biometric screening through remote photoplethysmography (rPPG), wearable device data, and behavioral health signals. These are earlier in adoption but growing.

Data source	What it captures	Actuarial strength	Current adoption
Electronic health records	Diagnoses, labs, vitals, visit history	Deepest clinical detail, highest mortality impact per RGA studies	Growing rapidly since 2024
Prescription histories (Rx)	Medication usage patterns	Structured, widely available, strong condition proxies	Widely adopted
Medical claims	Procedure and diagnosis codes, utilization	Broad coverage, good for utilization patterns	Widely adopted
Contactless biometric screening (rPPG)	Heart rate, respiratory rate, blood pressure indicators	Real-time physiological data without equipment	Early-stage adoption
Wearable device data	Continuous heart rate, activity, sleep	Longitudinal behavioral data	Pilot programs

The mortality slippage problem

Actuaries care about digital health data primarily because of mortality slippage. When an accelerated underwriting program skips the paramedical exam or attending physician statement, it inevitably lets through some risks that traditional full underwriting would have caught. The gap between expected and actual mortality in an accelerated program is slippage, and it directly affects pricing adequacy and reserve sufficiency.

RGA's 2025 research on this subject used a method comparing full underwriting decisions against accelerated underwriting decisions, with and without digital underwriting evidence. The methodology treats the full underwriting decision as a mortality surrogate, which is an imperfect but industry-accepted approach. They tested three DUE sources: medical claims, LabPiQture (their proprietary lab results database), and electronic health records.

The findings were clear. Each source individually reduced mortality slippage. EHRs showed the largest single-source impact, and they also increased the decision rate, meaning more cases could be completed without requesting additional evidence. But the real value appeared when sources were combined. Using all three DUE sources together produced a greater mortality improvement than any single source alone.

This matters for actuaries setting assumptions on accelerated underwriting blocks. If a carrier's AUW program uses only prescription data, the mortality assumption needs to account for the slippage that EHRs or claims data would have caught. RGA's research suggests that the mortality impact varies by carrier based on their specific book composition and underwriting rules, so general industry averages are a starting point, not a substitute for carrier-specific analysis.

What Gen Re's 2024 survey tells actuaries

Gen Re's 2024 U.S. Individual Life Accelerated Underwriting Survey, covering 38 carriers and representing billions in face amount, provides a useful benchmark. Eighty-two percent of participating carriers had implemented accelerated underwriting in some form. The average throughput rate, meaning the percentage of applications qualifying for the accelerated path, was around 59 percent.

On mortality experience, the survey found that carriers with more mature AUW programs and broader digital data integration reported better mortality outcomes. The survey also documented a continued expansion of eligibility limits, with several carriers raising the face amount threshold for accelerated underwriting eligibility, a move that requires actuarial confidence in the underlying data's ability to identify risk at higher coverage amounts.

Munich Re's parallel research, published in late 2024, noted that the AUW landscape had begun to stabilize in terms of program structure, but digital health data tool usage continued to expand. The stabilization is worth noting for actuaries building long-term assumptions. Programs aren't getting redesigned every year anymore; they're getting better data inputs plugged into established frameworks.

How digital health data changes mortality tables

Traditional mortality tables are built on large populations with relatively coarse segmentation. Digital health data enables finer risk stratification, but it also introduces complications that actuaries need to work through.

The first is selection effects. Applicants who consent to EHR data sharing may differ systematically from those who decline. If healthier applicants are more willing to share (because they have nothing to hide), then the mortality improvement attributed to EHR data may partly reflect selection rather than superior risk identification. Separating genuine predictive value from selection bias requires careful study design, and most published research hasn't fully addressed this.

The second complication is data consistency. EHR data quality varies significantly by provider, region, and health system. Lab values might be reported in different units. Diagnosis codes might be applied inconsistently. An actuarial model trained on one carrier's EHR data pipeline may not transfer cleanly to another carrier sourcing from different health systems.

The third is the question of longitudinal versus point-in-time assessment. Traditional underwriting captures a snapshot. Digital health data, especially from EHRs and prescription histories, can capture trends over time, like worsening A1C levels or escalating pain medication usage. This temporal dimension is valuable but requires actuarial models that can incorporate trajectory, not just current state.

Mortality assessment approach	Data granularity	Temporal depth	Key limitation
Traditional (exam + APS)	High clinical detail at one point	Snapshot only	Slow, expensive, applicant friction
Accelerated (no exam, questionnaire + Rx)	Moderate	Limited history	Mortality slippage risk
AUW + single DUE source	Moderate to high	Varies by source	Incomplete risk picture
AUW + combined DUE sources	High	Multi-year history possible	Data consistency across sources
Continuous underwriting (emerging)	Potentially very high	Ongoing	Privacy, regulatory, model maturity

The Society of Actuaries perspective

The Society of Actuaries has been tracking mortality trends through its Quarterly Mortality Monitoring Reports, with the most recent U.S. population report published in early 2026 covering data through Q3 2025. While these reports focus on population-level mortality rather than underwriting-specific data, they provide the baseline that actuarial mortality assumptions ultimately reference.

The SOA's Product Development Section published research in August 2024 examining accelerated underwriting mortality slippage monitoring trends. The research highlighted that carriers are moving beyond simple pass/fail AUW audits toward more sophisticated monitoring frameworks that track mortality outcomes by data source, underwriting path, and applicant characteristics. This shift reflects growing actuarial sophistication in how digital health data's impact on mortality is measured and managed.

For actuaries working on valuation or pricing, the practical takeaway is that mortality assumptions for AUW blocks need to be segmented by the digital data sources used. A carrier using only Rx data in its AUW program will have a different mortality profile than one using Rx plus EHR plus medical claims. Treating all AUW business as a single mortality class is increasingly inadequate.

Where contactless vitals screening fits

Camera-based vital sign measurement through rPPG is the newest entrant in this space. The technology uses a smartphone camera to detect subtle skin color changes caused by blood flow, extracting heart rate, respiratory rate, and blood pressure indicators from a short video recording. A 2022 hospital-based trial published in the Journal of Clinical Monitoring and Computing demonstrated 96 percent agreement between rPPG respiratory rate measurements and standard clinical methods across 963 patients.

For actuaries, the interesting question about rPPG isn't whether it can replace a full paramedical exam. It can't, at least not yet. The interesting question is whether it adds incremental predictive value when layered on top of existing digital data sources. An applicant's resting heart rate and respiratory rate, captured without any equipment, could serve as an additional data point in risk scoring models. Elevated resting heart rate, for instance, is an established cardiovascular risk marker in epidemiological research.

Companies like Circadify have developed contactless vital sign measurement capabilities for this use case, bringing rPPG-based screening to the insurance industry as a frictionless data collection method. The actuarial validation work is still early, but the trajectory looks familiar. Prescription data followed the same path a decade ago, and EHRs are going through it now.

Practical considerations for actuarial teams

Actuaries evaluating digital health data for mortality assumptions should consider several practical factors.

Data access and consent rates vary. Not every applicant will consent to EHR sharing, and consent rates differ by demographics, geography, and product type. Actuarial models need to account for the population that opts out and the potential adverse selection implications.

Vendor dependencies are real. The digital health data ecosystem is concentrated among a few providers. LexisNexis, RGA, Milliman, and a handful of specialty vendors control much of the data pipeline. Carrier actuaries should understand the specific data sources, coverage rates, and quality characteristics of their vendor's products rather than assuming all "EHR data" is equivalent.

Regulatory considerations are evolving. State insurance regulators are paying closer attention to algorithmic underwriting and the data sources that feed it. Colorado's AI governance law, which went into effect in 2024, requires insurers to test for unfair discrimination in automated decisioning. Actuaries need to ensure that mortality models using digital health data can be explained and tested for bias.

Model validation requirements will increase. As digital health data moves from a nice-to-have supplement to a core underwriting input, regulators and rating agencies will expect more rigorous validation of the mortality assumptions built on this data. Actuaries should be building the validation infrastructure now rather than retroactively.

Frequently asked questions

How much does digital health data actually reduce mortality slippage?

RGA's research shows that the reduction varies by data source and carrier, but combining multiple sources, particularly EHRs with medical claims and prescription data, produces the most meaningful improvement. The exact magnitude depends on the carrier's specific book, underwriting rules, and population mix. RGA recommends carrier-specific analysis rather than relying on industry averages.

Should actuaries treat accelerated underwriting mortality differently from traditional?

Yes. Gen Re's 2024 survey data and RGA's mortality research both indicate that AUW mortality profiles differ from traditional full underwriting, and the magnitude of that difference depends on the digital data sources employed. Actuaries pricing or reserving for AUW blocks should segment mortality assumptions by underwriting path and data sources used.

What's the biggest risk of relying on digital health data for mortality assumptions?

Data availability bias. The applicants for whom digital health data is available and complete may differ systematically from those for whom it's sparse or absent. Building mortality assumptions primarily on the data-rich population without adjusting for the data-poor population can lead to mispriced risk. This is an area where actuarial judgment still matters more than model output.

How will contactless biometric screening affect actuarial mortality models?

It's early. The technology works, and the clinical validation is building, but actuarial-grade mortality studies using rPPG data as a predictive variable haven't been published yet. The most likely near-term path is as an incremental data point within existing risk scoring frameworks rather than a standalone mortality predictor.

Carriers and actuarial teams interested in how contactless vital sign screening fits into digital underwriting workflows can explore emerging solutions at circadify.com/industries/payers-insurance.