The Scientific Foundation for Predictive AI Population Analysis

EXECUTIVE SUMMARY

In December 2025, three major studies identified significant developments across medical research, regulatory policy, and AI validation standards reinforced the scientific foundation underlying InterGen Data's approach to predictive AI.

This document synthesizes these studies and demonstrates how InterGen Data's SOLOMON AI platform aligns with emerging evidence about population-level prediction, regulatory acceptance of de-identified data, and best practices for evaluating predictive models.

Three Supporting Scientific Studies

JAMA

Geographic Data Predicts Health Outcomes

JAMA Network Open research demonstrated that neighborhood socioeconomic status during young adulthood strongly predicts premature mortality decades later. This validates InterGen Data's foundational premise: geographic and socioeconomic attributes, when properly analyzed, generate actionable health and wealth predictions and insights.

Food and Drug Administration (FDA)

De-Identified Population Data Accepted

The FDA's December 2025 guidance eliminated the requirement for identifiable patient data in Real-World Evidence submissions. This regulatory shift opens the door for InterGen Data's Bayesian cohort-based inference model, which generates individual-level predictions from de-identified population data without ever handling personally identifiable information.

The Lancet

Validation Standards Codified

The Lancet Digital Health's STRATOS initiative published definitive guidance on evaluating predictive AI, establishing that clinical utility is the most important performance domain, and identifying 13 commonly used metrics as "improper." InterGen Data's validation methodology directly aligns with The Lancet's recommended framework.

I: JAMA NETWORK OPEN RESEARCH

Your ZIP Code Predicts Your Health

Multiple studies published in JAMA Network Open and related journals have established that geographic and socioeconomic factors are powerful predictors of health outcomes, independent of individual clinical characteristics.

Key Research Findings

• –  Low household income increases mortality risk by 76%

• –  Combined individual and neighborhood disadvantage nearly triples mortality risk (HR =

  2.76)

• –  Neighborhood socioeconomic status during young adulthood predicts premature mortality decades later

• –  Effects are most pronounced among women and persist independent of income levels

• –  COVID-19 eliminated the "Latino Paradox" by exposing structural health inequities

  InterGen Data Alignment

  InterGen Data's SOLOMON AI platform analyzes 320+ million data attributes including geographic, socioeconomic, financial, and demographic indicators. Our predictive models incorporate the exact factors JAMA research identifies as predictive: household income patterns, neighborhood characteristics, employment stability, and life event trajectories. The JAMA research validates that our data foundation is scientifically sound.

II: FDA REAL-WORLD EVIDENCE GUIDANCE

De-Identified Data Gains Regulatory Acceptance

On December 15, 2025, the FDA eliminated a major barrier to using Real-World Evidence (RWE) in regulatory submissions. The agency announced it will now accept RWE without requiring identifiable individual patient data.

What This Unlocks

• –  National cancer registries (NCI SEER data)

• –  Insurance claims databases

• –  Electronic health record networks

• –  Hospital systems databases

• –  Population health datasets with millions of de-identified records

  InterGen Data Alignment

  InterGen Data's architecture was built from the ground up for de-identified population-level inference. Our Bayesian methodology generates individual-level predictions through cohort correlation, without ever processing personally identifiable information. This FDA guidance validates our privacy-preserving approach and opens doors for healthcare applications that were previously constrained by data privacy requirements.

THE INTERGEN DATA METHODOLOGY

1. Client provides de-identified attributes (no names, DOBs, SSNs, or account numbers)

2. SOLOMON matches individuals to historical cohorts using 320M+ attributes

3. Bayesian inference generates probability predictions for 95 health and wealth life events

4. Client receives predictions mapped to their internal UIDs

5. InterGen Data never sees personally identifiable information

III: THE LANCET DIGITAL HEALTH VALIDATION STANDARDS

The Science of AI Model Evaluation

In December 2025, The Lancet Digital Health published a landmark article by the STRATOS initiative evaluating 32 performance measures for predictive AI in healthcare. Their findings establish a new framework for validating clinical prediction models.

Key Findings

• –  13 of 32 commonly used metrics are "improper"

• –  F1 Score is the only metric that fails BOTH key scientific characteristics

• –  Clinical utility is the most important performance domain

• –  Calibration plots are essential for external validation

• –  Net benefit with decision curve analysis should be standard reporting

  The Recommended Framework

InterGen Data Alignment

InterGen Data's validation methodology directly aligns with the The Lancet's recommended framework. We report AUROC with confidence intervals, Bayesian calibration assessments, and decision-focused clinical utility measures. Our predictions are designed for action, not just statistical performance, which is exactly what The Lancet identifies as the most important domain.

IV: THE CONVERGENCE

Why These Three Developments Matter Together

Individually, each of these developments would be significant. Together, these developments illustrate an emerging consensus across three critical domains: scientific research, regulatory policy, and AI evaluation standards.

Each independently supports a core pillar of InterGen Data's methodology—population-level data as a predictive signal, privacy-preserving inference using de-identified data, and rigorous evaluation based on real-world utility rather than simplistic accuracy metrics.

This convergence reduces uncertainty around the viability of population-scale predictive intelligence and signals that the broader ecosystem is moving toward approaches similar to those used by InterGen Data.

What this Means for Enterprise Adoption

For enterprise organizations in healthcare, insurance, and financial services, these developments remove several historical barriers to adopting predictive AI Including: (1) scientific literature now supports the predictive value of population-level socioeconomic data, (2) regulators are increasingly accepting de-identified datasets for evidence generation, and (3) new validation standards emphasize real-world decision value rather than theoretical model accuracy.

Together, these shifts make population-scale predictive intelligence both more feasible and more defensible for regulated industries.

The Competitive Implications

These developments create a new competitive landscape in predictive AI:

• –  Vendors using improper metrics (F1, accuracy alone) will face scrutiny

• –  Privacy-invasive approaches become competitively disadvantaged

• –  Geographic and socioeconomic data becomes a core requirement

• –  Validation standards compliance will be demanded by enterprise buyers

• –  Adoption led by regulated industries (healthcare, insurance, financial services)
  

APPENDIX: CITATIONS

• JAMA Network Open: Multiple studies on neighborhood socioeconomic status and health outcomes, including cardiovascular mortality and premature death research.

• FDA Guidance: FDA News Release, December 15, 2025. FDA Eliminates Major Barrier to Using Real-World Evidence in Drug and Device Application Reviews.

• Lancet Digital Health: Van Calster B, Collins GS, Vickers AJ, et al. Evaluation of performance measures in predictive artificial intelligence models to support medical decisions: overview and guidance. Lancet Digit Health 2025; 7: 100916
  

THE FUTURE OF PREDICTION IS NOW

InterGen Data is a life event prediction AI intelligence company that delivers foresight into 95 major health and wealth events: what may happen, when, and the financial impact. This predictive year-over-year forecasting and intelligence helps organizations drive more personalized engagement and improve outcomes across the client journey.

For more information contact:

Robert Kirk

Founder & CEO Robert.kirk@intergendata.com

Muriel Petri

Chief Revenue Officer muriel.petri@intergendata.com