Real-time vital signs data: a hidden source of value in medical AI

If data represents the next gold rush for health care, a vast treasure trove of it slips away every day. The increased enthusiasm for AI has led to significant investments in novel solutions for health care, with data coming from a variety of sources such as medical charts, imaging, literature, guidelines, and the like. A largely untapped source of valuable data is staring health care practitioners like myself right in the face: monitors that track vital signs.

As an entrepreneur working with the flow of vital sign data for the past couple years, I’m increasingly convinced of its importance. Captured every second by monitors in hospitals, vital signs have immense potential to improve patient care and provide value from AI that investors have been hoping for but have not yet seen. I believe this potential is why BD (Becton, Dickinson and Company) recently acquired Edwards Lifesciences’ critical care unit for $4.2 billion.

As a hospitalist, I’ve observed how continuous monitoring, once reserved for patients in intensive care units, is now expanding to include broader swaths of patients. This shift is driven by advancements in hardware, making monitors smaller, more comfortable, and more affordable.

Rather than taking a blood pressure measurement once every four hours or checking oxygen saturation during rounds, continuous monitoring of vital signs offers immediate insights into a patient’s condition. When health starts to deteriorate and the body’s balance is interrupted, changes in vital signs reveal how the body is trying to compensate. Real-time physiological data from monitors recording blood pressure, oxygen saturation, heart rate, and temperature captures detailed patterns and trends that go beyond simple numerical readings: they include complex signals that need to be interpreted before clinical decisions are made. Depending on the setting, such as the operating room or the post-anesthesia care unit, other parameters may also be monitored.

These data are now abundant and free-flowing in hospitals, yet underused by practitioners. They are also largely overlooked by researchers and companies working on health care artificial intelligence and machine learning.

The sensors used in hospitals are much more accurate than consumer wearables like Oura or Apple Watch. While those devices have roles to play in personal health, hospital-grade monitors offer the depth of data necessary for clinical decision-making. Real-time physiological data capture the subtleties of patient conditions that no human could detect. With sophisticated modeling, they could identify major problems before they happen. Events such as infections, blood clots, and strokes occur frequently in hospitals, and early detection could make a significant difference.

Almost every outcome I care about as a physician can be correlated to a patient’s vital signs.

Where I think most of the value will be added is from the concept of “always on” clinical trials, a concept I recently heard about from Julie Yoo and Vijay Pande, both general partners at Andreessen Horowitz, on A16z’s excellent Raising Health podcast. “Always on” clinical trials refer to a continuous, real-time infrastructure that allows for on-demand analysis of patient data to identify outcomes retrospectively or prospectively. In every hospital, many organic clinical trials could be happening daily, but countless data points are flashing by unused. For these data to become meaningful, they not only need to be collected and stored appropriately, but also need to be tied to two things: precise timing on intervention and outcomes.

This is where vital sign monitors come in. Not only do they provide a solid source of information for determining outcomes, but the continuous nature of their data collection also makes them the perfect backbone for always-on clinical trials.

Creating value from continuous vital sign monitoring will come from tying real-time physiological data to relevant data points in the medical chart, tailored to specific models and desired outcomes. This approach can pave the way for always-on trials, continuously running and yielding valuable insights. Imagine being able to use AI to sift through vast amounts of data to instantly identify outcomes for specific subsets of patients who were given a particular drug in the hospital. This capability is within reach and represents an exciting frontier for AI in medicine.

The potential for using AI to continuously assess vital sign data is vast, and would represent a fundamental shift in patient care and medical research. Yet there are significant challenges to realizing this potential. Collecting, storing, standardizing, and effectively using this kind of data is a daunting task, as I am now learning through my own company and research. Robust data security measures would have to be put in place to protect patient information. Creating the necessary infrastructure would be one of the hardest challenges, requiring access to monitors and seamless integration with existing hospital systems. Developing a sustainable business model that incentivizes investment and addresses the costs of implementation and maintenance would also be crucial.

Despite these challenges, I believe that the integration of AI and real-time vital sign data in hospital settings holds great promise for creating significant value and improving patient outcomes. Much of this value will be created in hospital care, which accounts for more than 30% of health care expenditures, the largest contributor to costs.

Vital signs have been used to monitor health for more than 2,000 years. Taking advantage of advances in monitoring and data analysis can create new roles for them in predicting — and resolving — health problems.

Julio La Torre, M.D., M.B.A., is a practicing hospitalist physician, a co-founder and CEO of AiroSolve, and a recent graduate of the UCLA Biodesign Accelerator fellowship.