margins in the marginalia

There’s a phrase that keeps surfacing in medical literature about patients with rare or undiagnosed conditions: they “fail” treatments.

Not the treatment fails them. They fail the treatment.

It’s such a strange inversion. And the more I sit with it, the more it feels like a diagnostic error that happens before any clinical one — a problem framing failure. Someone living with an uncharacterized condition spends years accumulating evidence. Filing from specialist to specialist. Tracking what worsens things, what helps, which doctors close loops and which ones open them back up. And at some point they figure out that if their symptoms are controlled, doctors won’t have reason to investigate what’s actually wrong. So managing the symptoms becomes a survival strategy. Which suppresses the signal. Which makes the underlying problem harder to find. Which keeps the odyssey going.

Nobody designed that feedback loop. But it runs anyway.

fixes that fail

In systems dynamics, there’s an archetype called “Fixes that Fail.” A problem symptom appears. Feeling pressured by urgency, someone applies a quick fix. The fix addresses the symptom, not the structure. This relieves enough pressure that the root problem never gets investigated. The system’s capacity to correct itself atrophies. The symptom returns. The fix is applied again, harder. The pattern compounds.

What strikes me is how cleanly the diagnostic odyssey fits this shape. The rare disease patient is living inside a “Fixes that Fail” loop — not because any individual clinician is failing them, but because the structure of the system produces that behavior. The quick fix is symptom management. The fundamental problem, the one that keeps getting deferred, is understanding what’s actually happening at a level the system wasn’t built to look at.

You could draw the causal loop. Ambiguous presentation → clinical skepticism → fewer tests ordered → more ambiguous data → reinforced skepticism. A balancing loop somewhere: eventually urgency peaks enough to force action. A hospitalization. A crisis. The symptom stabilizes temporarily. The structure remains.

who counts as the beginning

I keep thinking about what biomedical research optimizes for, and where patients fit in that picture. The usual sequence: molecule → mechanism → model organism → clinical trial → patient. The patient shows up at the end, as a test subject for something already decided. But there’s another sequence — one design thinking has been arguing for decades — that goes: learn about reality first, then build mental models, then abstract those into innovations, then realize them. Research → Analysis → Synthesis → Realization.

In that order, the patient isn’t at the end. They’re the beginning. They’re the reality you’re supposed to learn about.

That’s not just a philosophical preference. It changes what counts as data. It changes whose knowledge is treated as valid. It changes what questions get asked.

A journey map drawn from the perspective of someone navigating the rare disease system would look less like a pipeline and more like a labyrinth. Lots of entry points. Walls you can’t see coming. Occasional open spaces where someone accidentally helped. The pain points wouldn’t be in the clinical steps — they’d be in the gaps between them. The years. The financial sting of a negative test result. The neurologist who said it was “nice” to finally have confirmation that something was actually wrong. The grammar of that sentence.

the genome is just one node

Rare disease might be the place where a certain assumption in biomedical research breaks down most visibly. The assumption is something like: the body is the system, the patient is the body, and everything else — the care ecosystem, the family structure, the insurance, the fifteen years of being told nothing is wrong — is a confounder to be controlled for.

But if you’re the patient, none of that is a confounder. That’s the system. The genome is just one node.

The iceberg model in systems thinking puts mental models at the bottom — the hardest layer to surface, the highest leverage when you can reach it. The mental model I keep noticing in biomedical research isn’t wrong exactly, just narrow. It defines the unit of analysis as the body, and everything else as noise. Rare disease, with its small populations and its irreducibly human complexity, has a way of making that narrowness visible. The people at the margins of the distribution — the ones whose presentations don’t fit the standard picture — have often been collecting data about the system’s structure for years. Nobody’s asking them to draw it.

Group model building, in systems thinking, is essentially a method for doing exactly that: getting diverse stakeholders into a room, surfacing their different hypotheses about what’s driving a shared problem, and building a collective map from the disagreements. When it works, it finds causal structures that none of the individual participants would have reached alone. The people closest to the problem — the ones most affected — tend to surface connections the researchers didn’t know to look for.

Rare disease patients have already built that model. Privately, out of necessity. The diagnostic odyssey is, among other things, a years-long participatory modeling exercise that nobody else is in the room for.

a question I haven’t resolved

In the systems literature on food inequity, researchers using the “Fixes that Fail” archetype found that the structural force driving the feedback loop wasn’t actually inside the food system — it was mass incarceration, poverty, the accumulated weight of racialized policy. The food system’s boundary had been drawn too narrowly to see it.

What’s the rare disease equivalent? What structural force, outside the biomedical system’s traditional boundary, keeps the diagnostic odyssey running? Funding incentives that favor tractable mechanisms over weird edges? A publication ecosystem that rewards novel findings over patient-centered ones? The mental model that scientific knowledge and patient knowledge are categorically different things, and that one can eventually substitute for the other?

I don’t know. But it feels like a question worth actually sitting with, rather than designing around.

The margin, in the design sense, is where you find out what the center gets wrong. There might be something similar happening here. The patients at the edges of the distribution — the ones whose conditions are too rare, too complex, too poorly characterized to fit the standard research pipeline — are carrying information about the system that the system doesn’t know how to receive.

What would it look like to start there, and work inward?

in the margins…

Benninger et al., “Fixes that Fail” (Am J Community Psychol, 2021) — A food systems paper, which is precisely why it belongs here. The archetype travels. Anywhere a symptomatic fix relieves enough pressure to prevent root cause investigation, the loop runs. The rare disease parallel isn’t metaphorical — the feedback structure is nearly identical. Worth reading for the method too: qualitative interviews mapped to a systems archetype, revealing causal structures the researchers didn’t anticipate going in.

Hovmand, Community Based System Dynamics (2014) — The textbook behind group model building. The argument is simple and quietly radical: the people most affected by a system hold knowledge about its structure that no outside researcher can fully reconstruct. The method is a way of getting that knowledge into the room. Rare disease, working with small populations by necessity, is well-positioned for this. The sample size problem might be an argument for deeper engagement, not against it.

Naumann et al., pedestrian safety and complex systems (Inj Prev, 2020) — Shows what happens when you bring non-traditional stakeholders into a causal mapping process: the model changes. Substantially. Participatory approaches don’t just add buy-in — they add information that no single vantage point could supply. Useful here as a worked example of group model building in a public health context with a clear behavior-over-time problem to explain.

Senge, The Fifth Discipline (1990) — The original home of the systems archetypes, including “Fixes that Fail” and “Shifting the Burden.” Still the clearest explanation of why quick fixes and fundamental solutions exist in tension, and why the quick fix tends to win in the short run. The rare disease system is full of shifting-the-burden structures that are worth naming.

Kumar, 101 Design Methods (2013) — The design innovation process — Research, Analysis, Synthesis, Realization — is in here. The sequence is the point. Most biomedical research runs it in reverse, moving from mechanism to patient rather than patient to mechanism. The book is a catalog of methods, but the underlying logic is what matters: start with the people, not the problem you’ve already decided to solve.