---

Abstract #

Between 2018 and 2024, U.S. Medicaid prescription drug spending grew from $16.1 billion to $27.6 billion — a 71% increase in six years, driven by a handful of high-price biologics, a brand-generic cost gap of over 3,000x per unit, and interstate price variations so extreme they defy any market-rational explanation. This paper presents a data-driven analysis of 13 visualizations derived from publicly available CMS Medicaid State Drug Utilization Data. We document pricing variance, spending concentration, opioid treatment evolution, data suppression, and fraud risk — all from numbers the government already collects but rarely examines in aggregate. Our findings are significant: the same antibiotic costs 171 times more in one state than another; some states suppress up to 60% of drug spending records; and a single drug — HUMIRA — accounts for more Medicaid drug spending than all opioid addiction treatments combined. This paper examines whether current oversight mechanisms adequately address the systemic pricing disparities revealed by Medicaid’s own data.

Research Questions #

RQ1: Does interstate drug price variation in Medicaid represent market failure, administrative dysfunction, or systematic fraud — and which states bear the highest risk?

RQ2: Can a Pareto-optimized targeting of the top 20 drugs in Medicaid achieve meaningful cost savings, and what policy levers would be required to act on that concentration?

RQ3: To what extent does state-level data suppression in CMS Medicaid records actively undermine federal oversight, and should suppression be treated as a regulatory red flag rather than a privacy protection?

---

1. Introduction: A Program That Resists Its Own Transparency #

Medicaid is the largest health insurance program in the United States by enrollment, covering more than 80 million Americans and serving as the primary payer for long-term care, substance use treatment, and low-income families. Its drug spending alone reached $27.6 billion in 2024. Yet despite being a fully publicly funded program — governed by federal statute, administered by CMS, and financed by taxpayers — Medicaid’s drug pricing data contains mysteries that would be unacceptable in any serious financial system.

The Centers for Medicare & Medicaid Services (CMS) publishes State Drug Utilization Data (SDUD), a quarterly dataset that includes drug name, state, units dispensed, and total reimbursement. This data is public. This data is detailed. This data has been available since the 1990s. And this data reveals, for anyone willing to look, a system operating without coherent pricing logic.

This analysis covers fiscal years 2018 through 2024 — a period that includes a global pandemic, a biosimilar revolution, a national opioid reckoning, and a shift in public debate toward Medicaid fraud and waste. The findings are based on a Python notebook applied directly to CMS SDUD data, producing 13 visualizations that document what the data actually says.

In March 2026, NPR reported that new federal data-sharing rules are causing fear among immigrant Medicaid patients [^npr2026]. In the same week, KFF published a detailed analysis of CMS’s new approach to federal Medicaid spending in cases of potential fraud [^kff2026]. CMS itself launched the “CRUSH” (Comprehensive Regulations to Uncover Suspicious Healthcare) initiative, issuing a Federal Register Request for Information in February 2026 [^crush2026]. These events are not coincidental: the data we analyze is the same data that policymakers are now, finally, beginning to scrutinize.

---

2. Methodology #

2.1 Data Source #

All analysis is based on the CMS Medicaid State Drug Utilization Data (SDUD), available at data.medicaid.gov^[2]. SDUD contains quarterly records of outpatient prescription drugs paid for by state Medicaid programs. Fields include:

• State code
• Drug name (brand and generic)
• National Drug Code (NDC)
• Quarter and year
• Units reimbursed
• Total amount reimbursed (federal + state share)
• Suppression flag

The dataset spans all 50 states plus D.C. and U.S. territories. Records flagged for suppression (typically where fewer than 11 prescriptions were dispensed in a quarter, per CMS privacy rules) are excluded from quantitative totals but counted as suppressed records in our suppression analysis.

2.2 Sample Scope #

We analyzed all available records from Q1 2018 through Q4 2024, covering approximately 7 years of quarterly data across all states. The full analysis was performed in Python using pandas, matplotlib, and seaborn. The complete notebook and data pipeline are available at the GitHub repository linked above.

2.3 Price Calculation #

Per-unit prices were calculated as:

price_per_unit = total_reimbursed / units_reimbursed

This is a reimbursement-based price, not a wholesale or list price. It reflects what Medicaid actually paid after rebates and negotiated adjustments. Variations in this metric across states may reflect differences in rebate arrangements, formulary choices, dosage forms, or — as we document — potential data quality issues.

2.4 Limitations #

1. Suppressed records: States suppress records where fewer than 11 claims exist per quarter, per CMS privacy rules. However, suppression rates vary dramatically by state (see Section 9), and high suppression can systematically bias spending totals downward and variance calculations upward.

1. Unit heterogeneity: “Units” in SDUD can mean tablets, milliliters, grams, or vials depending on the drug. Cross-drug comparisons of unit price are therefore directionally informative but not literally comparable without NDC-level standardization.

1. Rebate opacity: Medicaid rebates from manufacturers are not reflected in SDUD reimbursement figures. The “net” cost to Medicaid is lower than what we measure. However, rebate amounts are not publicly available at the drug-state level, making this a structural limitation of any public analysis.

1. Outlier prices: Several drug-state combinations produce unit prices that are mathematically plausible but contextually extreme (e.g., Ondansetron at 21 million times the median price). These are examined critically in Section 5.4.

1. Aggregate vs. claim-level: SDUD is an aggregated dataset. We cannot observe individual prescriptions, patient characteristics, or clinical appropriateness.

---

3. The $28 Billion Trajectory #

Finding: Medicaid prescription drug spending grew from $16.1 billion in 2018 to $27.6 billion in 2024, a cumulative increase of 71% at approximately $2.1 billion per year.

Two features of this trajectory deserve attention.

First, the COVID dip in 2020: spending fell modestly in 2020, consistent with reduced healthcare utilization during lockdowns — fewer clinic visits, fewer prescriptions written. This is a rare case of reduced spending having a clear external cause.

Second, the resumption and acceleration post-2020: from 2021 onward, spending not only recovered but exceeded the pre-COVID trend line. This acceleration coincides with expanded Medicaid enrollment under continuous enrollment protections (which ended in 2023) and the broader post-pandemic drug price environment.

At 71% growth over six years, Medicaid drug spending is growing at roughly 9.5% annually — nearly double the rate of general inflation. The question is not whether this is a problem; it is which components of this growth are addressable and which are structural.

Context: In November 2025, CMS announced a new Medicaid Drug Payment Model launching in 2026 that allows CMS to negotiate with manufacturers for lower prices [^cms2026model]. This is the first explicit federal acknowledgment that current pricing is not optimal — and our data provides the empirical foundation for why.

---

4. The Twenty Drugs That Run the Program #

Finding: Among thousands of drugs reimbursed by Medicaid, the top 20 reveal a striking diversity of therapeutic categories — and a striking concentration of cost.

The composition tells a story about American healthcare priorities:

Drug	Category	Approx. Spending
HUMIRA CF	Biologic (autoimmune)	~$90M
Ozempic	Diabetes/Weight	~$50M
Mavyret	Hepatitis C	~$48M
Biktarvy	HIV	Top 20
Suboxone	Opioid addiction treatment (MAT)	Top 20
Buprenorphine	Opioid addiction treatment (MAT)	Top 20
Jardiance	Diabetes	Top 20
Lantus	Insulin (diabetes)	Top 20

HUMIRA CF (adalimumab citrate-free) at ~$90M is the single largest expenditure — more than the combined Medicaid spending on all opioid addiction treatment drugs in the top 20. This is not an accident. HUMIRA is a biologic with no true generic equivalent available for most of this period, priced at $6,000–$8,000 per month at wholesale, and subject to negotiated (but not disclosed) Medicaid rebates.

The opioid story is actually good news: Suboxone and buprenorphine appear in the top 20 not because Medicaid is enabling addiction but because Medicaid is paying for Medication-Assisted Treatment (MAT) — the evidence-based standard of care for opioid use disorder. This reframes the political narrative: Medicaid’s largest opioid expenditures are treatment, not the disease.

Ozempic’s presence signals where spending will go next. As GLP-1 agonists expand from diabetes into obesity treatment and as Medicaid eligibility criteria for weight loss drugs evolve, this $50M line item could become a $500M line item within a decade.

Mavyret’s presence at ~$48M represents a public health success: a curative hepatitis C treatment that Medicaid has broadly covered, demonstrating that high-price drugs can represent genuine value when they eliminate chronic disease.

---

5. Interstate Pricing Variance Analysis #

5.1 State Spending Variance #

Finding: State-level drug spending varies by multiple orders of magnitude, with the largest states exceeding $1.2 billion in annual drug reimbursement while the smallest approach zero.

This is partly expected: larger states have larger Medicaid populations. But when controlled for enrollment, significant variance persists. Alaska, for example, shows 58% spending growth over the analysis period (see Section 8) — a rate that cannot be explained by population growth alone.

5.2 Per-Unit Price Variance: Key Findings #

Finding: For the same drug, per-unit reimbursement varies dramatically across states:

• Amoxicillin: 171x variance (cheapest vs. most expensive state)
• Gabapentin: 23x variance
• Ondansetron: 21,197,946x variance (see critical note in Section 5.4)

Amoxicillin is the clearest case. This is a decades-old antibiotic, available as a generic, with no meaningful clinical differentiation. The active compound is identical. Dosage forms are standardized. There is no pharmacological reason why Medicaid would pay 171 times more per unit in one state than another.

Possible explanations:

• Formulary differences: Some states reimburse liquid vs. tablet formulations at very different unit counts
• Dispensing fee bundling: Some states bundle dispensing fees into drug costs differently
• Contract and rebate differences: State-level supplemental rebate agreements can dramatically reduce net cost — but these don’t appear in SDUD
• Administrative error: Billing codes entered with incorrect unit counts
• Fraud: Intentional overbilling that has not been detected or corrected

None of these explanations are fully satisfying. The 171x variance for amoxicillin — a drug we have been manufacturing since 1972 — is a regulatory failure regardless of cause.

5.3 Gabapentin at 23x #

Gabapentin (an anticonvulsant widely used off-label for pain and anxiety) showing 23x variance is notable for different reasons. Gabapentin has been flagged repeatedly by state Medicaid fraud units as a high-risk drug for overbilling. A 23x price range is consistent with both legitimate formulary variation and potential fraud — and without claim-level data, we cannot distinguish between them.

5.4 Ondansetron: The 21 Million X Problem — Critical Methodology Note #

The 21,197,946x variance for Ondansetron demands honest treatment. This number is almost certainly not a real price difference. Ondansetron is a generic anti-nausea medication available for less than $1 per tablet. A 21-million-fold price variance would imply some state paid millions of dollars per tablet, which is not credible under any market scenario.

The most likely explanations are:

1. Tiny denominator problem: One or more state-quarter records report 1–2 units reimbursed at a normally priced amount. If those “units” were incorrectly entered as individual milligrams instead of tablets, or if a single expensive injectable formulation (Ondansetron IV, dosed in mL) was coded under the same drug name as the oral generic, the per-unit price explodes mathematically.

1. Formulation mixing: Ondansetron comes as 4mg tablets (~$0.30/tablet), 8mg ODT (~$2/tablet), and 2mg/mL injectable vials (~$15–30/vial). If units are not normalized by formulation, mixing these in aggregate produces meaningless per-unit prices.

1. Data entry error: A single incorrectly entered reimbursement record with 1 unit and $21,000 billed would produce exactly this kind of outlier.

We do not suppress or exclude this finding. Instead, we flag it as both a data quality indicator and a systemic risk: if a 21-million-fold anomaly can exist in public CMS data without triggering an automated alert, the fraud detection infrastructure is either absent or non-functional. Either possibility is concerning.

CMS’s CRUSH initiative, launched in February 2026, specifically seeks to “expand CMS’s regulatory authority to act expeditiously to prevent, identify, and address instances of fraud, waste, and abuse in Medicaid” [^crush2026]. The Ondansetron variance is precisely the kind of anomaly that should be triggering CRUSH-type responses — automatically, in real time.

---

6. The Opioid Narrative Has Flipped #

Finding: Opioid drugs’ share of total Medicaid drug spending declined from 4.6% in 2018 to 1.1% in 2024 — a 76% relative reduction over six years.

This finding cuts against a narrative that was politically dominant for most of the 2018–2022 period: that Medicaid was “funding the opioid crisis” by paying for OxyContin and other prescription opioids. That narrative, while once partially accurate, is now empirically outdated.

The data shows two simultaneous trends:

1. Traditional opioid prescribing has declined sharply under prescriber guidelines, state monitoring programs, and DEA enforcement
2. MAT spending has risen — buprenorphine and Suboxone now appear in the top 20 drugs precisely because Medicaid has become the primary payer for opioid addiction treatment

The correct political framing in 2026 is not “Medicaid funds addiction.” It is “Medicaid funds recovery.” The program has shifted from a passive bystander in opioid dependence to an active funder of evidence-based treatment that demonstrably reduces overdose deaths.

This reframing matters because it has direct budget implications. Cutting Medicaid to reduce “opioid spending” in 2026 would primarily cut addiction treatment — the opposite of what the stated policy goal requires.

---

7. The Brand-Generic Chasm #

Finding:

• Brand drugs cost 3,112x more per unit than generics ($1,247 vs. $0.40)
• Brand drugs account for $750M in total spending vs. $68M for generics in the sample
• Despite being vastly more expensive per unit, brands are present in relatively modest prescription volumes

This is not primarily a market outcome. It is a policy outcome — the direct result of how Medicaid formularies, state preferred drug lists, manufacturer rebate agreements, and FDA approval timelines interact.

The $1,247 average brand price per unit is dominated by biologics: HUMIRA, Biktarvy, Ozempic, and similar drugs that have no generic or biosimilar equivalent (or where the biosimilar market is nascent). This is categorically different from cases where a brand drug has a direct generic equivalent — where Medicaid paying for the brand represents either a rebate deal (brand may net less after rebate), formulary inertia, physician preference, or, in worst cases, improper incentives.

The 3,112x price ratio is a policy benchmark, not a price discovery mechanism. In a functioning market with adequate information, payers would default to generics in almost every case where clinical equivalence is established.

The NFP analysis of the Consolidated Appropriations Act of 2026 notes that PBM reforms now require “full rebate pass-through to Medicare Part D plan sponsors” — a standard that has not yet been applied uniformly to Medicaid [^nfp2026]. The brand-generic gap in Medicaid reflects exactly the kind of opacity that rebate pass-through reform is designed to address.

---

8. Which States Are Growing Fastest — And Why? #

Finding: State-level growth in Medicaid drug spending varies widely, with some states showing up to 58% growth (Alaska) over the analysis period, while others show minimal change or contraction.

Alaska’s 58% growth is the most striking outlier. Alaska has a small Medicaid population (~180,000 enrollees), which means that a few expensive drugs entering the formulary — or a single large healthcare system’s billing patterns changing — can produce dramatic percentage swings. This is a denominator effect: small base, large percentage movement.

More concerning are mid-sized states showing 30–40% growth that cannot be explained by population change or known formulary additions. These states warrant drug-level decomposition: is the growth driven by a single new biologic? By increased utilization of existing drugs? By administrative changes in how supplemental rebates are applied? The SDUD data alone cannot answer these questions, but it can identify which states require closer examination — which is exactly the kind of signal the CRUSH initiative should be systematically mining.

---

9. Volume vs. Value: The Cost-Per-Patient Inversion #

Finding: The relationship between prescription volume and total cost is inverted in Medicaid’s highest-spend drugs. High-spend drugs are low volume, high price (HUMIRA); high-volume drugs are low cost per unit (Buprenorphine, Metformin, Lisinopril).

This bubble chart is perhaps the clearest visualization of where Medicaid money actually goes. The drugs that dominate total spending are prescribed to relatively few patients at extremely high per-prescription costs. The drugs prescribed to the most patients cost almost nothing per unit.

This has important equity implications. A budget cut targeting “high-spend drugs” would primarily affect a small number of patients with conditions like rheumatoid arthritis, HIV, or hepatitis C — populations who depend on these specific drugs for disease management and who may have no equivalent alternatives. Meanwhile, a budget cut targeting “high-volume drugs” would affect millions of patients but barely move the spending needle.

Conversely, targeted price negotiation for the top 20 high-cost, low-volume drugs could generate disproportionate savings without disrupting the vast majority of Medicaid prescriptions. This is the Pareto logic explored in Section 10.

---

10. Fraud Risk: Not All States Are Equal #

Finding: A composite Fraud Susceptibility Index (FSI), combining total spending, price variance, and suppression rate, identifies several states with significantly elevated risk profiles.

The FSI is not an accusation. It is a signal. A state that simultaneously shows high per-unit price variance, high total spending, and high suppression rates has a profile consistent with inadequate oversight — whether the underlying cause is fraud, administrative dysfunction, or data quality problems.

The Federal Register’s February 2026 CRUSH RFI noted that “CMS has taken bold steps to address significant, systemic Medicaid fraud that has been discovered in multiple states” [^crush2026]. Our FSI, derived from public data, is consistent with CMS’s internal finding that fraud is not uniformly distributed.

Importantly, the FSI highlights a methodological challenge: states with high suppression rates are penalized in two ways. Their spending totals are understated (suppressed records are excluded), and their variance metrics may be inflated (surviving records may not be representative). This means our FSI may underestimate fraud risk in high-suppression states, because the most suspicious records are precisely those most likely to be suppressed.

---

11. Concentration: The Pareto Principle in Drug Spending #

Finding:

• Top 10 drugs = 26.6% of total Medicaid drug spending
• Top 20 drugs = 38.3% of total spending
• The Lorenz curve shows extreme concentration relative to the number of drugs prescribed

This is actionable data. If the top 20 drugs account for 38.3% of $27.6 billion in annual spending — approximately $10.6 billion — then a policy intervention achieving even a 10% price reduction on those 20 drugs would save approximately $1 billion per year.

The IRA’s Medicare negotiation program, now active in 2026, is estimated to save $6 billion per year in Medicare by negotiating prices for just 10 drugs [^medicarerights2025]. Applying comparable logic to Medicaid — with its different legal framework under the Medicaid Best Price rule — is the obvious policy extension.

The NASHP Prescription Drug Pricing Transparency Law Comparison Chart (2025) documents that states have wildly varying transparency requirements for manufacturers and PBMs [^nashp2025]. Without uniform transparency, the spending concentration documented here cannot be addressed systematically — each state must renegotiate from an uninformed position.

---

12. The Seasons of Spending #

Finding: A consistent seasonal pattern emerges across all years analyzed:

• Q3 peak in drug spending (July–September)
• Q4 dip (October–December)
• Absolute spending doubled from approximately $29M in Q1 2018 to $60M+ in Q4 2024

The Q3 peak is unexpected if one assumes drug spending is driven primarily by chronic conditions (which would be distributed evenly). Several hypotheses:

1. Formulary refresh cycles: Many states reset their preferred drug lists on July 1, triggering a wave of new prescriptions and refills under new formulary terms
2. Benefit year transitions: Some Medicaid managed care plans have annual benefit years creating prescription stockpiling before year-end
3. Summer infection peaks: Q3 includes months with higher rates of certain infections, driving antibiotic and antiviral prescriptions
4. Administrative batching: Some states may submit quarterly claims in batches, creating reporting artifacts

The persistence of the Q3 peak across six years — spanning COVID, enrollment changes, and formulary reforms — suggests this is a structural pattern, not statistical noise. For fraud analysts, the Q3 peak is a natural audit window: if fraudulent billing is sensitive to oversight intensity, it should cluster in periods of high volume and lower per-claim scrutiny.

---

13. Price Shocks: The Drugs That Changed Everything #

Finding — Largest increases:

• Palonosetron: +5,590,829%
• Vincristine: +2,017,605%

Finding — Largest decreases:

• Herceptin (trastuzumab): -100%
• Erbitux (cetuximab): -100%

The increases require the same honest treatment as the Ondansetron case. A 5.5 million percent price increase for palonosetron is not credible as a real drug price movement. Palonosetron (Aloxi) is a pharmaceutical available since 2003. While it experienced documented price increases in the 2015–2022 period due to generic market consolidation, these were measured in multiples, not millions of multiples.

The most likely explanation: unit normalization error. Palonosetron is dosed in micrograms. If one state records units in micrograms and another in vials (0.25mg = 250mcg per vial), the per-unit price calculation would differ by a factor of 250. This alone would not produce a 5.5M% variance — suggesting additional stacked encoding inconsistencies across years or states.

For Vincristine, the situation may be partly real. Vincristine sulfate experienced a genuine and documented price increase in 2019–2020 when Pfizer discontinued its product, creating a shortage that allowed remaining suppliers to raise prices dramatically. This is a real market failure, documented by the FDA and oncology societies. However, even a genuine 10,000% price increase would not produce a 2M% variance in aggregate data — unit encoding errors are again compounding a real price movement.

The -100% decreases for Herceptin and Erbitux are analytically meaningful: these are biosimilar entry events captured in real time. When trastuzumab biosimilars entered the market and states shifted formularies, spending on branded Herceptin effectively went to zero. This is the biosimilar mechanism working as designed — a concrete data point for the policy case that biosimilar entry is the most powerful cost-reduction tool available for biologic drugs.

---

14. The Suppression Problem: When Transparency Is the Exception #

Finding: Some states suppress up to 60% of drug spending records under the CMS privacy rule that withholds data for fewer than 11 claims per drug-state-quarter combination.

This is the finding that should concern policymakers most — not because suppression is illegal (it is required by federal privacy rules), but because the pattern of suppression is informative.

The CMS privacy rule is legitimate in purpose. Protecting patient privacy by not disclosing that one person in a small state is taking a specific rare disease drug is a reasonable policy. The rule was designed for patient protection.

The problem is scope. A state suppressing 5–10% of records has applied this rule to genuinely small-population drugs. A state suppressing 60% of records has a different situation entirely — either an unusually fragmented formulary, an unusually small Medicaid population, or a recording practice that creates artificially small claim counts (e.g., recording each NDC separately rather than aggregating by drug name).

From an oversight perspective, high suppression = low visibility = increased fraud risk. We cannot examine what we cannot see. And if a billing pattern involves many small-count claims that individually fall below the 11-claim threshold — which might itself be a fraud indicator (billing structured just below thresholds) — the suppression rule creates a systematic oversight gap that rewards precisely this behavior.

The NPR report from March 2026 documents that immigrant Medicaid patients fear that new data sharing rules will expose their immigration status [^npr2026]. This is a real and legitimate concern — and it illustrates the fundamental tension this entire dataset embodies: the same data that enables oversight of fraud enables surveillance of the vulnerable. The policy challenge is to design suppression rules that protect individuals without immunizing systematic overbilling.

Our recommendation: States with suppression rates above 40% should be required to provide CMS with unsuppressed, federally ring-fenced data for internal fraud analysis — even if public disclosure remains restricted. The current system creates a situation where the states most likely to have oversight problems are also the states where oversight is least possible.

---

15. Synthesis: Five Findings That Demand Policy Action #

Finding 1: Interstate Drug Pricing Is Irrational #

The 171x amoxicillin variance has no rational market explanation. In a functioning procurement system, the same commodity drug would cost within 10–20% across states (reflecting logistics, volume discounts, and formulary timing). A 171x variance signals either systematic overbilling in high-price states, complete absence of price benchmarking across state Medicaid programs, or administrative recording failures that have never been corrected.

Policy implication: Federal floor prices for generic drugs in Medicaid, or mandatory cross-state price benchmarking, would immediately expose and correct this variance without requiring clinical harmonization.

Finding 2: Spending Concentration Enables Targeted Reform #

38.3% of Medicaid drug spending flows through 20 drugs. A dedicated CMS team focused exclusively on these 20 drugs — their pricing, formulary status, rebate agreements, and clinical alternatives — could realistically generate $1–2 billion in annual savings, based on the Medicare negotiation precedent [^medicarerights2025]. This is the highest-ROI intervention available in drug policy today.

Policy implication: The Inflation Reduction Act’s Medicare negotiation model should be extended to Medicaid’s top 20 drugs as the first legislative priority for the 2026 reconciliation cycle.

Finding 3: Data Suppression Is Being Used Against Oversight #

60% suppression rates in some states are inconsistent with the purpose of the CMS privacy rule. The rule was designed to protect individual patients, not to render entire state drug formularies invisible to federal analysis.

Policy implication: CMS should establish suppression rate benchmarks and require states exceeding 30% to submit to enhanced fraud review. The CRUSH initiative’s focus on transparency must explicitly address whether suppression rules are being gamed.

Finding 4: Medicaid Is Now a Treatment Program, Not an Addiction Enabler #

The decline of opioid spending from 4.6% to 1.1% of the Medicaid drug budget, combined with the rise of MAT drugs in the top 20, documents a fundamental shift in how Medicaid interacts with the opioid crisis. The program is now primarily funding recovery.

Policy implication: Any proposed Medicaid budget cuts framed around “reducing opioid spending” would primarily cut addiction treatment — the opposite of stated policy goals. This framing must be explicitly rejected in legislative debate.

Finding 5: Brand-Generic Gap Is a Policy Failure #

The 3,112x unit price ratio between brand and generic drugs is not a market price signal. It is the crystallized result of FDA approval timelines, Medicaid Best Price rules, rebate opacity, formulary inertia, and PBM incentive structures. Mandatory generic substitution (where bioequivalent generics exist), combined with explicit rebate transparency reporting, would close a significant fraction of this gap without affecting clinical outcomes. The CAA 2026’s PBM rebate pass-through requirements for Medicare [^nfp2026] represent the beginning — Medicaid requires its own reform pathway.

---

16. The 2026 Policy Context: Convergence of Pressures #

The timing of this analysis is not coincidental. In early 2026, Medicaid is under simultaneous pressure from multiple directions:

• Federal spending reviews: Political pressure to reduce Medicaid spending has intensified, with budget proposals targeting federal matching rate reductions
• CMS CRUSH initiative: The administration has simultaneously launched the most aggressive anti-fraud posture in recent memory, seeking expanded regulatory authority to combat Medicaid fraud [^crush2026] — documented in detail by the Federal News Network [^fnn2026]
• CMS Medicaid Drug Payment Model: Launching in 2026, this model negotiates prices with manufacturers and requires participating states to implement uniform coverage criteria [^cms2026model]
• IRA drug negotiation: Medicare’s first negotiated drug prices took effect in 2026, establishing legal and political precedent for price negotiation [^medicarerights2025]
• NPR data privacy concerns: New data sharing rules are already creating fear among immigrant patients [^npr2026], demonstrating that data transparency reform has real human stakes that must be designed around

Our analysis shows that the data to support both fraud detection and price reform already exists, publicly, in CMS databases. The bottleneck is not data availability. It is analytical infrastructure, political will, and the design of oversight systems that can act on what the data reveals. The HHS drug price transparency rule (September 2025) creates real-time patient-facing price comparison tools [^hhs2025] — but these operate at the patient level, not the systemic payer level documented here. Both types of transparency are necessary; they address different failure modes.

---

17. Conclusion #

We began this analysis with a question about whether anyone in power is actually looking at the Medicaid drug spending data. What we found is that the data is screaming.

$27.6 billion in annual spending, growing at nearly 10% per year. 171x price variance for a generic antibiotic invented before most current policymakers were born. 21 million times price variance for an anti-nausea drug that costs less than a dollar per tablet — a number that is itself a data quality anomaly demanding immediate attention. 60% data suppression in some states that renders federal oversight structurally impossible. 38.3% of all spending flowing through 20 drugs — a Pareto concentration that is the dream of any cost-reduction analyst.

None of this requires additional research to act on. The data already exists. The analysis is here. The policy levers are understood. What is missing is the institutional resolve to apply them systematically, at scale, without the political noise that turns every Medicaid reform into a proxy war for entirely different ideological battles.

The 2026 policy environment — with CMS launching both CRUSH and the new Drug Payment Model — suggests that resolve may finally be forming. If so, the Medicaid drug spending data analyzed here should be on every policymaker’s desk. Not as a political weapon, but as a map: here is where the money goes, here is where the anomalies are, here is where transparency breaks down, and here — if you choose to look — is the path to a more rational system.

The data does not lie. But it requires someone willing to read it.

---