Clinical Variation: When Differences in Care Become Differences in Outcomes

Clinical variation isn’t the problem—lack of trusted, contextual insight is. Here’s how CMOs can turn variation into a catalyst for learning, alignment, and measurable improvement.

Every healthcare leader knows clinical variation exists.

The real challenge comes when trying to uncover the instances where variation reflects thoughtful, patient-centered judgment—and when it quietly signals risk, waste, or missed opportunity.

For more than four decades, research from the Dartmouth Atlas of Health Care has demonstrated two- to three-fold differences in utilization and treatment patterns across hospitals and clinicians. These differences cannot be explained by patient acuity, demographics, or illness severity alone.

What is new and ever-changing are the pressures surrounding it. Health systems are being asked to deliver more consistent outcomes, improve access and flow, reduce unnecessary utilization, and support a burned-out workforce and all at the same time.

The larger the hospital, the more significant the average cost savings and length-of-stay reduction opportunity. According to an Advisory Board analysis, the average excess cost for hospitals with over 1,000 beds is $58.9 million and average excess length of stay is 20,000 bed days. Additionally, on average hospitals in the Northeast have the greatest length of stay (LOS) reduction opportunity and hospitals in the West (Mountain and Pacific regions) have the biggest cost savings opportunities.

For Chief Medical Officers, such clinical care variation sits at the center of these tensions. Addressing it is not about standardizing medicine. It is about creating the conditions where clinical teams can see clearly, learn from one another, and act with confidence.

Why Clinical Variation Persists—Even in High-Performing Systems

Most variation is not driven by bad medicine or lack of effort. It is driven by context—or rather the absence of it.

Clinicians make thousands of decisions every day, often under time pressure, incomplete information, and competing priorities. Over time, practice patterns develop based on training, experience, risk tolerance, and local norms.

Without meaningful feedback, those patterns harden. They feel invisible from the inside, even when they create real downstream effects.

At the system level, clinical variation is often hard to see clearly because:

• Data remains fragmented. Operational, clinical, patient satisfaction and billing data live in different places, built for different purposes.
• Definitions differ. Length of stay, throughput, utilization, or “timeliness” may be calculated differently across reports, eroding trust and everyone speaking their own language.
• Comparisons lack context. Apples-to-apples views—by acuity, diagnosis, or patient mix—are difficult to assemble.
• Feedback loops are delayed. Annual or quarterly summaries come too late to support learning or course correction. It’s a lagging indicator and change can’t be addressed proactively in the here and now.
• Conversations become personal. Without shared evidence, discussions about performance can feel subjective or punitive. Performance reviews such as looking solely at relative value units (RVUs) without proper context lead to frustrations between clinicians and administrators.

The result is a familiar stalemate: leaders know variation matters, clinicians want to do the right thing, and everyone feels constrained by the data and tools intended to help.

The Hidden Cost of Unexamined Clinical Variation in Healthcare

Not all variation is harmful. Some reflect appropriate clinical judgment or local adaptation. The challenge is knowing which is which.

When unwarranted clinical variation goes unexamined, it can quietly affect:

• Patient experience: Longer waits, inconsistent care pathways, avoidable delays.
• Clinical outcomes: Missed opportunities for best practices to spread.
• Capacity and flow: Small differences in decision-making can compound into system-level bottlenecks.
• Clinician trust: Data that feels inaccurate or unfair is quickly dismissed.
• Leadership credibility: Calls to “do better” without specifics rarely lead to change.

These effects are rarely caused by any one individual. They emerge from systems that lack shared visibility into how care is actually delivered.

Create a System and Reframe the Goal: From Reducing Variation to Supporting Learning

The most effective organizations do not approach variation as something to eliminate. They approach it as something to understand.

That shift matters.

When the systems are defined and the goal becomes learning rather than enforcement, conversations change. Instead of asking, “Why are you an outlier?” teams can ask, “What’s different here—and what can we learn from it?” “Do we have the right process and systems in place to reach the end goal?”

This reframing requires three things:

• Shared definitions: Clinicians and leaders need confidence that metrics reflect real workflows and clinical reality.
• Meaningful comparisons: Performance should be viewed in context by diagnosis, acuity, site, and role so differences are interpretable and fair.
• Timely, trusted feedback: Insights need to arrive close enough to practice to support reflection, not retrospection.

As patient safety leader Peter Pronovost, MD, PhD, has long emphasized, “Most medical errors are not caused by bad people; they’re caused by bad systems”—a principle reinforced through decades of patient safety research and highlighted by the Agency for Healthcare Research and Quality.

The same principle applies to clinical variation. When organizations shift the focus from eliminating differences to understanding the systems that shape decision-making, the conversation changes. Variation becomes a signal for learning—pointing to where processes, data, or workflows may be getting in the way of consistent, high-quality care.

This is where many systems struggle, not because they lack data, but because they lack contextualized data to answer the ‘why.’

Create a Shared View of Addressing Clinical Variation Performance

The Institute of Medicine and National Academies have long promoted the concept of a learning health system—one where data collection and care delivery are integrated so that insights about variation feed back into practice and lead to continuous improvement.

At d2i, we start from a simple belief: meaningful improvement begins with shared understanding.

Our role is not to tell clinicians how to practice or tell administration how to run their business. We see our role to shape how clinical and operational leaders can better uncover patterns that are otherwise buried—across various dashboards, people, sites, and time—using data that clinicians recognize as accurate and relevant.

In practice, this means helping organizations:

• Unify data from multiple sources into a coherent, clinician-ready view.
• Apply consistent logic that reflects how care is actually delivered.
• Enable fair, peer-based comparisons that spark curiosity rather than defensiveness.
• Surface variation early enough to support dialogue, coaching, and local problem-solving.

When variation is visible and trusted, it becomes easier to have productive conversations. We want to spark conversations grounded in evidence, not opinion.

What a Different Reality Looks Like

Health systems that make progress on clinical variation often describe a similar shift:

• Performance discussions feel less adversarial and more collaborative.
• Clinicians recognize themselves in the data and engage with it.
• Leaders move from broad directives to specific, actionable questions.
• Best practices spread organically, informed by real examples from peers.
• Improvement efforts focus on removing obstacles, not assigning blame.

In this environment, variation becomes a signal—not a verdict. It points teams toward where learning is needed and where success already exists.

This case study with Emergency Physicians Professional Association (EPPA) details the impact and improvement on clinical care variation. Medical Director of Quality, Dr. Peter Currie noted, “d2i was able to provide us with more robust clinical data around things like resource utilization, antibiotic and opiate prescribing, and imaging utilization, as well as that next level of complexity that answers, ‘Are you providing appropriate care to a patient?’”

Moving Forward: Pinpointing Clinical Care Variation, Together

Addressing clinical variation is not a project with an end date. It is an ongoing leadership responsibility, made harder by complexity and easier by clarity.

For CMOs and clinical leaders, the opportunity is not to mandate uniformity, but to create the conditions where consistent, high-quality care can emerge. And a process that’s supported by data clinicians trust and conversations that respect their expertise.

At d2i, we don’t replace clinical judgment, but give it better context. We don’t prescribe solutions, but help teams see what is already there—and decide together what to do next.

If you want to learn how we can make your data matter more, get in touch with us.