Medicine has traditionally treated disease as something static.
A patient receives a diagnosis. That diagnosis is recorded in the chart. Treatment is prescribed. Follow-up visits occur periodically to assess progress.
This approach works reasonably well for certain acute problems. An infection is treated. A fracture heals. A surgical issue is corrected.
But most modern healthcare does not revolve around acute problems.
Most healthcare is about managing conditions that evolve over time.
Hypertension progresses gradually. Diabetes requires continuous adjustment. Digestive disorders fluctuate. Autoimmune diseases wax and wane. Mental health conditions shift with life circumstances.
These conditions are not static diagnoses. They are trajectories.
Yet the healthcare system still treats them as if they were fixed labels.
To manage disease effectively, we need a model that reflects how illness actually behaves. That model is what we call a problem graph.
The Limits of Diagnosis-Centered Medicine
In traditional healthcare systems, diagnoses serve as the primary way we organize medical information.
A patient is labeled with hypertension, diabetes, Crohn's disease, or depression. That diagnosis becomes the anchor for treatment decisions, documentation, and billing.
But diagnoses often obscure the most important question in medicine.
What is happening to the patient right now?
Two patients with the same diagnosis may be in very different clinical states.
One patient with hypertension may be newly diagnosed with borderline readings. Another may have years of uncontrolled blood pressure and evidence of organ damage.
Both share the same diagnosis. Their clinical realities are very different.
The same is true across medicine. Diagnoses provide useful categories, but they do not describe how a condition is evolving.
What clinicians actually manage is not the diagnosis itself.
They manage the state of the problem over time.
From Diagnoses to Problem Trajectories
A problem graph represents the evolving state of a medical problem.
Instead of treating disease as a fixed label, the problem graph models how a condition progresses across time and how clinical decisions influence that trajectory.
Each problem exists within a series of possible states.
For many chronic conditions, those states might include:
- Provisional recognition of a possible problem
- Confirmed diagnosis with initial treatment
- Active management and medication adjustment
- Stable control
- Escalation due to worsening disease
- Resolution or long-term stability
Patients move between these states as their health evolves.
For example, a patient with hypertension may progress through several stages:
- Elevated blood pressure is first detected.
- Additional readings confirm the diagnosis.
- Medication is started.
- Blood pressure improves but requires adjustment.
- Control stabilizes.
- Later, lifestyle changes allow medication reduction.
Each transition represents a meaningful clinical change. These transitions are what clinicians actually manage.
The problem graph captures this movement.
Why Time Matters in Medicine
Many of the most important clinical insights emerge only when we observe change over time.
A single blood pressure reading does not diagnose hypertension.
One elevated blood glucose does not define diabetes.
One episode of abdominal pain does not reveal a chronic gastrointestinal condition.
Patterns across time tell the real story.
Symptoms recur.
Measurements trend upward or downward.
Treatments produce responses that unfold over weeks or months.
Traditional visit-based healthcare captures only intermittent snapshots of these changes.
Between visits, the system often has little visibility into how the patient's condition evolves.
Problem graphs provide a way to represent the patient's condition continuously.
Instead of isolated data points, clinicians can see the structure of the disease trajectory.
Connecting Decisions to Disease Evolution
Another advantage of problem graphs is that they connect clinical decisions to changes in patient state.
When clinicians start a medication, adjust a dose, or recommend a lifestyle change, those interventions are not isolated actions.
They are attempts to move the patient from one state of the problem graph to another.
For example:
- Starting an antihypertensive medication aims to move a patient from uncontrolled blood pressure to controlled blood pressure.
- Adjusting therapy in inflammatory bowel disease aims to move a patient from active inflammation to remission.
Each intervention is a step within the trajectory of the disease.
By representing these transitions explicitly, healthcare systems can better understand how decisions influence outcomes.
Over time, the system can learn which decisions are most effective for different types of patients.
The Role of Continuous Observation
Problem graphs work best when patient information accumulates continuously rather than episodically.
In a visit-based model, clinicians often reconstruct the trajectory of a disease from memory.
Patients describe what happened between appointments. Laboratory tests offer occasional measurements. Symptoms are summarized retrospectively.
This process is inherently incomplete.
When care is asynchronous and patient state is continuously updated, the trajectory becomes much clearer.
Symptoms can be recorded as they occur. Measurements can be tracked over time. Treatment responses can be observed rather than inferred.
The problem graph evolves as new information arrives.
This produces a far more accurate understanding of how disease is progressing.
Why Problem Graphs Matter for Chronic Care
Chronic diseases require continuous management.
Blood pressure must be monitored. Medications must be adjusted. Symptoms must be interpreted in context.
Traditional healthcare systems often manage these conditions through occasional visits separated by months.
Problem graphs allow clinicians to manage the disease trajectory directly.
Instead of asking, "What happened during this visit?" the system asks, "Where is this patient within the progression of this problem?"
This shift changes how care is delivered.
Clinical decisions become part of a structured progression rather than isolated actions. Monitoring becomes proactive rather than reactive.
Most importantly, clinicians gain a clearer understanding of how each patient's condition evolves.
From Individual Patients to Learning Systems
Problem graphs also create a powerful foundation for learning.
When the states of disease and the transitions between them are represented explicitly, healthcare systems can begin to study those transitions at scale.
Which interventions move patients toward stability?
Which patterns predict worsening disease?
Which early signals indicate that treatment should change?
Because the trajectory of the disease is modeled directly, these questions become measurable.
Over time, healthcare systems can refine how diseases are managed and improve outcomes across entire populations.
The Future of Longitudinal Care
Medicine has long focused on diagnosing diseases.
But the real work of healthcare is managing how those diseases evolve.
Problem graphs provide a framework for doing exactly that. They allow healthcare systems to represent disease as a trajectory, connect decisions to outcomes, and observe how patient conditions change over time.
In earlier essays in this series, we explored why asynchronous care improves quality, why the patient state is the true atomic unit of care, and how healthcare solutions aren't solving the right problem.
Problem graphs extend these ideas by modeling how medical problems evolve across time. In our next article, we explore how decision bundles structure and scale clinical reasoning.
Together, these concepts form the foundation of a healthcare system designed for longitudinal care.
Because when we understand disease as a trajectory rather than a label, medicine becomes not just reactive treatment, but continuous management of health itself.