What we learned from a failed translation: a humility case study
Medical research earns its value when it improves care beyond the study site. Yet the journey from promising evidence to routine practice is rarely a straight line. A treatment, screening pathway, digital tool or model of care can perform well in controlled conditions and still fail to deliver meaningful benefit in everyday services.
This case study describes an anonymised quality improvement initiative involving a hospital-based intervention for people at risk of avoidable readmission after discharge. The intervention had encouraging research results, but its implementation produced little measurable improvement. Its failure became a valuable lesson in humility, implementation science and genuine partnership.
For health translation networks such as Brisbane Diamantina network, the experience reflects a central responsibility: connecting researchers, clinicians, consumers and communities early enough to test whether an innovation is useful, acceptable and workable in the setting where it is intended to live.
A promising intervention meets a complex service
The intervention combined a discharge checklist, follow-up telephone call and medication review for adults with a chronic condition. In the original research environment, it improved patients’ understanding of warning signs and appeared to reduce short-term hospital use. The evidence was carefully collected, and the clinical rationale was persuasive.
A multidisciplinary team therefore prepared to translate the model into several hospital and community services. Leaders expected that training staff, supplying a script and monitoring completion rates would be sufficient. The project was described as an evidence-based pathway, which subtly positioned local teams as recipients of a finished solution rather than partners in adapting care.
The first warning sign was that implementation activity looked stronger than patient outcomes. Staff completed forms and recorded calls, but patients often received fragmented advice from several professionals. Community providers did not always receive timely information, and carers were sometimes absent from the discharge conversation. The process existed, yet the intended experience of coordinated care did not.
The assumptions that went untested
The team assumed that the main barrier was inconsistency. If every clinician followed the same steps, the intervention would produce the same benefits observed in the trial. This assumption overlooked differences in staffing, language needs, transport access, digital literacy, appointment availability and the clinical complexity of patients in routine care.
Researchers also assumed that patients would welcome a follow-up call. Interviews later showed a more complicated picture. Some people valued the contact, while others received calls during work, felt uncomfortable discussing medications by phone or preferred a trusted community clinician. Several Aboriginal and Torres Strait Islander participants described previous experiences of services making decisions without sufficient cultural understanding or continuity.
There was a further error in how success was defined. The project measured whether the pathway had been delivered, but it did not initially examine whether patients understood the plan, could obtain their medicines or knew whom to contact. Fidelity to the procedure was mistaken for fidelity to the purpose.
What the data revealed
After six months, readmission rates had not changed. Completion statistics looked respectable, but interviews, observation and record reviews exposed gaps hidden by the dashboard. Calls were sometimes rushed, medication lists were copied forward without reconciliation, and the intervention was least reliable for people with unstable housing, limited English or multiple health conditions.
| Original belief | Evidence from practice | More useful interpretation |
|---|---|---|
| A standard script would create consistency | Staff adapted or shortened it under time pressure | The workflow needed redesign around real constraints |
| A phone call would reach most patients | Contact details were outdated and some patients preferred face-to-face support | Access and communication preferences were part of the intervention |
| Completion meant effective delivery | Documentation was present, but understanding varied | Patient-reported outcomes were essential |
| The research model could be reproduced unchanged | Services had different roles, resources and referral pathways | Adaptation was necessary, with core functions protected |
| More training would solve the problem | Staff understood the model but lacked time and cross-service coordination | Organisational conditions mattered as much as knowledge |
The team initially responded by proposing refresher training. That response was understandable, but it delayed the more difficult question: what if the intervention was poorly matched to the system? A second review brought clinicians, patients, carers, pharmacists, community providers and implementation researchers together to examine the failure without assigning blame.
Humility changed the inquiry
Humility in health translation is practical rather than ceremonial. It means treating the original evidence as strong but bounded, acknowledging what the research did not test and allowing frontline knowledge to challenge expert expectations. The team stopped asking why staff had failed to follow the model and began asking what the model demanded from staff and patients.
This shift changed the evidence-gathering process. Researchers observed discharge meetings, listened to recorded patient feedback, mapped referral pathways and invited people who had disengaged from the service to describe what happened. A consumer partner explained that the project’s language focused on preventing readmission, while patients often focused on getting through the first week at home.
The intervention was then separated into its essential functions and flexible features. The essential functions were a clear medication plan, an understandable escalation pathway and reliable connection with a responsible clinician. The telephone script, timing and staff role were treated as adaptable. Some services moved to pharmacist-led reviews; others used community health workers or combined phone contact with an in-person appointment.
Building translation around shared learning
The revised approach used small tests of change rather than a single large rollout. Teams tested communication methods with different patient groups, reviewed outcomes every few weeks and included measures of burden, trust and comprehension. A failed test was recorded as useful information, not as evidence that a service had performed badly.
Governance also became more inclusive. Patients and carers were involved in interpreting findings, while clinicians could identify operational risks before they became implementation failures. Researchers documented adaptations so that future evaluations could distinguish between a genuinely ineffective intervention and an intervention that had been delivered in an unsuitable form.
This approach aligns with the broader work of health translation partnerships, where clinical innovation depends on relationships across hospitals, universities, research institutes and communities. Translation is stronger when evidence moves in both directions: research informs practice, and practice reveals the conditions under which research can be trusted.
Practices that protect against overconfidence
A failed implementation should produce changes in behaviour, funding decisions and evaluation design. The following practices can help teams translate evidence with greater care:
- Test assumptions with patients, carers and frontline staff before selecting an implementation strategy.
- Measure meaningful outcomes such as understanding, access, continuity and experience alongside completion rates.
- Define the intervention’s core purpose separately from the procedures used in the original research.
- Include equity measures so that average results do not conceal poor outcomes for priority populations.
- Agree on stopping rules and learning reviews before the rollout begins.
The case also shows why psychological safety matters. Staff are more likely to report workarounds, missed contacts and unintended consequences when they will not be punished for revealing system problems. Honest reporting improves patient safety and gives researchers a more accurate account of implementation fidelity.
Turning a failed translation into better care
The intervention did not fail because evidence was irrelevant. It failed because the team treated evidence as a finished package rather than a starting point for local learning. The strongest result was not a dramatic improvement in readmission rates; it was a clearer understanding of the relationships, resources and choices required for the intervention to work.
Humility does not weaken ambition. It makes ambition accountable to patients’ lived experience. For researchers and health services, that means designing translation as an iterative partnership, publishing null findings and adaptation lessons, and recognising that a community’s refusal or hesitation may contain important evidence.
Health professionals, researchers and community partners can help make future innovations safer and more useful by sharing implementation lessons, strengthening co-design and supporting evaluation that reflects real-world care. Explore the work of Brisbane Diamantina Health Partners and contribute to a culture where learning from failure improves outcomes for patients, families, carers and communities.
