The drug program keeping free clinics alive is on the chopping block.

And most people outside of healthcare have never heard of it.

The 340B Drug Pricing Program requires pharmaceutical manufacturers to sell outpatient drugs to eligible safety-net providers at dramatically reduced prices - sometimes 25 to 50 percent below market. For Federally Qualified Health Centers, rural health clinics, and disproportionate share hospitals, 340B is not a perk. It is often the financial engine that keeps the doors open.

Now, a combination of forces is threatening to shutter that engine. The One Big Beautiful Bill Act, signed in July 2025, cuts an estimated $911 billion from Medicaid over the next decade. Separate regulatory changes are moving 340B program oversight from HRSA to CMS. And proposed contract pharmacy restrictions continue moving through the courts. Each threat, taken alone, is survivable. Together, they constitute an existential challenge for the safety-net providers that depend on 340B m…

Read more

Medicaid work rules could hollow out your FQHC patient base. That is not speculation. That is the documented outcome from every state that has tried this before.

The One Big Beautiful Bill Act, signed into law on July 4, 2025, restructures who qualifies for Medicaid entirely. The community engagement requirements demand that most non-disabled adults ages 19 to 64 document at least 80 hours per month of work, job training, education, or community service to keep their Medicaid coverage. States must adopt implementation procedures by December 31, 2026. Enrollment renewals for expansion enrollees shift from annual to every six months starting the same date. The CBO projects the law reduces federal Medicaid spending by $326 billion over ten years. It also projects 5.3 million more Americans will be uninsured by 2027.

Those 5.3 million people are not slackers. They are not gaming the system. Most of them are already working. The research from Arkansas, the only state to have fully tested thi…

Read more

Lung-RADS was never designed for AI. That is the problem.

For more than a decade, radiologists have relied on a 1-to-4 categorical scoring system to determine who needs follow-up after a chest CT. The system works in the sense that it lowered the false positive rate from the 27.6 percent seen in the original National Lung Screening Trial down to roughly 10.6 percent. That was meaningful progress. The American College of Radiology deserves credit for building a standardized framework at a time when the field had none.

But Lung-RADS was designed for human radiologists reading films one by one, applying clinical judgment to place a finding into a box. It was never designed to interface with AI systems that produce probabilistic outputs - continuous numerical scores representing degrees of malignancy risk across thousands of scans simultaneously. The categorical architecture that made Lung-RADS practical in 2014 is exactly what makes it a bottleneck in 2026.

The result is a structural mismatch between how AI sees lung findings and how our clinical workflows are built to receive them. And patients - particularly patients in underserved communities served by Federally Qualified Health Centers - are paying the price.

1. The Architecture of an Outdated System

Lung-RADS works by placing every lung finding into one of four categories. Category 1 or 2 means benign or probably benign - no immediate action needed. Category 3 means a short-interval follow-up CT in three to six months. Category 4 is suspicious - additional workup warranted.

The problem is the category borders. Any scoring system that draws hard lines around continuous biological phenomena will create a gray zone at the margins. In lung nodule assessment, that gray zone is enormous.

The original NLST data illustrates the stakes. Of 649 confirmed true-positive lung cancers identified during the trial, 86 - roughly 13 percent - were missed entirely under Lung-RADS criteria. These were real cancers that a categorical system classified as not requiring urgent action. At the other end, 1.7 percent of all screened participants underwent an invasive procedure for a lesion that turned out to be benign. Each unnecessary biopsy carries a real complication profile: pneumothorax in 15 to 20 percent of cases, hemorrhage, infection, and months of patient anxiety.

Lung-RADS version 1.1 improved specificity further, and newer versions have refined the criteria for airway nodules. But the fundamental architecture has not changed. Categories still have borders. Borders still create gray zones. Gray zones still produce both false positives and missed cancers.

What changes when you replace a category with a probability is not just the number - it is the clinical action that follows.

Lung-RADS reduced the NLST baseline false positive rate from 27.6% to 10.6%, but still missed 13% of confirmed lung cancers in the trial.

2. The False Positive Crisis: Numbers That Have Real Bodies

Let us talk about what false positives actually cost - not in dollars, but in bodies.

The 18 to 40 percent false positive rate range cited in the literature is a statistical range. In the context of lung cancer screening, it means that a meaningful fraction of every patient cohort that gets flagged for follow-up is being told they might have cancer when they do not. For an affluent, insured patient with a supportive care team, that experience is frightening but manageable. For a patient at an FQHC - uninsured or on Medicaid, working two jobs, without reliable transportation, navigating a clinic where their primary care physician is managing 2,000 active patients - a false positive is not just anxiety. It is lost wages, missed work shifts, potential immigration fears, and a follow-up appointment that may never happen.

The data on follow-up failure is stark. Approximately 52 percent of patients with solitary pulmonary nodules do not receive appropriate follow-up care, according to a 2025 cohort study. In a separate analysis, only 56.8 percent of patients with incidental pulmonary nodules received guideline-concordant follow-up at a tertiary care center.

And the disparity by race is not subtle. Among patients with incidental pulmonary nodules requiring follow-up, only 49.1 percent of Hispanic patients and 55.1 percent of Black patients were formally notified of their finding - compared to 79.5 percent of White patients. Non-White patients had significantly lower rates of both ordering and adherence to follow-up imaging across all studies.

This is not a clinical problem. It is an infrastructure problem compounded by a categorization problem. When a system uses categories instead of probabilities, the clinical signal that should trigger navigation action is weak, ambiguous, and easily lost in a noisy primary care workflow.

3. AI Probability Scoring: What Changes When You Replace Categories with Certainty

The scientific argument for continuous probability scoring over categorical classification is well established. The clinical argument is simpler: numbers travel better than labels.

A Lung-RADS category 3 finding means "short-term follow-up recommended." It is a label with a recommended action attached. A malignancy probability score of 73 percent means something more specific: there is a high likelihood that this nodule is malignant, and every hour of delay in the follow-up pathway is a quantifiable risk. One format triggers a reflexive workflow. The other triggers a conversation.

Recent AI systems trained on large screening datasets have demonstrated the performance gains that are possible when categorical logic is replaced with continuous scoring. A study published in JMIR Formative Research in 2026 found that real-world AI tools demonstrated sensitivity rates of 56.4 to 95.7 percent and specificities of 71.9 to 97.5 percent - with AUROC values between 0.89 and 0.99, compared to a mean radiologist AUC of 0.81.

Oatmeal Health's LungAI system outputs a continuous 0-to-100 malignancy probability score for every chest CT. At equivalent sensitivity to Lung-RADS, it reduces false positives by one-third - specificity improves from 82 percent to 88 percent. At equivalent specificity, it catches 3 percent more cancers that Lung-RADS would miss. (Pre-FDA clearance; not yet commercially available.) The score travels with the patient into the EHR, into the care navigation workflow, and into the shared decision-making conversation.

AI probability scoring can reduce false positives by up to one-third while simultaneously catching 3 percent more cancers compared to Lung-RADS categorical scoring.

The Vanderbilt University Medical Center study published in March 2026 put an economic value on this improvement. Researchers modeling the lifetime payer cost of AI-assisted lung nodule risk stratification found an incremental cost-effectiveness ratio of $4,485 per life year gained. In health economics, any ICER below $50,000 per life year is considered highly cost-effective. At $4,485, AI-based nodule risk stratification is not just effective. It is one of the most cost-effective diagnostic interventions in modern oncology.

4. The FQHC Dimension: Why the Stakes Are Highest Where Infrastructure Is Thinnest

The gap between clinical evidence and clinical practice is always widest at the edges of the healthcare system. For lung cancer screening and nodule management, those edges are FQHCs and community health centers.

Only 5 percent of the estimated 14.5 million screening-eligible Americans who receive care at FQHCs actually receive a preventive lung CT scan. The national screening adherence rate is below 16 percent. At FQHCs, it is a fraction of that. A 2026 study documenting improvement efforts at a resident-run FQHC found that baseline monthly lung CT completion was 1.2 scans per month. After targeted outreach and EHR care gap implementation, that number improved to 12.3 scans per month - still a small fraction of what the eligible population requires.

The barriers are well documented: limited IT infrastructure, thin staffing, no dedicated nodule tracking programs, and a patient population facing transportation, language, work schedule, and immigration-related barriers to follow-up. In this environment, every ambiguous clinical signal is an opportunity for a finding to get lost.

A continuous AI probability score addresses several of these barriers simultaneously. A score of 78 percent routes differently in a care navigation algorithm than a Lung-RADS 3 label. High-probability findings can be automatically escalated to a patient navigator. Lower-probability findings can be appropriately scheduled without consuming the scarce bandwidth of a pulmonologist or specialist. The workflow becomes intelligent because the input is precise.

Only 5% of the 14.5 million screening-eligible FQHC patients receive a lung CT. A 2026 FQHC study saw monthly completions jump from 1.2 to 12.3 scans per month after systematic outreach - still far below population need.

5. The Business Case: CPT 0721T and the Revenue Model for Adopters

The clinical case for AI probability scoring is strong. The financial case - which is what actually drives health system adoption decisions - is increasingly compelling as the reimbursement infrastructure matures.

CPT code 0721T (Quantitative CT tissue characterization, with interpretation and report) carries a Medicare reimbursement rate of $600 to $700 per scan. This is a Category III temporary code currently available for AI-based lung CT characterization tools that have received appropriate FDA clearance. For health systems and FQHCs that are already performing lung cancer screening CTs, CPT 0721T represents incremental revenue for the AI analysis layer applied to scans they are already reading.

The math at scale matters. An FQHC performing 20 lung screening CTs per month generates roughly $13,000 per month in incremental CPT 0721T revenue at $650 per scan. A community health system running a lung cancer screening program across four sites at 100 CTs per month generates $65,000 in monthly AI-augmented revenue. This is not theoretical - it is the current reimbursement landscape for FDA-cleared AI tools applied to lung CT.

The cost-effectiveness study from Vanderbilt reinforces this from a payer perspective. An ICER of $4,485 per life year gained means that payers - Medicaid MCOs, Medicare Advantage plans, and self-insured employers - have a clear economic rationale to cover and encourage AI-assisted nodule risk stratification. As this evidence base matures, the path from Category III temporary code to permanent Category I CPT billing becomes more plausible.

6. The Workflow Transformation: From Categorical Flag to Actionable Signal

The deepest change that AI probability scoring brings to lung nodule management is not about the score itself. It is about what the score enables downstream.

Categorical systems produce categorical responses. A Lung-RADS 3 finding triggers a recommendation: follow-up CT in three to six months. That recommendation lands in a radiology report, gets transmitted to a PCP, and from there enters the administrative processing queue of a practice that may or may not have dedicated tracking infrastructure. In FQHCs, most do not. The tracking gap is where lung cancers go from Stage I to Stage IV.

A continuous probability score is a structured data point that an EHR can act on automatically. A probability score above a defined threshold - say, 60 percent - triggers a patient navigator assignment. A score above 85 percent triggers an expedited pulmonology referral. A score below 15 percent triggers an extended follow-up interval rather than a reflexive three-month rescan. The algorithm handles the triage logic. The humans handle the conversations that require clinical judgment.

This is not theoretical. Research published on medrxiv documents exactly this dynamic in FQHC settings: AI-powered clinical decision support tools closed the lung cancer screening gap by automatically identifying unscreened high-risk patients, navigating them to care, and generating structured risk outputs that integrated directly with care coordination workflows.

Deep Dive: The Three-Scenario Model for FQHC Lung Nodule AI Adoption

For health system executives considering AI probability scoring, the decision is operational and financial as much as clinical. Here is a structured scenario model for a mid-size FQHC running a lung cancer screening program at 30 CTs per month.

Scenario A: No AI Integration (Status Quo)

Lung-RADS categorical scoring by radiologist. False positive rate: approximately 12 percent. Follow-up completion rate: approximately 55 percent. Missed high-risk nodules per year: estimated 3 to 4 based on the 13 percent miss rate applied to population prevalence. CPT 0721T revenue: $0. Malpractice exposure from missed findings: accumulating as AI becomes the evidence-based standard of care.

Scenario B: AI Probability Scoring with Manual Triage

AI score accompanies every radiology read. Radiologist reviews score and applies clinical judgment to follow-up recommendation. False positive reduction: approximately 30 percent (from ~12% to ~8%). CPT 0721T revenue: $19,500 per month ($234,000 per year at $650 per scan). Cancer detection improvement: 3 percent additional vs. Lung-RADS. Follow-up improvement: modest, as routing remains manual.

Scenario C: Full AI Integration with Automated Care Navigation

AI score routes directly to care navigation logic in EHR. Scores above 70 percent trigger automatic patient navigator outreach. Scores above 85 percent trigger same-week pulmonology referral. Follow-up completion target: 80 percent (up from 55%). CPT 0721T revenue: $19,500 per month ($234,000 per year). Projected reduction in Stage IV diagnoses at 30 scans per month: meaningful - every 1 percent improvement in early detection shifts approximately 2 to 3 patients per year from Stage IV to Stage I or II.

What This Means For You

The shift from categorical Lung-RADS scoring to AI probability-based triage is not a distant future event. The evidence is here. The reimbursement pathway exists. The tools are in active deployment. What differs across organizations is the speed of adoption - and the patients who fall through the gap in the meantime.

• For FQHC executives and community health center leaders: The follow-up failure crisis is not a staffing problem you can solve with more resources. It is a signal clarity problem. A continuous probability score gives your navigators a precise, actionable number instead of a clinical label that gets lost in a busy PCP workflow. Evaluate AI lung CT tools now - CPT 0721T revenue can offset the integration cost.

• For health system administrators and CMOs: As AI probability scoring becomes the evidence-based standard, systems running Lung-RADS-only workflows face growing malpractice exposure for missed nodules that an AI system would have caught. The standard of care is changing. Align your radiology protocols before the liability gap widens.

• For radiologists and pulmonologists: The transition to probability scoring does not eliminate clinical judgment - it elevates it. The AI handles the triage logic. You handle the complex cases, the anomalies, and the patient conversations that require human expertise. Systems that integrate AI probability scores report faster read times and reduced burnout on routine low-probability findings.

• For healthcare investors and founders: The lung nodule AI market is not crowded at the FQHC end. Most commercial deployment has targeted large academic medical centers and high-volume radiology groups. The 14.5 million screening-eligible FQHC patients represent a vast underserved market with a clear reimbursement pathway and a compelling clinical need.

• For policy advocates: The disparity data is unambiguous. At 49.1 percent notification for Hispanic patients and 55.1 percent for Black patients versus 79.5 percent for White patients, the nodule follow-up gap is a civil rights issue dressed in clinical language. AI-powered navigation is one of the few interventions that can close this gap at scale without requiring proportional increases in clinical staffing.

Key References

• American Cancer Society Cancer Statistics 2026: 229,410 estimated new lung cancer cases, 124,990 estimated deaths in the US.

• NLST and Lung-RADS Performance Analysis (PubMed, multiple studies): Lung-RADS reduced baseline false positive rate from 27.6% to 10.6%, but 13% of confirmed cancers were missed under categorical criteria.

• Rate of Incidental Lung Nodule Follow-Up (PMC, 2025): Only 56.8% of patients with incidental pulmonary nodules received guideline-concordant follow-up; disparity by race documented (49.1% Hispanic, 55.1% Black, 79.5% White notification rates).

• Vanderbilt University Medical Center / Optellum Cost-Effectiveness Study (PLOS ONE, March 2026): AI-assisted lung nodule risk stratification achieved an ICER of $4,485 per life year gained.

• AI Lung Nodule Detection in Clinical Practice (JMIR Formative Research, 2026): FDA-cleared AI tools demonstrated real-world sensitivity 56.4-95.7% and specificity 71.9-97.5%, AUROC 0.89-0.99 vs. radiologist AUC 0.81.

The gap between when a lung nodule is identified and when a patient receives definitive care is not primarily a radiology problem. It is a signal design problem. Categories were the best signal we had in 2014. Probability scores are the best signal we have in 2026. The question is not whether to make the transition - the evidence demands it. The question is whether health systems and policymakers will move fast enough to protect the patients who cannot afford to wait.

The lung cancer death rate is falling. It will fall faster with better signals.

What would change in your nodule management workflow if every chest CT came back with a probability score instead of a category?

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts. Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise. Subscribe at jonathangovette.substack.com

Every week, physicians and their staff spend an average of 14 hours per provider navigating prior authorization requests. That is more than two full working days lost to paperwork. Not seeing patients. Not making diagnoses. Filling out forms.

Now the rules have changed - and the data is coming in.

In January 2026, the CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F) went fully into effect. It requires covered health plans to respond to urgent prior authorization requests within 72 hours, standard requests within 7 calendar days, and to publicly report their turnaround times, denial rates, and appeal rates beginning March 31, 2026. It also mandates standardized FHIR APIs that allow providers to submit and track prior auth requests electronically.

The rule forced payers to build infrastructure they had resisted for years. AI companies immediately built on top of it.

Health systems that have deployed AI-powered prior authorization tools on those new payer APIs are reporting 60 to 70 percent reductions in average decision time. What used to take 3 to 5 business days is resolving in 4 to 8 hours. And in some cases, in under one hour.

Here is why this matters far beyond the headline.

Prior authorization delays are not just an administrative annoyance. They cause real, documented clinical harm. And they fall hardest on the patients who can least afford it.

1. The Numbers That Demand Attention

The 2026 AMA Prior Authorization Physician Survey makes for stark reading.

95 percent of physicians say prior authorization delays access to necessary care. 92 percent say it negatively affects clinical outcomes. 79 percent report that patients abandon treatment entirely because of authorization challenges. And 26 percent of physicians - more than one in four - report that prior authorization has led to a serious adverse event for a patient, including hospitalization, permanent impairment, or death.

These are not edge cases. These are systemic.

The burden does not fall evenly. Specialists and behavioral health providers spend an average of 24 minutes processing a single manual prior authorization - three times longer than the 8 minutes it takes a generalist, according to the 2025 CAQH Index. That differential reflects the clinical complexity of specialist requests and the fact that many insurers deliberately design specialty PA pathways to be opaque and burdensome.

Across all provider types, physicians are completing an average of 40 prior authorizations per week. For a physician working a standard 40-hour week, that alone accounts for 9 to 10 hours of administrative time, before accounting for staff support, appeals, peer-to-peer reviews, and resubmissions.

The financial damage is compounding. Manual prior authorization transactions cost providers an average of $10.97 per transaction. A fully electronic prior authorization costs $5.79 - nearly half. Across the volume of PA transactions processed annually in the US, CAQH estimates that universal electronic processing would save $494 million per year. That is money currently absorbed by provider organizations in staff time, fax infrastructure, and delayed revenue.

📊 The 2025 CAQH Index found US healthcare avoided $258 billion in administrative costs in 2024 through electronic transactions. But prior authorization remains one of the highest-friction processes in the system.

2. What the CMS Rule Actually Changed

CMS-0057-F is not just a speed mandate. It is an infrastructure mandate.

Before the rule, payers had no legal obligation to respond to prior authorization requests within any defined window. A health plan could sit on a request for 10 business days or longer. They faced no public accountability for their denial rates or how often appeals overturned their initial decisions.

The rule changed three things simultaneously. First, it set response time floors that are legally enforceable - 72 hours for urgent requests, 7 calendar days for standard requests. Second, it mandated public reporting of PA performance data starting March 31, 2026. For the first time, a health system administrator can look at the public data and know which payers are routinely overturning their own denials on appeal - a clear signal that the initial denial was unjustified. Third, it required standardized FHIR APIs for PA submission and tracking. This is the technical infrastructure that AI companies immediately began building on.

The compliance picture remains mixed. Only 12 percent of clinicians and 7 percent of office administrators report that they consistently receive PA decisions within the new timeframes, according to Cohere Health's 2025 National Provider Survey. That gap between mandate and reality is where the AI opportunity lives.

3. What AI Is Actually Doing

The AI tools entering the prior authorization space in 2026 are not digitizing the old paper process. They are replacing the logic of the process itself.

They predict approval likelihood before submission. By analyzing CPT code, diagnosis code, clinical documentation, and payer-specific criteria, AI models can estimate with high confidence whether a request will be approved or denied - before the physician has even signed the order.

They auto-populate clinical documentation from the EHR. Instead of requiring a staff member to read through a chart and manually copy relevant information into a payer portal, AI pulls the relevant elements and structures them to match payer-specific criteria. This alone eliminates the most time-consuming part of the PA workflow.

They flag missing information before submission. Most denials are not because the treatment is clinically inappropriate. They are because the request arrived missing a piece of documentation the payer requires. AI catches those gaps at submission time, not after a 7-day wait ends in a denial.

They route urgent cases to expedited review pathways automatically. A patient with active cancer waiting for imaging authorization is not the same as a patient scheduling a routine MRI three months out. AI systems now recognize clinical urgency signals and trigger the correct escalation pathway without staff intervention.

The early performance data is striking. R1 RCM reports that 68 percent of all authorization orders are cleared in under one hour and nearly 97 percent within one day, with an auth-related denial rate below 1 percent. Optum's AI-powered Digital Auth Complete system projects 45 percent fewer manual touches, 80 percent greater document bundling efficiency, and a 96 percent first-pass approval rate. Broader industry deployments report 83 percent reductions in handling time.

📊 A Surescripts analysis found that electronic prior authorization cut the median time from request to decision by 13 hours - a shift that, for patients awaiting cancer imaging or urgent referrals, can determine whether a diagnosis happens this week or next month.

4. The Radiology Problem Is Acute

For radiologists, pulmonologists, and imaging centers, prior authorization is not a background administrative issue. It is a frontline clinical bottleneck.

Advanced imaging services - CT, MRI, and PET scans - face some of the highest prior authorization rates of any service category in outpatient medicine. And when imaging is delayed, diagnoses are delayed. In cancer care, that delay is not measured in inconvenience. It is measured in stage at diagnosis.

Up to 28 percent of radiology claims are initially denied, with missing or delayed prior authorizations cited as the top reason, according to 2025 industry data. The average denied imaging claim costs a provider $118 in rework - before accounting for the lost revenue from an unfilled scanner slot or the clinical risk of a delayed result.

For Oatmeal Health's work in FQHCs, this plays out daily. Patients who are high-risk for lung cancer - smokers, former smokers, patients with occupational exposures - are being identified by AI and navigated toward low-dose CT screening. When those referrals hit a prior authorization wall, patients who already face systemic barriers to care disappear. They do not reschedule. They do not call back. They are lost to follow-up.

The CPT 0721T code for AI-assisted lung CT malignancy risk scoring adds a layer of clinical documentation that directly supports prior authorization requests by demonstrating that a screening decision was algorithmically supported. But even with that clinical evidence package, payers do not always move fast.

5. The Payer Side of the Equation

In June 2025, roughly 60 health insurers pledged to streamline and reduce prior authorization requirements in response to pressure from the Trump administration and bipartisan congressional interest in PA reform legislation.

The AMA's 2026 survey measured how that pledge landed: Only 33 percent of physicians believe the insurer pledge will make a meaningful difference in actual care delivery.

That skepticism is warranted. Health insurers have a financial incentive to delay and deny high-cost procedures. Prior authorization is one of their primary tools for managing that exposure. Public pledges to streamline it have a long history of being quietly walked back once political pressure subsides.

What is different in 2026 is the regulatory backstop. The CMS rule creates public reporting obligations that make denial behavior visible. A payer that publicly pledges to reduce prior authorization delays while its public denial rate data shows 40 percent of standard imaging requests taking longer than 7 days cannot sustain that contradiction indefinitely.

The real risk in 2026 is a two-tier system. Large health systems with IT budgets and revenue cycle teams can deploy AI-powered PA tools and capture the efficiency gains. FQHCs, independent physician practices, and rural critical access hospitals - which face the same prior authorization volume but lack the infrastructure to automate it - fall further behind.

📊 Across individual and group markets, roughly 3 out of every 4 health plans now report using AI for prior authorization approvals. The technology exists. The access gap is on the provider side, not the payer side.

Deep Dive: The Financial Model

The case for investing in AI-powered prior authorization has never been clearer.

A health system processing 10,000 prior authorization requests per month at the manual cost of $10.97 per transaction is spending approximately $110,000 per month on PA administration alone. At the electronic rate of $5.79 per transaction, the same volume costs $57,900 - a direct savings of $52,000 per month, or $624,000 per year, purely on transaction cost reduction.

Layer in denial prevention. If 28 percent of radiology requests are initially denied and each denial costs $118 in rework, a practice processing 500 imaging PAs per month is absorbing roughly $16,520 per month in avoidable rework costs. AI systems that reduce that denial rate to below 5 percent eliminate more than $13,000 per month in overhead.

For imaging centers, an empty CT scanner slot that went unfilled because a prior authorization did not arrive in time costs an average of $400 to $1,200 per scan in lost revenue. A system that reduces average PA decision time from 3 days to 4 hours keeps more slots filled.

Why Medicaid Is the Hardest Problem

The CMS-0057-F rule applies to Medicaid managed care plans and Medicare Advantage. It does not apply to fee-for-service Medicaid, which remains governed by state-specific regulations. This creates a compliance patchwork that AI platforms must navigate state by state.

For FQHCs, whose patient populations are disproportionately enrolled in Medicaid managed care, the practical implication is that even full compliance with the federal FHIR API mandate may not materially improve PA workflows if the relevant state Medicaid program is still running a fax system for fee-for-service members.

The Equity Gap No One Is Talking About

Medicaid patients - who are disproportionately lower-income, Black, Hispanic, and from rural communities - face higher prior authorization denial rates than commercially insured patients. A 2024 Kaiser Family Foundation analysis found that Medicaid managed care plans denied prior authorization requests at rates 2 to 3 times higher than Medicare Advantage plans for equivalent clinical requests.

When a commercially insured patient's imaging prior authorization is denied, they often have the resources and navigational capital to file an appeal or pay out of pocket. When a Medicaid patient's imaging prior authorization is denied at an FQHC, the most common outcome is that the patient does not get the scan.

AI-powered prior authorization tools that operate at the FQHC level - integrated directly into EHR workflows, trained on Medicaid payer criteria, and capable of auto-assembling appeals packages - have the potential to close that equity gap. But the FQHCs that need them most are the least likely to have the IT infrastructure to deploy them.

What This Means For You

• FQHC executives and community health center leaders: Identify your highest-volume PA service categories - imaging, behavioral health referrals, specialty prescriptions - and map them to the payers creating the most friction. Use the new CMS public reporting data to document which payers are out of compliance. That documentation supports both advocacy and contract negotiation.

• Health system administrators and CMOs: The financial case for AI-powered PA is now computable. If your system is processing more than 2,000 prior authorizations per month on mostly manual workflows, you are leaving six-figure annual savings on the table.

• Radiologists and pulmonologists: The combination of CPT 0721T AI risk scoring documentation with AI-powered PA submission creates a clinical-administrative package that is genuinely hard to deny. Take your PA denial and scan delay correlation data to your revenue cycle team.

• Healthcare investors and founders: The CMS FHIR API mandate created a new addressable infrastructure layer. The winning companies will have the deepest EHR integrations, the highest payer coverage, and the most training data for approval prediction models.

• Policy advocates and HRSA program officers: A health center that deploys AI-powered prior authorization is providing measurably better clinical access to its Medicaid population. Connect the funding mechanism to the technology deployment.

Key References

• CMS Interoperability and Prior Authorization Final Rule CMS-0057-F

• AMA 2026 Prior Authorization Physician Survey

• 2025 CAQH Index

• Naviant AI PA Processing Analysis

• Optum Digital Auth Complete

Prior authorization has been a known problem for decades. What is different in 2026 is the regulatory forcing function. CMS mandated the infrastructure. AI companies built the workflows. The data is now showing what is possible when both exist simultaneously.

The organizations that move now will capture compound advantages: lower administrative costs, higher first-pass approval rates, faster scan turnaround, and better clinical outcomes for the patients who most need them.

The organizations that wait will keep losing 14 hours per provider per week to a process that has already been solved.

If prior authorization delays have ever put your patients at real clinical risk, reply to this email or reach me at Jonathan@oatmealhealth.com.

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts.

Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise.

Subscribe at jonathangovette.substack.com

Health AI just had its biggest funding quarter of the decade.

Seven point four billion dollars poured into healthcare AI startups in Q1 2026 alone. That is up from $5.9 billion the prior quarter. Nineteen separate mega-rounds of at least $100 million closed. Eight new unicorns emerged in a single quarter - the most in nearly four years.

And the money is not going where most people think.

If you picture Silicon Valley generalists funding flashy chatbots and diagnostic apps that will take years to reach a patient, you are looking at the wrong era. The Spring 2026 funding surge is different in character. Investors are writing nine-figure checks to companies already embedded inside hospital workflows, already reducing denials, already cutting documentation time, already partnering with the EHR systems that control access to every provider in the country.

This is not speculative capital. This is deployment capital.

And for health system executives, FQHC operators, radiologists, and healthcare investors, understanding which bets are winning - and which problems still have no solution - is the single most important intelligence task of the year.

1. The Funding Numbers That Actually Matter

The headline figure is $7.4 billion in Q1 2026, but the structure of that number matters more than the total.

Sixty percent of all Q1 capital came from those 19 mega-rounds. That means a small number of category leaders are capturing most of the oxygen. Smaller companies are raising, but the power law is brutal - the top cohort is pulling away at a pace that smaller AI vendors will not close in a single fundraise cycle.

Eight new healthcare AI unicorns in a single quarter is historically notable. To put it in context, the prior four years averaged fewer than two new healthcare AI unicorns per quarter. The acceleration reflects two forces converging at once: clinical validation data is finally maturing enough to justify large valuations, and the enterprise sales cycle for health AI has compressed dramatically now that hospital IT teams have experience integrating AI vendors.

The M&A market is moving alongside venture capital. Acquisitions in healthcare AI rose 47% in Q1 2026 to 56 deals. The most telling was DeepHealth's $269 million purchase of Gleamer - a transaction that the buyers attributed explicitly to Gleamer's network of more than 700 hospital contracts. Buyers are paying for distribution now, not just technology.

📊 Healthcare is now absorbing nearly half of all vertical AI spending globally - and enterprise AI outlays in healthcare are projected to triple to $1.5 trillion over the next five years.

This is the backdrop against which every deal in this funding surge must be understood. Healthcare is not just one of many sectors adopting AI. It is becoming the defining commercial battleground for enterprise AI deployment.

2. Abridge: The Ambient Documentation Benchmark

No single company captures the current moment better than Abridge.

The ambient AI clinical documentation company has now raised over $850 million across two consecutive rounds. First came the $250 million Series D in February 2025, co-led by Elad Gil and IVP with participation from NVIDIA Ventures, Bessemer, and a roster of strategics including CVS Health Ventures. Then, just four months later, the $300 million Series E, led by Andreessen Horowitz and joined by Khosla Ventures, pushed the company's valuation to $5.3 billion.

At that pace - two mega-rounds in four months - investors are not waiting for additional proof. They are front-running a category they believe is winner-take-most.

The deployment numbers back the bet. When Abridge announced its Series D, it had exceeded 100 health system deployments. By the Series E just four months later, that count had grown to more than 150 - a 50% increase in enterprise deployments in a single quarter. The platform now operates across 55 specialties and 28 languages, with EHR integration that covers both outpatient and inpatient settings.

The clinical impact data is real and getting harder to dismiss. After 30 days using ambient AI scribes, ambulatory burnout dropped from 51.9% to 38.8%. More than 84% of physicians reported a positive effect on patient communication. Abridge's own client data shows 60 to 70% reductions in clinician burnout at deployed health systems. Ninety percent of clinicians who start using Abridge continue using it daily.

The documentation burden problem is colossal. Physicians spend between two and three hours on EHR documentation for every eight hours of patient care. That is not a minor friction point - it is the number one driver of physician burnout, career attrition, and the kind of distracted medicine that increases diagnostic errors.

When ambient AI removes that burden reliably, across 55 specialties and in 28 languages, at 150+ health systems, investors stop asking whether the market is real. They start asking how much of it one company can capture.

📊 AI Clinical Documentation attracted $1.23 billion in total investment in the funding cohort analyzed, with Abridge alone accounting for $850 million across two rounds.

3. Revenue Cycle AI: Where the Money Follows the Denials

If ambient documentation is where healthcare AI is restoring the humanity of medicine, revenue cycle AI is where it is fighting a $43 billion financial hemorrhage.

The American Hospital Association estimates that hospitals now spend $43 billion annually just trying to collect payments that insurers legally owe for care already delivered. That figure represents an industry-wide cash flow crisis that has no manual solution at the volumes modern health systems operate.

The denial rate data tells the story in one number: initial claim denial rates hit 11.65% in 2025 - meaning more than one in every nine claims submitted is rejected on first pass. Medicare Advantage plans saw a 4.8% spike in denial rates from 2023 to 2024. The average coding-related claim denial dollar amount rose 126% in a single year.

This is the problem Waystar is solving with its AltitudeAI platform - and why the company is one of healthcare's most closely watched revenue cycle players post-IPO.

Waystar's results with deployed clients are striking: $15.5 billion in claims prevented from being denied, 95% time savings on appeal package generation, and approximately 99% clean claim and first-pass acceptance rates. In March 2026, the company announced an expanded collaboration with Google Cloud specifically to accelerate agentic AI capabilities - building toward what the company calls the "autonomous revenue cycle."

What Waystar is describing is not incremental improvement to revenue cycle management. It is the elimination of the manual prior authorization and denial management workflow entirely, replaced by AI systems that learn from downstream payment outcomes and continuously improve upstream claim submission quality.

📊 Hospitals spent $18 billion just in 2025 overturning claim denials - costs that AI-powered revenue cycle platforms are targeting directly.

The implications for safety-net providers are particularly urgent. FQHCs and rural health centers operate on margins so thin that a 2 to 3 percentage point improvement in clean claim rates can be the difference between viable operations and closure. Revenue cycle AI is not a luxury tool for large health systems - it is increasingly an existential operational requirement for organizations that cannot afford to leave 15% of submitted revenue on the table.

4. Patient Communication AI: The Safety-Net Opportunity

The third major funding category in this surge addresses a problem that does not show up on any revenue cycle dashboard: patients who never follow up, never show up, never complete the care episode that was supposed to prevent the expensive downstream crisis.

Hippocratic AI raised $141 million in a Series B in January 2025 at a $1.64 billion valuation, bringing total raised to over $400 million. The company builds AI agents for patient-facing communication - chronic care management, post-discharge follow-up, care navigation, appointment reminders, and pre-visit preparation.

The company has inked partnerships with more than 50 large health systems, payers, and pharma clients, and built over 1,000 distinct use cases for patient-facing AI agents.

What distinguishes Hippocratic's positioning is its explicit focus on safety-net settings. FQHCs face staff-to-patient ratios that make human-only care navigation mathematically impossible. A community health center with 15,000 attributed patients and a navigation team of three people cannot proactively touch every patient who missed a screening, left a script unfilled, or skipped a follow-up.

AI agents that operate at unlimited scale, 24 hours a day, in multiple languages, at a cost structure that fits FQHC grant funding - that is not a nice feature. That is infrastructure.

5. The Adoption Gap That Investors Are Not Talking About

Every investor deck in health AI 2026 leads with the same chart: the S-curve of adoption, clinical validation data, total addressable market in the hundreds of billions.

What the decks leave out is the single statistic that should sober every buyer and builder in the space: only 3% of healthcare executives have deployed agentic AI in live clinical workflows.

Forty-three percent are piloting or testing. Three percent have actually deployed.

That gap is not a technology problem. The tools work. The clinical data is real. The ROI cases are validated. The deployment gap is a change management, data infrastructure, and governance problem that no amount of venture capital directly solves.

Less than one-fifth of enterprise healthcare data is ready for AI integration without major cleanup. That means the average health system trying to implement one of the funded tools in this report will spend more time preparing its own data than it will spend integrating the vendor's product.

Health systems that have built internal AI evaluation frameworks - with clinical champions, compliance pathways, IT integration playbooks, and outcome measurement protocols - are the real competitive advantage story of 2026. They can absorb new tools in 30 to 60 days rather than the 12 to 18 month integration cycles that characterized health IT deployments a decade ago.

📊 43% of healthcare executives are piloting agentic AI. Only 3% have deployed it in live workflows. The gap between testing and doing is where competitive advantages will be won or lost.

6. Deep Dive: The Economics of Health AI Deployment at Scale

Understanding the financial architecture behind these funding rounds illuminates why the market is consolidating the way it is.

The Abridge Financial Model

Abridge prices on a per-clinician subscription basis, typically in the range of $1,200 to $2,000 per clinician per year for enterprise deployments. At 150 health systems with an average of 500 deployed clinicians each, the ARR implied by current deployment is between $90 million and $150 million - with significant expansion runway as deployments scale within each health system.

The Series E $5.3 billion valuation at estimated ARR implies a 35x to 60x revenue multiple - aggressive even by software standards, but defensible if ambient documentation becomes mandatory clinical infrastructure within three years.

The investment thesis is simple: once a health system deploys ambient AI at scale, switches costs are enormous. Clinicians build workflows around it. EHR integrations deepen. Retraining 3,000 physicians to use a different product takes 18 months minimum. The company that wins the initial enterprise sale and demonstrates outcomes owns that account for a decade.

The Revenue Cycle Economics

Waystar's commercial model is even more compelling for investors. Revenue cycle AI charges a percentage of recovered revenue - meaning the business scales directly with outcomes. If Waystar's tools prevent $15.5 billion in denied claims for its client base, and charges 2 to 4 percent of recovered value, the implied revenue is $300 million to $600 million.

The self-learning model creates compounding advantage. Every denial outcome - paid, appealed, adjusted - becomes training data that improves the next prior authorization submission. A health system that has been on Waystar for three years has a system that has learned from millions of its specific payer mix and claim patterns. A new competitor cannot replicate that institutional learning quickly.

The FQHC Revenue Model for Patient Communication AI

For safety-net settings, the economics of patient communication AI are built around value-based care contracts and per-member-per-month capitation. Hippocratic AI's agents for chronic care management generate CPT-billable encounters when they document qualifying patient interactions, creating a reimbursement stream that partially offsets subscription costs.

For FQHCs operating under Section 330 grant funding, the operational calculus is different: patient communication AI reduces no-show rates, increases preventive screening completion, and improves quality metrics that directly affect HRSA performance scores and PPS cost-per-visit reimbursement. Even a 5 percentage point reduction in no-show rates at a 15,000-patient FQHC, at an average revenue of $250 per visit, represents $375,000 in annual recovered value.

What This Means For You

For FQHC executives and community health center leaders: The tools that your resource-constrained organization needs most - patient navigation AI, prior authorization automation, ambient documentation - are now commercially mature and increasingly being designed with safety-net economics in mind. The question is not whether to evaluate these tools but how to build the internal capacity to do so without pulling staff off direct care. Prioritize one problem that causes the most operational pain, find a vendor with a published FQHC use case and outcomes data, and run a 90-day pilot with a predefined measurement framework before committing to enterprise terms.

For health system administrators and CMOs: The AI implementation gap is now a competitive risk, not just an efficiency opportunity. Health systems with deployed ambient documentation are retaining physicians at materially higher rates and reducing documentation error rates that affect coding accuracy and revenue. The window to be an early mover is closing - 150 health systems are already live on Abridge alone. Build your AI governance infrastructure now so you can absorb tools in 30 to 60 days when you make a vendor decision.

For radiologists and pulmonologists: The funding surge includes radiology AI - DeepHealth's $269 million Gleamer acquisition signals that the radiology AI consolidation is accelerating. Workflow AI that integrates directly into PACS and flags study priorities, identifies incidental findings, and auto-populates structured reports is moving from optional to expected at competitive radiology groups. Understanding the CPT reimbursement codes available for AI-assisted reads (including CPT 0721T for lung CT malignancy scoring) is an immediate revenue and documentation priority.

For healthcare investors and founders: The category winners in ambient documentation and revenue cycle AI are largely set. The next wave of value creation is in the application layer - disease-specific AI tools, payer-side AI for population health, AI for prior authorization coordination between providers and payers, and safety-net-specific platforms that combine care navigation, documentation, and quality reporting. The white space is in distribution to under-resourced settings that the current unicorns are not optimized to serve.

For policy advocates: The funding surge is creating a two-speed healthcare system. Health systems with capital and IT capacity to absorb AI tools are accelerating. Safety-net providers without that infrastructure are falling further behind. HRSA and CMS have an opportunity to create AI readiness funding streams within existing grant mechanisms - specifically Section 330 FQHC supplemental grants - to close the adoption infrastructure gap before it compounds into a quality and equity chasm.

Key References

• Healthcare AI Funding Hit $7.4B in Q1 on Mega Rounds; 8 New Unicorns Emerge - citybiz.co (April 2026)

• Abridge Secures $300M Series E Led by a16z to Pioneer a New Paradigm of Care Intelligence - abridge.com / fiercehealthcare.com (2025)

• Hippocratic AI Banks $141M Series B, Hits Unicorn Status - fiercehealthcare.com (January 2025)

• Waystar AI Tools Prevent $15B in Denied Claims, Cut Denial Work 90% - stocktitan.net / waystar.com (2026)

• Health System Denials and Underpayments Still Rising in 2026 - revecore.com / AHA data (2026)

Closing

The spring 2026 health AI funding surge is not a signal that AI has solved healthcare. It is a signal that investors have identified where the problems are measurable, where the ROI is defensible, and where the deployment infrastructure is finally mature enough to support enterprise-scale adoption.

Seven billion dollars in a single quarter does not buy change. It buys tools. The change still requires the hard work: the clinical champions, the governance frameworks, the data cleanup, the training, and the organizational will to actually restructure how care gets delivered.

The honest question remains: are health systems building internal AI capacity fast enough to absorb all of this technology, or are we funding tools that will sit unused because the change management infrastructure does not exist?

Funding is not adoption. Adoption is not outcomes. Outcomes are the only thing that matters for patients - and patients at FQHCs and safety-net providers are still waiting for the industry to close the gap between what is being built and what is being deployed where they receive care.

What is your organization doing to close that gap? Reply and let me know - I read every email.

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts. Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise. Subscribe at jonathangovette.substack.com

There is a sentence you will not hear in most budget debates: "The states already got cut."

Not threatened. Not proposed. Cut.

On July 4, 2025, President Trump signed the One Big Beautiful Bill Act into law. Buried inside it - alongside the Medicaid work requirements and ACA subsidy rollbacks that grabbed headlines - were three provisions that permanently rewrote the rules of Medicaid financing for every state in the country. Provider taxes, the mechanism that 49 states have used for decades to draw down billions in federal matching funds, are now frozen, capped, and headed for a forced phase-down in the states that expanded Medicaid under the ACA.

The full impact is not theoretical. It is already landing on state budget offices, hospital CFOs, and community health center administrators who spent the spring of 2026 doing math that did not add up.

This is not a future policy risk. This is the fiscal reality of right now.

1. What Provider Taxes Actually Do - and Why 49 States Use Them

To understand what was cut, you need to understand the mechanism.

States are required to fund a share of their Medicaid programs. That "state share" typically comes from general revenue - tax dollars. But starting in the early 1990s, CMS approved an alternative: states could assess a tax on hospitals and other providers, use those proceeds to draw down a larger federal Medicaid match, and then redistribute a portion back to providers through higher reimbursement rates.

This is not a loophole. It is a CMS-approved, legally structured financing mechanism that has been in place for more than 30 years.

The math is compelling. A state assessing a $1 provider tax might draw down $2 or more in federal funds at the standard matching rate. The state nets revenue. The provider nets higher reimbursement after the tax. Medicaid patients get better-funded care. Federal dollars flow into states that would otherwise struggle to maintain program solvency.

According to KFF, provider taxes accounted for 18 percent of the non-federal share of Medicaid spending in state fiscal year 2026 enacted budgets. All states but Alaska have at least one provider tax in place. Forty-one states have three or more.

The most sophisticated versions are Managed Care Organization (MCO) taxes - assessments on managed care plans participating in Medicaid. New York's MCO tax generates $3.7 billion in state savings per year. California's generates $7.5 billion annually.

2. What the Law Changed - and How Quickly

The One Big Beautiful Bill Act imposed three distinct provider tax restrictions, each hitting a different category of state.

Restriction 1: The Moratorium on New Taxes

Effective July 4, 2025, no state may create any new provider tax of any kind, nor increase the rate of any existing provider tax beyond the level in effect on that date. This applies to every state, regardless of expansion status.

The implications are profound. States can no longer respond to Medicaid funding shortfalls by broadening their provider tax base. During the 2008 financial crisis, several states turned to provider taxes as a countercyclical tool to maintain Medicaid solvency without raising general taxes. That option is now permanently closed.

Restriction 2: Expansion State Phase-Down

For the 22 states that have adopted the ACA Medicaid expansion and have provider taxes above the new 3.5 percent threshold, the law requires a phased reduction. Starting in 2028, those states must reduce their provider tax safe harbor by 0.5 percentage points per year until they reach the new 3.5 percent ceiling, arriving there by 2032.

The current safe harbor is 6 percent of net patient revenues. The new ceiling cuts that almost in half.

Restriction 3: MCO Tax Acceleration for Recent Waivers

States whose MCO tax waivers were approved within two years of April 2026 face an accelerated compliance deadline. California and New York - the two largest Medicaid programs in the country - must comply by January 1, 2027. That gives them approximately six months from now.

3. The CBO Accounting - and What It Means for States

The Congressional Budget Office estimated that the provider tax provisions in the OBBBA will produce $191 billion in gross federal savings from the moratorium and expansion state phase-down combined. Including uniformity changes, total gross federal Medicaid savings from provider tax restrictions reach $225.7 billion over ten years.

To be explicit about what that number means: $225.7 billion that would have flowed to states as federal matching funds will instead stay in Washington.

📊 On a dollar basis, the largest estimated cuts fall to California ($112 billion) and New York ($63 billion) over ten years. Twenty states will lose at least 5 percent of their total Medicaid budgets.

CBO modeled how states would respond. Its central scenario: states replace approximately half of lost revenues through other state resources - tax increases or cuts to other programs. The other half gets absorbed through Medicaid spending reductions: lower provider payment rates, fewer covered services, or tighter eligibility criteria.

That 50/50 split is a projection, not a guarantee. In practice, states facing simultaneous pressure from Medicaid, education, and infrastructure spending rarely have the political will to raise general taxes. The realistic expectation is that the Medicaid reduction share ends up closer to the full amount.

The Pew Charitable Trusts reported in January 2026 that the OBBBA's Medicaid changes are already compounding state budget pressures, with some states facing up to $450 million in added costs and lost revenue in the current fiscal year. States were told to model scenarios that assume no new federal relief.

4. What States Will Actually Do

When state Medicaid programs face sudden financing shortfalls, they have a limited toolkit. The options are not abstract - they are the same three levers that states have reached for every time federal Medicaid funding has tightened.

Lever 1: Provider Rate Reductions

Medicaid payment rates are already below Medicare and commercial rates at most safety-net providers. A state facing a several-hundred-million-dollar shortfall will cut provider rates first, because it is faster than changing eligibility rules and avoids the political visibility of removing specific populations from coverage.

For an FQHC billing the Prospective Payment System rate, even a 5 percent rate reduction can erase the margin on a patient visit. For a rural hospital where Medicaid represents 30 to 40 percent of payers, a 10 percent Medicaid rate cut is the difference between marginal solvency and operational crisis.

Lever 2: Optional Benefit Cuts

Medicaid covers a set of mandatory benefits defined by federal law and a separate category of optional benefits that states may choose to offer. Dental for adults, vision, personal care services, and certain home-and-community-based services are optional. They are also heavily used by the disabled, the elderly, and the chronically ill.

When revenue is tight, states eliminate optional benefits. They are legally defensible, harder to communicate to the public, and disproportionately harm people who already have few alternatives.

Lever 3: Eligibility Tightening

The OBBBA also contains separate provisions - work requirements, more frequent eligibility redeterminations, and income documentation requirements - that will reduce Medicaid enrollment independent of the provider tax changes. The provider tax restrictions compound those enrollment reductions by shrinking the state's financing capacity precisely when demand on the safety net is rising.

CBO estimates 1.2 million more people uninsured by 2034 attributable to the provider tax provisions alone. That figure does not include the additional coverage losses from work requirements and redetermination changes.

5. The FQHC Equation - Silver Lining, Significant Risk

FQHCs sit in a complicated position under the OBBBA.

The legislation included one meaningful protection: FQHCs, Rural Health Clinics, and Certified Community Behavioral Health Clinics are exempt from the new $35 per-visit cost-sharing requirement that applies to Medicaid expansion adults with incomes between 100 and 138 percent of the federal poverty level, effective October 1, 2028. That is a real win. FQHC advocates fought hard for it, and it was not in the original House-passed version.

But cost-sharing exemptions protect patients from out-of-pocket costs. They do not protect FQHCs from provider rate cuts.

The Commonwealth Fund and Pew both project that FQHCs could lose between one-fifth and one-quarter of their total revenue as state Medicaid programs reduce payment rates in response to the provider tax restrictions. That is not a marginal efficiency issue. That is a structural solvency threat.

The Michigan Primary Care Association has estimated that clinics in that state alone will lose $94 million per year in reimbursement under the OBBBA's combined Medicaid provisions. Michigan is a large-state example, but the proportional impact will be similar in any ACA expansion state with a large FQHC network.

FQHCs are also disproportionately dependent on Medicaid. On average, 40 to 50 percent of FQHC revenue comes from Medicaid. Some safety-net clinics in expansion states with high Medicaid penetration exceed 60 percent. When the state cuts provider rates, FQHCs feel the cut across a majority of their patient base simultaneously.

The math is straightforward and grim. An FQHC seeing 20,000 Medicaid visits per year at a $280 per-visit PPS rate generates $5.6 million in Medicaid revenue. A 15 percent rate cut removes $840,000 from that clinic's annual operating budget. For a clinic running on 3 to 5 percent margins, that is not a budget adjustment. That is a potential closure trigger.

6. Deep Dive: The State-Level Cascade

The mechanism that makes provider tax cuts particularly damaging is their compounding effect across the Medicaid ecosystem. It is not just that states lose revenue. It is that the same law is cutting multiple revenue streams simultaneously.

The OBBBA also included significant reductions to Disproportionate Share Hospital (DSH) payments - the federal dollars that compensate hospitals for serving high volumes of uninsured and Medicaid patients. Safety-net hospitals facing both a DSH cut and a provider rate reduction from a financially constrained state are getting hit from two directions at once.

Scenario A: The Expansion State With a Large MCO Tax

Take a state like California or Illinois. They have MCO taxes currently above the new 3.5 percent threshold. They generate hundreds of millions to billions per year through those taxes, which they use to fund higher payment rates for safety-net hospitals and FQHCs.

Under the OBBBA, they must begin reducing that tax starting in 2028 - and California faces the accelerated January 2027 compliance deadline for its MCO tax waiver. As the tax phases down, the state loses federal matching funds. It then has to choose: replace the revenue with general fund dollars (competing with education, infrastructure, and public safety), reduce the supplemental payments to safety-net providers, or reduce Medicaid eligibility.

In California's case, $7.5 billion per year in MCO tax revenue is at risk over the phase-down period. That is not a line item that can be quietly absorbed.

Scenario B: The Non-Expansion State With Frozen Taxes

Non-expansion states like Texas, Florida, and Georgia are not subject to the phase-down. Their existing provider taxes are frozen at current levels with the 6 percent safe harbor maintained. They are not compelled to cut Medicaid financing the way expansion states are.

But they still face the moratorium. If a recession hits and Medicaid enrollment surges, they cannot respond by creating new provider taxes to maintain program solvency. Their financing flexibility is permanently capped.

Scenario C: The Safety-Net Hospital in an Expansion State

Urban safety-net hospitals in expansion states face the most direct exposure. Their Medicaid revenue base is large - often 40 to 60 percent of total patient revenue. Their DSH payments are being cut. And as the provider tax restrictions force states to reduce supplemental payments, the enhanced rates that hospitals have relied on to cross-subsidize uncompensated care will erode.

The Commonwealth Fund's 2026 modeling projects that the combined effect of the OBBBA's Medicaid provisions will increase demand for state and local general fund support to keep safety-net hospitals operational. But that demand increase is arriving precisely when state budgets are absorbing the revenue losses from the provider tax restrictions.

What This Means For You

For FQHC executives and community health center leaders: Model your Medicaid revenue exposure at multiple rate reduction scenarios - 5 percent, 10 percent, and 15 percent. Identify which programs are cross-subsidized by Medicaid margin and which would be cut first. Build the case now for local government and philanthropic support to bridge the gap during the phase-down period. The cost-sharing exemption you won is real, but it does not protect your operating budget.

For health system administrators and CFOs: The provider tax phase-down does not begin until 2028 in most expansion states, but California and New York face January 2027 compliance. If you operate in those states, the scenario planning window is now, not in two years. Model your supplemental payment exposure and your DSH payment trajectory together - the combined hit is what will define your FY2027 and FY2028 financial position.

For radiologists and pulmonologists: Medicaid rate cuts will affect reimbursement for diagnostic imaging in high-Medicaid markets. The states most likely to reduce rates are the same expansion states with the highest rates of uninsured-to-Medicaid conversion. If you depend on Medicaid volume, watch state legislative sessions in California, New York, Illinois, and Michigan for signals about where rates are heading.

For healthcare investors and founders: The provider tax restrictions will accelerate consolidation among community health centers and safety-net hospitals. Under-capitalized single-site FQHCs with high Medicaid concentration are the highest closure risk. Health system acquirers and nonprofit networks with diversified payer mix will gain bargaining power. Watch for grant-driven bridge financing opportunities in states where governors are trying to prevent closures before the phase-down hits.

For policy advocates: The moratorium on new provider taxes is the most permanent and underreported provision in the OBBBA. It eliminates a countercyclical tool that states have used in every Medicaid crisis since the 1990s. If the economy softens and Medicaid enrollment rises, states will be navigating the next coverage crisis with one hand tied behind their backs. That case needs to be made to Senate Finance Committee members now, before the 2028 phase-down begins and the political moment to negotiate modifications passes.

Key References

• Congressional Budget Office cost estimates for the One Big Beautiful Bill Act's Medicaid provider tax provisions (2025): $225.7 billion in gross federal savings over 10 years

• KFF: "Which States Might Have to Reduce Provider Taxes Under the Senate Reconciliation Bill" - identifying 22 affected expansion states including CA, NY, IL, and 19 others

• Commonwealth Fund: "How New Limits on State Provider Taxes Will Affect Medicaid Funding" (December 2025) - projects FQHCs lose 20-25% of revenue

• Pew Charitable Trusts: "New Federal Medicaid Policies Compound State Budget Pressures" (January 2026) - states face up to $450M in added costs and lost revenue

• Source: House Reconciliation Bill Would Restrict Medicaid Provider Taxes, Threatening Over $40 Billion in Annual Federal Matching Funds Across 49 States

The Medicaid provider tax restrictions in the One Big Beautiful Bill Act are already law. The moratorium is active. The phase-down clock has started for 22 states. And California and New York are six months from an accelerated compliance deadline that threatens billions in annual Medicaid financing.

The political window to address this was in the Senate, during the markup. That window closed on July 4, 2025.

What remains is implementation, advocacy for state-level workarounds, and the hard work of preparing safety-net providers for a financing environment that is structurally leaner than anything they have planned for.

The gap between what states can spend on Medicaid and what the safety net costs to operate is going to widen. The question for every FQHC executive, health system CFO, and state Medicaid director reading this is not whether that gap will appear. It is how wide it gets before someone closes it.

Reply to this email and tell me: how is your organization modeling the provider tax impact? I read every response.

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts. Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise. Subscribe at jonathangovette.substack.com

Not because they no longer qualify. Because the system processing their eligibility cannot keep up.

The House-passed reconciliation law includes provisions that most healthcare leaders have not read closely enough: a mandate that states conduct eligibility redeterminations every six months for Medicaid expansion adults beginning December 31, 2026, the elimination of continuous eligibility protections put in place after the pandemic, and $990 billion in gross Medicaid and CHIP federal spending cuts over the next ten years - the largest Medicaid cuts in American history.

The headlines focus on work requirements. The real danger to children is quieter. It is buried in renewal cycles, paperwork deadlines, and a bureaucratic infrastructure that was never designed to process 35 million families every six months without losing some of them.

The scene above plays out every day in community health centers across America: a child who came in for a well visit whose coverage has already lapsed - not because the family did anything wrong, but because a renewal notice went to the wrong address.

1. The Program That Protects Working Families' Children

CHIP - the Children's Health Insurance Program - was created in 1997 to fill the gap between Medicaid and private insurance. It covers children in households that earn too much for Medicaid but too little to afford employer or marketplace coverage. In most states, eligibility extends to 250% of the federal poverty level. In several states, it reaches 400%.

As of February 2026, 35.7 million children were enrolled in CHIP or the children's track of Medicaid. That is nearly half of all Medicaid and CHIP enrollment in the country.

📊 As recently as the start of 2025, that number stood at 37.4 million. By October 2025 it had fallen to 36.4 million. By February 2026, it was 35.7 million. The program has shed 1.7 million children in just 14 months - and the largest policy shock has not even arrived yet.

The enrollment decline below shows the trend before the reconciliation law's provisions take effect.

The Consolidated Appropriations Act of 2023 gave children an important protection: 12-month continuous eligibility, meaning states had to keep children enrolled for a full year regardless of changes in family income or paperwork status. The reconciliation law eliminates that protection and replaces it with 6-month redetermination cycles starting December 31, 2026.

That single change doubles the number of renewal checkpoints a child must successfully pass through every year. And research consistently shows that more checkpoints mean more procedural disenrollment.

2. What the Reconciliation Law Actually Changes

The headline provision for Medicaid and CHIP coverage loss is the work reporting requirement for expansion adults - CBO estimates it will cause 4.8 million people to lose coverage by 2034 alone. But for children specifically, the mechanism is different and in some ways harder to fight.

Children cannot work. Their eligibility does not depend on employment. They lose coverage because their parents cannot navigate a renewal system designed with administrative friction built in.

📊 The CBO projects the reconciliation law will increase the number of newly uninsured Americans by 1.3 million in 2026, rising to 5.2 million by 2027, 8.6 million by 2029, and 10 million by 2034.

The chart below shows that trajectory.

The $990 billion in gross Medicaid and CHIP cuts over ten years represents the largest single reduction to these programs since their creation. Three-quarters of total coverage losses are projected to come from Medicaid and CHIP cuts specifically. Children represent nearly half of all Medicaid and CHIP enrollment. The math is not subtle.

Key provisions that directly affect children's coverage:

Section 44108 mandates 6-month redetermination cycles for expansion populations, replacing 12-month continuous eligibility. The Medicaid continuous eligibility provisions that protected children from coverage lapses during the pandemic unwinding are eliminated. CHIP federal matching rate floors face pressure in states with high uninsured rates. And the per capita cap conversion shifts financial risk to states, which then pass it to providers through rate compression.

The paperwork above is not a metaphor. It is the literal barrier between an eligible child and their healthcare. Families in low-wage jobs with variable schedules, without access to printers or stable mailing addresses, face a renewal process that drops them when life gets in the way.

3. Procedural Disenrollment: The Real Mechanism of Loss

There is a specific term for what happens when eligible children lose coverage because of administrative failure rather than actual ineligibility. It is called procedural disenrollment. And it is the number one driver of coverage loss in every state that has implemented more frequent redeterminations.

The evidence from the pandemic unwinding of continuous enrollment is unambiguous. When states accelerated Medicaid renewals between 2023 and 2025, millions of enrollees were terminated for procedural reasons - missed notices, outdated contact information, forms that were submitted but not processed correctly. Studies from KFF and Georgetown's Center for Children and Families found that a significant proportion of those disenrolled were still legally eligible at the time of their termination.

📊 Arkansas ran a work requirement pilot in 2018. Within six months, 18,000 people lost coverage. The employment rate did not change. The vast majority of those disenrolled were either already working or qualified for an exemption. Paperwork failed them, not policy.

Children do not self-report. They do not file their own renewals. Their coverage depends entirely on their parent or guardian successfully navigating a state system - often in a language they are not fully fluent in, often while working two jobs, often without time off to make phone calls during state office hours.

Moving from 12-month to 6-month renewal cycles doubles the number of opportunities for procedural failure per year. For families already operating at the edge of administrative capacity, that is not a marginal increase in burden. It is a structural guarantee of higher disenrollment rates.

The families above are not falling through the cracks because of fraud or ineligibility. They fall through because the renewal system was never resourced to serve them at the pace the law now requires.

4. What This Means for FQHCs and Safety-Net Providers

For community health centers and FQHCs, the CHIP coverage cliff is not an abstraction. It arrives as a change in the payer mix sitting at the front desk.

Approximately 30% of FQHC patients are children. When those children lose CHIP coverage, their visits do not disappear. They shift. They move from covered visits reimbursed at the Prospective Payment System rate to self-pay visits on the sliding fee scale. That shift is enormous in revenue terms.

📊 The FQHC PPS rate for a covered pediatric visit runs approximately $280 per encounter. A self-pay sliding scale visit at the same clinic generates roughly $40. One child losing coverage costs an FQHC $240 per visit in net revenue.

The comparison below puts that gap in concrete terms.

For an FQHC seeing 200 pediatric patients per week, a 15% coverage loss from CHIP disenrollment translates to roughly $1.8 million in annual revenue reduction. That is staff. That is hours. That is capacity to serve the remaining patients.

The clinical impact compounds the financial one. Children who lose coverage skip well-child visits. They miss vaccine doses. They fall behind on developmental screenings. They arrive in emergency departments six months later with conditions that cost ten times more to treat than prevention would have.

For pediatric subspecialty practices, imaging centers that do pediatric work, and hospitals serving children, the downstream referral effect is equally real. When primary care access collapses for low-income children, specialty referrals dry up and ED volumes climb.

5. The Return-on-Investment Argument Nobody Is Making Loudly Enough

There is a fiscal argument against cutting children's coverage that rarely makes it into reconciliation debates because it requires thinking across a longer time horizon than a 10-year budget window.

Children with consistent health coverage have better educational outcomes, lower rates of adult hospitalization, lower lifetime Medicaid costs, and higher lifetime earnings. CHIP has one of the strongest return-on-investment profiles of any federal health program. The savings from cutting it today generate costs that land decades later - not inside the CBO's scoring window.

📊 A single ER visit for a preventable condition in an uninsured child costs roughly $2,000. Annual preventive care for that same child, including vaccinations and well visits, costs approximately $300. A hospitalization for a condition that reached crisis stage without preventive care averages $15,000.

That cost comparison plays out below.

The reconciliation law saves money on paper by reducing federal CHIP and Medicaid spending. But it shifts those costs - to state budgets, to hospital uncompensated care pools, to FQHC sliding-scale revenue, and ultimately to higher adult healthcare utilization from children who grew up without coverage.

This is not a partisan talking point. It is documented health economics with a 20-year evidence base.

6. The Policy Contradiction

There is a direct contradiction at the center of this legislation that deserves to be named clearly.

The same reconciliation law that eliminates continuous eligibility protections for children also includes provisions intended to reduce long-term healthcare costs. Early detection. Preventive care. Value-based payment models. AI-powered diagnostics.

All of those interventions assume a patient population that has consistent access to primary care. A child who loses CHIP coverage at age four and regains it at age six has a two-year gap in preventive care. Vaccinations missed. Vision and hearing screenings not done. Developmental delays not flagged. The cost of closing those gaps when coverage returns is higher than the cost of maintaining it continuously would have been.

The preventive care moment above only happens when a child has coverage. Disrupt the coverage and the visit does not happen.

Deep Dive: Financial Model for an FQHC Serving 500 Pediatric Patients Monthly

Let us make the financial stakes concrete for a community health center operating at a realistic scale.

Baseline assumptions: - 500 pediatric Medicaid/CHIP patients per month - Current payer mix: 85% CHIP/Medicaid covered, 15% self-pay - Average covered visit revenue (PPS): $280 - Average self-pay visit revenue: $40 - All 500 patients continue to seek care regardless of coverage

Scenario A: Status Quo

Monthly covered visits: 425 x $280 = $119,000 Monthly self-pay visits: 75 x $40 = $3,000 Monthly pediatric revenue: $122,000 Annual: $1,464,000

Scenario B: 15% Coverage Loss from CHIP Disenrollment

Monthly covered visits drop to 350 (70%), self-pay rises to 150 (30%) Monthly covered revenue: 350 x $280 = $98,000 Monthly self-pay revenue: 150 x $40 = $6,000 Monthly pediatric revenue: $104,000 Annual: $1,248,000

Annual revenue loss: $216,000 on the pediatric panel alone. For a health center running 2-3% operating margins, that is a structural deficit, not a manageable variance.

The conversation above needs to happen in every FQHC boardroom this year. The timeline is known. The mechanics are documented. The financial impact is calculable. What is missing is the urgency.

What This Means For You

• FQHC executives and health center CEOs: Model your pediatric Medicaid/CHIP volume now. Know exactly how much annual revenue is at risk if 15-20% of your pediatric patients experience a coverage gap. That model should be in your board materials before the end of Q3 2026.

• Health system administrators and CFOs: The 6-month redetermination mandate starts December 31, 2026. You have roughly six months to build or expand enrollment navigator capacity. Every navigator you hire before December is cheaper than the uncompensated care you absorb in Q1 2027.

• Pediatricians and primary care providers: Start proactively checking coverage status at every well-child visit. A lapsed CHIP enrollment caught at the visit is recoverable. A six-month gap in care is not. Treat coverage verification as a clinical protocol.

• Radiologists and imaging centers doing pediatric work: Your referral pipeline depends on primary care visits that depend on coverage. If CHIP disenrollment accelerates in your market, expect a downstream reduction in referred pediatric imaging. Factor that into your 2027 volume projections.

• Policy advocates: 12-month continuous eligibility for children was a hard-won protection with a strong evidence base. Its elimination is not a technical adjustment. It is a policy choice with predictable consequences. State-level continuous eligibility legislation is the immediate backstop - several states have already moved to protect it.

• Healthcare investors and founders: The enrollment navigator market is going to grow. The technology to automate Medicaid/CHIP renewal tracking, send proactive reminders to families, and flag at-risk enrollees before coverage lapses is underbuilt and in demand. There is a real business here.

Key References

• Georgetown University Center for Children and Families, "Medicaid and CHIP Cuts in the House-Passed Reconciliation Bill Explained" (2025) - detailed breakdown of provisions affecting children's coverage.

• CBO coverage estimates via Georgetown CCF (August 2025) - projects 1.3M newly uninsured in 2026, rising to 10M by 2034.

• CMS Medicaid/CHIP Enrollment Data Highlights, February 2026 - 35.7M children enrolled in CHIP and children's Medicaid.

• Georgetown CCF, "New CHIP Protections Are In Effect Now Despite Congressional Efforts to Eliminate Them" (July 2025) - documents the continuous eligibility rollback.

• KFF Medicaid Enrollment and Unwinding Tracker - documents procedural disenrollment patterns from 2023-2025 unwinding.

The families who depend on CHIP are not political abstractions. They are the parents working the early shift at a warehouse who cannot take time off to call a state eligibility office. They are the kids who show up for a well visit and find out their coverage lapsed three weeks ago because the renewal form went to the wrong address.

The law is signed. The timeline is set. The damage is coming on a schedule that every provider serving low-income children can plan around right now.

The question is not whether this happens. It is whether your organization is in a position to absorb it, or to actively protect the families in your panel before December 31.

What is your pediatric Medicaid/CHIP volume, and have you run the coverage loss scenarios yet? Reply directly - I read every response.

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts.

Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise.

Free subscribers get one post per month. Paid subscribers get full weekly access.

Subscribe free at jonathangovette.substack.com

18% of eligible Americans get lung cancer screening.

At FQHCs, it is under 6%.

Let that sink in. Lung cancer will claim an estimated 124,990 American lives in 2026. It kills more Americans than colorectal, pancreatic, and breast cancer combined. An estimated 229,410 new cases will be diagnosed this year. 85% will be found at a late stage, when 5-year survival collapses to under 9%. Catch it at Stage I and survival reaches 77-82%.

The math on early detection is not complicated. The execution is where the system has failed - and the communities it has failed most are the ones least equipped to absorb that failure.

1. The Screening Crisis: Why 94% Fall Through

Low-dose CT screening for lung cancer carries a Grade B USPSTF recommendation. Coverage is mandated. Eligibility criteria are clear. This should be a solved problem. It is not.

Nationally, fewer than 16% of eligible Americans have ever received lung cancer screening. At FQHCs and community health centers, that rate falls below 6% - frequently below 3% at high-volume safety-net sites. Research published in 2025 found that fewer than 5% of the 14.5 million eligible underserved Americans receive their annual no-cost lung cancer screening.

The chart below shows the screening rate gap across different care settings - from urban academic centers to rural hospitals.

The barriers are structural, not motivational. A 2021 PMC study found that 72% of FQHCs reported patients faced coverage gaps as a primary barrier, 58% cited preauthorization requirements, and 73% flagged out-of-pocket cost concerns despite screening technically being free. Only 51% had adequate access to specialty providers needed to manage abnormal findings.

One resident-run FQHC demonstrated what fixing the infrastructure problem looks like: after implementing targeted patient outreach and EHR care gap optimization in early 2025, monthly completed screening scans jumped from 1.2 to 12.3 per month. A tenfold increase driven entirely by better clinical decision support - not by adding staff.

The table below puts the screening gap in context across the full spectrum of care settings.

The screening gap is not a patient compliance problem. It is an infrastructure problem. And AI is uniquely positioned to close it.

2. What Lung-RADS Gets Wrong

Even when patients do get scanned, the standard tool for reading those scans has real limits that rarely get discussed.

Lung-RADS - the Lung Imaging Reporting and Data System - assigns risk categories based on nodule size thresholds with some accommodation for growth rate. On paper: systematic and standardized. In practice: a blunt instrument applied to a precision problem.

The core issue is that size and growth alone are poor predictors of cancer. Shape, texture, density, spiculation, patient history, smoking pack-years: none of these are formally weighted in the standard Lung-RADS calculation. A nodule with an 89 on actual malignancy probability looks identical to one scoring 41 if both fall in the same size bucket.

The real-world consequences: two radiologists reading the same scan disagree on the Lung-RADS risk category roughly one-third of the time. False positive rates run up to 40% - unnecessary biopsies, extra imaging, patient anxiety for nothing. False negatives cluster around 8-10% - real cancers missed, sent home, found 18 months later at Stage III.

Lung-RADS was designed to standardize reporting. It was not designed to maximize diagnostic accuracy. In a world before AI, it was the best available option. That world is over.

3. The AI Difference: A 0-100 Malignancy Score

The difference between a binary risk category and a continuous malignancy probability is not cosmetic. It is clinical.

Oatmeal Health's Pre-FDA AI layers a continuous 0-100 malignancy probability score onto each nodule inside existing CADe and PACS workflows. No new login. No workflow change. The score integrates directly into the read.

A continuous score eliminates the cliff problem. Under Lung-RADS, a nodule just above or below a size threshold receives dramatically different management - even when the underlying biology is nearly identical. A 0-100 score treats each nodule as the individual risk object it actually is.

It also enables systematic triage. When appointment slots are limited - and at FQHCs and busy pulmonology practices, they always are - the nodule scoring 94 should not wait behind the one scoring 47. AI makes that prioritization systematic instead of anecdotal.

Recent research shows integrated AI systems achieving AUC values of 0.98 internally and 0.945 on independent validation cohorts - outperforming both radiologists and leading AI models. AI can detect malignancy up to one year earlier than radiologists for indeterminate and slow-growing nodules.

AI-assisted CADe and CADx systems demonstrate 30-75% reductions in scan review time and 30-50% faster reporting. Read time drops from 5-10 minutes to 1.5-3 minutes per case. In a practice reading 6,000 studies per year, that is hundreds of recovered hours.

📊 AI-assisted lung nodule assessment has demonstrated 30-75% reductions in scan review time, translating to hundreds of recovered hours annually in high-volume radiology practices.

4. CPT 0721T - The Revenue Case for Adoption

Clinical outcomes are compelling. Health system administrators need a financial case too.

CPT code 0721T covers AI-driven CT tissue characterization for lung nodule malignancy assessment. It is a Category III code assigned to New Technology APC 1508, with Medicare reimbursement in the $601-$700 range per procedure.

This is real billable revenue attached to a workflow improvement that also reduces read time. For a practice already reading high volumes of lung CTs, adopting AI malignancy scoring under CPT 0721T is net-positive from day one: revenue increases while per-scan time cost decreases.

The table below shows before-and-after impact across the key factors that matter to administrators and radiologists.

The chart below shows how CPT 0721T revenue scales with scan volume.

For FQHCs specifically, CPT 0721T reimbursement changes the sustainability equation entirely. FQHCs operate on thin margins - typically 3-5% operating surplus in good years, frequently in the red. Any screening program needs a reimbursement pathway. AI malignancy scoring has one that most FQHC leaders have not yet found.

📊 CPT 0721T Medicare reimbursement of $601-$700 per procedure creates a net-positive financial case for AI lung nodule assessment at virtually any scan volume.

5. The Pulmonologist Perspective: Smarter Follow-Up

For pulmonologists managing patients with known lung nodules, a 0-100 score changes clinical decision-making in concrete ways.

The most common scenario: a Lung-RADS 3 nodule - intermediate risk, standard recommendation is 6-month follow-up CT. Under standard Lung-RADS, all Lung-RADS 3 nodules look the same. AI scoring distinguishes a probability of 12 from a probability of 67 within the same Lung-RADS category.

For the nodule scoring 12: safe to extend follow-up to 12 months, reduce radiation exposure, free up appointment slots. For the nodule scoring 67: shorten follow-up, consider tissue sampling, escalate urgency.

This restores clinical nuance that Lung-RADS removed in the name of standardization. Standardization that erases meaningful differences between patients is not a feature. It is a limitation.

6. The FQHC Model: Why the Gap Is Largest There

FQHCs serve 32 million patients annually across 1,400 organizations and 14,000 service sites. They are the primary care safety net for low-income adults, racial and ethnic minorities, uninsured patients, rural communities, and patients experiencing homelessness.

These are also the patients with the highest lung cancer risk. Smoking rates are significantly higher in low-income populations. Occupational and environmental exposures concentrate in communities served by FQHCs. Delayed primary care access means more patients arrive with advanced disease.

The chart below makes the survival gradient visible - and the cost of finding cancer late undeniable.

The FQHC lung cancer screening gap is not a medical mystery. It is a resource allocation problem. Most FQHCs lack automated eligibility identification, patient navigation capacity, and the diagnostic tools to prioritize abnormal findings appropriately once scans are obtained.

AI addresses all three. Automated patient identification surfaces eligible patients from the EHR without manual chart review. AI-guided navigation creates systematic outreach touchpoints. AI malignancy scoring ensures that when a high-risk nodule is found in a patient who may struggle to return for follow-up, it is flagged immediately - not buried in a queue.

📊 FQHCs serve 32 million patients annually, including the highest-risk, lowest-screened populations for lung cancer - yet fewer than 5% of eligible FQHC patients receive annual lung cancer screening.

Deep Dive: The FQHC Financial Model

Let us model what AI integration looks like financially for a mid-sized FQHC.

Baseline: 15,000 active patients. Smoking prevalence of 22%. Eligibility rate among smokers: approximately 40%. That gives roughly 1,320 eligible patients. At 5% screened: 66 scans per year. Zero AI integration. Zero CPT 0721T revenue.

Scenario A: Navigation Only

Implement AI-driven EHR identification and patient outreach. No change to the reading workflow.

Result: screening rate increases to 25% - 330 scans per year. A fivefold improvement based on published 2025 intervention data. But no CPT 0721T revenue without the malignancy scoring component. The program depends on grant funding to sustain itself.

Scenario B: Full AI Integration

Add AI malignancy scoring billed under CPT 0721T alongside navigation.

Result: 330 scans at $650 average CPT 0721T reimbursement = $214,500 in new annual revenue. Plus 30-50% reduction in per-scan radiologist time. Plus reduction in unnecessary follow-up from false-positive Lung-RADS reads.

The infographic below shows the revenue difference across the three scenarios.

Full AI Integration does something Navigation Only cannot: it creates a self-sustaining revenue model. CPT 0721T revenue funds the integration. Better diagnostics improve efficiency. Higher efficiency enables more volume. More volume generates more revenue. A flywheel - not a grant-dependent program that evaporates every three years.

What This Means For You

• FQHC CEOs and CMOs: The screening gap is an infrastructure problem, not a patient engagement problem. Automated EHR identification plus AI malignancy scoring creates a sustainable, reimbursement-backed program. The technology exists. The CPT code exists. The question is whether you move before your neighboring health system does.

• Health system administrators: CPT 0721T pays $601-$700 per scan for AI-assisted nodule assessment. At 1,000 lung CT reads per year, that is up to $700,000 in net-new annual revenue on a tool that also cuts radiologist read time by 30-50%. Build the business case now.

• Radiologists and radiology practices: AI malignancy scoring is a force multiplier - faster reads, better accuracy on difficult nodules, reduced missed-cancer liability, and CPT 0721T revenue layered onto your existing workflow. The practices integrating AI now will hold a quality and efficiency advantage over those that wait.

• Pulmonologists: A 0-100 score gives back the clinical nuance Lung-RADS removed. Use it to safely extend follow-up on low-probability nodules and urgently escalate high-probability ones. Stop treating every Lung-RADS 3 the same way.

• Healthcare investors and founders: CPT 0721T is live. Medicare is paying. The FQHC and health system market is actively looking for solutions. The infrastructure build is happening now - the window to enter is open.

• Policy advocates: Mandates without operational support do not move screening numbers. The policy lever that works is funding for clinical decision support infrastructure at FQHCs: automated identification, navigation, and AI diagnostic scoring. Advocate for that specifically.

Key References

• American Cancer Society, Lung Cancer Statistics 2026. 229,410 new cases and 124,990 deaths estimated. cancer.org

• SEER Cancer Statistics, National Cancer Institute. Stage-based 5-year survival rates: Stage I 77-82%, Stage IV 7-9%. seer.cancer.gov

• Closing the Lung Cancer Screening Gap in FQHCs with AI-Powered Clinical Decision Support, medRxiv 2025. Documents tenfold improvement in monthly screening completions with EHR-based AI intervention. medrxiv.org

• Integrated AI Nodule Detection and Diagnosis for Lung Cancer Screening, arxiv 2025. AI achieves AUC 0.98 internally, 0.945 on independent cohort, detects malignancy up to one year earlier. arxiv.org

• CMS New Technology APC 1508. CPT 0721T and 0722T reimbursement $601-$700 per procedure for AI-driven CT tissue characterization. cms.gov

Every day without a systematic screening program at your health system is a day a Stage I patient becomes Stage IV.

The technology is here. The reimbursement is here. The question is whether the will is too.

What is stopping your organization from acting? Reply and tell me - I read every response.

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts.

Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise.

Free subscribers get one post per month. Paid subscribers get full weekly access.

Subscribe free at jonathangovette.substack.com

The enhanced premium tax credits that had made marketplace coverage affordable since 2021 - first through the American Rescue Plan, then extended through the Inflation Reduction Act - quietly expired the night before. No dramatic legislative failure. No headline vote. Just a cliff, four years in the making, that Congress watched approach and, when December came, did not prevent.

The result is what health economists predicted and healthcare operators feared. Monthly premiums for marketplace enrollees jumped 58 percent overnight. Average deductibles hit $3,786 for 2026 - up more than $1,000 from the year before. An estimated 4.8 million Americans who were insured through the marketplace at the end of 2025 are now uninsured. For providers, hospitals, FQHCs, and anyone who works on the delivery side of healthcare, the question is not whether this is affecting your patient panel. It already is. The question is how much, and how fast it will compound.

The scene above reflects what safety-net providers are experiencing right now: rising patient volume, thinner margins, and an uninsured caseload that is growing faster than funding can absorb.

1. The Architecture of the Enhanced Tax Credits

The enhanced premium tax credits created under the American Rescue Plan Act of 2021 were not a small adjustment. They fundamentally restructured who could afford marketplace coverage.

Before 2021, ACA subsidies were means-tested on a sliding scale that cut off sharply at 400 percent of the federal poverty level - roughly $55,000 for a single adult in 2025. Anyone earning above that threshold received nothing. The enhanced credits eliminated that cap entirely, extending subsidies to households well above 400 percent FPL. They also deepened the subsidy for lower-income enrollees, making silver plans essentially free for people near the poverty line.

The result was enrollment growth that no previous ACA administrator had achieved. By 2025, 24.3 million people had signed up for marketplace plans - more than double the 11.4 million enrolled in 2020. Of those, 22.4 million - or 92 percent - were receiving enhanced premium tax credits. The average enrollee was saving roughly $800 per year compared to pre-ARPA costs.

The chart above shows the enrollment trajectory: a near-doubling between 2020 and 2025 driven by the enhanced credits, followed by a collapse back toward pre-expansion levels when those credits expired.

The Inflation Reduction Act extended these credits through the end of 2025. That extension came with a built-in deadline: January 1, 2026 was circled on every policy analyst's calendar as the day the cliff arrived - unless Congress acted. Congress did not act in time.

2. What the Numbers Actually Look Like

The human toll of the subsidy expiration becomes clearest through individual premium numbers. In 2025, the average marketplace enrollee paid $113 per month in net premiums after their enhanced credit. In 2026, that number jumped to $178 per month - a 58 percent increase. For many enrollees, particularly those in their 50s and 60s who face age-rated higher premiums, the monthly bill crossed a threshold where coverage stopped making financial sense.

The deductible story is equally striking. The average marketplace deductible grew 37 percent - from $2,759 in 2025 to $3,786 in 2026 - the steepest single-year increase in ACA history. For silver plans, the 2026 average deductible is $5,304. For bronze plans, it reaches $7,186.

This created a rational but painful shift in plan selection. Enrollment in bronze plans jumped from 30 percent to 40 percent of total plan selections between 2025 and 2026. People kept coverage but accepted dramatically higher out-of-pocket exposure to do it. The insurance technically exists, but the effective financial protection it provides in a real healthcare crisis is far thinner.

The average deductibles above - $5,304 for silver and $7,186 for bronze - reflect how the plan tier shift compounds financial exposure for people who stayed enrolled but moved to cheaper, higher-deductible plans.

The overall enrollment drop has been severe. Marketplace effectuated enrollment fell from roughly 22.3 million in 2025 to an estimated 17.5 million in 2026 - a loss of nearly 5 million covered lives. The Urban Institute estimates that 4.8 million of those individuals became uninsured. They did not transition to Medicaid. They did not find employer coverage. They joined the uninsured population outright.

📊 4.8 million people became newly uninsured when ACA enhanced premium tax credits expired January 1, 2026 - the largest single coverage loss event in ACA marketplace history.

3. The Legislative Graveyard

This did not have to happen. And it still does not have to continue.

The enhanced premium tax credits did not expire because they lacked political support. They expired because that support was never concentrated enough to force action before December 31, 2025.

In December 2025, the Senate voted on competing approaches to extend the credits. The Lower Health Care Costs Act, which would have extended enhanced PTCs through 2028, failed to reach the 60-vote threshold needed to advance. With the Senate gridlocked, the credits lapsed on January 1, 2026.

On January 8, 2026, the House passed HR 1834 - a bipartisan bill to extend the enhanced PTCs - by a vote of 230 to 196, with 17 Republicans joining Democrats. Lawmakers from both parties acknowledged at the time that the bill's Senate prospects were uncertain. As of this writing, it has not passed the Senate.

The House reconciliation bill that passed on May 22, 2026 - the "One Big Beautiful Bill Act" - did not include ACA subsidy extension. It instead added new marketplace provisions that tighten eligibility verification and eliminate passive re-enrollment. The CBO estimates these changes alone would reduce marketplace enrollment by an additional 2.2 million people on top of the 4.8 million already lost.

A bipartisan Senate working group is reportedly negotiating an alternative path. But negotiations are not law. Every month the credits remain expired is another month that 4.8 million people remain uninsured, and another month that providers absorb the financial consequences of that gap.

📊 The House reconciliation bill's marketplace provisions would cut enrollment by an additional 2.2 million - compounding the 4.8 million already uninsured from the subsidy cliff.

4. The FQHC and Safety-Net Provider Impact

The coverage loss does not evaporate. It redistributes.

When people lose insurance, they do not stop getting sick. They delay care. They present later with more advanced conditions. They use emergency departments for primary care. And when they finally need services, they turn to the providers who cannot say no: FQHCs, community health centers, safety-net hospitals, and free clinics.

The image above reflects the daily operational reality at most FQHCs: high patient volume, limited staff, and a financial model that was already under pressure before the coverage cliff added to it.

The financial math for these providers is severe. U.S. providers face $32.1 billion in lost revenue in 2026 from the subsidy expiration, along with an additional $7.7 billion in new uncompensated care demand. Of that $7.7 billion, $2.2 billion falls on hospitals, $1.0 billion on office-based physicians, $1.5 billion on prescription drug spending, and $3.1 billion on other care categories.

The infographic above ranks the uncompensated care burden by provider type. Hospitals and other care services carry the largest share of the projected $7.7 billion in new demand.

For FQHCs specifically, this adds to an already deteriorating financial picture. Health center net margins fell from positive 1.6 percent in 2023 to negative 2.1 percent in 2024. Section 330 funding covers a fraction of per-patient costs. The subsidy cliff sends more uninsured patients through the door at exactly the moment these organizations are least positioned to absorb them.

The geographic concentration makes this worse. Texas, Florida, Georgia, South Carolina, Tennessee, Alabama, and Mississippi - states that never expanded Medicaid under the ACA - face the largest increases in uninsured populations from the subsidy cliff. In these states, the gap between marketplace coverage and no coverage is a gap between having insurance and having nothing, because Medicaid expansion never filled it for adults above poverty-level income.

📊 FQHC net margins fell from +1.6% in 2023 to -2.1% in 2024. The uninsured surge from ACA subsidy expiration hits these organizations during an already critical financial stretch.

5. Cancer Screening in the Crosshairs

Preventive care is the first casualty of coverage loss. Cancer screening is where the damage becomes irreversible.

The research on this connection is not theoretical. Among adults 45 and older, 71 percent of those without insurance report being behind on colorectal cancer screening, compared to 33 percent of insured adults. Uninsured women are behind on breast cancer screening at twice the rate of insured women - 63 percent versus 30 percent. Cost is the primary reason cited by uninsured adults for skipping screenings. Among insured adults, cost does not even rank in the top four reasons.

The chart above shows the screening gap by cancer type and insurance status. The uninsured consistently lag the insured by roughly double - a gap that widens as coverage disappears.

For lung cancer specifically, the stakes are highest and the math is starkest. Lung cancer screening rates are three times higher among the insured than the uninsured. Low-dose CT scanning can reduce lung cancer mortality by 20 percent among high-risk adults. Every eligible high-risk adult who loses insurance is a person who stops getting screened.

And here is the clinical reality that drives everything at Oatmeal Health: stage I lung cancer has a 77 percent five-year survival rate. Stage IV survival is 9 percent. The difference between catching it early and catching it late is not incidental. It is the difference between a survivable diagnosis and a terminal one.

When 4.8 million people lose coverage, the gap between who gets screened and who does not widens dramatically. Some of those people were exactly the high-risk patients - older adults with smoking histories, in their 50s and 60s - who were most eligible for LDCT. Some of them are now going unscreened. That cost will not appear in the 2026 statistics. It will appear in the 2028 and 2029 stage-at-diagnosis data, when the missed cancers finally surface.

📊 Lung cancer screening rates are 3 times higher among insured adults than uninsured. With 4.8 million people newly uncovered, a wave of delayed diagnoses is building - it will show up in stage-at-presentation data in 2027 and beyond.

6. The Financial Modeling for Health System Leaders

The $32.1 billion headline revenue loss figure is worth unpacking for health system leaders who need to translate policy into operational planning.

For a mid-size hospital system serving a marketplace-heavy patient population, the math starts with the payer mix. If 8 to 10 percent of your commercially insured volume came from marketplace plans, and 20 to 25 percent of those patients have shifted to uninsured status, you are looking at meaningful revenue pressure concentrated in specific service lines.

Emergency department volume is a leading indicator. When outpatient coverage disappears, patients do not stop needing care. Primary care deferral becomes ED presentation. Conditions that could have been managed with a $150 office visit become $2,500 emergency encounters - with far lower collection rates.

For radiology departments and imaging centers, the impact shows up in referral volumes. Preventive imaging, screening CTs, and elective diagnostic studies are heavily insurance-dependent. When coverage lapses, referrals for these services drop. A lung cancer screening program that took years to build can drain quickly when the insurance backing it disappears.

The comparison above puts the scale in perspective: $32.1 billion in lost revenue is the direct hit from coverage disappearing. The $7.7 billion in new uncompensated care demand is the indirect hit from those same people still needing services, just without the ability to pay.

📊 Providers face a combined $39.8 billion financial shock in 2026: $32.1 billion in lost revenue plus $7.7 billion in new uncompensated care demand.

Deep Dive: The Geographic and Population Concentration Risk

The national headline numbers obscure significant geographic concentration. Coverage loss from the subsidy cliff is not evenly distributed across America. It is heavily concentrated in specific states, specific income bands, and specific provider markets.

Non-Expansion States Carry the Largest Burden

The 10 states that have not expanded Medicaid face compounding risk. Marketplace coverage was the only coverage option for adults with incomes between 100 and 400 percent FPL in these states. When enhanced PTCs expired, residents lost access to affordable insurance with no Medicaid pathway below them. Texas alone, which has the highest uninsured rate in the country and has not expanded Medicaid, accounts for a disproportionate share of the projected 4.8 million newly uninsured.

For FQHCs and safety-net hospitals in these states, the combination of Medicaid non-expansion and ACA subsidy expiration creates a coverage environment as hostile as anything seen in the past decade.

The 400-500% FPL Cliff

KFF analysis found that people with incomes between 400 and 500 percent of the federal poverty level - who made up just 3 percent of 2025 marketplace enrollees - accounted for 27 percent of the enrollment drop in 2026. Their plan sign-ups fell 44 percent, representing over 321,000 people concentrated in a narrow income band. These are working adults, often in their 50s and early 60s, who do not qualify for Medicaid and just lost subsidized coverage. They are also prime candidates for lung cancer screening - older adults with smoking histories who now face $5,000-plus deductibles if they buy coverage and no insurance at all if they walk away.

The image above represents what happens to communities when preventive care becomes unaffordable: deferral, delay, and eventual presentation at far greater acuity.

What the Senate Window Actually Means

A bipartisan Senate working group is negotiating. Healthcare leaders should track this closely, because a mid-year or retroactive extension is possible. If the Senate acts before 2027 open enrollment, some of the 4.8 million enrollment losses could be recoverable through a special enrollment period.

But "possible" is not "certain." Operational planning cannot wait for political resolution that may never arrive. The scenario planning starts now.

What This Means For You

• FQHC executives and community health center leaders: Model your 2026 patient panel for uninsured volume increases of 15 to 25 percent above 2025 baseline, especially in non-Medicaid-expansion states. Review your Section 330 grant renewal strategy - the funding formula matters more when uncompensated care is rising. Prioritize patient navigation resources for marketplace re-enrollment if the Senate acts; getting patients back on coverage quickly has both clinical and financial impact. Evaluate slide-fee schedule utilization for patients who dropped marketplace coverage but may qualify for discounted FQHC services.

• Health system administrators and CMOs: Track ED volume week-over-week by payer source. An increase in uninsured and self-pay encounters is the earliest operational signal of the coverage cliff's impact on your system. Review your DSH payment eligibility - if your uncompensated care ratio is increasing, your DSH qualification may be strengthening even as your margins compress. Do not cut preventive and screening service lines as a short-term margin response; the downstream clinical and liability consequences of interrupted screening programs are severe.

• Radiologists and pulmonologists: Lung cancer screening referral volumes for uninsured patients will decline. Work with your FQHC partners now to understand what alternative navigation pathways exist for high-risk patients who lost coverage. Stage-at-diagnosis distributions will likely worsen in 2027-2028 as a downstream result of reduced screening in 2026. CPT 0721T billing remains viable for insured patients - work with AI vendors who can identify the eligible high-risk patients who still have coverage and prioritize those patients for screening.

• Healthcare investors and founders: The coverage cliff creates specific financing pressure points for FQHC-adjacent businesses. Revenue models that depend on marketplace insurance as a payer source need contingency planning. Safety-net health IT and revenue cycle companies face both headwinds (lower reimbursement volume) and tailwinds (higher demand for uncompensated care management tools). Legislative risk is real and near-term - a Senate extension would materially change the market dynamics described here.

• Policy advocates: The window is narrow. Senate action before 2027 open enrollment is the critical intervention point. The 400-500 percent FPL population is politically legible - working adults who lost coverage due to legislative inaction, not benefit fraud or gaming. That framing matters for advocacy with Senate moderates.

Key References

• Urban Institute (2025): "4.8 Million People Will Lose Coverage in 2026 If Enhanced Premium Tax Credits Expire" - estimates 7.3 million total coverage loss, 4.8 million becoming newly uninsured

• KFF (2026): "The Average Marketplace Deductible Grew by About $1,000 Per Person in 2026" - comprehensive 2026 enrollment, premium, and deductible data

• Fierce Healthcare (2025): "Providers Face $32.1B in Lost 2026 Revenue If Enhanced ACA Subsidies Expire" - provider financial impact modeling by category

• ASTHO (2026): "ACA Enhanced Premium Tax Credits: Legislative Developments in 2025 and 2026" - full legislative timeline including Senate failures and House passage of HR 1834

• Prevent Cancer Foundation (survey): Insurance status and cancer screening completion rates among insured and uninsured adults

• American Lung Association: State-level lung cancer screening coverage requirements and insurance eligibility criteria

The ACA subsidy cliff is not a hypothetical. It happened. The credits expired. 4.8 million people lost coverage. Providers are already absorbing the financial and clinical consequences.

What remains uncertain is how long this lasts. The Senate has tools and, apparently, the political incentive to act - the bipartisan House vote in January showed that will exists somewhere. Whether that will materializes into law before the 2027 open enrollment window is the central policy question for everyone running a healthcare organization right now.

The math does not wait for political resolution. Operational planning needs to happen now. Patient panels need to be modeled for coverage loss. Screening programs need to prepare for thinner insurance backing. Safety-net providers need to resource for higher uninsured volume.

What are you doing in your organization to plan for the 4.8 million coverage loss? Reply to this email - I read every response and want to understand what operators and clinicians are seeing on the ground.

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts.

Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise.

Free subscribers get one post per month. Paid subscribers get full weekly access.

Subscribe free at jonathangovette.substack.com

The Medicaid work requirements debate has consumed most of the oxygen in health policy circles this year. But quietly, a second cliff has already landed. The enhanced premium tax credits that made ACA marketplace coverage affordable for more than 22 million Americans expired at the end of 2025. Congress did not renew them.

That decision - largely underdiscussed in the headlines - is now rippling through every part of the healthcare system. FQHCs are seeing it. Emergency departments are starting to feel it. And cancer screening programs, which depend entirely on patients staying covered, are about to record numbers that look a lot like the pre-ACA era.

1. How We Got Here: The Enhanced Premium Tax Credit Story

The Inflation Reduction Act of 2022 did something the original Affordable Care Act could not fully accomplish. It made marketplace coverage genuinely affordable for low- and moderate-income households. The enhanced premium tax credits it introduced reduced average premiums by approximately $800 per enrollee per year. For a family of four earning $50,000 annually, that difference is the line between enrolling and going uninsured.

The results were immediate and measurable. ACA marketplace enrollment climbed steadily from 14.5 million in 2022, to 16.3 million in 2023, to 21.4 million in 2024, reaching 22.3 million effectuated enrollees in 2025. That is not a coincidence. That is what happens when you make healthcare affordable.

📊 ACA marketplace enrollment reached 22.3 million in 2025, the highest in program history - driven entirely by the enhanced subsidy structure.

But the IRA's enhanced credits had a built-in expiration: December 31, 2025. Renewal required congressional action. The House-passed reconciliation bill did not include that renewal. ACA gross marketplace subsidy spending was cut by $131.8 billion over ten years. The credits are gone.

The enrollment data tells the story in a single number. Monthly effectuated marketplace enrollment is projected to fall to approximately 17.5 million people in 2026 - a drop of nearly 5 million people from the prior-year peak.

2. The Coverage Cliff in Real Numbers

KFF estimates that average annual marketplace premium payments will more than double for formerly subsidized enrollees - from $888 in 2025 to $1,904 in 2026. That is a 114 percent increase for people who were already carefully managing their household budgets.

📊 Marketplace premiums increased 114 percent on average for formerly subsidized enrollees in 2026 - from $888 to $1,904 annually.

The Urban Institute projects that 4.8 million more people will become uninsured as a result. That figure excludes the Medicaid impacts. When you add in the OBBBA's Medicaid cuts - which the Congressional Budget Office estimates will cause 7.6 million people to lose Medicaid - and the American Medical Association's total estimate of 11.8 million coverage losses comes into focus.

These are not abstract statistics. They are specific people in specific communities. People who earn too much for Medicaid in states without expansion. People who are self-employed contractors. People who work part-time and have no employer coverage. People who, for four years, finally had a pathway to affordable insurance. Now they do not.

📊 The AMA estimates 11.8 million Americans will lose health care coverage from combined marketplace and Medicaid policy changes.

For many, the fallback is not getting coverage elsewhere. It is going without.

3. The FQHC Pressure Cooker

When Americans lose coverage, they do not disappear from the healthcare system. They shift into it. FQHCs - federally qualified health centers - serve patients regardless of insurance status. They use sliding fee scales tied to income. They are the infrastructure of last resort for low-income populations.

That is exactly why the current policy environment is so dangerous for them.

The National Association of Community Health Centers (NACHC) estimates that combined Medicaid and marketplace coverage losses will generate approximately $7 billion per year in additional uncompensated care costs for community health centers. That is not spread evenly across a thousand facilities. That is concentrated pressure on facilities that are already operating on thin margins, in communities that are already underserved.

📊 NACHC estimates the combined coverage losses will generate $7 billion per year in additional uncompensated care for community health centers.

Up to 5.6 million FQHC patients are directly affected by the policy changes. These patients generate Medicaid revenue when insured, but when uninsured they generate sliding-fee or uncompensated care. The difference in per-visit revenue can be several hundred dollars per encounter. At scale, across millions of visits, that is an existential financial problem.

Section 330 grant funding - the federal support backbone for FQHCs - does not scale automatically to absorb coverage shocks. When CMS distributes grants, it does so based on prior-year service data and grant applications. There is no automatic adjustment mechanism for sudden surges in uninsured patient volume.

FQHC executives need to understand this is not a future risk. It is already arriving.

4. Cancer Screening Will Take a Direct Hit

The relationship between insurance status and cancer screening is one of the most documented and consistent findings in health services research. Insured patients get screened. Uninsured patients do not - not because they do not want to, but because the system creates barriers at every step.

The data is stark. For colorectal cancer screening, insured patients have a 69.7 percent up-to-date rate. Uninsured patients: 30.7 percent. That is a 39-point gap. For breast cancer screening, insured patients average 77.5 percent compliance versus 67.8 percent for uninsured. For cervical cancer, 86.2 percent versus 78.5 percent.

📊 Colorectal cancer screening compliance is 69.7 percent among insured patients versus 30.7 percent among the uninsured - a nearly 40-point gap driven directly by coverage status.

Lung cancer screening follows the same pattern. Low-dose CT screening is covered under Medicare and most ACA-compliant plans. But access depends on a referral, a primary care relationship, and an imaging center. All three are more accessible to insured patients. Uninsured high-risk smokers - exactly the population FQHCs serve - are far less likely to be identified, navigated, and scanned.

This is where the stakes become most concrete. Oatmeal Health's core mission is catching lung cancer earlier in communities that need it most. When we talk about Stage I survival at 77 percent versus Stage IV at 9 percent, we are talking about what happens when a high-risk patient gets a scan in time versus what happens when they do not.

Coverage loss does not reduce the number of people who get cancer. It reduces the number of people who find out in time to do something about it.

5. Hospital Safety-Net Strain and DSH Cuts

Hospitals serving low-income populations receive Disproportionate Share Hospital (DSH) payments to help offset uncompensated care costs. These payments were designed to shrink as ACA coverage expanded - on the logic that more insured patients would mean less uncompensated care.

That logic has now been inverted. Coverage is contracting, but DSH payments were scheduled for major reductions anyway. Federal Medicaid DSH payments face reductions of more than 50 percent in 2025, 2026, and 2027 - approximately $8 billion per year in cuts.

📊 Federal Medicaid DSH payments face reductions of more than 50 percent in 2025-2027, roughly $8 billion per year, even as uninsured patient volumes are rising.

The result is a pressure sandwich. Uninsured patients are increasing. DSH payments are decreasing. The gap falls on hospital balance sheets - and ultimately on the communities these hospitals serve.

Safety-net hospitals that cannot absorb the uncompensated care burden will face their own operational crises. Reduced services, closed units, and reduced access to specialty care are the predictable downstream consequences.

6. What the Senate Must Decide

The Senate still has a window to act. Seventeen House Republicans voted with Democrats during the reconciliation debate to extend enhanced PTCs - a signal that the political consensus against renewal is not monolithic. A Senate amendment or stand-alone bill extending the credits could reverse the trajectory.

The critical deadline is November 2026, when open enrollment for plan year 2027 begins. For insurers to price 2027 plans with enhanced subsidies, congressional action needs to come well before then - ideally by August 2026 to allow time for actuarial filings and marketplace updates.

📊 The Senate has a narrow window: action by August 2026 is needed so insurers can price 2027 plans with restored subsidies before November open enrollment.

The political calculation is shifting. As enrollment data for 2026 becomes clearer, as FQHC leaders report patient volume surges, and as hospital systems flag rising uncompensated care in quarterly reports, the real-world consequences of inaction become harder to ignore.

The question for Senate leadership is whether a partial extension - even two or three years - can attract enough bipartisan support to pass before the damage compounds.

Deep Dive: The Financial Cascade for Safety-Net Providers

The NACHC's $7 billion per year estimate deserves a closer look. It is not a worst-case projection. It is a central estimate based on the combined effect of marketplace and Medicaid coverage losses hitting FQHC patient populations.

For a mid-size FQHC serving 40,000 patients, the financial model plays out in three scenarios.

Scenario A: Enhanced PTCs Restored in 2026

If Congress acts and enhanced credits are restored for the 2027 plan year, patient coverage stabilizes. Some enrollees who dropped in the interim may not immediately re-enroll, producing a modest increase in uncompensated care - estimated at approximately $500 million nationally. Financially, most FQHCs can absorb this with existing sliding fee and grant funding.

Scenario B: Two-Year Extension, Partial Restoration

A two-year extension buys time but does not fully restore the 2025 enrollment baseline. Many of the 4.8 million who dropped will not automatically re-enroll. Outreach programs will be needed. NACHC modeling suggests additional uncompensated care of approximately $3.5 billion nationally under this scenario.

Scenario C: No Action, Cliff Sustained

Without any extension, coverage losses continue to compound through 2027 and into 2028. Work requirements kick in for Medicaid in late 2026. The combination produces the full $7 billion additional uncompensated care burden modeled by NACHC. For individual FQHCs, this represents an operating margin pressure they cannot absorb through grant offsets alone.

The difference between Scenario A and Scenario C is $6.5 billion per year - a gap that ultimately falls on the patients least able to absorb it.

What This Means For You

• FQHC executives and administrators need to model the uninsured patient surge in their specific service areas now. Do not wait for Q3 data. The leading indicators - enrollment data, sliding fee applications, urgent care volume - are available today. Build the scenario analysis before the wave arrives.

• Health system CMOs and CFOs should revisit uncompensated care reserve assumptions. If your hospital modeled 2026 DSH revenue at 2024 levels, your model is wrong. DSH cuts are scheduled, and uninsured patient volume is rising. The combination requires an updated financial plan.

• Radiologists and pulmonologists running lung cancer screening programs should expect LDCT referral volumes to dip from newly uninsured patients. High-risk uninsured patients who qualify under Medicare or state programs still need active navigation to reach care. The pathway requires effort that was not required before.

• Healthcare investors and founders should recognize that safety-net providers are entering a period of significant financial stress. Companies that reduce costs, improve billing efficiency, or close the coverage gap for underserved populations are addressing a real and growing need. The market is not contracting - it is fragmenting.

• Policy advocates have a narrow window. The August 2026 timeline for Senate action is not a guideline - it is a deadline. Advocacy organizations, hospital systems, and FQHC networks have data. The enrollment drop numbers, the uncompensated care projections, and the cancer screening gap data are documented. They need to be in front of Senate offices now, not in September.

Key References

• Urban Institute: "4.8 Million People Will Lose Coverage in 2026 If Enhanced Premium Tax Credits Expire" (September 2025) - projects marketplace coverage loss under subsidy expiration.

• KFF: "ACA Marketplace Premium Payments Would More than Double on Average" - documents the 114 percent premium increase from $888 to $1,904 annually for formerly subsidized enrollees.

• AMA: "Changes to Medicaid, the ACA and Other Key Provisions of the One Big Beautiful Bill Act" - estimates 11.8 million total coverage losses from combined Medicaid and marketplace policy changes.

• NACHC/Multiple Sources: Estimates $7 billion per year in additional FQHC uncompensated care from combined coverage losses, affecting up to 5.6 million FQHC patients directly.

• PMC/Contemporary OB-GYN and peer-reviewed literature: Documents 39-point gaps in colorectal cancer screening by coverage status (69.7 percent insured vs. 30.7 percent uninsured).

The ACA subsidy cliff is not hypothetical. It is not a projection. For millions of Americans, it is already the reality they woke up to on January 1, 2026.

The consequences are arriving in stages. First, coverage termination notices. Then, deferred care and skipped screenings. Then, advanced-stage diagnoses and emergency department visits that could have been prevented. The lag between policy change and clinical outcome is long enough that the full impact will not be visible for 12 to 18 months - but the trajectory is already locked in.

The Senate window is real. But healthcare operators cannot run their organizations on the hope of legislative action. Scenario planning, uncompensated care reserves, community outreach programs, and FQHC-embedded screening pathways need to be in place before the patient surge arrives - not after.

What is your organization's plan for 2026 uninsured patient volume? I would genuinely like to hear how health systems, FQHCs, and imaging programs are approaching this. Reply directly to this email.

About the Author

Jonathan Govette is the Co-Founder and CEO of Oatmeal Health, an AI lung cancer diagnostic company catching cancers earlier in the communities that need it most. Oatmeal uses AI to identify unscreened high-risk patients, navigate them to care, and score every lung CT for malignancy risk - billed under CPT 0721T. Stage I survival is 77%. Stage IV is 9%. We work in FQHCs because that gap is largest there.

Jonathan writes daily about radiology, pulmonology, AI diagnostics, health policy, hospital operations, and healthcare startups.

Subscribe to stay ahead of healthcare's most important shifts.

Weekly deep-dives on AI, radiology, health policy, FQHCs, and the business of healthcare - written for operators, clinicians, and investors who want the signal, not the noise.

Free subscribers get one post per month. Paid subscribers get full weekly access.

Subscribe free at jonathangovette.substack.com

Everyone is focused on work requirements. And yes, those matter enormously. More than 10 million people are projected to lose Medicaid coverage by 2034 under the One Big Beautiful Bill Act. That story deserves every column inch it is getting.

But buried inside the same legislation are Medicaid financing changes that could hit safety-net hospitals even harder than the coverage loss headline - and they are operating on a faster timeline, with less public attention, and with no relief valve built in for the hospitals most exposed.

Disproportionate Share Hospital payments - the federal and state dollars that help hospitals stay financially viable when they serve large numbers of Medicaid and uninsured patients - are being cut by $8 billion per year. Those cuts took effect October 1, 2025. By the end of this fiscal year, the damage is already accumulating. Over three years, the total reduction reaches $24 billion.

Read more

That is the share of American community health centers operating with 90 days of cash on hand or less. Not a strategic reserve. Not a planning cushion. A survival window. And in 2026, that window is being compressed from both sides at once.

The One Big Beautiful Bill Act, signed into law in July 2025, set in motion the largest structural shift to Medicaid in decades. Nearly $930 billion in federal cuts over ten years. Work requirements for adults ages 19 to 64 effective December 31, 2026. Six-month eligibility redeterminations starting January 1, 2027. For FQHCs - which depend on Medicaid for roughly 43 percent of total operating revenue - this is not a policy abstraction. It is a scheduled financial event with a known impact date.

At the same time, Congress extended the Community Health Center Fund at $4.6 billion for FY 2026 - the largest single CHCF increase in a decade. Advocates worked hard for tha…

Read more

Medicare just split physician payments into two classes. Here's which track you want to be on.

For the first time in Medicare's history, the program is operating a formally bifurcated physician payment system — one track for providers participating in advanced alternative payment models with value-based incentives, and one track for providers remaining in the traditional fee-for-service structure with quality performance adjustments. The financial gap between the two tracks is widening, and the trajectory makes clear which direction Medicare payment policy is heading.

📊 Physicians in advanced APMs — including ACOs, bundled payment models, and condition-specific value-based arrangements — receive a 3.5 percent payment bonus on top of their base Medicare rates, are exempt from MIPS quality reporting requirements, and have access to shared savings opportunities that can significantly exceed their base Medicare revenue. The combined financial benefit of APM participation — bonus, shared sa…

Read more

CMS just dropped a bombshell on Medicare Advantage plans. Here's who wins and who loses.

On April 2, CMS finalized sweeping changes to the Star Ratings methodology for Medicare Advantage plans effective for the 2027 rating year — changes that affect how plans are evaluated, how quality bonus payments are calculated, and which plans are positioned to grow or shrink their Medicare Advantage market share.

📊 Medicare Advantage Star Ratings are not an academic exercise. They are directly tied to quality bonus payments that add billions of dollars annually to MA plan revenue, and they are increasingly influential in beneficiary plan selection decisions. Plans with 4 or 5 stars receive quality bonuses and can use them to offer richer benefits that attract more enrollees. Plans that fall below 4 stars lose the bonus and the competitive advantage it funds. The stakes of the Star Rating methodology are enormous.

The 2027 changes include several significant methodological shifts. CMS is reducing t…

Read more

CMS just assembled an 18-member dream team. Here's who made the cut and what it means for healthcare policy.

On April 10, HHS and CMS announced the formation of a new Healthcare Advisory Committee — an 18-member body with representation from health systems, physician groups, payers, technology companies, patient advocates, and academic researchers. The committee's mandate is to provide ongoing input to CMS leadership on payment policy, quality measurement, and healthcare delivery transformation.

📊 Advisory committees at CMS are not new, and their influence on actual policy outcomes varies significantly depending on the political environment and leadership priorities. What distinguishes this committee from previous iterations is the explicit inclusion of technology sector representatives alongside traditional healthcare stakeholders — a composition that reflects the current administration's emphasis on technology-driven healthcare transformation.

The committee's formation under MAHA poli…

Read more

CMS just killed the clipboard. And that's just the beginning.

On April 9, CMS unveiled the first fully digital Medicare enrollment and benefits management platform — eliminating paper-based processes for beneficiary enrollment, coverage verification, and benefits navigation that have been the administrative foundation of Medicare since 1965. The launch represents the most significant modernization of Medicare administrative infrastructure in decades.

📊 The scale of the administrative burden being addressed is significant. Medicare processes approximately 60 million beneficiary enrollments, coverage changes, and benefits inquiries annually. Paper and phone-based processes that require manual data entry, physical document handling, and human intervention at multiple steps create cost, delay, and error rates that a digital platform can eliminate. The administrative cost savings from digitization — estimated by CMS at several hundred million dollars annually — are real and recurring.

For be…

Read more