A shared grammar: a counterfactual dialogue between disease and illness

An enduring divide separates the biomedical account of disease from the phenomenological understanding of illness. In this article, I develop a two-level counterfactual model that treats these domains as parallel and formally comparable modes of reasoning: the Clinical Counterfactual (CC) for causal hypotheses and the Experiential Counterfactual (EC) for goal-directed patient claims. Grounded in Biostatistical Theory and operationalized through Homeostatic Property Cluster theory, the framework fixes background conditions for counterfactual assessment and anchors diagnosis in causal-mechanistic disruption rather than population averages alone. An Integration Condition evaluates coherence between levels, and an intervention is favored only if at least one admissible scenario exists in which removing the dysfunction resolves the symptom and enables the realization of the patient's valued activity. Practically, I outline a structured documentation workflow in which CCs and ECs are articulated in parallel and linked by a deviation-analysis protocol that is activated when the two levels diverge. The protocol treats disagreement as diagnostically informative, prompting focused re-examination of causal assumptions, value priorities, and evidential gaps. Conceptually, the framework avoids reducing reasons to causes. It employs counterfactual difference-making as a shared analytical language while preserving distinct explanatory roles at each level. Unlike patient-centered outcome approaches that assess values mainly post hoc, this model integrates patient goals ex ante into causal reasoning itself, with patient-reported outcomes serving as downstream validation rather than as the integrative glue. The result is a disciplined and transparent heuristic for clinical reasoning, particularly under conditions of uncertainty, contestation, and incomplete knowledge, that renders causal inferences, patient goals, and normative assumptions explicit and mutually testable.