Why cooperation fails in health systems: governance autoimmunity as a structural diagnosis, and what it changes about designing collaboration

Introduction and objectives Cooperation in global health fails repeatedly in ways money and evidence do not explain: reforms meeting safety, cost and outcome criteria are still rejected. We test a structural explanation. Governance systems defend themselves as immune systems do, and under three conditions those defences begin attacking beneficial change. The objective is to state that mechanism precisely enough to test, and to identify where cooperation can still be designed to succeed. Methodology Complex System Immunity Theory treats biological immune systems and governance systems as two instances of one abstract architecture. The correspondence is formal — a structure-preserving mapping — so immune dynamics yield falsifiable predictions about governance rather than illustrative comparisons. Governance Genetic Engineering specifies the organisational instance: defences become crude under stakeholder proliferation, objective misalignment, and the absence of any authoritative correction acting on the governing apparatus itself. Predictions were tested against a retrospective case series of health system reforms and comparable public sector failures. Results and discussions Cases separate on the predicted variable, not on merit, scale or evidence. Integration attempted inside an existing governance structure (Torbay) was dismantled; an equivalent community model built alongside it (Frome) survived while holding authority over nothing the incumbent governed. This gives a threshold condition: a parallel structure is tolerated until it acquires authority over a function the incumbent governs, then attacked. Cooperation designed around a blocked centre defers rejection rather than escaping it. The framework states its own limit — failure by unencodable novelty is not autoimmunity, distinguished by whether the defences acted at all. Conclusions Cooperation failure in health is partly a design fault with an identifiable structure, not only a shortage of will or funding. Repair must therefore restore the ability to tell beneficial change from threat, rather than add further control. Prevention in systems still simple is cheaper and likelier to succeed than intervention where the pattern is established. This is the rationale for our maxillofacial cancer pathway work in Uganda, treated as a design window. The underlying papers are open access on SSRN; we invite scrutiny and collaboration.