I have become increasingly interested in the information healthcare observes but never records.
Healthcare generates an extraordinary amount of data. Every laboratory test produces measurements. Every imaging study creates findings. Every prescription reflects a clinical decision. Every diagnosis becomes part of the medical record. We have spent decades building systems to store, organize, retrieve, and analyze these artifacts.
Yet much of what physicians learn about a patient never appears in any structured form.
Consider a routine consultation. A patient mentions that they have stopped taking their evening walks because they become tired more easily than they used to. A spouse mentions increasing forgetfulness. A daughter describes subtle changes in mood or behavior. A patient admits that they are struggling to follow a medication regimen despite understanding its importance. None of these observations are unusual. In fact, they happen every day in clinics around the world.
What is interesting is that they often influence clinical judgment far more than we realize. A physician hearing that a patient has stopped walking may think differently about disease progression. A family member's observation about memory may influence the differential diagnosis. A discussion about medication adherence may completely change the interpretation of an otherwise confusing laboratory result.
Yet despite their importance, most healthcare systems are not really designed to preserve these observations. They are designed to preserve diagnoses, procedures, prescriptions, and test results. The observations themselves often remain trapped inside conversations, human memory, and clinical intuition.
Why this matters for personalized medicine
That distinction may sound subtle, but I suspect it becomes increasingly important as healthcare moves closer to personalized medicine.
For years, personalized medicine has largely been framed as a biological challenge. The assumption has been that the more we understand a patient's biology, the more effectively we can personalize care. Advances in genomics, molecular diagnostics, biomarker discovery, precision therapeutics, and imaging have all moved healthcare closer to that vision.
Yet biology alone rarely tells the entire story.
Two patients with similar diagnoses, similar laboratory values, and even similar genetic profiles can experience very different outcomes. One recovers quickly while another develops complications. One follows treatment successfully while another struggles despite receiving the same care. One remains independent while another gradually loses function. The differences are often visible long before they appear in a laboratory result or imaging study.
Those differences are frequently phenotypic.
A phenotype is the observable expression of health and disease in an individual. It is not simply the condition listed on a chart. It is the way that condition manifests in a real human being over time. It includes symptoms, behaviors, responses to treatment, functional ability, cognitive changes, recovery patterns, resilience, and countless other signals that together describe how a person's health is evolving.
The challenge is that healthcare has traditionally been far better at capturing events than trajectories. We capture diagnoses. We capture prescriptions. We capture laboratory values. We capture imaging findings. What we often fail to capture are the countless observations that explain how those events are actually experienced by the patient.
The consultation as a source of phenotype data
This is one reason why I find ambient clinical documentation so interesting.
Most discussions around ambient scribing focus on productivity. The value proposition is straightforward. Reduce administrative burden. Save physicians time. Generate documentation automatically. Allow clinicians to focus on patients instead of keyboards. Those benefits are real, and they are already creating value in clinical practice.
What interests me is something slightly different.
For the first time, we have systems capable of listening to the conversation itself. Not just the diagnosis that emerges from the consultation or the prescription that gets written afterwards, but the actual exchange between patient and physician. That conversation contains an extraordinary amount of information about how health is changing.
Patients describe symptoms before those symptoms become diagnoses. They describe limitations before those limitations become measurable in a test result. They talk about behavioral changes, treatment challenges, cognitive concerns, functional decline, social isolation, recovery patterns, and the countless small adjustments they make as they adapt to disease. Much of this information has historically existed only in the room where the conversation took place.
The consultation is not just a documentation event. It is one of the richest sources of phenotypic information in healthcare.
A phenotype is a trajectory
Importantly, none of these observations are usually significant on their own. A patient mentioning increased fatigue during a consultation may not change the course of treatment. Neither does a comment about reduced physical activity, disrupted sleep, mild memory concerns, or changes in appetite. Viewed in isolation, these observations often appear anecdotal.
But healthcare is rarely about isolated observations. Viewed across months and years, these observations begin to tell a story. They reveal patterns. They reveal progression. They reveal adaptation. Most importantly, they reveal trajectories.
Two patients may have the same diagnosis and receive the same treatment, yet one improves while the other deteriorates. One remains independent while the other gradually loses function. One responds well to therapy while the other struggles with adherence, side effects, or complications. The diagnosis may be identical, but the lived experience of the disease is completely different.
A diagnosis is an event.
A phenotype is a trajectory.
The infrastructure personalized medicine will need
As healthcare becomes increasingly data-rich, I suspect one of the biggest opportunities lies in recognizing that many of the signals we need are already being observed every day. The challenge is not necessarily collecting more information. The challenge is learning how to preserve, connect, and learn from observations that have historically existed only as human memory and conversation.
This is also why I increasingly think phenotype will become a critical component of personalized medicine. If personalized medicine is ultimately about understanding the individual rather than the average patient, then phenotype becomes difficult to ignore. It sits between the biology of disease and the lived experience of the person. It provides context that laboratory values alone cannot provide and helps explain why patients with seemingly similar clinical profiles often follow very different paths.
This is where Aether's work on longitudinal health graphs becomes important. Reports, prescriptions, vitals, imaging studies, and clinical conversations should not remain disconnected fragments. They should become part of a living picture of how a patient is changing over time.
That is also why I see phenotype as the next layer after the personal health record. A record helps organize what happened. A timeline helps show when it happened. A health graph helps connect what happened. But phenotype helps us understand what those changes mean for the person.
Personalized medicine ultimately requires understanding the patient, not just the disease. Phenotypes are how we begin to capture that understanding.
Healthcare has spent decades building systems to record what happened to a patient. The next generation of healthcare infrastructure may be focused on understanding how a patient is changing. That shift may sound subtle, but I suspect it is profound.
The future of healthcare will not be built solely on better diagnostics, better drugs, or better AI models. Those advances matter enormously, but they are only part of the story. Equally important is our ability to maintain continuity across time, connect fragmented observations, and build a more complete understanding of the individual behind the diagnosis.
Because ultimately, health is not a collection of records. It is the accumulation of observations unfolding over time. And the more we understand those observations, the closer we get to understanding the person.