Access, automation, and the patient in the middle

In rural and urban clinics across the country, something unusual is happening. The patient sitting in front of a doctor is no longer the only one determining what treatment they receive. This week's health news offers three clear examples of a system undergoing structural drift—not toward better care, but toward a patchwork of rules, algorithms, and geographic accidents that increasingly decide who gets what, and when.

For example, KFF Health News documented how Medicare's AI testing on prior authorization is creating delays and errors for both patients and doctors. Prior authorization—the requirement to get insurance approval before receiving a treatment—has long slowed care. Automating it promised speed. Instead, according to the reporting, it has introduced a new bottleneck: the algorithm. Doctors report that the AI system denies claims that a human reviewer might approve, and the process to appeal takes just as long as the original authorization. The patient is caught between two systems that do not talk to each other, and neither is accountable to the physician in the room.

Then there is the work requirement trap. STAT News outlined the Catch-22 for disabled workers hidden in Medicaid's new work requirements—people whose illness or disability prevents them from working, but whose benefits now depend on proving they are trying. The eligibility rule creates a logical impossibility for the sickest patients. Again, the patient is subject to a decision made outside the clinic, by people who may never meet them.

What ties these stories together is not that they are all bad outcomes. It is that they reveal a health system where the locus of decision-making has drifted away from the clinical encounter. Medication access is now determined by state borders and shipping logistics. Treatment approval is now determined by algorithms trained on historical data, which may encode historical bias. Eligibility is now determined by work capacity, not medical need. In each case, the patient becomes a subject of the system rather than its center.

This is not a crisis of individual stories. It is a slow crisis of architecture. The system was never perfect—rural patients have always faced longer waits, poor patients have always faced more barriers—but the coordination between doctor and patient has historically been the anchor point. That anchor is now surrounded by competing systems: insurance systems, regulatory systems, state systems, algorithmic systems. Each one adds a layer of abstraction between the person who is sick and the person who might help.

The work ahead is not to eliminate any single rule or technology. Algorithmic review is not inherently wrong; disabled patients should not be forced to work. The problem is that each system operates independently, without clear feedback from outcomes. A doctor does not know whether an AI denial harmed a patient. A disabled patient does not know which part of the bureaucracy is keeping them from treatment. States do not measure whether their abortion bans are simply displacing care rather than reducing it.

The patient—the actual person in the exam room—has become invisible to the system that is supposed to serve them. That is not a failure of any single policy. It is a failure of integration. And it will not be fixed by adding one more layer of coordination on top of the existing ones. It will be fixed only when someone starts asking, at every decision point, whether the system is making it easier or harder for the patient in front of the doctor.