Implementation science and translating evidence into practice — branching viva

Opening (candidate)

My first move is to reframe this from a dissemination problem into an implementation problem. Emailing and posting a guideline is dissemination — the passive spread of information — and it changes behaviour almost nothing. The gap is one of implementation: only 52% of eligible children receive the first antibiotic within an hour, and closing that needs active, tailored strategy. I would define implementation science as the scientific study of methods to promote the systematic uptake of research findings into routine care, then apply the knowledge-to-action cycle: identify the gap, adapt the bundle to context, assess the barriers with Cabana and CFIR, map each barrier to a strategy from the ERIC compilation, deploy with fidelity monitoring, measure the implementation outcomes with Proctor and the impact with RE-AIM, and plan sustainability from the outset — disaggregating reach by subgroup so an equity gap cannot hide. [1] [9]

Branch A — Why the gap exists

Examiner: We have a good guideline. Why is it not being followed? [3]

Candidate: Because a guideline existing is dissemination, and dissemination is only the first link in a chain that can break anywhere. Cabana's framework lets me diagnose where it breaks, in three tiers. Knowledge: clinicians may not be aware of the hour-one target or familiar with the weight-based doses. Attitudes: some may disagree it applies to a well-looking febrile infant, lack self-efficacy in escalation, or be caught in the inertia of previous practice. Behaviour and external: the workflow may lack pre-mixed antibiotics, the triage system may not flag sepsis early, or pharmacy turnaround may be the bottleneck. Morris showed evidence takes about seventeen years to reach routine practice, and even then a minority is fully adopted — so a gap like this is the expected default, not a surprise. I would not assume the answer; I would assess it with structured staff interviews. [3] [2]

Branch B — Frameworks

Examiner: There are a lot of frameworks. Which do you use, and why? [4]

Candidate: I use them by the job they do. Before I choose a strategy I use CFIR, a determinant framework with five domains — intervention characteristics, outer setting, inner setting, characteristics of individuals, and process — to understand what will help or block adoption in this department. To diagnose the specific guideline barriers I use Cabana, as I just described. To follow the sequence I use the knowledge-to-action cycle. And to evaluate afterward I use RE-AIM — Reach, Effectiveness, Adoption, Implementation and Maintenance — to estimate the public health impact. The trap is treating them as interchangeable; CFIR tells me what to assess, RE-AIM tells me how to judge, Cabana tells me why staff are not following the guideline, and the cycle tells me what order to do it in. [4] [5] [3]

Branch C — Strategy selection

Examiner: So what do you actually do? [8]

Candidate: I map each diagnosed barrier to a strategy from the ERIC compilation — Powell's seventy-three discrete strategies clustered into families. If the barrier is awareness, I use educational outreach and clinical opinion leaders. If it is attitude and inertia, I recruit local champions and build consensus. If it is environmental — no pre-mixed antibiotics, weak triage flags — I restructure: a sepsis trolley, an opinion-led escalation pathway, and decision-support reminders visible in the record. I build a parsimonious bundle of two or three matched strategies, not a pile, because Grimshaw's synthesis showed tailored strategies matched to barriers outperform untargeted ones, and an over-bundled approach fatigues the team. I deploy with fidelity monitoring from day one. [8] [9]

Branch D — Measurement

Examiner: How will you know if it worked? [7]

Candidate: I measure three layers, and the implementation layer is the one that is usually missed. Proctor separated implementation outcomes — acceptability, adoption, appropriateness, feasibility, fidelity, cost, penetration, sustainability — from the service and clinical outcomes. So I track adoption: did the department take up the bundle? Fidelity: is it delivered as designed, with the antibiotic given within the hour? Penetration: how far across the eligible population? Sustainability: will it last after the lead rotates? And alongside the clinical outcome — did children do better? — I measure reach with RE-AIM, and I break reach down by subgroup. The single most neglected measure is fidelity; without it I cannot tell a therapy that does not fit from a therapy delivered badly. [7] [5]

Branch E — Equity and sustainability

Examiner: The average improved. Are you done? [1]

Candidate: No. An improving average is the most common way equity silently fails — Indigenous, migrant or disadvantaged children may be the subgroup that never reached the hour-one target, and an average that hides that has widened a gap under cover of success. I make reach-by-subgroup an explicit aim and measure from day one. And I am not done until adoption, fidelity and sustainment targets are met equitably. Sustainability needs a named owner, ongoing data, and the practice embedded into policy, order sets and orientation so it survives staff turnover. Aarons showed implementation leadership — a leader who is proactive, knowledgeable, supportive and perseverant — measurably predicts which units sustain. So I assess leadership as a prognostic factor and protect the gains structurally, not with hope. [1]

Close

I would reframe this as an implementation problem, diagnose the barriers with Cabana, assess the context with CFIR, map each barrier to a strategy from ERIC, deploy a parsimonious bundle with fidelity monitoring, measure implementation outcomes with Proctor and impact with RE-AIM, disaggregate reach by subgroup so an equity gap cannot hide, and plan sustainability with a named owner, data and policy levers before I ever think about spread. The honest caveat is that the field's predictive power is maturing, frameworks organise thinking without guaranteeing success, and context frequently defeats a copied intervention — so I tailor, measure, and adapt rather than assume. [1] [9] [7]