Adoption Dynamics

Why Healthcare Transformation Programs Die Between the Pilot and the Scale

Adoption follows predictable dynamics. Everett Rogers’ diffusion of innovation model (2003, fifth edition of research originating in 1962) describes a pattern so consistent across industries, technologies, and organizational contexts that it qualifies as a social law: new practices spread through populations in a characteristic S-curve, with distinct adopter segments that differ not in degree of enthusiasm but in kind of motivation. Understanding these segments — and the lethal gap between two of them — is the difference between a pilot that generates an impressive report and a transformation that actually changes how care is delivered.

The pattern is not complicated. What is complicated is that most healthcare transformation leaders either do not know it exists or know it intellectually but do not build their implementation plans around it. The result is a graveyard of pilot programs that showed excellent results in two units and then quietly died during “scale-up.”

Rogers’ Adopter Categories: Five Different People Requiring Five Different Arguments

Rogers identified five adopter categories based on decades of diffusion research across agriculture, education, public health, and technology. The categories are not arbitrary percentiles. They reflect fundamentally different psychological orientations toward innovation, and each requires a different adoption argument.

Innovators (approximately 2.5% of a population) are risk-tolerant experimenters who adopt new practices because the novelty itself is rewarding. They are intrinsically motivated — they do not need evidence that the innovation works; they want to find out. In a health system, these are the physicians who volunteer for every new EHR module, the nurses who seek out pilot programs, the quality improvement staff who read implementation science journals for pleasure. They are useful for initial feasibility testing but dangerous as a basis for adoption projections, because their motivation is orthogonal to what drives the other 97.5%.

Early adopters (approximately 13.5%) are opinion leaders who adopt based on vision and strategic potential. They see the innovation’s trajectory and want to be identified with it. Their motivation is partly intrinsic but substantially social — they gain status from being seen as forward-thinking. In healthcare, these are the department chiefs who champion new clinical pathways, the nurse managers who volunteer their units for transformation pilots, the medical directors who present early results at conferences. Early adopters are critical because they provide the social proof that enables subsequent adoption. But they are a trap, because their enthusiasm can be mistaken for evidence that the organization is ready to adopt.

Early majority (approximately 34%) are pragmatists who adopt based on proven results and peer evidence. They do not want to be first. They want to see that someone like them — in a setting like theirs, with patients like theirs, with constraints like theirs — has implemented the innovation and achieved measurable results. Their question is not “is this exciting?” but “does this work, and will it work here?” In healthcare, these are the majority of clinicians: competent, busy, risk-aware professionals who will change practice when shown compelling evidence from a credible peer, and not before.

Late majority (approximately 34%) are skeptics who adopt primarily because non-adoption has become socially or professionally untenable. They are motivated by peer pressure, institutional mandate, or the dawning realization that they are now the outlier. In healthcare, these are the clinicians who adopt new protocols when it becomes clear that everyone else on the unit already has, or when the quality committee starts tracking compliance. They are not irrational — they are risk-averse and have usually been burned by previous innovations that were championed enthusiastically and abandoned quietly.

Laggards (approximately 16%) are tradition-bound individuals who adopt only when the old way is no longer available or when non-adoption creates an immediate personal crisis. Their reference point is the past. In healthcare, these are the clinicians who use paper workarounds years after the EHR transition, who maintain informal parallel processes that bypass the new system, and who adopt the new practice only when the old one is physically impossible. They are not lazy or stupid. They have a high threshold for change because their experience has taught them that most changes are transient and most promises of improvement are not kept.

The critical insight is not the percentages — those are approximations. The critical insight is that the argument that persuades an early adopter (vision, potential, strategic positioning) does not merely fail to persuade the early majority — it actively repels them. Pragmatists interpret visionary enthusiasm as evidence that the innovation has not been subjected to rigorous scrutiny. The pilot champion’s passion is not transferable. It is, for pragmatists, a warning sign.

The Chasm: Where Healthcare Transformation Goes to Die

Geoffrey Moore’s “Crossing the Chasm” (1991, revised 2014) identified the gap between early adopters and early majority as not merely a delay in the adoption curve but a fundamentally different problem. Moore was writing about technology markets, but the dynamic applies with brutal precision to healthcare transformation.

The chasm exists because early adopters and early majority have incompatible definitions of evidence. Early adopters accept the pilot as proof of concept. Early majority demands proof of generalizability. Early adopters are willing to tolerate incomplete workflows and workarounds because they are invested in the vision. Early majority expects a complete, polished implementation that integrates into their existing workflow without requiring heroic effort. Early adopters are motivated by being ahead of the curve. Early majority is motivated by not falling behind. The same innovation, the same organization, the same implementation team — but two groups that require fundamentally different evidence, different messaging, and different support structures.

In healthcare, the chasm manifests with depressing regularity. A health system pilots a new clinical pathway, decision support tool, or care model on two or three units. The pilot units are selected because they have willing champions — typically early adopters who volunteered or were recruited based on their enthusiasm. The pilot produces strong results: better outcomes, higher efficiency, positive staff feedback. Leadership greenlights scale-up. And then the program stalls.

It stalls because the implementation model that worked for the pilot — high-touch support from the project team, a local champion who troubleshoots in real time, direct attention from executive sponsors, tolerance for workflow disruptions during the learning curve — does not scale. The remaining units do not have volunteer champions. Their managers did not ask for this change. Their staff are already managing three other “priority” initiatives. The evidence from the pilot site is dismissed as “that’s their population” or “they had more resources” or “their champion made it work, and we don’t have a champion.”

This is not resistance in the traditional sense. It is pragmatism. The early majority is asking a perfectly reasonable question: “Show me evidence that this works in a setting like mine, not in the pilot site that was hand-picked and lavished with attention.” When the implementation team cannot answer that question — because they have only pilot data from atypical settings — adoption stalls at 10-15%. The chasm has claimed another program.

The Sepsis Screening Example: Anatomy of a Chasm Failure and Recovery

A 450-bed regional health system deploys a clinical decision support (CDS) tool for early sepsis screening across its 12 inpatient units. The tool integrates with the EHR to flag patients meeting SIRS criteria, prompts a nurse-driven assessment, and escalates to the rapid response team when the screen is positive. The evidence base for early sepsis identification is strong (Rivers et al., 2001, and subsequent bundles from the Surviving Sepsis Campaign). The tool itself is well-designed.

The pilot (months 1-6). Two units are selected: a 30-bed medical unit and a 20-bed step-down unit. Both unit managers volunteered. The medical unit manager is a clinical nurse leader with a master’s degree in nursing informatics who has published on sepsis identification. The step-down manager is a 15-year veteran who chairs the hospital’s quality council. Both are textbook early adopters — opinion leaders, professionally invested in being identified with innovation. The implementation team provides dedicated support: a nurse informaticist embedded on each unit for the first month, weekly check-ins, real-time workflow adjustments, and direct access to the CMIO for clinical questions. Results at six months: 94% screening compliance, 30% improvement in time-to-antibiotic, 18% reduction in sepsis-related ICU transfers. Leadership presents the results at the board meeting. The CEO announces system-wide rollout.

The scale-up (months 7-12). The remaining 10 units receive a standardized implementation package: a two-hour training session, an e-learning module, a quick-reference card, and a go-live date. No embedded informaticist — the project budget assumed the pilot model would not need to scale. No local champions — the unit managers received the rollout as a mandate, not an invitation. The project team holds monthly check-in calls that are attended by project staff and largely ignored by unit leaders who have other priorities.

At six months post-go-live, screening compliance across the 10 scale-up units is 8%. Not 8% improvement — 8% compliance. Nurses on three units report they were not aware the tool had been activated. Two units developed workarounds that bypass the CDS alert entirely. One unit manager openly describes the tool as “another thing IT pushed on us.” The 30% improvement in time-to-antibiotic observed in the pilot has not materialized. The program is at risk of being labeled a failure and defunded.

Root cause analysis. The implementation team, to their credit, conducts a structured assessment rather than blaming “resistance.” They use Damschroder et al.’s Consolidated Framework for Implementation Research (CFIR, 2009) to diagnose the failure. CFIR organizes implementation determinants into five domains: intervention characteristics, outer setting, inner setting, characteristics of individuals, and implementation process. The assessment reveals:

• Inner setting: The pilot units had implementation climate (leadership engagement, learning orientation) that the scale-up units did not. The pilot units’ managers had created readiness; the scale-up units’ managers had received a mandate.
• Implementation process: The pilot used an adaptive, champion-driven model. The scale-up used a standardized, push model. The mismatch between the innovation’s complexity and the implementation support was the primary failure.
• Intervention characteristics: The CDS tool required workflow adaptation for each unit’s specific patient flow, handoff patterns, and staffing model. The pilot units had co-designed their workflows with the informaticist. The scale-up units received a generic workflow that did not fit.

The chasm-crossing strategy (months 13-24). The team redesigns the implementation with four elements targeted at the early majority:

Peer-to-peer evidence. Instead of project team presentations, nurses and charge nurses from the pilot units present their experience — including initial skepticism, workflow challenges, and how they resolved them — directly to the scale-up units. The message shifts from “leadership says do this” to “we did this, here’s what actually happened, here’s what we’d do differently.” Greenhalgh et al. (2004), in their systematic review of innovation diffusion in health service organizations, identify peer influence and observability as among the strongest predictors of adoption — stronger than training, incentives, or mandates.

Workflow adaptation for each unit. The fidelity-adaptation tension — implementing the program as designed versus adapting it for local context — is resolved by defining a “hard core” (the screening criteria and escalation protocol are non-negotiable) and an “adaptive periphery” (the workflow for when and how the screen is performed can be customized by unit). This distinction, articulated by Damschroder et al. (2009) within CFIR, prevents both rigid implementation that ignores local reality and unconstrained adaptation that guts the intervention.

Manager accountability for adoption metrics. Screening compliance becomes a standing item on each unit manager’s monthly performance review. Not as a punitive metric — as a management responsibility. Managers who need help receive coaching from the implementation team. Managers who demonstrate sustained non-engagement receive escalation through the nursing leadership chain. The early majority responds to institutional signals about what is expected and measured (Proctor et al., 2011).

De-implementation of the old screening process. The paper-based sepsis screening tool that predated the CDS system is physically removed from the units. The old order set is deactivated in the EHR. The workaround pathways that bypassed the CDS alert are closed. This is the step most implementation programs skip, and it is often the most important. People do not adopt a new process while the old process remains available. The old process is familiar, comfortable, and does not require learning. As long as it exists, it will be used. De-implementation — the deliberate, systematic removal of the old way — requires explicit attention because it is not automatic. People have habits, muscle memory, workarounds, and sometimes professional identity attached to the current process. Removing it requires the same change management discipline as implementing the new one.

Results at 24 months. Screening compliance across all 12 units reaches 78%. Time-to-antibiotic improvement across the system is 22% — below the pilot’s 30% but clinically significant and operationally sustainable. Three of the 10 scale-up units exceed 85% compliance. The program is no longer dependent on the original two champions.

Implementation Science: The Structured Approach to Crossing the Chasm

CFIR (Damschroder et al., 2009) provides the structured assessment framework for understanding what must be true for adoption to succeed in a given context. It is not a checklist — it is a diagnostic instrument. The five domains (intervention characteristics, outer setting, inner setting, characteristics of individuals, implementation process) each contain constructs that can be assessed before implementation begins, identifying which factors are favorable and which require intervention.

The practical value of CFIR for operators is that it converts “why isn’t this working?” from an unanswerable question into a structured diagnostic. When adoption stalls, CFIR provides the framework to determine whether the problem is the intervention itself (too complex, poor evidence, not adaptable), the external environment (regulatory barriers, payer requirements, competitive dynamics), the internal environment (leadership engagement, implementation climate, readiness for change), the individuals (knowledge, self-efficacy, motivation), or the implementation process (planning, execution, evaluation). Different diagnoses require different interventions. Treating a leadership engagement problem with more training is as misguided as treating a training problem with more leadership messaging.

The fidelity-adaptation tension deserves special attention because it is where implementation programs most commonly err in both directions. Pure fidelity — implementing the innovation exactly as designed regardless of local context — fails because healthcare settings differ in patient populations, workflows, staffing models, EHR configurations, and organizational culture. What works on a 30-bed medical unit in a teaching hospital may not work on a 12-bed unit in a critical access hospital, not because the clinical evidence is different but because the operational context is different. Pure adaptation — allowing each site to modify the innovation freely — fails because the modifications may remove the active ingredients that make the intervention effective. A sepsis screening protocol adapted to “fit the workflow” by eliminating the time-to-antibiotic escalation trigger has been adapted into ineffectiveness.

The resolution is the hard core / adaptive periphery distinction. Define the non-negotiable elements that constitute the intervention’s mechanism of action. Everything else is adaptable. This requires understanding the intervention well enough to distinguish what it is from how it is delivered — a distinction that many implementation teams never make explicit.

Integration Points

Human Factors Module 4: Loss Aversion and Adoption. Adoption of any new practice requires accepting a guaranteed short-term loss (learning curve, workflow disruption, temporary productivity decline) for an uncertain long-term gain (better outcomes, greater efficiency — eventually). This is precisely the structure that Kahneman and Tversky’s prospect theory predicts humans will reject. Loss aversion — the finding that losses are psychologically weighted roughly twice as heavily as equivalent gains — means that a new workflow must deliver substantially more than it costs in disruption to be perceived as worthwhile. The early majority, being pragmatists, perform this calculation explicitly. If the implementation team frames adoption as “this will improve outcomes” without acknowledging and mitigating the short-term cost, the early majority hears “this will make my shift harder for reasons that may not pan out.” Implementation plans that do not explicitly address the loss-aversion calculation are plans that assume humans are rational utility maximizers. They are not.

Human Factors Module 6: Product Design and Adoption Friction. The clinical decision support tool in the sepsis example succeeds or fails partly on its clinical evidence — but day-to-day adoption is determined by its interaction design. Alert fatigue, workflow interruption, cognitive load during high-acuity moments, and integration with existing EHR patterns (HF M6 concepts) determine whether the tool is used or worked around. A well-evidenced intervention delivered through a poorly designed interface will be de-adopted by the early majority regardless of the evidence. Product design determines the adoption friction coefficient: how much effort is required to use the new way versus the old way. When the new way is harder, the old way wins — not because clinicians reject evidence, but because they reject unnecessary friction during already-demanding work.

Product Owner Lens

What is the workforce problem? Healthcare transformation programs consistently succeed in pilot and fail at scale — not because the intervention lacks evidence, but because the implementation model that works for early adopters (champion-driven, high-touch, vision-based) does not work for the early majority (evidence-driven, workflow-integrated, peer-validated). Most organizations do not recognize this as a predictable, diagnosable dynamic and instead attribute scale-up failure to “resistance” or “culture.”

What system mechanism explains it? Rogers’ diffusion of innovation describes five adopter categories with fundamentally different motivations. Moore’s chasm identifies the gap between early adopters and early majority as a qualitative shift in what constitutes persuasive evidence. Damschroder’s CFIR provides the diagnostic framework for assessing implementation context. The interaction of these three frameworks explains why pilots succeed (early adopters in favorable contexts with high-touch support) and scale-up fails (early majority in variable contexts with standardized support).

What intervention levers exist? Peer-to-peer evidence from pilot sites (not project team presentations), workflow adaptation within defined fidelity boundaries (hard core / adaptive periphery), manager accountability for adoption metrics, deliberate de-implementation of old workflows, and CFIR-based diagnostic assessment of implementation context before scale-up begins.

What should software surface? (a) Adoption curve tracking: compliance with the new practice by unit over time, plotted against the expected S-curve to identify units stuck in the chasm. (b) Implementation context dashboard: CFIR domain scores by unit, highlighting which factors are favorable and which require intervention. (c) De-implementation monitoring: usage rates of deprecated workflows, workaround detection, and old-process persistence alerts. (d) Peer evidence library: searchable repository of unit-level implementation narratives — what worked, what was adapted, what failed — accessible to units preparing for adoption.

What metric reveals degradation earliest? The adoption velocity — the rate of compliance increase per unit per month. A unit that reaches 20% compliance in the first month and 22% in the second has stalled. The velocity, not the level, is the early signal. A unit at 15% but accelerating is healthier than a unit at 40% that has plateaued. Velocity decline triggers proactive intervention before the unit settles into a stable low-adoption equilibrium that becomes progressively harder to disrupt.

Warning Signs

These indicators suggest an adoption program is approaching or stuck in the chasm:

• Pilot results are strong but scale-up compliance is below 15% after three months — the implementation model is champion-dependent and does not transfer
• The implementation team describes non-adopting units as “resistant” without diagnosing why — attribution to resistance is a sign that structured assessment has not been done
• Scale-up units received the same implementation package regardless of local context — fidelity without adaptation signals a failure to distinguish the hard core from the adaptive periphery
• The old workflow remains available alongside the new one — de-implementation has not been addressed, and the path of least resistance leads back to the old way
• Adoption metrics are tracked at the system level but not the unit level — system averages mask the bimodal distribution where pilot units are at 90% and scale-up units are at 10%
• Champions from pilot units are now “burned out on this project” — the early adopters have been over-relied upon and are withdrawing, leaving no peer evidence mechanism for the early majority
• Executive sponsorship has moved on to the next initiative — the attention span of leadership is shorter than the adoption timeline, and the program loses organizational priority before it crosses the chasm
• Training completion is high but behavioral adoption is low — knowledge transfer has occurred but workflow integration has not; people know what to do but the system does not support doing it
• The program has no explicit de-implementation plan for the process it replaces — the most commonly missing element in healthcare transformation programs, and the one most predictive of stalled adoption