A therapy, device, or diagnostic succeeds only when the patient, care team, and delivery setting are ready to carry its clinical capability into a reliable result.
Start with the failure
A technically sound product can fail for several different reasons. A therapy may meet biological resistance. A device may be difficult to use. A diagnostic may produce an answer that never changes care. A service may reach the wrong patient too late. The first task is to locate the failure before proposing another product.
The useful classification is practical: biological, technical, behavioral, operational, institutional, financial, informational, regulatory, or distributional. More than one may be present. Each has different evidence, owners, remedies, and limits.
A patient outcome can be limited by biological, technical, behavioral, operational, institutional, financial, informational, regulatory, or distributional failure.
Treatment readiness
Readiness asks whether a patient is biologically, clinically, operationally, and behaviorally prepared for the intervention. Relevant measures may include organ function, inflammation, disease burden, prior treatment, medication interactions, nutrition, caregiver support, and the likelihood that the patient can complete the required work.
A marker can be associated with response without causing it. A score can predict a result without telling anyone how to improve it. Readiness research must keep prediction, causation, and practical control separate. It must also state who owns the data, how often it changes, and whether measuring it improves a decision.
A measurable precursor is tested for causal relevance and safe modifiability before it is used to prepare a patient for treatment.
Patient state engineering
Sometimes the opportunity is not a new therapy. It is a bounded intervention that changes the conditions under which an existing therapy operates. Examples worth studying include prehabilitation, medication adjustment, nutritional support, immune or metabolic conditioning, microbiome modification, and behavioral preparation.
Preparing the patient is not the same as treating the disease. The distinction matters clinically, commercially, and legally. The research question is narrower: which measured state is modifiable, how quickly can it change, how long does the change last, and does changing it improve a patient result that matters?
Companion conditioning
Companion diagnostics help identify likely responders. Companion conditioning asks whether an unsuitable or weakly responsive patient can be prepared to respond. Response monitoring then checks whether the intended change occurred. Adaptive conditioning changes the preparation when the measured response is inadequate.
This remains a category hypothesis. It needs evidence for timing, safety, clinical acceptance, payment, and whether the approach should be tied to a drug, a therapeutic class, or a mechanism. Any claim that conditioning expands eligibility or improves persistence requires direct testing.
A patient is assessed, prepared through a bounded intervention, checked for response, treated, and monitored for adaptation.
Orchestrate the treatment
Some outcomes depend on sequence and timing as much as on the individual components. A care pathway may measure, prepare, treat, observe, revise, and treat again. Biomarkers may trigger transitions. Exceptions may require a clinician to depart from the expected path.
The orchestration record should state what is measured, who decides, which transition follows, the permitted time window, the evidence left behind, and the recovery path. AI may help with synthesis and monitoring. Clinical authority and accountability remain with qualified people.
Measurements and accountable decisions govern the timing, sequence, transitions, exceptions, and revision of a treatment pathway.
Work with the installed base
Established therapies already have regulatory status, manufacturing, evidence, reimbursement, distribution, clinician familiarity, and patient demand. Improving their performance may be more useful than competing with them.
A serious opportunity screen looks for an expensive therapy or procedure with a meaningful failure population, a measurable failure mechanism, a modifiable state, and a credible enabling intervention. It then asks who gains economically, whether the intervention can reach the patient, and whether partnership is more sensible than independent commercialization.
An established intervention with meaningful failure is screened for a measurable mechanism, modifiable condition, enabling intervention, and credible beneficiary.
Find the patient in the real workflow
Epidemiology estimates a population. It does not show that eligible patients can be found, reviewed, referred, reached, and treated. Patient discovery requires a machine-readable phenotype, available data signals, clinical review, a lawful outreach route, and a care setting able to act.
The discovery workflow should record false positives, missed patients, review burden, referral loss, patient choice, and treatment conversion. It must not allow a manufacturer’s economic interest to distort clinical judgment.
What would make us stop
Do not proceed when the proposed state is not measurable, the marker is too weak to guide care, the condition cannot be changed safely, the support burden exceeds the available workforce, the recovery path is inadequate, or the added process does not improve a meaningful outcome. A large treatment market does not repair a weak mechanism.
| Question | Evidence needed | Reason to stop |
|---|---|---|
Where does failure occur? |
Observed pathway and outcome data |
Failure cannot be located |
Can readiness be measured? |
Repeatable measure tied to a decision |
Signal is unstable or unactionable |
Can the state change? |
Safe intervention with a useful time window |
Change is unsafe, slow, or immaterial |
Does the result improve? |
Credible comparison and patient outcome |
Process improves without patient benefit |