Root-cause reasoning: where a useful clinical instinct outruns the evidence

The instinct is a sound one

Treating a symptom without asking what produces it is poor practice, and every branch of medicine teaches otherwise. Diagnostic reasoning is built around identifying an underlying process rather than around suppressing the presentation. Where a single correctable cause exists, finding it is straightforwardly better than managing the consequences indefinitely.

The framing that distinguishes root causes from symptoms is therefore not an alternative to conventional reasoning but a restatement of it. The disagreement is about what happens when no single cause can be found rather than about whether looking for one is worthwhile at all. Framing conventional practice as uninterested in causes is a rhetorical move rather than an accurate description of how diagnosis is taught.

Chronic conditions rarely have one cause

Most long-term conditions arise from an interaction between inherited susceptibility, environment and accumulated physiological change over time. In that structure there is no single lever, and each contributing factor accounts for a modest share of the overall picture. A framework that assumes one underlying cause will keep searching, and searching generates findings whether or not a cause exists.

Incidental abnormalities are common in any broad testing programme, and each one invites explanation and further investigation. This is the mechanism by which extensive testing can produce a compelling narrative that is largely constructed from noise.

Mechanistic plausibility is not evidence

A proposed causal chain can be biochemically coherent at every step and still fail to describe what happens in a person. Biology contains redundancy and feedback, so interfering with one step frequently produces compensation rather than the predicted downstream effect. The history of medicine includes many interventions that were mechanistically sound and turned out to be ineffective or harmful in trials.

This is precisely why trials exist, and why the strength of a mechanism does not substitute for evidence of effect. Explanations built entirely from mechanism are persuasive in proportion to their internal coherence rather than in proportion to their accuracy.

The testing cascade

Broad panels increase the number of results returned, and the probability that at least one falls outside its range rises with each additional test. Each flagged result then requires interpretation, and interpretation in the absence of a clinical question is unusually difficult. Following up incidental findings has costs, including further procedures, exposure to imaging and the anxiety of an unresolved abnormality.

Guidelines on test ordering exist substantially because of this cascade rather than because of the price of individual tests. Ordering tests to answer a specific question is what makes results interpretable, and that is a difference in method rather than in thoroughness.

What the useful version looks like

The defensible version of root-cause reasoning takes a detailed history, considers contributing factors broadly and tests to answer specific questions. It also accepts that a proportion of presentations will not resolve into a single identifiable cause, which is an honest position. Time spent on history and on the person's own account of their symptoms is genuinely valuable and is often what patients report missing.

That value does not require a distinct theoretical framework, and it does require appointment structures that allow it. For anyone weighing up an extensive testing programme, the useful question is what would actually change in their care depending on how each result came back.