There’s Nowhere to Hide From Big Data

Healthcare data can identify inappropriate care patterns through several mechanisms, and generative AI is enhancing these capabilities in significant ways.

Traditional Data Identification Methods

Statistical benchmarking can identify outliers (not the kind of outliers we aim for).

Let’s use cesarean section rates as an example. Why?

The new head of the FDA is Dr. Marty Makary. He has been vocal for a long time about high c-section rates in the U.S. as one are of a specific specialty care where care may not be appropriate, and why that matters so much.

His critiques are part of a larger mission to reveal how systemic issues in healthcare often drive unnecessary—and sometimes dangerous—medical interventions. His three books fundamentally challenge how we think about healthcare quality, as he highlights a crucial distinction between traditional “quality metrics” and “appropriate care.” I think of this as part of the next stage of value-based care.

So given his new and influential role at the FDA, I think his work is worth knowing, and I want to share it with you. I believe Dr. Makary’s work writ large is really a window into what to expect given a growing ability and motivation to use data to improve care by surfacing providers who may be delivering sufficiently “inappropriate” care. Let’s define that for this article as care that’s avoidable, unnecessary, and/or preventable.

I’m listening to his third book, Blind Spots: The Global Rise of Unnecessary Health Care and What We Can Do About It. He again (as he did in my favorite of his, The Price We Pay), uses c-sections as a key example of where we can improve in appropriate care. Dr. Makary points out that nearly one in three births in the United States is delivered by C-section, and that many of these surgeries occur without clear medical justification.

He argues that these high rates are not random; they are symptomatic of a health system structured around financial incentives and defensive medicine, where procedures are often favored over less invasive options even when the risks may outweigh the benefits.

Here are some ways he highlighted the problem:

1. Data-Driven Analysis: Makary compiled data and statistics to highlight the disparity between medically necessary and elective—or even avoidable—cesarean deliveries. By contrasting how often c-sections occur with the actual clinical indications for surgery, he built a case that many cesarean deliveries are driven more by systemic pressures and physician preferences and incentives than by patient need. This evidence-based approach bolstered the public and professional discussion about the real costs—both in terms of patient safety and economic waste—of over-medicalization in childbirth .

2. Public Advocacy and Communication: As a physician (surgeon) himself, he knew how hard it would be to change entrenched behaviors. Makary embraced public platforms. He appeared on podcasts, participated in interviews, and wrote op-eds to explain how financial incentives and defensive practices influence high c-section rates. In these engagements, he reached a broad audience, including fellow healthcare professionals, hospital administrators, policymakers, and expectant mothers. He wanted a more informed discussion on when cesareans are truly necessary and to push for a culture that prioritizes patient outcomes over procedure volumes and physician convenience.

3. Engaging Multiple Stakeholders: Dr. Makary’s outreach was collaborative. He engaged with a diverse network of reform advocates to call for a systemic reevaluation of how obstetrical care is delivered. He aimed to influence both clinical practice guidelines and hospital policies, thereby encouraging safer, more judicious use of cesarean deliveries. In his second book, The Price We Pay, he goes into depth about work engaging various specialty societies to gain their leadership’s buy-in to address key opportunities in a specialty’s care and what a key lever this is to pull. (I highly recommend this book, if you are looking for a great read or listen!)

4. Challenging the Financial and Cultural Norms: At the heart of his critique is the recognition that financial structures and cultural practices within medicine push toward more surgical intervention. He surfaces the underlying causes, such as fee-for-service models and the pervasive fear of litigation among doctors. By laying out these systemic drivers, he made an urgent call for reform—not merely a reexamination of clinical indications, but a broader overhaul of incentive structures in healthcare.

How Generative AI Enhances Detection

Generative AI significantly improves potentially inappropriate care detection through these capabilities:

Pattern Recognition at Scale: AI can simultaneously analyze thousands of variables across millions of records, identifying subtle patterns that human analysts might miss. For instance, it might detect that a provider's C-section rate spikes specifically for patients with certain insurance types or on particular days of the week, suggesting non-medical decision-making factors.

Natural Language Processing: AI can analyze clinical notes, discharge summaries, and documentation to identify inconsistencies between stated medical necessity and actual patient conditions. It can flag cases where documentation doesn't support the level of intervention provided.

Predictive Modeling: Rather than just identifying problems after they occur, AI can predict which providers are likely to deviate from evidence-based care based on emerging patterns in their practice data.

Multi-dimensional Analysis: AI can correlate clinical decisions with financial incentives, scheduling patterns, facility capacity, and other contextual factors to identify potential conflicts of interest or inappropriate care drivers.

Evolving Potential for Data Access and Decision-Making Frameworks

Policymakers can access this data through state health departments, CMS databases, and all-payer claims databases. They use analytics dashboards to identify system-wide trends and create targeted interventions. For example, if AI identifies geographic clusters of high C-section rates, policymakers might implement regional quality improvement initiatives or adjust reimbursement policies.

Payers can integrate multiple data streams including claims, prior authorizations, and member outcomes data. In value-based care contracts, they establish shared savings arrangements where providers are rewarded for meeting quality benchmarks while controlling costs. AI helps payers identify which providers consistently deliver evidence-based care versus those driven by volume-based incentives. You can imagine how much more important it inherently becomes, in this paradigm, to be able to defend the decisions you make as a provider of any kind.

Providers in VBC contracts receive regular performance reports showing their metrics against benchmarks. Some AI-powered clinical decision support tools can alert them in real-time when proposed treatments deviate from evidence-based guidelines, helping them make better decisions while understanding the financial implications of their choices.

Three Hypothetical Examples of the Potential Negative Impact on Profit-Driven Providers

Example 1: Surgical Intervention Penalties A orthopedic surgeon historically performed knee arthroscopies at rates 40% higher than peers, generating substantial fee-for-service revenue. AI analysis revealed that many patients had conditions better treated with physical therapy. Under new VBC contracts, the surgeon faces financial penalties for unnecessary procedures and must demonstrate medical necessity through enhanced documentation. The surgeon's income drops significantly as they're forced to recommend conservative treatments first, aligning with evidence-based care but reducing procedural volume.

Example 2: Diagnostic Over-Testing Consequences A cardiology practice routinely ordered stress tests and echocardiograms for low-risk patients, generating significant revenue from ancillary services. AI pattern analysis identified that this practice's testing rates were 3x higher than peers treating similar patient populations, with no corresponding improvement in outcomes. Under shared savings arrangements, the practice now faces reduced capitation payments and must justify testing with clinical decision support tools. Their revenue model shifted from volume-based testing to outcomes-based care management.

Example 3: Specialty Referral Restrictions A primary care practice maintained high referral rates to specialist colleagues within their health system, often bypassing conservative management options. AI analysis showed patients were referred for conditions that evidence-based guidelines suggest should be managed in primary care and other conservative practitioners initially. Under new risk-sharing contracts, the practice must demonstrate they've exhausted appropriate primary care interventions before referrals. This reduces the practice's ability to generate referral-based revenue sharing with specialists while requiring them to develop more comprehensive primary care capabilities.

In each hypothetical case, data transparency and AI-enhanced analytics expose the gap between profit-maximizing behavior and evidence-based care, forcing providers to align clinical decisions with patient outcomes rather than revenue optimization. This creates short-term financial pressure but ultimately supports better patient care and system-wide cost control.

Redefining Quality: Beyond Outcomes to Appropriateness

Let’s make this even more relatable.

Most healthcare providers understand quality through the lens of outcomes, not appropriateness. Here are some surgical examples:

Did the surgery go well?

Did the patient recover?

Did we avoid complications?

Did we avoid a readmission in the first 30 days?

In other words, this traditional view of quality focuses on how well we as providers perform the care we've already decided to deliver. It's retrospective, outcome-focused, and assumes the intervention was warranted in the first place.

Makary's concept of "appropriate care" flips this paradigm. Appropriate care asks more fundamental questions such as:

Should this care have been provided at all?

Did this amount of care need to be provided?

It examines whether treatments follow evidence-based best practice guidelines, whether interventions are truly necessary, and whether care is avoidable, unnecessary, or preventable.

The difference is profound.

Traditional quality might celebrate a surgeon with a low complication rate for cesarean sections. Appropriate care would question why that surgeon's c-section rate is 40% when evidence suggests rates above 15% rarely improve outcomes.

Traditional quality applauds the dermatologist who performs flawless MOHS procedures. Appropriate care could include identification of a reasonable average number of “stages” performed per surgery. (In Mohs micrographic surgery, each thin layer that the surgeon removes and examines is typically referred to as a "stage." Essentially, the procedure is performed in multiple stages. During each stage, the surgeon excises a very thin layer of tissue, maps it, and then examines the entire surgical margin under a microscope to check for cancer cells. If any cancerous cells remain at the edges, another stage is carried out, removing one more layer.)

Dr. Makary contributed to this article, with great visuals demonstrating that the high outliers had nearly a 4 stage average.

1.7 vs. 4.

The Perfect Storm: AI, Data Integration, and Administrative Priorities

What makes this shift urgent for providers to understand rather than merely interesting is the convergence of three powerful forces: artificial intelligence, comprehensive data integration, and an Administration with a self-reported focus on using data to achieve better health outcomes at lower costs.

We now have the technological capability to collect, integrate, and analyze vast amounts of healthcare data in real-time. AI can now identify patterns across millions of patient encounters, comparing provider practices not just within specialties but across entire health systems and geographic regions. What once required expensive, time-consuming studies can now be accomplished with algorithmic analysis of existing electronic health records, insurance claims, and patient outcomes data.

These analyses are becoming more sophisticated, more accessible, and more focused on appropriateness rather than just traditional quality metrics. And they are being given a big boost of oxygen by an Administration with CMS goals around data integration and the use of sophisticated analytics to change how healthcare is paid for an delivered.

The Transparency Revolution: When Hiding Becomes Impossible

The implications for healthcare providers are staggering. Consider my own profession of physical therapy. Some practitioners schedule patients for visits far exceeding evidence-based guidelines, often driven by reimbursement structures that reward volume over value. Previously, this practice might go relatively unnoticed in FFS and checked to a degree by managed care.

When therapists want to know when they will really need to change their practice, know that AI can identify patterns across thousands of PT episodes, flagging providers whose treatment durations consistently exceed peers treating similar conditions with similar outcomes.

There are organizations contracting with therapists in episodes of care, where they are rewarded financially for valuable care that also helps avoid unnecessary tests and procedures like MRI’s and surgical interventions.

Here’s the good news for providers whose care may be undervalued in the current system AND who individually provide high-value care to the best of their ability. Sophisticated analytics can show those who hold the funds where care patterns SHOULD be different and why that is better for everyone.

Looking again at PT—data can surface high variability and low value of care that could be avoided with different utilization patterns. For example, if patients diagnosed with low back pain and related conditions go on to have high rates of advanced imaging and procedures without better outcomes compared to expected, or worse outcome and higher spend longitudinally, it’s an opportunity that can be surfaced to those holding the financial risk for care.

A top solution? Physical therapy.

Even better? Physical therapy where patients are accessing care most of the time: at their primary care provider, in urgent care centers, and in emergency rooms. And then track the data to identify the true value of that redirected care. That puts physical therapists in a place of new negotiating power with all stakeholders.

Two Additional Examples

Antibiotic Prescribing: Despite decades of education about antibiotic resistance, inappropriate antibiotic prescribing remains rampant. Studies consistently show that 20-30% of outpatient antibiotic prescriptions are unnecessary, often prescribed for viral infections or conditions that would resolve without treatment and/or that couldn’t benefit from the antibiotic.

AI systems can now analyze prescribing patterns in real-time, identifying providers whose antibiotic use significantly exceeds peers treating similar patient populations. When this data becomes transparent to patients, payers, and health systems, providers who continue inappropriate prescribing practices face not just professional embarrassment but potential financial consequences.

Side note: Dr. Makary has a chapter on microbiome damage from antibiotics in his most recent book, calling out antibiotic overprescribing risks.

Cardiac Interventions: Percutaneous coronary intervention (PCI) with stent placement for stable coronary artery disease is an example of how technological capability, financial incentives, and patient expectations can complicate appropriate care decisions. Several major trials have challenged traditional approaches: the COURAGE trial (2007) showed PCI plus optimal medical therapy was not superior to medical therapy alone for reducing death or heart attacks in stable CAD patients; ORBITA (2017) found PCI didn't significantly improve exercise capacity or symptoms compared to placebo in stable single-vessel disease; and ISCHEMIA (2020) demonstrated that initial invasive strategy didn't reduce major cardiac events compared to conservative management in stable ischemic heart disease.

While these studies are debated within cardiology, they raise important questions about utilization patterns. Data analytics can now identify providers whose intervention rates consistently deviate from utilization patters generally aligned with evidence-based guidelines, adjusted for patient complexity and symptom severity. As comparative data becomes more transparent, the dramatic appeal of "opening blocked arteries" must be weighed against evidence showing when less invasive approaches achieve similar outcomes.

The Double-Edged Sword: Risk and Opportunity

This shift toward appropriate care presents both significant risks and unprecedented opportunities for healthcare providers, depending on how their current practices align with evidence-based guidelines.

The Risk Providers whose practices rely on high-volume, procedure-heavy approaches face existential challenges. When transparent data shows that your patient visit frequency, procedure rates, or intervention complexity consistently exceeds peers without corresponding outcome improvements, justification becomes nearly impossible.

The risk extends beyond individual providers to entire specialties and health systems. Organizations that have built business models around high-volume, questionably appropriate care face fundamental restructuring pressures when payers and patients can easily identify more appropriate alternatives.

The Opportunity Providers who deliver high-value, evidence-based care—even if currently potentially undercompensated—have unprecedented opportunities to demonstrate their worth. Primary care physicians who prevent hospitalizations through excellent chronic disease management, physical therapists who achieve superior outcomes and help prevent downstream, unnecessary higher costs, and specialists who use the least invasive effective approaches and operate on the right patients at the right time can now prove their value with data.

This represents a potential rebalancing of healthcare economics. In this paradigm, reimbursement more closely aligns with actual value rather than procedure volume or complexity. Providers who have long felt undervalued may finally see financial recognition that matches their clinical contributions.

Accountable Care Takes on New Meaning

The concept of accountable care is evolving beyond simple shared savings models to encompass true appropriateness accountability. When AI can identify that one provider consistently achieves similar outcomes with fewer interventions, less expensive approaches, or shorter treatment durations, the definition of "accountable" expands dramatically.

This accountability extends to patient transparency. Expect more user-friendly, intuitive platforms where patients can easily compare providers not just on traditional quality metrics but on appropriateness measures: which cardiologist recommends the fewest unnecessary procedures, which dermatologist achieves excellent cancer outcomes with the least invasive approaches, which physical therapist’s intervention result in better long-term pain resolution at lower longitudinal costs.

A Call to Action: Prepare Now

Every healthcare provider must recognize that this transformation isn't optional—it's inevitable. The convergence of AI capability, data integration, and administrative priorities around value-based care means that practice patterns invisible just years ago are now transparent and comparable.

For providers whose current practices align with evidence-based appropriate care: This represents validation and opportunity. Document your approaches, understand your data, and prepare to demonstrate your value. Your evidence-based practices are about to become competitive advantages.

For providers whose practices may not withstand appropriateness scrutiny: The time for adjustment is now, not later. Review your utilization patterns against peer benchmarks and evidence-based guidelines. Identify areas where volume or financial incentives may have driven practices beyond appropriate levels. Begin transitioning toward more evidence-based approaches before external pressure forces reactive rather than proactive changes.

For all providers: Engage with your data. Understand how your practices compare to peers and evidence-based benchmarks. Prepare for increased transparency and accountability. Most importantly, refocus clinical decision-making around patient value rather than financial optimization.

The era of hiding inappropriate care behind individual clinical judgment, patient complexity, or "that's how we've always done it" is ending. Data-driven appropriate care analysis makes practice patterns transparent, comparable, and ultimately, indefensible when they consistently deviate from evidence-based guidelines.

The question isn't whether this transformation will happen—it's how you are positioned for it. In an era where data speaks louder than tradition, it’s the perfect time to evaluate risks and opportunities in a new way.