Healthcare improvement is no longer a “copy the intervention and roll it out” game. The core purpose of this article is to show how improvement work and implementation work can be deliberately fused to respond to fast-moving, high-uncertainty healthcare environments, where context changes mid-flight and where AI is becoming unavoidable. The author proposes “improve-mentation” as a practical, research-informed approach that synergistically combines improvement science, implementation science, and real-world change experience, with a built-in expectation of iteration so the change can be adapted to the evolving context across private and public systems and across countries (Ovretveit, 2026).
The problem the paper targets is blunt: even when an intervention is proven effective somewhere, that does not make it effective everywhere, and “telling people what works” rarely changes practice. The article frames this as a generalisability and translation challenge under complexity, emphasizing that evidence of efficacy in one setting is not evidence of effectiveness in another, especially for social and organisational interventions that interact with local workflows, cultures, resources, regulations, and incentives (Ovretveit, 2026).
To make that argument operational, the paper foregrounds why implementation so often stalls: healthcare itself is complex because multiple professions must coordinate changes; innovations are complex because they often come as multi-component bundles; and implementation actions are complex because they involve parallel and sequential work across system levels, shaped by stakeholder values and “value complexity” that can trigger mistrust and conflict (Ovretveit, 2026). In other words: the barrier is rarely one missing checklist item, it is the system reacting exactly like a system (Ovretveit, 2026).
The article’s practical promise is that improve-mentation offers structured ways to handle that reality, not pretend it away. It describes four improve-mentation approaches and highlights the distinct purpose of each: the integrated improve-mentation framework (a five-step cycle that explicitly repeats as the situation changes), learning evaluation (where evaluation is designed as a feedback engine that enables adaptation), active implementation frameworks (models linking effective innovations, effective implementation, and enabling contexts to produce outcomes), and getting to outcome (a structured, accountability-focused approach that combines improvement and implementation methods across multiple steps) (Ovretveit, 2026).
Finally, the paper is explicit about why this matters now: data-rich clinical and management environments are expanding rapidly, and the author positions improve-mentation as increasingly relevant for safely incorporating AI into improvement work by supporting rapid cycles, context-aware feedback, and learning loops without losing accountability for outcomes (Ovretveit, 2026). If your organisation wants to move faster without “breaking reality” in the process, this piece argues that the route is not more enthusiasm, but better integration of the sciences of change, plus disciplined iteration. In short: improve-mentation is presented as a way to make improvement resilient to context, instead of hostage to it (Ovretveit, 2026).
References: Ovretveit, J. (2026). Improving complex systems with improve-mentation: Challenges and solutions. Frontiers in Health Services, 5, 1724893. https://doi.org/10.3389/frhs.2025.1724893
Mini dictionary:
Improve-mentation means combining improvement science and implementation science in a problem- and setting-specific way to achieve better outcomes through research-informed, effective, and timely change, with deliberate iteration so the change can be adapted to an evolving context (Ovretveit, 2026).
Complexity refers to why “simple” changes stay hard in healthcare, because difficulty is not only in the clinical task itself but also in the number of coordinated behavior changes required, the multiple professions involved, and the supporting system changes that must align for the intervention to work as intended (Ovretveit, 2026).
Context means the conditions around a change that shape whether it works and how it must be adapted, and the article treats context as something to actively measure and learn from, including influences from different levels of the healthcare system that affect implementation actions and outcomes (Ovretveit, 2026).
Generalisability means not assuming that evidence of efficacy in one setting proves effectiveness in all other settings, especially for social and organisational interventions whose effects vary across social and organisational environments, so local testing and adaptation are often necessary even when evidence looks strong elsewhere (Ovretveit, 2026).
Iterative adaptation is the principle that repeated testing and revision are not a weakness but a design feature for complex change, where rapid-cycle methods allow teams to adjust the solution as data and circumstances evolve and to build organisational capacity for future change (Ovretveit, 2026).
Integrated Improve-mentation Framework is a structured five-step cycle that operationalizes improve-mentation by defining the problem, choosing data that indicate the problem is solved, designing and implementing a solution, reviewing data and revising the solution, and repeating as the situation changes, with a later version explicitly emphasizing digital data for rapid cycles (Ovretveit, 2026).
Learning evaluation is an approach that blends quality improvement and implementation research to study healthcare innovations by building feedback loops that allow the intervention to adapt to ongoing contextual change, often with researchers playing a prominent facilitation and measurement role (Ovretveit, 2026).
References: Ovretveit, J. (2026). Improving complex systems with improve-mentation: Challenges and solutions. Frontiers in Health Services, 5, 1724893. https://doi.org/10.3389/frhs.2025.1724893