ResearchPod Summary
Alex: Welcome to another episode of ResearchPod. Today, we're examining a fundamental blind spot in implementation science: the assumption that changing behavior is a symmetric process.
Sam: So we've been treating the act of starting a new clinical practice the same way we treat stopping an old one?
Alex: That's the paper's central argument. Current behavioral theories are monolithic — they don't distinguish between acquiring new habits and inhibiting established ones. And that asymmetry matters enormously in practice.
Sam: If a hospital mandates stopping routine chest X-rays, but the theory offers no roadmap for unlearning, are they just setting clinicians up for failure?
Alex: That's exactly the puzzle. We've built a sophisticated toolkit for "doing more" — adoption, scale-up, spread. But we lack a comparable framework for the equally critical task of "doing less." The authors call this de-implementation, and their argument is that it requires a fundamentally different behavioral logic.
Sam: So how did they approach this? What's the method?
Alex: They used a Critical Interpretive Synthesis — essentially a structured review designed not just to aggregate findings, but to build theoretical insight from across a literature. They examined 66 papers, asking whether existing behavioral theories treat implementation and de-implementation as mechanistically distinct.
Sam: And the answer?
Alex: Largely no. Most theories are symmetric — they assume the same determinants drive behavior whether you're adding or subtracting a practice. Motivation, for instance, is treated as a unidirectional lever: increase it, and behavior changes. But the authors argue that the psychological pathways for inhibiting an established behavior are categorically different from those for initiating a new one.
Sam: That's a meaningful gap. If the mechanisms are genuinely distinct, then applying an adoption framework to a discontinuation problem isn't just incomplete — it's the wrong model entirely.
Alex: Precisely. And the authors found that most theories don't even acknowledge the distinction. Practitioners attempting to wind down low-value care are essentially navigating without a map built for that direction.
Implementing evidence-based healthcare requires both the adoption of new practices (implementation) and the abandonment of ineffective or harmful ones (de-implementation). While behavioral science has robustly informed how to increase desired behaviors, it is unclear whether these same theories apply to the reduction of undesired behaviors. This study used a Critical Interpretative Synthesis (CIS) to examine 66 papers across diverse fields—including psychology, education, and business—to determine if existing behavioral theories provide a distinct theoretical rationale for de-implementation compared to implementation.
The synthesis revealed three primary insights:
Lack of Theoretical Distinction: The vast majority of the 15 identified behavioral theories do not distinguish between the mechanisms for increasing and decreasing behavior. Many theories were applied exclusively to one direction or the other, rather than providing a unified framework that accounts for both.
Reliance on Substitution: A common strategy for de-implementation is to replace an undesired behavior with a substitute behavior. However, the researchers found no theoretical basis for this approach in the literature, nor were there systematic methods proposed for selecting appropriate substitutes. The focus remained on applying existing theories to increase the frequency of the new, substitute behavior.
Operant Learning Theory (OLT): OLT was the only theory that explicitly distinguished between the two directions, utilizing reinforcement to increase behavior and punishment to decrease it. However, the authors note that applying OLT in complex healthcare settings is challenging due to the delayed nature of clinical outcomes, ethical concerns regarding punishment, and the potential for unintended consequences.
As healthcare systems face increasing pressure to reduce low-value care, the lack of a clear theoretical foundation for de-implementation is a significant barrier. Current practices often rely on intuitive, ad-hoc strategies like behavior substitution without a rigorous framework for success. This study highlights a critical gap in behavioral science, suggesting that researchers must either adapt existing theories to account for the unique challenges of 'unlearning' or develop new, evidence-based strategies specifically for de-implementation.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.
Sam: It's like having a detailed map of a city that only shows one-way streets going forward. You can plan a route, but the moment you need to reverse course, the map is useless.
Alex: That captures it well. The one exception the review identified was Operant Learning Theory, which explicitly separates techniques for increasing versus decreasing behavioral frequency. It's the only framework that formally encodes directionality.
Sam: But operant conditioning is a fairly blunt instrument. Does it actually translate into the complexity of a modern clinical environment?
Alex: That's the tension. Operant Learning Theory does offer a mechanism — specifically something called Differential Reinforcement of Incompatible Behavior, or DRI, where you reinforce a substitute behavior that physically can't coexist with the one you're trying to eliminate. It's well-established in the behavioral literature. But the techniques that follow from it — including punishment-based contingencies — sit uncomfortably with professional autonomy norms in healthcare. The theory is directionally correct but practically constrained.
Sam: So the field is caught between models that are theoretically incomplete and the one model that has the right structure but generates implementation friction of its own.
Alex: That's the core tension. And it gets sharper when you look at what happens in practice. When researchers do attempt to reduce low-value care, they often default to substitution — replacing the outdated practice with something new. Which sounds reasonable. But the review found no systematic method for choosing that substitute. It's ad hoc. Clinicians are using substitution as a strategy without the theoretical grounding to know whether it's the right substitute, or why.
Sam: Which means they're accidentally doing DRI without knowing it — and without the framework to optimize it.
Alex: Exactly. The practice is there; the theory isn't. And that matters because without the theory, you can't predict when substitution will work, when it will fail, or how to design the replacement behavior to be genuinely incompatible with the one you're trying to eliminate.
Sam: What's the biggest structural gap the authors identify beyond the symmetric framing?
Alex: The near-total exclusion of cognitive psychology. The review found that the action-versus-inaction asymmetry — a well-documented phenomenon in the regret literature — is essentially absent from implementation science frameworks. In clinical settings, the perceived psychological cost of not acting is often higher than the cost of taking an unnecessary action, even when the evidence clearly supports restraint. A clinician who orders a test that turns out to be unnecessary faces a different kind of accountability than one who withholds a test and something goes wrong. That asymmetry creates a structural barrier to de-implementation that standard behavioral models simply aren't built to capture.
Sam: So the omission commission asymmetry — the sense that harm caused by inaction feels morally weightier than harm caused by action — is actively working against evidence-based restraint, and the frameworks don't account for it.
Alex: Right. And that's arguably the most clinically consequential gap. You can design the best de-implementation intervention in the world, but if the underlying model ignores the psychological cost of non-action, you're missing the mechanism that's generating the resistance in the first place.
Sam: So where does this leave the field?
Alex: The paper doesn't offer a new framework — that's explicitly outside its scope. What it does is make the case that one is needed, and that it has to do three things the current literature doesn't: encode directionality, provide a principled basis for substitution, and incorporate the cognitive asymmetries around omission and commission. Until that framework exists, de-implementation efforts are working with borrowed tools that weren't designed for the job.
Sam: That's a useful diagnostic, even without a prescription. It at least tells you where the map runs out.
Alex: It does. And for anyone designing de-implementation studies right now, the practical takeaway is probably this: if your behavioral theory doesn't distinguish between starting and stopping, treat that as a design limitation worth naming explicitly — not an assumption you can quietly carry forward. Thanks for listening to ResearchPod.