Resistance and De-implementation

Why Old Workflows Die Hard

Module 7: Change Readiness and Adoption Behavior Depth: Application | Target: ~1,500 words

Thesis: Resistance to change is not irrational — it is a predictable response to perceived loss, uncertainty, and workload increase, and de-implementation of old workflows is often harder than implementation of new ones.

The Operational Problem

A 280-bed community hospital rolls out a digital nursing assessment system to replace paper-based intake forms. The project plan allocates twelve weeks for training and go-live, with an additional four weeks of “stabilization.” At week sixteen, the project team declares success: 95% of assessments are being entered in the digital system.

But a unit-level audit at week twenty tells a different story. Forty percent of nurses on the medical-surgical unit are completing paper assessments in parallel with the digital entry — filling out the old form, then transcribing into the system. They are doing the work twice. The documentation burden has doubled, not decreased. Overtime on the unit has increased 18% since go-live. Two charge nurses have submitted transfer requests. The digital system is technically adopted. The paper system is not de-implemented. And the nurses doing parallel documentation are not being irrational. They are protecting themselves against a system they do not fully trust, using the only tool they know works.

The post-mortem reveals three failures. First, the paper forms were never physically removed from the units. They remained in the supply closets, available and familiar. Second, no sunset date was set for paper cessation — the project plan addressed digital adoption but said nothing about paper retirement. Third, the digital system lacked two fields that experienced nurses considered essential for shift handoff: a free-text “watch for” field and a pain reassessment timestamp. The design team had not consulted bedside nurses during requirements gathering.

The resistance contained valid feedback (missing fields that degraded clinical utility) and irrational persistence (parallel documentation “just in case” that doubled workload). Dismissing all resistance as the former misses design problems. Dismissing all resistance as the latter misses the loss-aversion dynamics driving it. The operator’s job is to distinguish the signal from the noise — and both are always present.

Resistance as Rational Behavior

From the employee’s perspective, organizational change presents a predictable asymmetry. The costs are certain and immediate: new learning curves, disrupted routines, increased cognitive load during the transition, temporary loss of competence in daily tasks. The benefits are uncertain and delayed: the new system might be better, might improve outcomes, might eventually reduce workload — but not today, and not for sure.

Kahneman and Tversky’s prospect theory (1979) predicts the result. Loss aversion — the empirically demonstrated tendency to weight losses roughly twice as heavily as equivalent gains — means that certain short-term costs will dominate uncertain long-term benefits in the employee’s decision calculus. This is not irrationality. It is the predictable output of human decision architecture operating on the information available. A nurse who has spent eight years mastering paper-based assessment does not need to be “resistant to change” to prefer the system in which she is competent over the system in which she is a novice. She is making a locally rational choice under uncertainty.

Ford and Ford (2009) argued that resistance should be understood not as an obstacle to be overcome but as a form of engagement — a resource that contains information about the change itself. Employees who resist are paying attention. They have identified something about the change that troubles them — and that something may be a legitimate design flaw, an unaddressed implementation gap, or a genuine threat to their ability to do their jobs. The silent employee who complies without objection may be the one leaders should worry about more: compliance without engagement produces adoption without commitment, which collapses under the first operational stress.

Three Types of Resistance

Resistance is not monolithic. It manifests in three distinct forms, each driven by a different mechanism and requiring a different response.

Cognitive resistance. The employee does not understand or does not believe in the change. They may lack information about why the change is happening, may not see how the new approach is better, or may have evidence from prior failed initiatives that organizational change promises rarely deliver. The response is informational: clear explanation of the rationale, transparent acknowledgment of what is uncertain, and credible evidence that the new approach works — ideally from peers or comparable settings, not from consultants or executives.

Emotional resistance. The employee understands the rationale but feels fear, anxiety, frustration, or grief. Grief is underappreciated in organizational change: people mourn lost routines, lost competence, and lost identity. A billing specialist who spent a decade mastering a legacy system is not just losing a workflow when the system is replaced — she is losing the expertise that defined her professional value. Bridges (2009) distinguished between change (the external event) and transition (the internal psychological process), arguing that unmanaged transition produces sustained resistance regardless of the quality of the change itself. The response to emotional resistance is acknowledgment: naming the loss, validating the difficulty, and providing time and support for the transition.

Behavioral resistance. The employee may or may not understand or accept the change, but their actions undermine it. This ranges from passive non-compliance (continuing old workflows, not attending training, letting deadlines slip) to active workarounds (building shadow systems, circumventing controls) to outright sabotage (deleting data, undermining adoption among peers). The response must match the severity: passive non-compliance may respond to clearer expectations and accountability; active workarounds often contain design feedback that should be harvested before they are shut down; sabotage requires direct intervention.

The most common leadership error is treating all resistance as behavioral — assuming that the problem is compliance and the solution is enforcement. This misses the cognitive and emotional layers, which are almost always present and which enforcement makes worse.

De-implementation: The Harder Half

De-implementation is the deliberate removal of existing practices, workflows, or tools. It is not the flip side of implementation — it is a distinct process with its own dynamics, its own failure modes, and its own literature. Prasad and Ioannidis (2014) documented the phenomenon of “medical reversal” — the adoption of clinical practices based on initial evidence, followed by the discovery that the practice offers no benefit or causes harm, followed by painfully slow abandonment. Their review of 146 medical reversals found that practices persisted for years or decades after contradicting evidence emerged. The problem was not that clinicians were unaware of the evidence. The problem was that de-implementation is structurally harder than implementation.

Norton, Kennedy, and Chambers (2017) proposed a framework for studying de-implementation as a distinct phenomenon, identifying five factors that make removal harder than adoption:

Habit. Established workflows become automatic behavior. Overriding an automatic behavior requires sustained cognitive effort — exactly the resource that is scarcest during a transition period when new workflows are also demanding attention. The nurse reaching for the paper form is not making a decision. She is executing a motor pattern encoded over thousands of repetitions.

Identity. People identify with their expertise in current processes. The revenue cycle specialist who has mastered ICD-10 coding in the legacy system derives professional identity from that mastery. Replacing the system does not just change her workflow — it threatens her status as the person who knows how things work.

Social validation. When everyone on the unit does it the same way, the practice is reinforced by social proof. Deviating from the group norm — even when the deviation is the officially sanctioned new process — creates social friction. The first nurse to stop using paper forms faces implicit pressure from colleagues who have not yet made the switch.

Sunk cost. Organizations and individuals have invested time, training, and political capital in the current system. Kahneman and Tversky’s escalation-of-commitment research (explored in HF M4) predicts that past investment in the old system creates irrational reluctance to abandon it, even when the new system is demonstrably superior. “We spent three years building this workflow” is not a valid argument for keeping it, but it feels like one.

Absence of forcing function. The old system usually still works. It may work less well, less efficiently, or with greater risk — but it works. Unlike a broken tool that must be replaced, a suboptimal workflow can persist indefinitely because it does not produce acute failure. It produces chronic underperformance, which is far less visible and far less motivating.

Design Principles for Managing Resistance and De-implementation

The hospital in the opening example violated most of these principles. Correcting the failures suggests a general protocol:

Acknowledge the loss. Change means loss — of competence, routine, identity, comfort. Leaders who dismiss this as weakness or inflexibility are ignoring the loss-aversion dynamics that HF M4 documents. Name the loss explicitly. “We know that many of you have spent years developing expertise with the current system, and that expertise is real and valued. The transition will be difficult, and we expect it to take time.”

Involve resisters in design. The nurses who resisted the digital system had identified a genuine design flaw — missing fields that degraded clinical utility. If the design team had included bedside nurses in requirements gathering, the flaw would have been caught before go-live, not after. Ford and Ford’s (2009) reframe — resistance as resource — is operationally actionable: the people most likely to resist are often the people with the deepest knowledge of current workflow, and their objections frequently contain design requirements.

Address cognitive, emotional, and behavioral resistance separately. Information campaigns do not resolve grief. Empathy does not fix ignorance. Accountability does not address fear. Each type of resistance requires its own intervention, and most change management programs default to information (town halls, FAQ documents) while ignoring the emotional and behavioral dimensions entirely.

Remove the old system. Do not let old and new workflows coexist indefinitely. Parallel systems create parallel work, and loss aversion ensures that the familiar system will win any competition with the unfamiliar one. In the hospital example, physically removing the paper forms from supply closets and setting a hard sunset date would have forced adoption rather than allowing indefinite dual documentation.

Set clear sunset dates. De-implementation requires a deadline — a specific date after which the old practice is not just discouraged but unavailable. Without a deadline, the absence of a forcing function (the old system still works) will sustain parallel operation indefinitely. The deadline must be communicated early, repeated often, and enforced without exception.

Warning Signs

Parallel workflows persist past the stabilization period. If staff are maintaining old processes alongside the new system more than four weeks post-go-live, de-implementation has failed. Measure this directly — do not rely on adoption metrics that track new-system usage without checking whether old-system usage has stopped.
Workarounds emerge immediately after go-live. Early workarounds are diagnostic: they identify gaps between the new system’s design and operational reality. Harvest the information before shutting down the workaround.
Resistance is dismissed as “change fatigue” without investigation. Change fatigue is real, but it is also a convenient label that allows leaders to avoid examining whether the resistance contains valid feedback. Every instance of resistance should be evaluated for cognitive, emotional, and behavioral components before being categorized.
The old system remains available “just in case.” This is a de-implementation failure waiting to happen. If the old system is still accessible, it will be used — not because staff are defiant but because loss aversion makes the familiar option feel safer.
Design teams did not include end users from the resisting group. If the people most likely to resist were not consulted during design, their resistance is more likely to contain legitimate design feedback.

Product Owner Lens

What is the workforce problem? Organizational change fails not because employees resist irrationally, but because change imposes certain costs for uncertain benefits, and de-implementation of old workflows is structurally harder than adoption of new ones. The result is parallel systems, doubled workload, accelerated burnout, and adoption metrics that mask operational reality.

What system mechanism explains it? Loss aversion (Kahneman and Tversky, 1979) weights certain short-term costs roughly twice as heavily as uncertain long-term gains. Habit, identity, social validation, sunk cost, and absent forcing functions (Norton et al., 2017) sustain old practices even after superior alternatives are available. Resistance contains three distinct components — cognitive, emotional, behavioral — that require different interventions.

What intervention levers exist? Acknowledge the loss explicitly. Involve resisters in design to harvest feedback. Address resistance types separately. Remove old systems rather than allowing coexistence. Set and enforce sunset dates.

What should software surface? Parallel-system usage: are users engaging both old and new workflows simultaneously? Workaround detection: are users building shadow processes outside the sanctioned system? Adoption vs. de-adoption tracking: new-system adoption rate paired with old-system cessation rate, not just adoption alone. Resistance pattern classification: which user groups show which resistance patterns, to target interventions.

What metric reveals degradation earliest? Parallel-system usage rate — the percentage of users engaging both old and new workflows after go-live. This metric rises before adoption metrics fall, because parallel use inflates adoption numbers (the new system is being used) while concealing de-implementation failure (the old system is also being used). A parallel-usage rate above 20% at four weeks post-go-live predicts sustained dual workload, accelerated burnout in affected roles, and eventual adoption collapse when the doubled burden becomes unsustainable.

Integration Hooks

HF Module 4 (Loss Aversion and Framing). Loss aversion is the primary cognitive mechanism driving resistance to organizational change. HF M4 establishes that losses are weighted roughly twice as heavily as equivalent gains — which explains why the certain costs of change (disrupted routines, lost competence, increased cognitive load) dominate the uncertain benefits (better outcomes, eventual efficiency) in employee decision-making. The framing effects documented in HF M4 also explain why change communication matters: a transformation initiative framed as “what you will gain” produces a different response than the same initiative framed as “what will change” — and the loss frame is what employees experience by default, because the status quo is their reference point. De-implementation triggers loss aversion more acutely than implementation because it involves the removal of something familiar rather than the addition of something new, placing the entire experience in the loss domain of prospect theory’s value function.

PF Module 4 (Milestone Planning and Execution). Grant-funded transformation programs typically plan milestones around implementation: system go-live, training completion, adoption targets. They rarely plan milestones around de-implementation: old-system sunset, parallel-workflow cessation, workaround elimination. This creates a structural gap in program execution. A program that reports 90% adoption of a new care coordination workflow has met its milestone — but if 35% of staff are still maintaining the old workflow in parallel, the program has not achieved the operational change it was funded to produce. PF M4’s milestone planning framework must account for de-implementation as a separate timeline with its own milestones, its own metrics, and its own risk factors. The de-implementation timeline is almost always longer than the implementation timeline, and grant programs that do not budget time and resources for it will report adoption while delivering parallel systems.