Cultivating a Culture of Continuous Innovation in Healthcare Facilities
- Posted by Greg Wahlstrom, MBA, HCM
- Posted in Article, Facilities, Sustainability & Resilience
Cultivating a Culture of Continuous Innovation in Healthcare Facilities
Innovation becomes a culture only when leaders convert ideas into safer care, better work, measurable value, and repeatable operating capability.
Stop asking for more ideas.
Healthcare organizations rarely lack ideas. They lack a disciplined path for deciding which problems matter, testing solutions without compromising care, learning from evidence, and stopping work that should not scale.
Build innovation as an accountable enterprise capability
A culture of continuous innovation is not a collection of pilots, a technology showcase, or an annual challenge. It is the organization’s ability to detect important friction, turn it into a testable proposition, protect patients and staff during experimentation, measure the result, and spread a better way of working.
The original 2024 article made several sound points: leaders should commit visibly, invite staff participation, support interdisciplinary work, invest in development, use technology thoughtfully, and build partnerships. Those principles remain necessary. They are not sufficient for the pressure health systems face in 2026. Labor constraints, fragile margins, rapid AI adoption, growing digital dependency, regulatory obligations, and high expectations for access and experience mean that innovation must be governed as part of the operating model.
The National Academy of Medicine describes a learning health system as one in which science, informatics, incentives, and culture are aligned for continuous improvement, innovation, and equity. That definition is useful because it shifts attention away from novelty. The objective is to make learning part of care delivery so that evidence is created, interpreted, and applied through routine work.
CMS applies the same discipline at a national portfolio level. The CMS Innovation Center’s 2025 strategic direction emphasizes measurable health outcomes and responsible stewardship, while its work on successor models and scaling pathways recognizes that evaluation, adaptation, and spread require different mechanisms. A local health system should use the same logic: each initiative needs a defined outcome, an evaluation plan, a scale pathway, and a stopping rule.
The executive task is to create an innovation portfolio that is connected to strategy, quality, finance, workforce, technology, and capital. The portfolio should include improvements that make today’s work more reliable, adaptations that respond to changing demand, and selected transformations that change the care model. It should also reveal opportunity cost. Every pilot consumes clinical attention, analytic capacity, change energy, integration work, security review, and leadership time. Approving one initiative can delay another even when the budgets sit in different departments.
Culture follows this operating reality. People contribute ideas when they believe leadership will listen, protect thoughtful experimentation, respond to evidence, and explain decisions. They disengage when pilots are announced but never evaluated, when failures are punished, when frontline burden is ignored, or when favored projects continue without proof. Trust grows when the process is visible and the rules apply consistently.
Begin with work that is failing people
Innovation should start with a clearly observed problem, not a vendor demonstration. The strongest opportunities usually sit where patients, clinicians, operations, and finance experience the same friction in different forms.
Combine operational data, safety events, patient feedback, staff observation, financial performance, equity analysis, and direct workflow study before defining the solution.
Patients cannot enter or navigate care at the right time.
Signals
Long waits, abandoned calls, referral leakage, denied authorizations, appointment churn, transportation barriers, and uneven digital access.
Executive test
Is the problem demand, capacity, scheduling logic, communication, coverage, clinical criteria, or a handoff between organizations?
Evidence and care delivery do not align reliably.
Signals
Variation in pathways, delayed diagnosis, medication risk, repeated tests, missed follow-up, alert fatigue, and preventable escalation.
Executive test
What decision is failing, for whom, under what conditions, and how will safety be protected while the new approach is tested?
People spend effort without producing proportional value.
Signals
Duplicate documentation, inbox overload, rework, manual reconciliation, unnecessary movement, role ambiguity, overtime, turnover, and moral distress.
Executive test
Does the innovation remove work, shift work, or add a new layer that will eventually become another burden?
Patients, information, supplies, and decisions move at different speeds.
Signals
Boarding, discharge delay, idle rooms, late starts, lost equipment, missing results, handoff defects, and avoidable length of stay.
Executive test
Which constraint governs the system, and will a local fix move the bottleneck rather than improve total flow?
Patients or staff do not understand, accept, or rely on the new model.
Signals
Low adoption, workarounds, complaints, unequal outcomes, consent confusion, opaque algorithms, and inconsistent escalation.
Executive test
Were the people affected involved early enough to change the design, or were they asked only to approve a finished solution?
Turn learning into a repeatable loop
A system is stronger than a program because it defines how work enters, moves, receives evidence, and exits. The loop below applies to a bedside process change, a new access model, an AI tool, a facility redesign, or a partnership.
Signal
Define the consequential problem and baseline. Identify who experiences it and what happens if nothing changes.
Sandbox
Design the smallest safe test that can challenge the core assumptions without exposing the enterprise.
Evidence
Measure outcomes, process reliability, burden, risk, cost, adoption, and variation across populations and settings.
Scale
Standardize the intervention, clarify ownership, integrate systems, train roles, and verify readiness at each site.
Sustain
Monitor drift, refresh the design, retire outdated work, and return new friction to the signal stage.
Signal: define the problem without embedding the answer
Many projects arrive already framed as a purchase: implement a platform, automate a task, add a command center, deploy an AI model, or open a new site. That framing narrows the search too early. A strong signal statement describes the observed failure, affected population, current baseline, operational context, and consequence. “Reduce avoidable discharge delay for medically ready patients who face medication, transport, or post-acute barriers” is more useful than “buy a discharge application.”
Executives should require direct observation alongside aggregate data. Walk the process with patients and staff. Review exceptions and near misses. Compare shifts, sites, language groups, payers, and clinical populations. Averages can conceal the exact conditions where an innovation will fail. The signal stage should end with an agreed problem owner and a measurable outcome, not a preferred product.
Sandbox: make experimentation safe and consequential
A sandbox is a controlled operating environment with explicit boundaries. It defines the included population, clinical criteria, duration, staffing, consent or communication requirements, data access, escalation route, rollback plan, and maximum tolerable risk. The goal is not to protect the project from criticism. It is to protect patients and staff while producing a decision-quality test.
The smallest test should still confront the hardest assumption. If an access model depends on patients completing digital intake, include people who need language, disability, or caregiver support. If an AI tool is intended to reduce documentation burden, measure total work after review and correction, not only draft-generation time. If a care pathway relies on evening pharmacy access, test it when those services are constrained.
Evidence: measure value as a balanced outcome
No single metric can establish that an innovation works. A faster process may transfer burden to another team. Higher use may reflect weak selection rather than benefit. Lower cost may be achieved by reducing access. A balanced evaluation should include patient outcome, safety, experience, equity, workforce burden, process reliability, total cost, and strategic relevance. Measures must be defined before the test begins so the team cannot choose favorable evidence afterward.
Use both quantitative and qualitative evidence. Data can show the size and distribution of an effect. Interviews, observation, and case review explain why the result occurred and whether it can travel. An improvement that depends on one exceptional manager or extra pilot staffing is not yet a scalable operating model. The evidence stage should produce a decision: stop, redesign, continue the test, scale selectively, or scale broadly.
Scale: reproduce capability, not merely technology
Scaling requires more than copying a configuration. Leaders must identify the active ingredients that cannot change and the local elements that should adapt. A scale package should include workflow, role design, training, data definitions, integration, escalation, controls, staffing assumptions, patient communication, equity safeguards, operating cost, and a readiness assessment. Sites should not be declared live because installation is complete.
Spread should occur in waves with observable entry and exit criteria. Early sites validate the implementation package. Later sites test whether the model works under different staffing, demand, infrastructure, and community conditions. A central team can support standards and learning, but the receiving operation must own results. Innovation ends when the new capability is accepted into normal governance, budgeting, performance review, and continuous improvement.
Sustain: manage drift, obsolescence, and new learning
Every intervention changes after launch. Staff develop workarounds, patient populations shift, software updates, regulations evolve, performance varies, and the original problem may move. Sustaining does not mean freezing the design. It means maintaining the outcome through controlled adaptation. The owner should monitor leading and lagging measures, investigate unusual variation, review incidents, and maintain a clear path for staff and patients to report friction.
Leaders should also retire work. Old forms, duplicate reports, temporary committees, parallel systems, and obsolete controls often remain after the new model launches. A closeout review should identify what stops, what transfers, what is archived, and what must continue. Capacity for future innovation is created partly by removing yesterday’s work.
Use gates that force evidence and tradeoffs
Decision gates create a common standard without making every initiative identical. Low-risk workflow improvements can move quickly. Higher-risk clinical, data, AI, capital, and vendor initiatives require deeper review. The rigor should follow the consequence of failure.
Problem gate
Confirm that the problem is consequential, strategically relevant, owned by an operating leader, and supported by a credible baseline. Separate the need from the proposed solution. Identify affected patients, staff, sites, and partners.
Would leadership still invest in solving this problem if the proposed technology or partner were unavailable?
Safety and feasibility gate
Classify clinical, privacy, cybersecurity, regulatory, financial, workforce, ethical, and reputational risk. Define the sandbox, accountable clinical owner, escalation path, stop criteria, and rollback plan. Confirm that the team has time, data, integration, and operational capacity.
Can the organization detect harm or failure early enough to intervene, and is one person clearly authorized to stop the test?
Evidence gate
Compare results with the preapproved measures and baseline. Review variation, unintended consequences, adoption, workarounds, staff burden, patient experience, total cost, and equity. Distinguish evidence produced with temporary pilot resources from performance under normal operations.
What evidence supports the result, what remains uncertain, and what finding would reverse the recommendation?
Scale gate
Approve the target population, sites, sequence, operating owner, implementation package, capital and operating budget, vendor obligations, workforce plan, and post-launch measures. Confirm that source systems, support teams, training, and local leaders are ready.
Is the organization scaling a proven capability, or amplifying an unresolved pilot with more exposure?
Sustain or retire gate
Transfer the initiative into routine governance and budget. Set performance thresholds, review cadence, change-control rules, incident pathways, and retirement criteria. Remove duplicate work and close pilot infrastructure that is no longer needed.
Who owns the result after the innovation team leaves, and what will cause the organization to redesign or retire the model?
Balance the portfolio across four jobs
An enterprise portfolio should not consist only of large transformations or only of local improvements. It needs a deliberate mix of work, with different time horizons, evidence standards, risk profiles, and funding mechanisms.
Make today’s care safer and easier
These initiatives reduce known defects, variation, delay, rework, and burden in established services. They usually have a clear baseline and can produce value quickly.
- Medication, handoff, diagnostic, and infection-prevention reliability
- Flow, scheduling, discharge, and supply improvements
- Documentation simplification and removal of duplicate work
- Standard work that protects clinical judgment rather than replacing it
Respond to changing demand and constraints
These initiatives modify the operating model as patient needs, labor availability, reimbursement, or community conditions change.
- Virtual, hybrid, mobile, home, and distributed care pathways
- Team-based care and role redesign
- Capacity pooling and cross-site coordination
- New access channels that preserve clinical appropriateness
Create a materially different capability
These bets may require new infrastructure, partnerships, skills, or economics. The uncertainty is higher, so the investment should be staged and the assumptions visible.
- New service lines, platforms, and care-delivery models
- Enterprise data and AI capabilities
- Major facility and digital redesign
- Partnerships that change where or by whom care is delivered
Build the capacity to improve repeatedly
These investments support the entire portfolio. Their value appears in faster, safer, and more reliable decisions across many initiatives.
- Implementation science, analytics, simulation, and design capability
- Patient and workforce co-design mechanisms
- Portfolio governance, evidence standards, and transparent reporting
- Reusable integration, privacy, cybersecurity, and change-control patterns
Fund by learning stage
Release resources in tranches. Provide small funds to validate the problem and design a safe test. Increase funding only when evidence and readiness justify the next gate. Separate pilot funding from the recurring operating budget required at scale. This reduces sunk-cost bias and makes it possible to stop weak initiatives without treating every stop as a failure.
Give people time, authority, and protection to learn
Leaders often describe innovation as everyone’s job while allocating no time, decision rights, or support. Culture becomes credible when participation is built into roles and managers are accountable for learning.
Frontline teams
Frontline staff see friction first, but speaking up can carry social and workload risk. Create simple routes to surface problems, participate in observation and testing, and receive feedback about decisions.
- Protect time for improvement and co-design.
- Recognize problem definition, not only winning ideas.
- Do not ask staff to absorb pilot work indefinitely.
- Close the loop with people who raised the issue.
Middle managers
Managers translate enterprise intent into daily conditions. They can protect testing and learning, or quietly suppress it when performance pressure leaves no capacity.
- Include improvement capability in manager expectations.
- Teach testing, measurement, facilitation, and change leadership.
- Provide escalation when local targets conflict with learning.
- Reward removal of low-value work and honest reporting.
Executives and board
Senior leaders set the boundaries, allocate scarce capacity, and determine whether evidence has authority. Their behavior defines whether the process is real.
- Model curiosity before advocacy.
- Ask about burden, safety, equity, and opportunity cost.
- Stop weak work visibly and without blame.
- Hold sponsors accountable for outcomes after scale.
The AHRQ learning health system workforce framework includes improvement and implementation science, informatics, engagement, and equity among the capabilities needed to reduce avoidable variation and support the use of evidence in practice. These are not specialist concerns alone. Executives need enough fluency to ask better questions, and operational teams need access to expertise when the consequence of failure is high.
Govern the lifecycle, not only the launch
AI compresses the time between idea and deployment, but it does not remove the need for clinical accountability, evidence, data governance, security, workflow design, and ongoing monitoring. It makes those disciplines more important.
Map the use and consequence
Define the specific decision or task, intended user, patient population, operating environment, and plausible failure modes. The NIST AI Risk Management Framework offers a voluntary structure for governing, mapping, measuring, and managing AI risk. Health systems should connect that structure to clinical quality, privacy, cybersecurity, legal, ethics, human resources, and vendor oversight.
Measure performance in the local environment
Vendor evidence is not a substitute for local validation. Examine data compatibility, workflow, accuracy, failure distribution, subgroup performance, human factors, override behavior, and downstream action. The purpose is not merely to show that the output resembles a benchmark. It is to determine whether the system supports a safer or more effective decision under actual operating conditions.
Design human oversight as work
“Human in the loop” is meaningful only when the person has time, information, authority, and skill to challenge the output. Define who reviews what, under which threshold, with what documentation and escalation. Measure correction and review burden. Automation that shifts hidden labor to clinicians may worsen the operating problem it was meant to solve.
Control change and monitor drift
Models, data, interfaces, clinical practices, and populations change. The FDA’s 2025 guidance on predetermined change control plans for AI-enabled devices reinforces the importance of anticipating modifications while maintaining safety and effectiveness. Even when a tool is outside device regulation, health systems should document material changes, retest when appropriate, and retain the ability to pause or roll back.
Report incidents and retire responsibly
Create simple reporting routes for unsafe, biased, misleading, or burdensome behavior. Combine technical alerts with user observation and case review. Retirement requires its own plan: preserve records, communicate the change, remove integrations, manage vendor and data obligations, and verify that staff do not continue using an unsupported workaround.
Adoption rises, but total discharge time does not improve
The pilot appears promising during the first two weeks. Clinicians open the dashboard, case managers receive earlier prompts, and leadership sees high use. At the evidence gate, the team finds little change in median discharge time. Staff report additional messages, and afternoon delays remain concentrated among patients needing medication access, transportation, home equipment, interpreter support, or post-acute placement.
Instead of labeling the pilot a success because adoption is high, the team returns to the friction map. Direct observation shows that the prediction is often accurate, but it does not change the constrained services that determine departure. Pharmacy cutoffs, transportation scheduling, documentation timing, and external placement remain the governing bottlenecks. Some patients receive repeated status questions without faster resolution, creating experience and trust concerns.
The case illustrates a central rule: an innovation can perform as designed and still fail to improve the system. Culture improves when teams are allowed to report that result, preserve the learning, and change direction without defending the original solution.
Measure the health of the portfolio and the results of the work
Executives need two views. The first shows whether the portfolio is moving with discipline. The second shows whether scaled initiatives are producing durable value. A dashboard should trigger a decision, not decorate a meeting.
| Domain | Executive measure | Evidence | Escalation question |
|---|---|---|---|
| Strategic alignment | Share of active resources tied to named enterprise priorities and defined outcomes | Approved portfolio map, owners, budgets, and expected value | Which work continues because of sponsorship rather than strategic relevance? |
| Problem quality | Initiatives with a verified baseline, affected population, and direct workflow observation | Signal brief, process data, patient and staff insight, equity analysis | Where has the solution been selected before the problem is understood? |
| Learning velocity | Median time from problem gate to decision-quality test, segmented by risk class | Gate dates, review delays, dependencies, and decision logs | Which approvals protect value, and which add waiting without reducing risk? |
| Safety and trust | Tests with clear clinical owner, stop criteria, incident route, and patient or staff communication | Risk classification, sandbox protocol, case review, reports | Can the organization detect harm, burden, or inequity before broad exposure? |
| Evidence quality | Initiatives assessed against preapproved balanced measures | Outcome, safety, experience, equity, workforce, cost, and adoption results | What important consequence is missing from the evaluation? |
| Portfolio decisions | Share stopped, redesigned, selectively scaled, broadly scaled, or retired | Decision records and reasons by quarter | Does the absence of stopped work signal excellent selection or weak governance? |
| Scale reliability | Sites meeting readiness and outcome thresholds after launch | Readiness review, implementation fidelity, local adaptation, post-launch trend | Where is deployment complete but operating capability not established? |
| Capacity created | Hours, steps, reports, systems, and committees removed through the portfolio | Verified work elimination and sustained workload change | Is innovation creating room for better work or adding permanent complexity? |
Install the operating system in 90 days
The first 90 days should not produce a new innovation department and a larger pilot list. They should create one visible portfolio, common decision rules, a small number of consequential tests, and evidence that low-value work can stop.
Days 0–30
See the system
- Name one accountable executive and a small interdisciplinary portfolio council.
- Inventory active pilots, innovation funds, vendor trials, AI tools, transformation work, and local improvement projects.
- For each item, record the problem, owner, population, risk, resources, measures, current gate, and next decision date.
- Identify duplicated work, orphaned pilots, missing clinical ownership, and initiatives without a defined outcome.
- Select three enterprise frictions using patient, workforce, quality, access, financial, and equity evidence.
- Publish the first portfolio review date, decision owners, and evidence expected at each gate.
Days 31–60
Set the rules
- Approve risk tiers and the five portfolio gates.
- Create standard one-page signal, sandbox, evidence, and scale briefs.
- Define minimum patient, workforce, equity, cost, and safety measures.
- Establish one route for staff and patients to surface friction and receive feedback.
- Assign review service levels so low-risk tests move quickly and high-risk work receives appropriate scrutiny.
- Stop or redesign at least one initiative that cannot meet the new standard.
Days 61–90
Run and learn
- Launch two or three bounded tests against the selected enterprise frictions.
- Hold decision reviews using preapproved evidence rather than presentation quality.
- Publish a portfolio scorecard that distinguishes activity, learning, and outcomes.
- Create a scale package for one proven capability and a closeout plan for one stopped effort.
- Integrate innovation decisions into capital, technology, quality, workforce, and operating reviews.
- Set the next quarterly portfolio reset and require owners to retire duplicate work.
What leadership must do next
Govern innovation as a portfolio. Start with consequential friction, test inside explicit boundaries, measure balanced outcomes, scale capability rather than software, and transfer ownership into routine operations.
The strongest signal of maturity is not the number of ideas launched. It is the organization’s ability to learn quickly, stop responsibly, remove low-value work, and sustain improvements patients and staff can feel.
References for implementation
- National Academy of Medicine: Leadership Consortium and Learning Health System
- National Academy of Medicine: Learning Health System Series
- CMS Innovation Center: 2025 Strategic Direction
- CMS Innovation Center: Innovation Models
- CMS: Successor Models, Scaling Pathways, and the Innovation Center
- AHRQ: Building the Learning Health System Workforce
- AHRQ: Approaches to the Quality Improvement Process
- AHRQ: SOPS Hospital Survey 2.0
- AHRQ: Engaging Patients in Digital Healthcare Innovation
- HHS: Artificial Intelligence Strategy and Implementation
- NIST: Artificial Intelligence Risk Management Framework
- FDA: Guidances with Digital Health Content
- FDA: Predetermined Change Control Plans for AI-Enabled Device Software Functions
