Behavioral and Biometric Data Combined: Better Risk Assessment?

Life insurance underwriting has always been about filling in a picture of someone's health. For decades, that picture came from two sources: what applicants said about themselves, and what a blood draw confirmed. Both had limits. Questionnaires invited misrepresentation. Lab panels captured a single moment. But now, two new categories of data are converging in underwriting workflows, and their combination is producing behavioral biometric data risk assessment models that outperform either data stream on its own. Behavioral signals from wearables and digital footprints, layered on top of objective biometric measurements from technologies like remote photoplethysmography, are starting to reshape how carriers think about mortality prediction.

"Wearable-derived physical activity data, when combined with traditional risk factors, improved all-cause mortality prediction accuracy with machine learning models achieving AUC values above 0.90." — Queen's University Belfast, UK Biobank study (2024)

What Behavioral Data Actually Means in Underwriting

The term "behavioral data" gets thrown around loosely in insurtech circles. In the context of life insurance risk assessment, it refers to digitally captured patterns of how people live, not what they report about how they live.

Physical activity is the most studied behavioral signal. The UK Biobank accelerometer sub-study tracked over 95,000 participants wearing wrist-mounted activity monitors for seven days. Researchers at Queen's University Belfast used this dataset to build machine learning mortality prediction models that incorporated step counts, sedentary time, moderate-to-vigorous physical activity minutes, and sleep duration alongside traditional demographic and medical variables. The combined model reached AUC values above 0.90, a meaningful improvement over models using traditional variables alone.

Activity intensity matters more than volume. A 2022 study published in Nature Medicine by Emmanuel Stamatakis and colleagues at the University of Sydney analyzed wearable data from 25,241 UK Biobank participants who reported no structured exercise. They found that brief bouts of vigorous intermittent lifestyle physical activity — as short as one to two minutes each — were associated with a 38 to 40 percent reduction in all-cause mortality risk and a 48 to 49 percent reduction in cardiovascular mortality risk. The finding is significant for underwriting because it suggests that granular activity pattern data captures risk signals invisible to questionnaire-level inquiries like "Do you exercise regularly?"

Sleep is another behavioral dimension worth watching. Duration, consistency, and timing of sleep all carry mortality associations in the epidemiological literature. Sleep data from wearables gives underwriters a continuous, objective signal — a far cry from asking someone to estimate how many hours they sleep per night on a form.

Digital engagement patterns — app usage, screen time, and browsing behavior — sit at the more controversial end of the behavioral spectrum. Some insurtech firms have explored these signals, but privacy concerns and regulatory scrutiny under frameworks like the NAIC Model Bulletin on AI have made carriers cautious about incorporating them into formal underwriting decisions.

Biometric Data: The Physiological Counterpart

Where behavioral data describes what a person does, biometric data describes what their body is doing at a physiological level. In the insurance context, this means objective measurements of cardiovascular, respiratory, and metabolic function.

Heart rate and heart rate variability remain the most actuarially validated biometric signals. RGA's collaboration with the University of Leicester on the UK Biobank dataset — analyzing over 407,000 participants — confirmed that resting heart rate was among the non-traditional factors that "dramatically improved the ability to differentiate mortality and morbidity risks." Low heart rate variability, which reflects reduced autonomic nervous system flexibility, has been independently linked to elevated cardiovascular mortality across multiple large cohort studies.

Blood pressure, respiratory rate, and oxygen saturation round out the core biometric measurement set. Remote photoplethysmography technology now captures these signals from a smartphone camera in under 60 seconds, without any physical device beyond the phone itself. Hospital validation studies published in PMC have confirmed that rPPG achieves strong agreement with clinical-grade instruments for respiratory rate measurement across a wide physiological range.

The advantage of biometric data over behavioral data is directness. A heart rate variability reading reflects actual autonomic function. A step count reflects behavior that correlates with cardiovascular fitness but does not measure it. Biometric data is also harder to game — you can wear a fitness tracker on your dog's collar, but you cannot fake your resting heart rate during a camera-based scan.

Why Combination Outperforms Either Alone

The case for combining behavioral and biometric data comes down to something simple: they measure different things, and those things are only loosely correlated.

Consider two applicants with identical resting heart rates of 72 bpm. One is sedentary, averaging 3,000 steps per day with minimal vigorous activity. The other averages 9,000 steps and engages in regular high-intensity activity. Their biometric snapshots look similar, but their behavioral profiles suggest very different risk trajectories. The sedentary applicant's cardiovascular fitness is likely declining. The active applicant is maintaining or building reserve capacity that protects against future cardiac events.

Now reverse the scenario. Two applicants with similar step counts — both averaging 8,000 steps per day — but one has a resting heart rate of 58 bpm with high HRV, while the other sits at 82 bpm with low HRV. Same behavior, different physiology. The second applicant may have an underlying cardiovascular condition that activity data alone would not flag.

Munich Re's fourth biennial Accelerated Underwriting Survey, conducted in fall 2024, documented growing carrier interest in combining data streams. The survey found that adding electronic health records to claims data changed underwriting decisions in approximately 12% of cases. When behavioral data from wearable sources was layered on top, the differentiation improved further — particularly for applicants in the "gray zone" between standard and preferred risk classes where traditional data provided insufficient separation.

Behavioral vs Biometric vs Combined: Data Characteristics

Characteristic	Behavioral Data (Wearables)	Biometric Data (rPPG/Sensors)	Combined Model
What it measures	Activity patterns, sleep, sedentary time	Heart rate, HRV, blood pressure, SpO2	Both physiological state and lifestyle patterns
Temporal coverage	Continuous (days to months)	Point-in-time or periodic	Longitudinal behavior + current physiology
Mortality prediction	Moderate (activity is a proxy for fitness)	Strong (direct physiological measurement)	Strongest (complementary signals reduce blind spots)
Fraud resistance	Low-moderate (device can be manipulated)	High (physiological signals difficult to fake)	High (cross-validation between streams)
Applicant friction	Requires wearable device and sharing consent	30-60 second camera scan	Moderate (two data collection points)
Regulatory clarity	Evolving (NAIC AI guidance, state biometric laws)	Evolving (same regulatory environment)	More complex (two data governance frameworks)
Best use in underwriting	Longitudinal risk trajectory, lifestyle classification	Instant risk snapshot, cardiovascular screening	Full-spectrum risk profiling
Data latency	Requires days-to-weeks of wear time	Available in under 60 seconds	Biometric available instantly; behavioral requires accumulation

Industry Applications: How Carriers Are Building Combined Models

Accelerated Underwriting Triage

The most immediate application is triage in accelerated underwriting programs. Carriers are using biometric data at the point of application to make an initial risk classification, then incorporating behavioral data from wearable integrations for applicants who fall into borderline risk categories. This two-stage approach reduces the number of cases requiring traditional underwriting review while maintaining risk selection accuracy.

Munich Re's Automated EHR Summarizer, built on their 2022 acquisition of Clareto, represents the data infrastructure side of this trend. The platform aggregates electronic health data and provides triage guidance to underwriters — identifying which cases need human review and which can proceed through automated pathways. Adding behavioral signals to this triage workflow creates a third verification layer that catches risks neither medical records nor biometric snapshots would surface alone.

Dynamic Underwriting and Post-Issue Monitoring

Some carriers are exploring models where behavioral data collected after policy issuance informs ongoing risk assessment. The logic is intuitive: if a policyholder's activity levels drop off substantially over six months, that trajectory says something about changing health status. Paired with periodic biometric check-ins — a quarterly rPPG scan, for instance — this creates a living risk profile rather than the static snapshot that traditional underwriting produces.

RGA's collaboration with the University of Leicester explicitly examined this forward-looking dimension. Dr. Tom Sherwood Maycock, an associate professor at Leicester, noted that "the extension into wearable technology through smartwatches and mobile phones is poised to further revolutionize the insurance industry by introducing real-time data tracking and more dynamic risk assessment."

Reinsurance Treaty Pricing

Reinsurers are watching this convergence closely. Combined behavioral-biometric data provides reinsurers with granular portfolio-level risk information that was previously unavailable. If a ceding company can demonstrate that its book of business includes policyholders with verified activity levels and favorable biometric profiles, the reinsurer can price the treaty more aggressively. This creates a financial incentive loop: carriers that collect better data get better reinsurance terms, which funds further data collection investment.

Current Research and Evidence

The evidence base for combined behavioral-biometric models in insurance is still young, but several large-scale studies provide the foundation.

The Queen's University Belfast UK Biobank study (2024) remains the most comprehensive wearable-mortality analysis to date. Lead researcher and team analyzed over 95,000 participants with seven-day accelerometer data, demonstrating that machine learning models incorporating wearable-derived features achieved substantially higher predictive accuracy than traditional actuarial variables alone. The study was published in Expert Systems with Applications.

Stamatakis et al. (2022) in Nature Medicine established that vigorous intermittent lifestyle physical activity — captured by wearables but invisible to questionnaires — carries independent mortality risk reduction. The dose-response relationship they documented (median 4.4 minutes per day associated with 26-30% all-cause mortality risk reduction) provides a concrete behavioral signal that underwriters can operationalize.

Munich Re and Klarity's UK Biobank collaboration assessed non-traditional variables including resting heart rate, walking pace, and grip strength for mortality differentiation. Their findings confirmed that "non-traditional variables dramatically improved the ability to differentiate mortality and morbidity risks," with the strongest gains coming from variables that captured both behavioral and physiological dimensions.

A 2026 medRxiv preprint from researchers using UK Biobank data developed interpretable machine learning models for 10-year cardiovascular risk prediction using lifestyle-based factors. The study demonstrated that behavioral data alone — physical activity, diet, sleep, smoking status — predicted cardiovascular disease, heart failure, and atrial fibrillation with clinically meaningful accuracy, particularly when processed through gradient-boosted tree models.

The Future of Combined Risk Assessment

The trajectory here is clear, even if the pace of adoption remains uncertain. Biometric measurement technology has already solved the friction problem — a 30-second smartphone scan captures cardiovascular data that once required a nurse visit. Wearable penetration continues climbing, with an estimated 1.1 billion connected wearable devices globally as of 2025. The missing piece is not data availability but data integration: building underwriting models that ingest both streams, weight them appropriately, and produce risk classifications that actuaries and regulators accept.

Three things will likely speed up adoption. Reinsurer validation studies comparing combined-model portfolios against traditional ones will give carriers the actuarial evidence they need to justify the spend. Regulatory clarity around wearable behavioral data in insurance decisions — the NAIC is actively developing guidance here — will reduce legal uncertainty. And platform consolidation will make it technically easier for carriers to receive biometric and behavioral data through a single integration instead of juggling multiple vendor relationships.

Carriers that figure out integration first will have a real underwriting edge — better risk selection, lower claims costs, sharper pricing for healthy applicants, and a data asset that compounds over time. Carriers that wait will end up on the wrong side of adverse selection, writing the risks that data-equipped competitors passed on.

Frequently Asked Questions

How does behavioral data differ from biometric data in insurance underwriting?

Behavioral data captures patterns of activity, sleep, and lifestyle through wearable devices and digital interactions. Biometric data measures physiological signals like heart rate, blood pressure, and oxygen saturation directly. Behavioral data tells you what someone does; biometric data tells you how their body is functioning. Both carry mortality risk information, but they measure different dimensions — which is why combining them produces stronger risk models than either alone.

Can behavioral data from wearables actually predict mortality?

Yes. The most robust evidence comes from the UK Biobank accelerometer sub-study, where researchers at Queen's University Belfast demonstrated that wearable-derived physical activity features improved all-cause mortality prediction when added to traditional risk variables. Separately, Stamatakis et al. (2022) showed in Nature Medicine that even brief bouts of vigorous daily activity — captured by wearables — were associated with 38-40% reductions in all-cause mortality risk.

What biometric signals are most valuable for underwriting risk assessment?

Resting heart rate and heart rate variability are the most actuarially validated biometric signals, with large-cohort evidence linking them to cardiovascular and all-cause mortality. Blood pressure and respiratory rate provide additional risk information. These signals can now be captured through remote photoplethysmography using a smartphone camera, eliminating the need for dedicated medical equipment.

Are there privacy concerns with combining behavioral and biometric data?

There are, and they are being actively addressed. The NAIC Model Bulletin on AI provides guidance on the use of algorithmic data in insurance decisions. State-level biometric privacy laws — most notably in Illinois, Texas, and Washington — add additional requirements around consent and data handling. Carriers building combined models need robust data governance frameworks that address both biometric and behavioral data streams, including clear applicant consent mechanisms and transparent data usage policies.

---

Solutions that combine behavioral and biometric data streams are already entering the market. Circadify's rPPG technology captures cardiovascular biometric data from a smartphone camera in under 60 seconds, providing the physiological measurement layer that complements wearable behavioral data for comprehensive risk assessment.