The following is a guest article by Daniel Tashnek, Co-Founder of Prevounce Health
Remote patient monitoring (RPM) has quietly harbored artificial intelligence (AI) for years, long before technology like ChatGPT made headlines. These were not the sophisticated large language models we know today. Rather, they were rule-based protocols that determined when to alert clinicians about concerning patient data.
Now, as modern AI transforms healthcare delivery, RPM sits at the center of a fascinating evolution that promises both tremendous opportunity and significant responsibility.
RPM Journey: From High-Touch to High-Scale
To understand AI’s current role in RPM, it helps to trace the technology’s path. Early RPM programs focused on high-acuity patients where there was a high short-term chance of a harmful (and expensive) event — think recently discharged heart failure patients with dedicated nurses monitoring their vital signs in near real time. These are relatively high-touch programs where human oversight is essential for safety and positive outcomes.
The landscape shifted during the COVID-19 public health emergency, when RPM quickly expanded to monitor medium-to-high risk patients at much larger scales. Instead of watching a handful of critical patients, programs began tracking hundreds of individuals.
Today’s RPM has evolved further toward chronic disease management, monitoring medium-risk patients where the goal is preventing future complications rather than managing immediate crises. A hypertensive patient’s stroke risk over five years presents a very different monitoring challenge than a transplant recipient’s six-month survival outlook. This change in scale has created the perfect environment for AI to demonstrate its value — not by replacing human judgment, but by enhancing efficiency and catching patterns that might slip by clinicians and care teams, particularly during busy days that are now essentially the norm.
Current AI Applications: Three Key Areas
Several areas have emerged where AI is already making a measurable impact within remote patient monitoring. Below are three key domains where AI is helping transform RPM into a more efficient, scalable, and proactive model of care and support.
Clinical Documentation and Efficiency
The most mature AI applications in RPM focus on what we might call “low-risk, high-impact” improvements to provider workflows. Solutions like automated transcription of patient encounters, structured data extraction from multiple input sources, and intelligent documentation support are saving clinicians and care teams significant time while generally improving data quality.
These tools excel at trend summarization and visualization, using pattern recognition to highlight subtle changes in vital signs and biometrics that busy clinical staff might miss. The key takeaway here isn’t that humans couldn’t spot these trends given enough time and attention. It’s that AI can consistently surface the most important information upfront, allowing more patients to be effectively monitored with the same resources.
AI is also making a difference on the compliance side of RPM documentation, particularly in helping check to make sure each code has been fully furnished while reducing the time spent on manual oversight. These systems can automatically check that clinical staff are spending the appropriate time with the right patients at the right intervals to meet documentation and billing requirements before any codes are billed.
Care Management Support
AI can also be used during the care management aspect of RPM to flag missed opportunities, such as topics that have not been addressed in recent visits and discussions but should be prioritized in future encounters. By automating these checks, AI not only supports accurate and defensible billing but also lowers administrative costs by reducing the need for manual chart audits and follow-ups. The result is a more efficient workflow that keeps programs compliant and more financially sustainable.
Another application of AI during care management is providing recommendations of social and community referrals and opportunities. While speaking to a patient who mentions they are having trouble finding healthy foods, the AI may flag this background information and then remind the care manager about the local Meals on Wheels program or that there is an upcoming nutrition class near the patient.
Predictive Analytics and Risk Stratification
This is where things get both exciting and complex. Consider traditional electrocardiogram (ECG) analysis systems that relied on hundreds of thousands of readings from a single data type. AI-powered RPM can integrate vital signs, clinical notes, patient-reported outcomes, and questionnaire responses into comprehensive risk assessments.
The real innovation lies in AI’s ability to identify previously unrecognized patterns. Perhaps certain combinations of patient responses that wouldn’t individually trigger alerts become clinically significant when analyzed over time. This creates an early warning system for patient deterioration and risk that extends far beyond what human analysis could achieve at scale.
Need for Human Oversight
Here’s where responsible AI implementation becomes crucial. These sophisticated predictive tools need to be deployed with particular care for high acuity patients, where human clinical judgment remains irreplaceable. The concern isn’t just about AI accuracy. It’s also about avoiding automation bias, where clinicians might lean too heavily on AI recommendations and reduce their own vigilance.
But for medium- and lower-risk patients who couldn’t feasibly receive intensive human monitoring given the availability of personnel and costs, the question shifts from “Is AI better than a human?” to “Is AI better than nothing?” At a population scale, having some intelligent monitoring is clearly preferable to no monitoring at all.
This distinction underscores a fundamental principle: AI in RPM should function as decision support, not decision-making. Every system should maintain clear human oversight, with clinicians retaining ultimate responsibility for validating information prior to making patient care decisions.
Challenges and Considerations
As AI becomes more embedded in remote patient monitoring and broader healthcare workflows, it brings not only promise but also new layers of complexity. Here are some of the most pressing challenges organizations must consider as they evaluate and integrate AI into their RPM programs.
The “Black Box” Problem
Even well-validated AI tools can still lack clarity in how they arrive at decisions. During testing of a summarization tool we have under development, we couldn’t force the AI to make obvious errors. In fact, what initially appeared to be AI errors often turned out to be the system correctly flagging human mistakes. While that may sound like a positive outcome — and in many ways it is — it underscores a deeper issue: Even high-performing AI will eventually get something wrong, and we may not know when it happens or understand why.
This unpredictability requires that providers maintain their clinical guard. AI can be convincingly wrong, and care teams must resist the temptation to become overly reliant on these tools, no matter how impressive their track record.
Vendor Selection and Due Diligence
The rapid growth of AI has attracted companies with impressive technology but limited healthcare experience. Just as early RPM saw wearable device businesses entering the market without understanding clinical workflows, today’s AI landscape includes vendors that may lack the expertise necessary for ongoing safe, effective, and compliant care delivery.
Providers with RPM programs must carefully evaluate not just the technology capabilities but also the healthcare experience and clinical understanding of potential AI partners.
Looking Forward: Pairing Innovation With Responsibility
The integration of AI into RPM represents a significant opportunity to improve care delivery, enhance provider efficiency, reduce risk, and extend monitoring capabilities to broader patient populations who would benefit from the service. However, realizing these benefits requires a measured approach that prioritizes patient safety, maintains human oversight, and addresses a host of concerns.
Success will largely depend on providers’ ability to balance the addition of innovation with responsibility, ensuring that AI serves as a powerful tool in support of human clinical judgment rather than a replacement for it.
The future of AI in RPM isn’t about choosing between human and artificial intelligence. It’s about thoughtfully combining both to create remote care monitoring systems that are more effective, more efficient, and more accessible than either could achieve alone.
About Daniel Tashnek
Daniel Tashnek is the Co-Founder of Prevounce Health, a healthcare software and services company that simplifies the provision of preventive medical services, chronic care management, and remote patient monitoring. Daniel is also a practicing healthcare attorney specializing in regulatory compliance, reimbursement, scope of practice, and patient care issues.
No comments:
Post a Comment