How We Built Our GenAI Solution for Remote Patient Monitoring

Behind the Scenes of Validic's Responsible, Scalable GenAI Development

Until fairly recently, the integration of generative AI in healthcare was more aspirational than practical. The gap between AI's promise and responsible clinical implementation remained significant — particularly where decisions directly impact patient outcomes.

In healthcare, the stakes couldn't be higher. When GenAI generates output, it isn't automatically factual, and inaccurate information can cascade throughout care delivery. Unlike other industries where AI errors might lead to inconvenience, in clinical settings they could affect diagnoses, treatment plans, and ultimately, patient lives.

Trust forms the foundation of every clinical relationship. In 2023, nearly 75% of healthcare organizations reported exploring or implementing AI technologies, yet fewer than 30% expressed high confidence in deploying these tools responsibly at scale. This gap represents both a challenge and an opportunity — because when built thoughtfully, GenAI can address healthcare's most pressing challenges: provider burnout, data overload, and the need for personalized interventions.

Building on a Foundation of Rich Patient Data

At Validic, we connect patients and providers through personal health devices — blood pressure monitors, weight scales, glucometers, and more — creating continuous streams of clinical measurements between office visits. This connectivity has transformed the data landscape from infrequent clinical snapshots to rich, longitudinal datasets revealing day-to-day patient conditions. A hypertensive patient might submit 60+ blood pressure readings in three months versus a single in-office measurement.

This data density creates both opportunity and challenge: unprecedented visibility into health trends alongside the challenge of helping providers process this abundance within existing clinical zworkflows.

Our Health IoT integration ensures reliable data pipelines across hundreds of devices and millions of connections. This longitudinal data creates the essential foundation for meaningful GenAI applications — while single measurements show status, patterns over time reveal trajectories and early warning signs that might otherwise remain invisible.

GenAI excels precisely at this pattern recognition — finding signals within noise and surfacing meaningful trends from thousands of measurements. Our unique position, with access to rich longitudinal data, enables us to develop AI solutions that reflect real-world patient conditions rather than point-in-time snapshots.

Our Development Approach: Pragmatic and Flexible

Developing healthcare GenAI solutions requires balancing immediate clinical needs with a rapidly evolving AI landscape. We created flexible architectural abstractions that allow adaptation as technologies mature—avoiding lock-in while delivering immediate value.

After evaluating multiple options, we selected AWS Bedrock for our initial implementation, providing access to foundation models from various vendors, including Anthropic's Claude and Amazon's offerings. This flexibility enables matching specific clinical use cases with appropriate underlying models while maintaining a consistent integration layer.

Before writing code, we validated our approach with current and prospective clients. These conversations confirmed our proposed use cases would genuinely enhance clinical workflows rather than adding complexity. Care teams specifically highlighted the value of summarizing longitudinal data trends and flagging potential intervention points—tasks that are time-consuming when performed manually.

Our implementation prioritizes:

• Rapid integration with existing workflows through established EHR connections

• Flexible model selection based on use case requirements

• Deployment architectures maintaining data privacy and security

• Performance monitoring ensuring response times align with clinical workflows

We've designed guardrails including confidence thresholds determining when insights should be presented, mechanisms preventing hallucinated content from reaching clinicians, and clear attribution of source data. Human-in-the-loop validation ensures clinicians maintain control over how insights are applied, acknowledging AI as a tool augmenting rather than replacing clinical judgment.

Embedded Where It Matters: Within Existing Workflows

Healthcare providers already navigate multiple systems daily, and adding another dashboard would create friction rather than benefit. Our fundamental approach: if GenAI insights aren't accessible within existing clinical workflows, their impact remains limited. The best insights are worthless if they require additional steps to access.

EHR integration leverages our established connectivity across major platforms, delivering GenAI insights directly where clinical decisions happen—contextualizing patient-generated data alongside other clinical information. This eliminates the "swivel chair" problem where clinicians must pivot between systems to gather a complete patient picture.

No new dashboards—just better decision support in existing workflows. For care managers reviewing daily RPM data, automatically generated summaries appear alongside raw measurements. For physicians preparing for patient visits, trend analyses appear within their standard chart review process, creating a seamless experience that respects their limited time.

Clinical Use Cases: Delivering Tangible Value

Our initial deployment focuses on three core use cases where longitudinal data, workflow integration, and GenAI capabilities intersect most effectively:

Face-up summaries provide immediate situational awareness tailored to each monitoring program. For diabetes management, these summaries highlight glycemic control patterns, while hypertension programs emphasize blood pressure trends across different times of day. This approach allows clinicians to grasp patient status within seconds rather than minutes.

Trend analysis examines both long-term trajectories and comparative periods, identifying whether recent measurements represent significant changes from baseline, seasonal patterns, or responses to interventions. This helps distinguish between normal variability and clinically meaningful shifts warranting attention.

Attention prioritization suggests which patients may benefit most from clinical intervention by considering trend velocity, pattern changes, and measurement consistency—identifying patients whose conditions may be subtly deteriorating or substantially improving. This intelligent triage helps care teams allocate limited time for maximum impact.

Early Results and What's Next

Feedback from clinical partners has revealed two distinct use patterns. First, clinicians find significant value in GenAI insights when preparing for specific patient interactions. These "chart preparation" scenarios allow providers to quickly absorb contextual information without manually reviewing dozens of measurements. With these features, we hope to hear, "I can now get the same understanding in 30 seconds that previously took 3-5 minutes of chart review."

Second, we discovered a need for "out of band" insight delivery for programs where providers don't routinely review individual charts. There's interest in proactively surfacing insights at both patient and population levels—prioritized notifications about changing conditions or program-level insights identifying broader trends across patient cohorts.

We're prioritizing "face-up insights" first, providing immediate workflow benefits with minimal disruption. This will be followed by prioritization tools and eventually expanded to include program-level analytics.

Client input drives our roadmap prioritization. Rather than pursuing technological capabilities for their own sake, we're focusing on workflows and use cases our partners identify as most valuable. This ensures our GenAI innovations address genuine clinical needs rather than creating solutions seeking problems.

Moving forward, we're focused on measuring impact—examining how these tools affect care team efficiency, clinical decision-making, and ultimately, patient outcomes. This evidence-based approach will help refine and expand our GenAI capabilities in ways that deliver meaningful clinical value.

Want to see our GenAI solution in action and play a key role in what's next? Email us at hello@validic.com.