The Crisis of Over-Protocolization in Healthcare

Modern healthcare systems thrive on precision, predictability, and compliance. Protocols and standardised pathways derived from years of evidence have long served as the backbone of patient safety and clinical consistency. Yet, in a world increasingly defined by uncertainty, complexity, and rapid technological advancement, this rigid protocol mindset can become a liability.

Historically, protocols emerged from necessity. The Flexner Report of 1910, for instance, revolutionised American medical education by enforcing structure and evidence-based approaches. It standardised excellence but also flagged a top-down culture where deviation from the norm was discouraged. Over time, this helped improve quality, but it also cultivated a fear of experimentation, especially in clinical workflows and healthcare delivery.

Health systems are typically excellent at scaling what is already proven, but ill-equipped to experiment with what is emerging. Innovation is treated as an afterthought, rather than a core operational function. As technologies evolve faster than regulatory or funding cycles, the gap between what’s possible and what’s implemented grows dangerously wide.

Enter “sandbox thinking.” Rather than replacing protocols, sandboxes precede them. These safe, flexible environments allow new technologies and models to be tested in isolation, refined through feedback, before being rolled out. Just as children learn by building and breaking in a sandbox, healthcare systems must rediscover how to safely experiment without risking real lives or reputations.

What is a Sandbox? Origins, Principles, and Use Cases

The concept of a sandbox originates in software developmen,t secure, contained environments where developers test new code without threatening the integrity of the broader system. But the idea has since expanded into industries like finance, policy, and even urban planning.

In fintech, for instance, regulatory sandboxes allow startups to trial innovations with real users under controlled conditions and regulatory oversight. This bridges the gap between innovation and compliance. In gaming, platforms like Minecraft allow users to build, destroy, and iterate within a rule-governed but open environment, driving creativity and problem-solving.

Core principles of a sandbox:

• Low-risk experimentation: Failure is acceptable, even encouraged, as a pathway to learning.
• Isolation: Tests occur in a shielded environment, preventing broader system disruption.
• Iteration: Results are used for continuous refinement.
• Supervision: A sandbox has boundaries, but within them, creative freedom thrives.

In healthcare, this concept has yet to be fully embraced. And it should be.

Why Healthcare Needs Sandboxes Now

• We stand at a turning point in global health. AI is rewriting diagnostics. Remote monitoring is redefining access. The metaverse is infiltrating rehabilitation. Yet many of these tools struggle to make it from pilot to practice.
• Why? Because healthcare’s adoption model is binary: either fully implemented or not at all. This model works when the cost of failure is catastrophic. But in today’s agile world, failing to experiment is the greater risk.
• Consider IBM Watson for Oncology. Promising at its launch, it faced immense backlash when outcomes didn’t meet expectations. But had Watson been tested in a simulation-based sandbox and adjusted iteratively based on clinician feedback, if that happened, it could have evolved more sustainably.

Similarly, many hospitals deploy AI-based triage systems or chatbot symptom checkers without meaningful testing against real-world complexity, language barriers, atypical patient behaviors, and comorbidities. Sandboxes would allow these tools to be pressure-tested across diverse simulated conditions, identifying blind spots long before they reach patients.

Healthcare must abandon the "wait until it's proven" mentality. It’s time to adopt "test until it's ready."

Lessons from Other Industries: What Healthcare Can Learn

Industries outside healthcare have embraced the sandbox ethos with transformative outcomes.

Aviation is a prime example. No pilot is licensed without extensive simulator hours, facing everything from routine landings to catastrophic engine failures in a safe, reproducible environment. Why doesn’t healthcare require similar conditions before rolling out high-stakes innovations? The risk in aviation is approximately one per million flights. In comparison, there are about 250,000 patient deaths due to medical error annually in the United States, where healthcare quality is considered high relative to other countries.

Gaming and EdTech lead in user-centered iteration. Games evolve through beta testing, hotfixes, and user analytics. Healthcare apps, by contrast, often launch with minimal end-user feedback and suffer low engagement. Imagine if EHR systems were sandboxed with clinicians to test UX design before national rollouts.

Policy labs in cities like Helsinki, Amsterdam, and Singapore have used urban sandboxes to trial congestion pricing, bike-sharing models, or climate initiatives, analysing real behavioral data before legislation. These experiments have led to better adoption, trust, and measurable outcomes.

“Healthcare has the stakes and scale. Now, it needs the sandbox tools.” 

Simulated Environments in Healthcare: A Case for Expansion

Simulation in healthcare is nothing new, but it’s narrowly applied. Most simulation centers are focused on clinical education: CPR drills, surgical rehearsals, or OSCEs for medical students.

Yet the potential is far broader.

Imagine a Hospital Sandbox Lab where:

• An AI scribe is tested across 20 simulated outpatient visits, adjusting its natural language processing to local dialects.
• A new nurse scheduling algorithm is run on a digital twin of an entire ward to measure burnout reduction.
• A VR platform for stroke rehab is tested on simulated patient avatars before being trialed in real patients.

Mini-case study: A recent simulation-based study from Stanford researchers offers a concrete example of sandbox-style innovation in healthcare. They developed a Patient Simulator using real-world electronic health record (EHR) vignettes. This simulator engaged with an AI triage agent in over 500 diverse, multi-turn clinical scenarios, including children, chronic disease patients, atypical presentations, and more

Key outcomes from the study: Clinician reviewers rated the simulated encounters as 97.7% consistent with real patient presentations.

The AI-derived case summaries were deemed 99% accurate by expert validation.

Crucially, the system also identified edge cases like non-verbal symptoms or atypical disease courses that commonly trip up real-world AI tools.

This study exemplifies sandbox thinking by employing large-scale simulations to test and iterate AI clinical tools before live deployment. It’s a procedural and ethical advance and any shortcomings were corrected in sandbox environments, not at patient bedsides.

Designing the Healthcare Sandbox: Principles and Requirements

To function, a healthcare sandbox must be more than an idea, it must be a structured system.

Key design pillars:

• Safety by design: The sandbox should be isolated from actual patient care, with simulated data and scenarios.
• Inclusive participation: Clinicians, designers, developers, regulators, and critically, patients must co-design the experiments.
• Rapid feedback loops: Every test cycle should result in learnings that guide the next.
• Regulatory harmony: Sandbox pilots can be paired with provisional regulatory support to shorten the path from test to deployment.
• Outcome pluralism: Success isn’t just improved clinical metrics, it’s usability, trust, scalability, and experience.

These sandboxes can also become training grounds for policy. For example, Qatar’s health authorities could prototype a national telehealth reimbursement model in a sandbox simulating provider behaviors, patient access patterns, and fraud scenarios before rollout.

The Ethical Advantage: A Better Way to Fail

Healthcare ethics has long been governed by “do no harm.” Ironically, this has often been interpreted as “never try unless you're sure.” But inaction, or delayed action, can be just as harmful.

Sandboxes offer a new ethical paradigm as one of proactive responsibility. Instead of trialing risky models on real patients, organisations can simulate consequences, build in guardrails, and co-design with patients themselves. It allows for mistakes to be made where no one is harmed and for progress to be shaped with empathy and foresight.

Failure in a sandbox is not only permissible, it’s ethical.

Reframing Leadership and Mindsets for Iteration

The success of any sandbox depends not just on space and tools, but on mindset especially among healthcare leaders. Healthcare executives are often risk-averse, trained to maintain operations, meet KPIs, and avoid litigation. But innovation leadership requires different muscles: comfort with ambiguity, rapid decision-making, and openness to failure.

To cultivate this:

• Institutions must invest in innovation literacy, training leadership teams on agile, design thinking, and sandboxing frameworks.
• Executive sponsors should champion 10–20% of their budget toward “controlled experimentation” each year.
• Boards should shift some of their metrics from static compliance toward dynamic learning.

During one of my trips, I was introduced to a compelling real-world model of sandbox thinking in the Netherlands. The Dutch Healthcare Authority (NZa) has implemented a national program known as the “small-scale experiments” policy framework, which allows healthcare providers and insurers to pilot innovative care models that don’t yet meet the traditional criteria for reimbursement or standard regulation.

Launched in 2019 and updated in 2024 to align more closely with broader healthcare laws, this initiative enables organisations to test new delivery models such as digital consultations, remote monitoring, or integrated home-based therapies within a controlled and legally sanctioned environment. These experiments operate with oversight and data collection requirements but enjoy temporary exemptions from full-scale compliance constraints.

For example, a provider might trial a telemonitoring program for chronic heart failure patients. Under normal rules, such a service might not qualify for reimbursement. But under this experimental framework, it can be tested with real patients, studied for safety and efficacy, and if successful, it can be used to build a case for national adoption.

By embedding experimentation within the regulatory architecture, the Dutch system exemplifies how policy and innovation can co-evolve, offering a viable template for other nations looking to institutionalise safe, real-world healthcare innovation.

Conclusion: The Future Is a Sandbox

The Sandbox Doctrine argues that health systems must shift from binary models of adoption toward iterative models of discovery. Protocols are essential, but they represent the end of a process, not its beginning.
In the age of AI, digital therapeutics, and hybrid care models, it is no longer enough to test innovations in theory. We must simulate them in context. We must build, break, and rebuild them before they ever touch a real patient.

What if your hospital had a sandbox unit?

What if your national health ministry maintained a simulation-based policy lab?

What if regulators started asking, “Where did you sandbox this?” instead of “Where did you pilot it?”

This isn't just a technical shift. It's a philosophical one. It demands humility, courage, and vision. And it starts with asking the right questions, not “Is this safe yet?” but “How can we learn safely?”

“The future of healthcare belongs not to those who avoid failure, but to those who design environments that make failure safe, meaningful, and transformative.”

Healthcare doesn’t just need new tools.

It needs new playgrounds.

And the courage to play in them.

References:

• Singhal, S., et al. (2024). Simulating Healthcare Interactions with Large Language Models. arXiv. Link: https://arxiv.org/abs/2506.04032
• Dutch Healthcare Authority (NZa) – Small-Scale Experiments Policy Framework (2025 Update) Source: MTRC Medical Technology Consultancy Link:  https://mtrconsult.com/news/dutch-healthcare-authority-updates-policy-framework-small-scale-experiments-2025
• Ross, C., & Swetlitz, I. (2017). IBM pitched its Watson supercomputer as a revolution in cancer care. It’s nowhere close. STAT News. Link: https://www.statnews.com/2017/09/05/watson-ibm-cancer/

--AHHM Issue 70--