What Actuaries Need to Know About Contactless Vitals Data

Actuaries evaluating contactless vitals data for life insurance underwriting are walking into genuinely unfamiliar territory. The technology behind it, remote photoplethysmography or rPPG, captures heart rate, heart rate variability, respiratory rate, and blood pressure indicators through a smartphone camera. No cuff, no finger clip, no nurse visit. An applicant looks at their phone for 30 to 60 seconds, and the system reads subtle color changes in facial skin caused by blood flow beneath the surface. The data comes back in near real-time. For actuaries accustomed to lab panels and paramedical exam results that have decades of mortality correlation behind them, the obvious question is: how credible is this data, and what can you actually do with it in a pricing model?

"rPPG-derived pulse rate showed strong agreement with ECG, with a mean absolute error of 1.061 bpm, root-mean-squared error of 2.845 bpm, and Pearson correlation of 0.962." — Clinical validation study published in PMC (2025), n = 817 samples

How rPPG Actually Produces Vital Sign Data

Before getting into actuarial applications, it helps to understand what the technology is measuring and how. rPPG works by detecting micro-variations in skin color that occur with each heartbeat. When blood pulses through capillaries near the skin surface, it changes how light reflects off the face. These changes are invisible to the naked eye, but a phone camera picking up 30 frames per second can detect them. Signal processing algorithms then extract a blood volume pulse waveform from the video, and from that waveform, the system derives several metrics.

Heart rate is the most straightforward extraction and the most thoroughly validated. Heart rate variability, which measures the variation in time between successive heartbeats, requires cleaner signal quality but carries significant clinical and actuarial value because HRV correlates with autonomic nervous system function, stress load, and cardiovascular risk. Respiratory rate comes from the modulation of the blood volume pulse signal by breathing. Blood pressure estimation is more complex and less mature, relying on pulse transit time proxies and machine learning models trained against reference measurements.

A 2025 review published in Frontiers in Digital Health by researchers examining the full scope of rPPG health assessment found that 81.4% of the bibliography in their systematic review was published between 2015 and 2025, reflecting how rapidly this research area has expanded. Heart rate, HRV, blood pressure, and respiratory rate are the four biomarkers most consistently measured with rPPG, with heart rate achieving the highest accuracy across studies.

What the Validation Data Actually Shows

Actuaries care about measurement error, and rPPG studies have gotten specific enough to be useful.

A large clinical validation published in PMC in 2025 tested rPPG pulse rate measurement against ECG across 817 samples. The mean absolute error was 1.061 bpm. Mixed-effects regression analysis showed that accuracy held steady across device types (iPhone and iPad), ambient lighting conditions, age groups, and gender. Disease-stratified analysis found no significant effect of hyperlipidemia or type 2 diabetes on rPPG accuracy, which matters for insurance applicant populations that skew toward those conditions.

A separate 2025 review in PMC examining contactless vital sign monitoring across multiple datasets reported that heart rate accuracy via rPPG achieved MAE between 0.23 and 5 bpm depending on methodology, with clinical datasets like BIDMC PPG reaching 98.3% accuracy for heart rate and 93.8% for respiratory rate.

Vital sign	Measurement method	Reported accuracy (MAE/correlation)	Actuarial relevance
Heart rate	Blood volume pulse extraction	MAE 0.23–5 bpm; r = 0.962 vs ECG	Direct cardiovascular risk indicator
Heart rate variability	Inter-beat interval analysis	Comparable to PPG sensors in controlled settings	Autonomic function, stress, cardiac risk
Respiratory rate	Pulse signal modulation	93.8% accuracy on clinical datasets	Pulmonary health, COPD indicators
Blood pressure	Pulse transit time + ML models	Less mature; active research area	Hypertension screening at application

The blood pressure piece is the one actuaries should watch most carefully. It is the metric with the most actuarial value for mortality prediction but also the one where rPPG accuracy is still catching up to clinical standards. Shen.AI, one of the companies in this space, registered a clinical trial (NCT06829615) specifically to validate their camera-based blood pressure measurements against reference standards, suggesting the industry itself recognizes this gap and is working to close it.

Where Contactless Vitals Fit in the Actuarial Data Stack

Here is the part that matters most for pricing and reserving. Contactless vitals data is not replacing the actuarial data sources that already work. It is filling a specific gap: providing real-time physiological data at the point of application, without the cost and friction of a paramedical exam.

RGA's 2025 research on digital underwriting evidence tested how different data sources reduce mortality slippage in accelerated underwriting programs. They found that electronic health records had the largest single-source mortality impact, followed by claims data and lab databases. Contactless biometric screening sits in a different part of the stack. It does not replace EHR or Rx data. Instead, it captures present-state physiological measurements that those historical data sources miss.

Think of it this way. EHRs tell you what a doctor diagnosed in the past. Prescription data tells you what medications someone takes. Claims data shows healthcare utilization patterns. Contactless vitals tell you what the applicant's cardiovascular system is doing right now, at the moment they apply. For an actuary building a mortality model, that is a genuinely different signal.

Data source	Time orientation	What it captures	Mortality signal type
Electronic health records	Historical	Past diagnoses, labs, visits	Known conditions, severity
Prescription histories	Recent historical	Current medications	Active condition management
Medical claims	Historical	Procedures, utilization	Healthcare engagement patterns
Contactless vitals (rPPG)	Real-time	HR, HRV, RR, BP indicators	Present cardiovascular state
Wearable data	Longitudinal	Activity, sleep, continuous HR	Behavioral and ongoing health

The actuarial question is not whether contactless vitals replace other data. They do not. The question is whether adding a real-time physiological snapshot at application improves mortality prediction beyond what historical data sources already provide. That is an empirical question, and it will take several years of experience data before actuaries can answer it definitively for their own books.

Credibility and Sample Size Concerns

Any actuary evaluating a new data source runs into the credibility problem. How much data do you need before you can assign credible mortality assumptions to it?

Gen Re's 2024 U.S. Individual Life Accelerated Underwriting Survey, covering 38 carriers, found that 82% had implemented accelerated underwriting with an average throughput rate of about 59%. But most programs have been running for less than a decade, which means the mortality experience data on digitally underwritten lives is still thin compared to the decades of experience behind traditional full underwriting. Contactless vitals, being even newer than EHR-based digital underwriting, have even less experience data behind them.

Pacific Life's 2026 Underwriting Outlook Survey, polling over 100 underwriting and insurance technology professionals, found continued growth in AI and digital data adoption but acknowledged that the talent pipeline for actuaries and underwriters who understand these tools remains a concern. This is a real problem. The data is arriving faster than the profession's capacity to evaluate it.

For actuaries at carriers considering contactless vitals, the practical approach is probably limited credibility weighting. Use the rPPG data as a supplementary risk signal alongside established digital underwriting evidence, and increase its weight in your models as experience accumulates. Do not build pricing assumptions on contactless vitals alone until you have enough policy-years to support it.

Regulatory Considerations Actuaries Cannot Ignore

The NAIC's Accelerated Underwriting Working Group published draft regulatory guidance in 2024 that specifically addressed non-traditional data in life insurance underwriting. Their scope includes biometric data captured through digital means, which encompasses contactless vitals. Actuaries need to be aware that state insurance regulators are actively developing frameworks for how this data can be used in underwriting decisions, and those frameworks may impose constraints on what variables can feed into pricing models.

The regulatory guidance flagged concerns about transparency, consumer consent, and potential disparate impact. For actuaries, the disparate impact question is the one to track most closely. If rPPG accuracy varies by skin tone, lighting conditions, or device quality, and those factors correlate with protected classes, regulators will have questions. The clinical validation data so far is encouraging on this front. The 2025 PMC study found no significant accuracy differences across demographic groups. But actuaries should not assume that finding will hold across all implementations and all populations.

State-level biometric data privacy laws add another layer. Illinois BIPA, Texas CUBI, and similar statutes in Washington state regulate how biometric data is collected, stored, and used. Actuaries may not deal with compliance directly, but they need to understand that data availability may vary by jurisdiction, which affects the uniformity of underwriting programs and the resulting mortality experience.

What HRV Data Means for Risk Stratification

Heart rate variability deserves its own discussion because it is the contactless vital sign with the most unexplored actuarial potential. HRV reflects autonomic nervous system balance, specifically the interplay between sympathetic (fight-or-flight) and parasympathetic (rest-and-digest) activity. Low HRV has been associated in clinical literature with increased cardiovascular mortality, diabetes complications, and overall mortality in both healthy populations and those with existing conditions.

For actuaries, HRV is interesting because it captures something that traditional underwriting inputs miss: a real-time indicator of physiological stress and autonomic health. Someone with controlled blood pressure on medication might still have depressed HRV, suggesting their cardiovascular system is under more strain than the BP reading alone would indicate.

The challenge is that HRV norms vary significantly by age, fitness level, and time of day. A single rPPG-derived HRV measurement at the point of application provides a snapshot, not a trend. Actuaries should be cautious about over-interpreting a single measurement. But as a screening tool that flags applicants for further review, HRV adds a signal that blood tests and medical records do not provide.

The Future of Contactless Vitals in Actuarial Work

The trajectory here is fairly clear. Camera-based vitals measurement is getting more accurate, more validated, and cheaper to deploy. The 2025 Frontiers in Digital Health systematic review noted that the field is expanding rapidly, with new studies addressing edge cases like varying skin tones, ambient lighting, and motion artifacts. As these technical challenges get resolved, the data quality will improve and so will actuarial confidence in it.

The real shift will come when carriers have enough policy-years of experience with contactless vitals to run proper A/B mortality studies: comparing books underwritten with and without rPPG data. That is probably three to five years away for early adopters, longer for the rest of the industry. Until then, actuaries should treat contactless vitals as a promising supplementary signal with limited credibility, not as a replacement for established data sources.

Solutions like Circadify are working on making contactless vitals capture available to insurers through SDK and API integrations, which would lower the barrier for carriers to start collecting this data and building the experience base that actuaries need.

Frequently Asked Questions

How accurate is rPPG for measuring heart rate compared to clinical devices?

Clinical validation studies show rPPG heart rate measurement achieves a mean absolute error of roughly 1 to 5 bpm compared to ECG, with Pearson correlations above 0.96. The 2025 PMC validation study with 817 samples found MAE of 1.061 bpm. Accuracy holds across different devices, lighting conditions, and demographic groups based on current evidence.

Can actuaries use contactless vitals data for pricing today?

Most actuaries would not assign full credibility to contactless vitals data for pricing at this stage. The technology is validated for measurement accuracy, but the mortality experience data connecting rPPG measurements to actual claims outcomes is still too thin. The practical approach is supplementary credibility weighting alongside established digital underwriting evidence like EHR and Rx data.

Does rPPG accuracy vary by skin tone or demographics?

The 2025 clinical validation study found no significant accuracy differences across age groups, gender, or conditions like hyperlipidemia and type 2 diabetes. Research on skin tone variation is ongoing, but early results are encouraging. Actuaries should monitor this closely because demographic accuracy differences could raise regulatory and disparate impact concerns.

What vital signs can rPPG measure that matter for underwriting?

rPPG reliably measures heart rate and respiratory rate. Heart rate variability, which correlates with cardiovascular risk and autonomic function, is measurable in controlled conditions. Blood pressure estimation is the most actuarially valuable metric but also the least mature. Actuaries should expect blood pressure accuracy to improve as clinical trials like NCT06829615 generate validation data.