FLORIDA AGENCY FOR HEALTH CARE ADMINISTRATION STATE OF FLORIDA v. ADMINISTRATOR FOR THE CENTERS FOR MEDICARE MEDICAID SERVICES SECRETARY DEPARTMENT OF HEALTH AND HUMAN SERVICES UNITED STATES OF AMERICA THE CENTERS FOR MEDICARE MEDICAID SERVICES DEPARTMENT OF HEALTH AND HUMAN SERVICES (2025)

Medicaid, the government-sponsored health-insurance program for those with limited resources or special medical needs, has been described as an exercise in “cooperative federalism.” Harris v. McRae, 448 U.S. 297, 308 (1980) (quoting King v. Smith, 392 U.S. 309, 316 (1968)). The program is jointly funded and administered by the states and the federal government. This case presents a question about how states may—and may not—foot their share of the Medicaid bill. For years, Florida has raised a substantial portion of its Medicaid contribution through an intricate arrangement called the “Directed Payment Program.” In 2023, the federal Centers for Medicare and Medicaid Services released a “Bulletin” that, while innocuous in name, was potentially dramatic in effect. The Bulletin advanced an understanding of federal law that Florida feared could jeopardize its Directed Payment Program and the billions in Medicaid funds it generated. Florida sued the federal government and moved to preliminarily enjoin the Bulletin's implementation. The district court denied Florida's motion and dismissed the case on the ground that the Bulletin wasn't a reviewable “final agency action” within the meaning of the Administrative Procedure Act.

We hold that the district court was right to deny Florida's preliminary-injunction motion but that it did so for the wrong reason. Contrary to the district court's conclusion, the Bulletin was a final agency action, and is therefore subject to judicial review. But Florida is unlikely to succeed on the merits of its challenge to the Bulletin and, accordingly, isn't entitled to the preliminary injunction it seeks. We therefore reverse the district court's dismissal of the complaint, affirm its denial of the preliminary injunction, and remand for further proceedings.

I

A

Medicaid funding is a group project. The federal government provides financial assistance “to states that choose to reimburse certain costs of medical treatment for needy applicants.” Ga., Dep't of Med. Assistance ex rel. Toal v. Shalala, 8 F.3d 1565, 1566 (11th Cir. 1993). A complex matching formula generates the “[f]ederal medical assistance percentage” in any particular state. 42 U.S.C. § 1396d(b). The federal share is always some percentage of a state's Medicaid expenditures—never less than 50%, but often higher, see id. § 1396d(b)(1). So the more of its own money a state spends on Medicaid, the more the federal government has to kick in.

As in many group projects, the participants’ incentives aren't always aligned. Because of the way the federal matching formula works, states are understandably motivated to inflate their own Medicaid spending numbers as a means of driving up federal contributions. One way states can do this—without internalizing the budgetary costs, so to speak—is by imposing so-called “provider taxes.” See U.S. Gov't Accountability Off., GAO-21-98, Medicaid: CMS Needs More Information on States’ Financing and Payment Arrangements to Improve Oversight 8–10 (2020). In broad strokes, here's how it works: (1) A state imposes a tax on healthcare providers; (2) the state uses those tax revenues to increase its own Medicaid contribution; (3) by virtue of its matching obligation, the federal Medicaid contribution likewise increases; and (4) those Medicaid dollars—now including the extra federal share—get returned to the same tax-paying providers through more generous Medicaid payments. A counterintuitive consequence of all this is that healthcare providers have an incentive to—and in fact do—lobby state and local governments to impose higher taxes. That's because although institutions subject to the provider tax will have to pay bigger tax bills, they expect to receive even more back once they get their Medicaid payments, as supplemented with an inflated federal share. In recent years, states have started relying more heavily on provider taxes to finance their Medicaid contributions, which, in turn, has driven dramatic increases in mandatory federal Medicaid outlays. See id. at 22–34.

In 1991, Congress adopted an important limit on states’ ability to juice federal contributions in this way—the so-called “hold harmless” rule. See 42 U.S.C. § 1396b(w)(1)(A)(iii); Medicaid Voluntary Contribution and Provider-Specific Tax Amendments of 1991, Pub. L. No. 102-234, § 2(a), 105 Stat. 1793, 1793–99. Under § 1396b(w), the Department of Health and Human Services must deduct from its federal-matching calculation certain kinds of state revenue. Chief among these mandatory deductions is any funding that the state obtained through “health care related taxes.” Id. § 1396b(w)(1)(A), (3)(A). On its own, that provision would require the exclusion of revenue raised through the sort of provider taxes that we've described. But the “health care related tax[ ]” ban has an exception: States may raise revenue through a healthcare-related tax so long as it is “broad-based.” Id. § 1396b(w)(1)(A)(ii). But, it turns out, the “broad-based” tax exception has its own exception: HHS must nonetheless deduct revenue from even a broad-based tax “if there is in effect a hold harmless provision (described in paragraph (4)) with respect to the tax.” Id. § 1396b(w)(1)(A)(iii).

The cross-referenced Paragraph (4) sets out four different ways of detecting whether a “hold harmless provision” is in effect. See id. § 1396b(w)(4). We'll return to all four in due course, but for now it's important to know about the method described in subparagraph (C), clause (i). According to (C)(i)—and the case basically rides on this language—“there is in effect a hold harmless provision with respect to a broad-based health care related tax” if “[t]he State or other unit of government imposing the tax provides (directly or indirectly) for any payment, offset, or waiver that guarantees to hold taxpayers harmless for any portion of the costs of the tax.” Id. § 1396b(w)(4)(C)(i). This much is clear and undisputed: By virtue of clause (C)(i), a state can't impose a provider tax on hospitals while also promising to promptly refund (i.e., “hold harmless”) those very same hospitals. Were things otherwise, states could secure practically unlimited extra federal matching funds using revenue that is little more than a short-term loan from the hospitals.

HHS's Centers for Medicare and Medicaid Services (or CMS) is concerned that states and providers have developed a sophisticated way to evade the hold-harmless rule. Things get pretty involved, so we'll try to simplify it using an illustrative example: Imagine that a (very small) state has four private hospitals, each of which provides care to a different number of Medicaid-insured patients. Hospital 1 opts not to take Medicaid insurance at all, so its Medicaid share is 0%. Hospitals 2 and 3 accept Medicaid, but see few low-income patients—let's say their Medicaid share is 20%. Hospital 4 both accepts Medicaid and receives many low-income patients—its Medicaid share is 80%. The state opts to boost its own Medicaid expenditures—and thus the federal match—by imposing a provider tax on all four hospitals. This is a “broad based” tax, so it doesn't get deducted from the federal-share calculation.

The tax is good for Hospital 4 and bad for Hospital 1. Hospital 4 will pay the tax but, in the end, will receive higher Medicaid payments; Hospital 1, by contrast, has to pay the tax but won't get anything in return. Hospitals 2 and 3, obviously are somewhere in between. The state could try to keep everyone happy by simply refunding the hospital tax—but that the hold-harmless rule prevents it from doing so. CMS worries that hospitals will end-run the hold-harmless rule's prohibition by redistributing their Medicaid payments among themselves as a means of offsetting the provider tax. So, for instance, Hospital 4 might transfer some of its new, provider-tax-enabled Medicaid-payment surplus to Hospital 1 to cover Hospital 1's provider-tax assessment. That way, even though both Hospital 1 and Hospital 4 are paying a new tax, they're both better off.

This case tests whether the hold-harmless rule prevents these sorts of private-to-private redistributions.

B

In its ongoing efforts to implement the hold-harmless rule, CMS has taken three regulatory steps that turn out to be important to our analysis.

First, in 2008, CMS revised its hold-harmless regulations, partly in response to a few congressional tweaks. See 2008 Final Rule, 73 Fed. Reg. 9,685 (Feb. 22, 2008) (codified at 42 C.F.R. §§ 433.54–433.70); see also Tax Relief and Health Care Act of 2006, Pub. L. No. 109-432, div. B, title IV, § 403, 120 Stat. 2922, 2994–95 (amending 42 U.S.C. § 1396b(w)(4)(C)). Neither the statutory amendment nor the agency's revisions materially altered the (C)(i) definition's text or its implementing regulations. Compare 42 C.F.R. § 433.68(f)(3) (1993), with id. § 433.68(f)(3) (2008). But in the rule's preamble, CMS explained that “[a] direct guarantee [under (C)(i)] would be found when a State payment is made available to a taxpayer or a party related to the taxpayer ․ in the reasonable expectation that the payment would result in the taxpayer being held harmless for any part of the tax.” 73 Fed. Reg. at 9,686; accord id. at 9,694. In CMS's view, “[a] direct guarantee does not need to be an explicit promise or assurance of payment”; “[i]nstead, the element necessary to constitute a direct guarantee is the provision for payment by State statute, regulation, or policy.” Id. at 9,694.

Second, in 2019, CMS proposed another revision to its hold-harmless regulations. See Medicaid Fiscal Accountability Regulation, 84 Fed. Reg. 63,722 (proposed Nov. 18, 2019). The proposed rule's preamble reiterated CMS's 2008 interpretation. See id. at 63,730–31, 63,734. As CMS had said then, in its view a hold-harmless arrangement exists if “taxpayers have a reasonable expectation that their forthcoming Medicaid payment (including any redistribution) ․ results in participating taxpayers being held harmless for all or a portion of the tax amount.” Id. at 63,734. But, CMS continued, it had “become aware” of some novel—and it believed unlawful—arrangements in which taxpaying providers were being “held harmless” by way of payments not from state or local governments, but rather from Medicaid Managed Care Organizations or other taxpayers. Id. CMS insisted that “[t]he fact that a private entity makes the redistribution payment does not change the essential nature of the payment, which constitutes an indirect payment from the state or unit of government to the entity being taxed that holds it harmless for the cost of the tax.” Id. at 63,735. Put simply, CMS said, the sort of taxpayer-to-taxpayer redistributions exemplified in our hospital hypo are forbidden. CMS explained that in its view “[s]uch arrangements” were “inconsistent with existing statutory and regulatory requirements prohibiting hold harmless arrangements.” Id. at 63,734 (emphasis added). Though CMS was proposing regulatory amendments, the changes, it said, did “not reflect any change in policy or approach, but merely codif[ied] currently prohibited practices.” Id. at 63,735.

Third, in 2021, CMS withdrew its 2019 proposed rule. See 2021 Proposed Rule Withdrawal, 86 Fed. Reg. 5,105 (Jan. 19, 2021). The terse notice explained that “based on the considerable feedback we received ․, we have determined it appropriate to withdraw the proposed provisions at this time.” Id. at 5,105. The notice explained that “the withdrawal of this proposed rule does not affect existing federal legal requirements or policy that were merely proposed to be codified in regulation.” Id.

C

The story underlying this case begins in earnest two years later, in 2023, when CMS circulated a document that it called an “Informational Bulletin.” See CMCS Informational Bulletin, Ctrs. for Medicare & Medicaid Servs., Health Care-Related Taxes and Hold Harmless Arrangements Involving the Redistribution of Medicaid Payments (Feb. 17, 2023), https://perma.cc/VW7XGX8V. In the Bulletin, CMS explained that it had “been approached by several states with questions regarding the statutory and regulatory requirements applicable to health care-related taxes.” Id. at 1. Citing the 2008 rule's preamble, the Bulletin insisted that the hold-harmless prohibition covers purely private redistribution arrangements because they “result[ ] in a reasonable expectation that the taxpaying hospitals, whether directly through their Medicaid payments or due to the availability of the redistributed payments ․, are held harmless for at least part of their health care-related tax costs.” Id. at 3–4. And the Bulletin warned that, should CMS discover forbidden private arrangements, the agency could “disallow” (i.e., claw back) a portion of the offending state's Medicaid funds. Id. at 5.

The Bulletin spelled trouble for Florida's Directed Payment Program, under which local governments can assess special fees on private hospitals. See Fla. Stat. § 125.01(1)(r). The governments pool those assessments in “Local Provider Participation Funds,” which they then transfer to the Florida Agency for Healthcare Administration. The Agency, in turn, uses that money to supplement Florida's Medicaid contribution, which, for reasons already explained, enables the state to collect additional federal matching funds. The extra federal funds in hand, the Agency redistributes the combined monies to entities known as “Managed Care Organizations,” which, in turn, return them to Florida hospitals.

At least 20 Florida counties and municipalities assess fees through the Directed Payment Program. One is Broward County, which imposes a special assessment on private hospitals. See Broward County, Fla., Code §§ 16-123 to -137 (2025). Assessments like Broward County's qualify as “broad based” taxes because (to simplify just a bit) they apply uniformly “to all providers in a class” in “the area over which the unit of government has jurisdiction.” 42 C.F.R. § 433.68(c)(2); see Broward County, Fla., Code § 16-127 (2025) (requiring “[t]he Assessment [to] be broad based”). And because the assessments are broad-based, they are presumptively permitted under § 1396b(w). According to Broward County's ordinance, local hospitals “requested enactment” of the assessment. Broward County, Fla., Code § 16-137 (2025).

The Directed Payment Program has been lucrative. In fiscal year 2023–2024, Florida expected to raise $1.1 billion in nonfederal Medicaid funding through Local Provider Participation Funds, which it planned to use to secure another $2 billion from the federal government. See Wallace Decl. ¶ 15, Dkt. No. 10-1. According to the state, this money is “integral to Florida's healthcare system.” Br. of Appellants at 6. Though the state doesn't concede that private redistribution arrangements exist in Florida, it acknowledges that CMS might determine that they do—and fears that those billions in provider-tax-based funds could therefore be at risk. See Wallace Decl. ¶ 28.

The state has cause for concern. A few months before it issued the Bulletin, CMS sent a letter to the state expressing concern that it might be harboring private hold-harmless arrangements and promising a future review of the Direct Payment Program. See Resp. in Opp'n Ex. A, Dkt. No. 22-2. Then, just days after releasing the Bulletin, CMS followed through and commenced a Financial Management Review of Local Provider Participation Funds. See Compl. Ex. B, at 1, Dkt. No. 1-8. According to the review letter, CMS asserted that there “appear to be pre-arranged agreements to redirect Medicaid payments away from Medicaid providers serving a high percentage of Medicaid beneficiaries to hospitals that do not participate in Medicaid or that serve a low percentage of Medicaid beneficiaries.” Id. at 2.

Displeased with the Bulletin, Florida sued under the Administrative Procedure Act. It alleged that the Bulletin misinterpreted the hold-harmless rule, that it was arbitrary and capricious, and that CMS had improperly issued it without the requisite notice and comment. The state then moved for a preliminary injunction, and CMS moved to dismiss.

Acting on the recommendation of a magistrate judge, the district court denied Florida's preliminary-injunction motion and granted CMS's motion to dismiss. The court based both decisions on its conclusion that the Bulletin wasn't a “final agency action” within the meaning of the APA's judicial-review provision, 5 U.S.C. § 704, and, therefore, that it lacked jurisdiction.

This is Florida's appeal.¹

II

We begin with jurisdiction. According to CMS, the Bulletin wasn't a final agency action and Florida's suit isn't ripe. We disagree.

A

The Administrative Procedure Act authorizes judicial review of “final agency action.” 5 U.S.C. § 704. In this Circuit, finality is jurisdictional. Nat'l Parks Conservation Ass'n v. Norton, 324 F.3d 1229, 1236 (11th Cir. 2003). Two conditions “generally must be satisfied for agency action to be ‘final’ under the APA.” U.S. Army Corps of Eng'rs v. Hawkes Co., 578 U.S. 590, 597 (2016). The action must (1) “mark the consummation of the agency's decisionmaking process” and (2) “be one from which rights or obligations have been determined, or from which legal consequences will flow.” Bennett v. Spear, 520 U.S. 154, 177–78 (1997) (citation modified). Nobody here disputes prong one—that the Bulletin consummates CMS's decisionmaking.² The disagreement is instead about prong two—whether the Bulletin determines “rights or obligations” or portends “legal consequences.” We'll lay out the doctrine and then apply it to the Bulletin.

1

CMS insists that the Bulletin is merely a “restatement of an already-existing policy or interpretation” of the sort that we (and others) have said “does not, on its own, determine any rights or obligations and imposes no legal consequences.” Clayton Cnty. v. FAA, 887 F.3d 1262, 1266–67 (11th Cir. 2018); accord, e.g., Indep. Equip. Dealers Ass'n v. EPA, 372 F.3d 420, 427 (D.C. Cir. 2004) (Roberts, J.) (observing that when an agency “merely restate[s] in an abstract setting—for the umpteenth time—[its] longstanding interpretation,” that restatement is not a final agency action); Gen. Motors Corp. v. EPA, 363 F.3d 442, 449–50 (D.C. Cir. 2004) (similar). The reason for that rule is clear enough: When an aggrieved party seeks judicial review of a mere restatement, it is, in effect, suing about the wrong thing—it should have challenged whatever action the agency took in which it first adopted the complained-of view. See Gen. Motors, 363 F.3d at 451.

But in arguing that the Bulletin is such a restatement, CMS seems to conflate finality with an altogether different doctrine—namely, about whether a particular agency action is “interpretive” or “legislative.” Under the APA, when an agency introduces a new rule, it must generally comply with the three-step notice-and-comment process. See 5 U.S.C. § 553; Perez v. Mortg. Bankers Ass'n, 575 U.S. 92, 96 (2015) (describing the steps). But there are exceptions: The notice-and-comment requirement doesn't apply to “interpretative rules, general statements of policy, or rules of agency organization, procedure, or practice.” 5 U.S.C. § 553(b)(A). We'll have reason to return to the vexing interpretive-legislative distinction later. For now, it's enough to know “that the critical feature of interpretive rules is that they are issued by an agency to advise the public of the agency's construction of the statutes and rules which it administers.” Perez, 575 U.S. at 97 (citation modified). While legislative rules “create[ ] new law, rights, or duties,” interpretive rules “only remind[ ] affected parties of existing duties.” Warshauer v. Solis, 577 F.3d 1330, 1337 (11th Cir. 2009) (citation modified).

As we understand it, CMS is making roughly the following argument: (1) Unlike legislative rules, interpretive rules merely reiterate existing obligations, id. at 1337; (2) when an agency merely reiterates existing obligations, its action doesn't have direct and appreciable legal consequences, see Clayton Cnty., 887 F.3d at 1266–67; (3) when an agency action doesn't have direct and appreciable legal consequences, it doesn't constitute final agency action, Bennett, 520 U.S. at 178; (4) courts lack jurisdiction to review non-final agency actions, Norton, 324 F.3d at 1236. From those premises, CMS seems to say, the conclusion follows that interpretive (i.e., non-legislative) rules aren't subject to judicial review. And because, CMS adds, the Bulletin is an interpretive rule, it isn't subject to judicial review.

Clever, but not right. In Perez v. Mortgage Bankers Association, the Supreme Court implicitly recognized that interpretive rules can be subject to judicial review. See 575 U.S. 92, 105–06 (2015). According to the decision there, “interpretive rules do not have the force and effect of law,” id. at 103—and yet, the Court said, should an agency revise a preexisting interpretive rule by way of a second interpretive rule, the second rule would be subject to APA review, id. at 105–06. Perez therefore “affirms that interpretive rules can be final” and importantly, “by implication, that the test for finality is independent of the analysis for whether an agency action is a legislative rule rather than an interpretive rule.” Cal. Cmtys. Against Toxics v. EPA, 934 F.3d 627, 635 (D.C. Cir. 2019).

CMS's deduction falters in premise No. 2. While the distinction between legislative and interpretive rules turns on a formal analysis of whether an agency really is simply interpreting the underlying statute, the distinction between final and non-final action is pragmatic. Contrast Perez, 575 U.S. at 97 (explaining that “interpretive rules” are “issued by an agency to advise the public of the agency's construction of the statutes and rules which it administers” and “do not have the force and effect of law” (citation modified)), with Hawkes, 578 U.S. at 599 (describing “the pragmatic approach we have long taken to finality” (citation modified)). Sure, an interpretive rule might accurately reflect a statute's best reading—but it can also be final because it triggers new real-world rights or obligations.

So, with an interpretive rule—as with any challenged agency action—we should inquire whether, as a practical matter, and given the particular statutory context, the rule has “direct and appreciable legal consequences.” Hawkes, 578 U.S. at 598 (quoting Bennett, 520 U.S. at 178). Or, as our D.C. Circuit colleagues have put it, we should “pragmatically focus on the concrete consequences the action has or does not have as a result of the specific statutes and regulations that govern it.” POET Biorefining, LLC v. EPA, 970 F.3d 392, 405 (D.C. Cir. 2020) (citation modified); cf. Appalachian Power Co. v. EPA, 208 F.3d 1015, 1021 (D.C. Cir. 2000) (observing that “an agency's [non-legislative-rule] pronouncements can, as a practical matter, have a binding effect”).

2

The Bulletin has practical, concrete legal consequences. To be sure, the Bulletin purports to “reiterate[ ]” CMS's preexisting position, as stated in the 2008 rule's preamble. Bulletin, supra, at 2–4. But it portends investigation and enforcement. It promises (ominously) that CMS will “inquire about potential redistribution arrangements and may conduct detailed financial management reviews of health care-related tax programs that appear to include redistribution arrangements.” Id. at 5. It insists that states must “make clear to their providers” that private redistribution arrangements are “not permissible.” Id. It announces that “CMS expects states to make available” “detailed information ․ regarding their health care-related taxes”—apparently including information about private-only arrangements. Id. It threatens clawing back Medicaid funding through “deferral or disallowance of federal financial participation.” Id. And it warns that, should CMS “discover[ ] the existence of impermissible financing practices,” it “will take enforcement action as necessary.” Id. The 2008 rule's preamble included none of these portents.

The Bulletin's demands that states make particular information available and notify healthcare providers of updated expectations have the ring of “appreciable legal consequences.” Hawkes, 578 U.S. at 598. Perhaps even more so, the Bulletin's discussion of deferral or disallowance—which could put billions in Medicaid funding at risk—signals final agency action because it could “force[ ]” Florida “to choose between costly compliance and the risk of prosecution at an uncertain point in the future.” Racing Enthusiasts & Suppliers Coal. v. EPA, 45 F.4th 353, 358 (D.C. Cir. 2022) (citation modified).

Three subsequent developments bolster our conclusion that the Bulletin constitutes final agency action. First, shortly after circulating the Bulletin, CMS initiated its Financial Management Review of Florida's Local Provider Participation Funds—and suggested that the state was harboring private hold-harmless arrangements. Second, more than a year after issuing the Bulletin, CMS circulated a new document announcing its intention to exercise its enforcement discretion not to take immediate action against preexisting private redistribution arrangements of the sort described in the Bulletin. See CMCS Informational Bulletin, Ctrs. for Medicare & Medicaid Servs., Exercise of Enforcement Discretion (Apr. 22, 2024), https://perma.cc/6ZE5-Y6DY. And third, CMS and Florida recently executed a new agreement—as a precondition of federal approval of one of the state's Medicaid Managed Care Programs—that imposes a new information-collection requirement: Florida must provide to CMS a copy of any agreements “regarding any arrangement among [healthcare] providers ․ relating to each locality tax or payments received that are funded by the locality tax.” Special Terms and Conditions, Florida Managed Medical Assistance Program 55 (Jan. 31, 2025), https://perma.cc/XZ5Z-JKSK. The point isn't that any of these documents is necessarily final agency action on its own; rather, the point is that together they serve to confirm that the Bulletin itself is driving fresh, practical legal consequences. See POET, 970 F.3d at 405 (holding that EPA's application of a guidance document in a subsequent letter “reinforces the Guidance's finality”).

Contrast the Bulletin with agency actions that courts have deemed mere restatements of policy. In Clayton County, we held that a letter from the FAA wasn't a final agency action. 887 F.3d at 1269. In so holding, we emphasized several considerations: The letter took the same position that the FAA had expressed in a policy statement two years earlier; the challenger knew about the earlier policy statement; and by eliciting the letter the challenger had transparently attempted to evade the applicable statute of limitations. See id. at 1266–70. Similarly, in Independent Equipment Dealers, the D.C. Circuit held that an EPA letter wasn't final agency action. 372 F.3d at 428. The EPA's letter likewise restated a “longstanding interpretation,” had no “concrete impact,” and “was purely informational in nature.” Id. at 427. To be sure, like the FAA and EPA letters, the Bulletin appears to restate a preexisting interpretation—but quite unlike those letters, the Bulletin is not “purely informational in nature.” Id. Rather, as already explained, it demands that regulated parties take action by collecting data, examining program participants, and enforcing CMS's view of the hold-harmless rule—and not so subtly hints at the dire consequences of non-compliance.

As we have said many times, for final-agency-action purposes, finality is understood pragmatically, and no single feature of the Bulletin is necessarily decisive. But taken together, the considerations we've emphasized make clear that in the Bulletin, CMS “has given the States their ‘marching orders’ and [CMS] expects the[m] to fall in line.” Appalachian Power, 208 F.3d at 1023. Accordingly, we hold that the Bulletin is final agency action subject to judicial review.³

B

Separately, but relatedly, CMS contends that Florida's challenge to the Bulletin isn't ripe. For many of the same reasons that we've found the Bulletin to be final agency action, we conclude that it is ripe for review.

Ripeness turns on (1) the fitness of the issues for judicial decision and (2) the hardship to the parties of withholding our consideration. Abbott Lab'ys v. Gardner, 387 U.S. 136, 148–149 (1967). In administrative law, ripeness usually—if perhaps not always—“overlaps with” finality. Ala. Power Co. v. U.S. Dep't of Energy, 307 F.3d 1300, 1311 n.10 (11th Cir. 2002); see Kristin E. Hickman & Richard J. Pierce, Administrative Law Treatise § 17.12 (7th ed. 2018 & Supp. 2025) (observing that courts often consider ripeness to be “interchangeable with ․ finality,” but that it also can “sometimes serve[ ] independent purposes”). Our analysis focuses in particular on the hardship to the plaintiffs of delaying review, whether judicial intervention would “inappropriately interfere” with the administrative process, and whether the case would benefit from further factual development. Ohio Forestry Ass'n, Inc. v. Sierra Club, 523 U.S. 726, 733 (1998).

Whatever the theoretical distinctions might be, the Bulletin is ripe for essentially the same reasons it is final. Delaying review will cause Florida hardship because the Bulletin “inflicts significant practical harm,” id. at 733, in that the state now faces a new demand to root out private hold-harmless arrangements. Review won't inappropriately interfere with CMS's decisionmaking process because, per the Bulletin's own terms, it represents the agency's refined, non-tentative views. Cf. id. at 735 (holding that a forest management plan was not ripe for review in part because “further consideration will actually occur before the Plan is implemented”). And no further factual development is necessary. This is principally a dispute over the meaning of a statute, which won't require us to “entangl[e]” ourselves “in abstract disagreements over administrative policies.” Abbott Lab'ys, 387 U.S. at 148.

* * *

The Bulletin is final, and Florida's challenge is ripe. We therefore have jurisdiction. On, then, to the merits of the state's preliminary-injunction request.⁴

III

To obtain the preliminary injunction it seeks, Florida must show that “(1) it has a substantial likelihood of success on the merits; (2) it will suffer an irreparable injury unless the injunction is granted; (3) the harm from the threatened injury outweighs the harm the injunction would cause the opposing party; and (4) the injunction would not be adverse to the public interest.” Dream Defs. v. Governor of the State of Fla., 57 F.4th 879, 889 (11th Cir. 2023) (citation modified). Here, we needn't proceed any further than the preliminary-injunction test's first prong because we conclude that Florida is unlikely to succeed on the merits. Johnson & Johnson Vision Care, Inc. v. 1-800 Contacts, Inc., 299 F.3d 1242, 1247 (11th Cir. 2002).

Florida argues that CMS's Bulletin is unlawful for three reasons: (1) that it exceeds CMS's statutory authority; (2) that it's arbitrary and capricious; and (3) that CMS issued it without proper process. For the following reasons, we reject all three contentions.⁵

A

1

In assessing CMS's statutory authority to issue the Bulletin, “[w]e start, as always, with the text.” Bufkin v. Collins, 604 U.S. 369, 379 (2025). Here, that is no mean feat. The Social Security Act, of which the hold-harmless rule is part, “is among the most intricate ever drafted by Congress.” Schweiker v. Gray Panthers, 453 U.S. 34, 43 (1981). As relevant here, the Act provides that CMS should deduct from the federal government's Medicaid contribution an amount corresponding to state revenue collected while a hold-harmless provision is in effect. See 42 U.S.C. § 1396b(w)(1)(A)(iii). Then comes the subsection at issue here—which, fair warning, is both bulky and a little clunky:

For purposes of paragraph (1)(A)(iii), there is in effect a hold harmless provision with respect to a broad-based health care related tax imposed with respect to a class of items or services if the Secretary determines that any of the following applies:

(A) The State or other unit of government imposing the tax provides (directly or indirectly) for a payment (other than under this subchapter) to taxpayers and the amount of such payment is positively correlated either to the amount of such tax or to the difference between the amount of the tax and the amount of payment under the State plan.

(B) All or any portion of the payment made under this subchapter to the taxpayer varies based only upon the amount of the total tax paid.

(C)(i) The State or other unit of government imposing the tax provides (directly or indirectly) for any payment, offset, or waiver that guarantees to hold taxpayers harmless for any portion of the costs of the tax.

(ii) For purposes of clause (i), a determination of the existence of an indirect guarantee shall be made under paragraph (3)(i) of section 433.68(f ) of title 42, Code of Federal Regulations ․

Id. § 1396b(w)(4) (emphasis added).

Happily, just one sentence really matters. Nobody disputes that county-level assessments under Florida's Directed Payment Program constitute “broad-based health care related tax[es]” within the meaning of the prefatory clause. And nobody contends that the Bulletin can be justified under subparagraphs (A) or (B), or that it addresses indirect guarantees, which are the subject of clause (C)(ii). Instead, everything pretty much rides on clause (C)(i)—which, to repeat, states that a hold-harmless provision exists when “[t]he State or other unit of government imposing the tax provides (directly or indirectly) for any payment, offset, or waiver that guarantees to hold taxpayers harmless for any portion of the costs of the tax.” CMS insists that (C)(i) is implicated when private “taxpaying providers” choose to “redistribut[e] Medicaid payments after receipt to ensure that all taxpaying providers receive all or a portion of their tax costs back.” Bulletin, supra, at 2. The question before us is whether CMS's interpretation fits the statutory language.

We can begin to clarify matters by diagramming clause (C)(i)’s single sentence, matching the two key verbs with their accompanying nouns. First up, “provides”: It is the “State or other unit of government” that does the “provid[ing].” This much seems obvious and uncontroversial. Next, “guarantees”: It is the “payment, offset, or waiver” that does the “guarantee[ing].” This too seems obvious—but it's hotly contested. Florida insists that there's a “grammatical link” “between the government, as the actor providing for something, and a guarantee, as the thing provided for.” Br. of Appellants at 49 (quoting Texas v. Brooks-LaSure, 680 F. Supp. 3d 791, 808 (E.D. Tex. 2023)). Therefore, Florida says, a hold-harmless provision under (C)(i) exists only when the state government “intentionally, deliberately, and with certainty causes the taxpayer's liability to be offset.” Id. We disagree. However intuitive, Florida's reading has no footing in the ordinary rules of syntax. Contrary to Florida's contention, the phrase “that guarantees to hold taxpayers harmless” is a restrictive relative clause that modifies the phrase “payment, offset, or waiver.” So grammatically, (C)(i) works in two steps: (1) The state or local government provides for a payment offset or waiver; and (2) the payment, offset, or waiver in turn “guarantees to hold taxpayers harmless” for at least part of the government-imposed healthcare tax.

With this clarification in hand, it seems clear to us that CMS has the better argument. CMS's position, stretching all the way back to 2008 and reiterated in the Bulletin, is that “[a] direct guarantee [under (C)(i)] would be found when a State payment is made available to a taxpayer or a party related to the taxpayer ․ in the reasonable expectation that the payment would result in the taxpayer being held harmless for any part of the tax.” 73 Fed. Reg. at 9,686; accord Bulletin, supra, at 4. In the ways that matter, CMS's interpretation tracks (C)(i)’s text. The statute requires that the government “provide[ ] ․ for any payment, offset, or waiver”; so too, CMS anchors its interpretation to a state “ma[king] available” a “payment.” And, crucially, the statute then asks whether the “payment” “guarantee[s] to hold taxpayers harmless”; so too, CMS asks whether “the payment would result in the taxpayer being held harmless.”

It's true, as Florida says, that “result in” doesn't mean exactly the same thing as “guarantee.” According to the state, to “guarantee” is to “ ‘make oneself answerable for (something) on behalf of someone else,’ ‘undertake to ensure for another,’ ‘serve as a warrant or guaranty for,’ ‘engage to ․ indemnify ․ loss,’ or ‘promise.’ ” Br. of Appellants at 48–49 (omissions in original) (quoting Guarantee, Random House Unabridged Dictionary (2d ed. 1993)). Perhaps CMS's phrase “[r]esult in” is more an approximation of “guarantee” than a jot-for-jot synonym, but it captures the same kind of conduct: If a follow-on consequence of a state disbursing Medicaid payments is that providers use those payments to offset taxes, then the Medicaid funds are “ensur[ing]” or “indemnify[ing]” providers. Put another way, the Medicaid payments “result in” a tax offset.

Florida seems to suggest that by using the word “guarantee,” (C)(i) imports something like an intent requirement, which a payment—an inanimate object—can't meet. We disagree, for two reasons. First, lifeless items can most certainly “guarantee” things. Money deposited as bail guarantees the accused's presence at trial—no matter who ponied up the cash or what that person was thinking. A title guarantees ownership of a car—no matter whether the previous owner had regrets about selling. Gravity guarantees that objects will fall—no matter who was thinking about that. So too, a “payment, offset, or waiver” can “guarantee” that a provider will be made whole after it is taxed regardless of whether the state intends to promise anything.

Second, Florida's intent-based interpretation would make (C)(i) the odd one out in an otherwise effects-based list. As the state recognizes, (A), (B), and (C)(ii) “all define effects-based tests.” Reply Br. at 21. But under Florida's reading, (C)(i) alone would turn on intent. See, e.g., Br. of Appellants at 2 (arguing that (C)(i) applies “only if the taxing government requires or promises that the taxpayer will be reimbursed for the tax”). Not dispositive, but weird. Under CMS's reading, by contrast, (C)(i) dovetails with (A), (B), and (C)(ii).

* * *

CMS's central insight—with which we agree—is that clause (C)(i) simply doesn't say that the state must guarantee anything. Rather, the state provides a payment, which itself, in turn, guarantees that a taxpayer will be held harmless. To the same effect, under the Bulletin, when healthcare providers redistribute state Medicaid payments to offset taxes, those payments result in the taxpayers being held harmless. If anything, CMS extends to states a discretionary boon the statute itself doesn't require. The Bulletin cabins its interpretation to instances in which the state or other unit of government has a “reasonable expectation” that its payment will be misused. The statute includes no such limitation.

In sum, we aren't persuaded by Florida's first-order textual arguments.

2

The state challenges the Bulletin on two additional textual bases. We find them no more persuasive.

First, Florida contends that CMS's interpretation in the Bulletin renders subparagraphs (A) and (B) and clause (C)(ii) superfluous. On CMS's reading, the state says, every circumstance covered by (A), (B), and (C)(ii) would also satisfy (C)(i). For instance, imagine that a county government levies a 10% tax on hospital revenue. The county also operates a grant program that reimburses residents up to 10% of the cost of their hospital stays. As a result, almost all the tax gets funneled back to the hospitals. This arrangement would qualify as an “indirect” guarantee under (C)(ii).⁶ Under CMS's reading of (C)(i), it would also constitute a “direct” guarantee, because the county would be providing for a payment—the reimbursement to residents—that guarantees to hold the hospitals harmless for a portion of the tax. Under Florida's interpretation of (C)(i), by contrast, this program wouldn't be a direct guarantee, because the state isn't itself doing the guaranteeing—whether the money flows to the hospitals depends on the actions of third-party patients.

Florida may well be right that CMS's interpretation generates some (even substantial) overlap. But—and it's an important but—§ 1396b(w)(4) isn't quite like other statutes with respect to which courts have relied on the presumption against superfluity. That canon has “special force” when “a statutory construction [ ] renders an entire subparagraph meaningless.” Pulsifer v. United States, 601 U.S. 124, 143 (2024). That issue doesn't exist here. CMS isn't interpreting (C)(i) in a way that swallows up the rest of the statutory provision. By the statute's own terms, each subparagraph (and clause) embodies a separate means of getting at the same thing: whether “there is in effect a hold harmless provision with respect to a broad-based health care related tax.” 42 U.S.C. § 1396b(w)(4). Because (A), (B), (C)(i), and (C)(ii) are all designed to identify hold-harmless arrangements—which come in different shapes and sizes—the overlap among them is unsurprising.⁷ At the same time, each subparagraph and clause is different in the sense that each identifies a different means—or, in the case of (C)(ii), defines a means—by which a hold-harmless arrangement might be accomplished.

Consider, in particular, the methods for sussing out direct and indirect guarantees. As we've said, to ascertain whether there is a direct guarantee, we must determine (1) whether the state provides a payment and (2) whether that payment guarantees that taxpayers are held harmless. By contrast, to detect an indirect guarantee, we ignore the mechanics of any funding arrangement and instead ask simply (1) whether provider taxes exceed a certain threshold and (2) whether a certain percentage of taxpayers recover a certain percentage of the tax. 42 U.S.C. § 1396b(w)(4)(C)(ii); 42 C.F.R. § 433.68(f)(3)(i). The direct and indirect tests may often arrive at the same place, but they travel different routes. So, in sum, while § 1396b(w)(4) admits some superfluity of coverage, it doesn't entail superfluity of method.

Second, and separately, Florida insists that, for (C)(i) to apply, the “ ‘unit of government imposing the tax’ [must] be the same ‘unit of government’ that ‘provides ․ for’ the guarantee.” Reply Br. at 24; see Br. of Appellants at 52–53 & n.11. We disagree. As an initial matter, the state is again mixing up which verbs go with which nouns. It's the payment—not the unit of government—that “ ‘provides ․ for’ the guarantee.” Moreover, and in any event, Florida's reading proves too much. Were it correct, a state could end-run the entirety of § 1396b(w)(4)(C)(i) simply by assigning—as Florida has—the taxing job to local governments, while retaining the payment job for itself.

3

Florida finally challenges CMS's interpretation on two extra-textual bases, neither of which convinces us.

First, the state contends that it would be absurd for the legality of its tax scheme to depend on private companies’ behavior. We don't think so. For one thing, federal law already requires Florida to collect information on Medicaid funding recipients, so the state shouldn't be operating in the dark. See, e.g., 42 C.F.R. § 433.74(a) (“[E]ach State must submit to CMS quarterly summary information on the source and use of all ․ health care-related taxes collected.”); id. (“Each State must also provide any additional information requested by the Secretary related to ․ any taxes imposed on, health care providers.”). For another, the way healthcare providers respond to and implement provider taxes isn't really “independent” behavior. Hospitals and clinics have close, ongoing contractual relationships with Florida's state healthcare agency. And Florida seems to have freely chosen to adopt its particular (provider-solicited) healthcare-related tax scheme so as to boost the federal government's contribution to state Medicaid coffers. “Medicaid offers States a bargain.” Medina v. Planned Parenthood S. Atl., 145 S. Ct. 2219, 2226 (2025). Part of the bargain is following the hold-harmless rule.

Second, Florida cites the presumption that, under the Spending Clause, U.S. Const. Art. I, § 8, cl. 1, “Congress must speak ‘unambiguously’ and ‘with a clear voice’ when it imposes conditions on federal funds.” West Virginia ex rel. Morrisey v. U.S. Dep't of the Treasury, 59 F.4th 1124, 1141 (11th Cir. 2023) (quoting Pennhurst State Sch. & Hosp. v. Halderman, 451 U.S. 1, 17 (1981). True, interpreting § 1396b(w)(4) isn't without its challenges. Even so, we are convinced that CMS's interpretation in the Bulletin is demonstrably superior to Florida's. Because § 1396b(w)(4) isn't ambiguous, the limitations imposed by the Spending Clause are satisfied.

* * *

Because the Bulletin is consistent with the statutory text, Florida is not substantially likely to succeed in its frontal challenge to CMS's interpretation.

B

Next up, Florida's contention that CMS acted arbitrarily and capriciously in adopting the Bulletin. Arbitrary-and-capricious review is deferential—an agency's decisions need only “be reasonable and reasonably explained.” FCC v. Prometheus Radio Project, 592 U.S. 414, 423 (2021). Nonetheless, “[a]n agency may not ․ depart from a prior policy sub silentio or simply disregard rules that are still on the books.” FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515 (2009). “[T]he requirement that an agency provide reasoned explanation for its action [ ] ordinarily demand[s] that it display awareness that it is changing position.” Id. And “[w]hen an agency changes course, ․ it must be cognizant that longstanding policies may have engendered serious reliance interests that must be taken into account.” Dep't of Homeland Sec. v. Regents of the Univ. of Cal., 591 U.S. 1, 31 (2020) (citation modified). Florida asserts that when CMS issued the Bulletin, it changed its position without acknowledging as much and, in the process, ignored important reliance interests.

CMS didn't change its position in the way that matters—that is, regarding the meaning of (C)(i). Rather, the Bulletin simply applies, using more specific language, an interpretation of (C)(i) that the agency embraced more than a decade and a half ago. In the 2008 Final Rule's preamble, CMS announced that “[a] direct guarantee [under (C)(i)] would be found when a State payment is made available to a taxpayer or a party related to the taxpayer ․ in the reasonable expectation that the payment would result in the taxpayer being held harmless for any part of the tax.” 73 Fed. Reg. at 9,686; accord id. at 9,694. Then, in the 2019 proposal, CMS—applying the same interpretation—insisted that “[t]he fact that a private entity makes the redistribution payment does not change the essential nature of the payment.” 84 Fed. Reg. at 63,735. Although CMS withdrew the 2019 proposed rule, both the proposed rule's preamble and the rescission notice made clear that CMS's positions didn't depend on whether the proposed regulations went into effect. See id. at 63,735 (explaining that the proposed rule did “not reflect any change in policy or approach, but merely codified currently prohibited practices”); 86 Fed. Reg. at 5,105 (“[T]he withdrawal of this proposed rule does not affect existing federal legal requirements or policy that were merely proposed to be codified in regulation.”). The Bulletin undoubtedly signaled a shift in enforcement priorities. But “a belief about how an agency is likely to exercise its enforcement discretion is not a serious reliance interest.” FDA v. Wages & White Lion Invs., L.L.C., 145 S. Ct. 898, 927 (2025)

Florida makes much of the fact that CMS withdrew the 2019 proposed rule. On Florida's telling, “the 2019 proposal's existence proves that the policy it articulated—and that the Bulletin later adopted—was new because, if the policy already existed, the 2019 proposal (and the later Bulletin) would have been unnecessary.” Reply Br. at 9. It's true that a withdrawn proposal probably isn't evidence of CMS's official views. Cf. CFTC v. Schor, 478 U.S. 833, 845 (1986) (“It goes without saying that a proposed regulation does not represent an agency's considered interpretation of its statute.”). But the converse—that the 2019 interpretation can't be CMS's view—doesn't follow. CMS was absolutely clear in 2019 that, as the agency saw it, the hold-harmless arrangements covered by the proposal were already forbidden and that the proposal was exegetical, not inventive. See, e.g., 84 Fed. Reg. at 63,735 (describing the proposal as “add[ing] clarifying language to the hold harmless definition” (emphasis added)). Nobody doubts that the 2008 preamble embodied CMS's official position, and in that document CMS advanced its reasonable-expectation interpretation of (C)(i). See 73 Fed. Reg. at 9,686; id. at 9,694. CMS applied that interpretation in the Bulletin. And never—not in 2008, 2019, 2021, or 2023—did CMS purport to exempt private-only arrangements from the hold-harmless rule.

As evidence that CMS harbored views inconsistent with the Bulletin, Florida cites two HHS Office of Inspector General statements, a decision of the HHS Departmental Appeals Board, and a brief email exchange about a telephone call that included the Director of CMS's Financial Management Group. But none of these is a relevant source for identifying CMS's official views.

An agency's “interpretation must at the least emanate from those actors, using those vehicles, understood to make authoritative policy in the relevant context.” Kisor v. Wilkie, 588 U.S. 558, 577 (2019). Both the Inspector General and the Appeals Board operate at some remove from the rest of HHS, and so we can't assume that their views are CMS’s views. Cf. Univ. of Med. & Dentistry of N.J. v. Corrigan, 347 F.3d 57, 60–61 (3d Cir. 2003) (“The Office of Inspector General of HHS, along with inspector generalships for other federal administrative agencies and departments, ․ are designed to be independent and objective units separate from their respective departments and agencies.” (citation modified)); Ariz. Health Care Cost Containment Sys. v. McClellan, 508 F.3d 1243, 1248 n.6 (9th Cir. 2007) (“The [Appeals Board] is a separate adjudicatory department within HHS that provides independent review of disputed decisions for many HHS programs.”). And the curt email referencing a phone call from the Financial Management Group Director is far too little to establish CMS's authoritative views. Like our sister circuit, “[w]e would marvel if a few casual communications in the guise of informal calls and a staff email constituted an agency's formal position.” UnitedHealthcare of N.Y., Inc. v. Lace-well, 967 F.3d 82, 95–96 (2d Cir. 2020); cf. Kisor, 588 U.S. at 577 (indicating that “an ‘informal memorandum’ recounting a telephone conversation between employees” is not an “authoritative pronouncement” (quoting N.Y. State Dep't of Social Servs. v. Bowen, 835 F.2d 360, 365–366 (D.C. Cir. 1987))).

In the end, the documents that do express CMS's official positions—the 2008 rule, the 2021 rescission, and the Bulletin—are all consistent. So, CMS didn't sub silentio change its mind. Florida's arbitrary-and-capricious argument is unlikely to succeed.

C

Last of all we briefly return, as promised, see supra at 15, to the distinction between “legislative” and “interpretive” rules. Importantly here, legislative rules require the promulgating agency to abide by notice-and-comment procedures, while interpretive rules don't. 5 U.S.C. § 553(b)(A). CMS released the Bulletin without notice and an opportunity for comment. Accordingly, if the Bulletin was a legislative rule, then it is procedurally invalid. But it wasn't, and so it isn't.

Legislative rules “create[ ] new law, rights, or duties,” while interpretive rules “do[ ] not modify or add to a legal norm based on the agency's own authority.” Warshauer, 577 F.3d at 1337. To distinguish a legislative from an interpretive rule, we consider two things. First, we look to “the agency's characterization of the rule”—though this is “not dispositive.” Id. (citation modified). And second, we consider that an “interpretative” rule should “simply state[ ] what the administrative agency thinks the statute means, and only remind[ ] affected parties of existing duties.” Id. (citation modified).

The Bulletin is an interpretive rule. For one thing, CMS characterizes the Bulletin as mere interpretation, not an announcement of new law. See Bulletin at 1 (saying that “this informational bulletin reiterates our longstanding position”). And for another, because the Bulletin properly and without novelty interprets § 1396b(w)(4)—at least with respect to the topics disputed in this litigation—the Bulletin “simply states what the administrative agency thinks the statute means,” and in fact does “only remind[ ] affected parties of existing duties.” Warshauer, 577 F.3d at 1337.

Of course, the mere fact that the Bulletin's challenged aspects are consistent with the statutory text isn't alone enough to make the Bulletin a valid interpretive rule. Although statutes have a “single, best meaning,” Loper Bright Enters. v. Raimondo, 603 U.S. 369, 400 (2024), at times “the best reading of a statute is that it delegates discretionary authority to an agency.” Id. at 395. So here, it's important that—with respect to the narrow question whether § 1396b(w)(4)(C)(i) can cover private hold-harmless arrangements—the statute doesn't confer on CMS any discretion, CMS doesn't purport to be exercising any such discretion, and the Bulletin simply and correctly transcribes a longstanding and accurate interpretation.

Because the Bulletin does “not have the force and effect of law,” and simply “advise[s] the public of the agency's construction of the statutes and rules which it administers,” it's an interpretive rule. Shalala v. Guernsey Mem'l Hosp., 514 U.S. 87, 99 (1995) (citation modified). Florida's argument that CMS evaded the notice-and-comment requirement is therefore unlikely to succeed.

* * *

We are unpersuaded by Florida's objections to the Bulletin. The state has therefore failed to make the requisite showing on the preliminary-injunction test's threshold likelihood-of-success prong. So while the district court was wrong to deny Florida's preliminary-injunction motion on jurisdictional grounds, its bottom line was correct: No preliminary injunction should issue.

IV

For the foregoing reasons, we hold that the Bulletin is final agency action, that Florida's challenge to it is ripe, and that Florida is unlikely to succeed on the merits. We therefore REVERSE the district court's decision to dismiss the case for lack of subject matter jurisdiction, but AFFIRM the denial of Florida's motion for a preliminary injunction and REMAND for further proceedings consistent with this decision.

Although it is a close question, I agree with the majority that under a pragmatic approach CMS’ Informational Bulletin constitutes final agency action under the Administrative Procedure Act, 5 U.S.C. § 704. See Abbott Laboratories v. Gardner, 387 U.S. 136, 149 (1967) (“The cases dealing with judicial review of administrative actions have interpreted the ‘finality’ element in a pragmatic way.”). I also agree that Florida's challenge to the Bulletin is ripe. I join Part I of the majority opinion except as to footnote 1, as well as Parts II, III, and IV, and add the following thoughts.

* * * * *

I do not believe we can say categorically, as the majority does in footnote 1, that in a preliminary injunction appeal we always conduct plenary review of legal issues. I recognize that there are Supreme Court and Eleventh Circuit cases that stand for such a proposition, and which conduct de novo review of legal issues when reviewing the grant or denial of preliminary injunctive relief. See, e.g., Smith v. Vulcan Iron Works, 165 U.S. 518, 525 (1897); Wood v. Fla. Dep't of Educ., 142 F.4th 1286, 1289 & n.1 (11th Cir. 2025). But, as I have tried to explain elsewhere, see Wood, 142 F.4th at 1294–96 ( Jordan, J., dissenting), there are many cases to the contrary, and the general (and better) rule is that on review of a preliminary injunction an appellate court reviews the substantial likelihood of success prong for abuse of discretion even if it turns on a legal issue. See, e.g., Ashcroft v. ACLU, 542 U.S. 656, 666 (2004) (concluding that the district court's determination as to likelihood of success on a First Amendment challenge to a federal statute “was not an abuse of discretion”); Doran v. Salem Inn, Inc., 422 U.S. 922, 934 (1975) (finding no abuse of discretion in the district court's grant of a preliminary injunction, and explaining that “[t]his is the extent of our appellate inquiry, and we therefore ‘intimate no view as to the ultimate merits of respondents’ contentions’ ”) (citation omitted).

The majority cites Eleventh Circuit cases spanning from 2012 to 2025 to support its view that plenary review applies. But a number of our earlier preliminary injunction cases conduct abuse of discretion review as to likelihood of success. And where there is an intra-circuit conflict, the earlier cases control. See Harris v. Lincoln Nat'l Life Ins. Co., 42 F.4th 1292, 1297 (11th Cir. 2022).

For example, in Butler v. D.A. Schulte, Inc., 67 F.2d 632, 635 (5th Cir. 1933), we noted “the rule that on appeals from interlocutory injunctions the merits will not ordinarily be considered, but only whether there has been an abuse of discretion in preserving the status pending hearing on the merits.” And Butler is not an outlier—there are other earlier cases holding that a district court's ruling on the substantial likelihood of success prong triggers deferential abuse of discretion review. See Wooten v. Ohler, 303 F.2d 759, 762 (5th Cir. 1962) (“[W]e do not review the intrinsic merits of the case as such. Rather, our inquiry is whether there has been an abuse of discretion. Our review of the probable merits does not go to the question of whether we would ultimately hold that the trial [j]udge was right or wrong, but only to the ascertainment of whether his action was within his broad range of discretion.”); Lea v. Vasco Prods., Inc., 81 F.2d 1011, 1011–12 (5th Cir. 1936) (“We cannot agree with appellants that the merits are before us for decision. All that we have here under case as made on the temporary application is whether the [district] court abused its discretion in restraining the defendants as it did.”).

As the Supreme Court's decision in Ashcroft, 542 U.S. at 664–65, makes clear, abuse of discretion review is appropriate even where substantial likelihood of success turns on a purely legal issue. A number of our more recent preliminary injunction cases conduct this same deferential review with respect to issues of law. See, e.g., Cafe 207, Inc. v. St. Johns Cnty., 989 F.2d 1136, 1137 (11th Cir. 1993) (addressing substantial likelihood of success on a First Amendment claim: “Whether the district court's determination of this point is right or wrong, the record before us indicates no abuse of discretion.”). Given all of this caselaw, we cannot (and should not) broadly declare that legal issues going to likelihood of success always trigger plenary review in preliminary injunction appeals.

To me, deferential review of the district court's preliminary assessment of the merits usually makes sense. When a district court, particularly on a close legal issue of first impression, analyzes whether the plaintiff has shown a substantial likelihood of success on the merits, it is not definitely deciding who wins and who loses; it is instead weighing the parties’ arguments and making a probabilistic determination about which side is likely to prevail. So, “[i]f the underlying constitutional [or legal] question is close, ․ [the appellate court] should uphold the injunction [or the denial of the injunction] and remand for trial on the merits.” Ashcroft, 542 U.S. at 664–65. See also Doran, 422 U.S. at 932 (“While we regard the question as a close one, we believe that the issuance of a preliminary injunction in behalf of respondents ․ was not an abuse of the District Court's discretion.”); Meccano Ltd. v. Wanamaker, 253 U.S. 136, 142 (1920) (“If ․ two [appellate courts] have expressed conflicting views, we cannot now declare which is right or undertake finally to decide the several issues involved upon their merits. The matter for review here is the action of the courts below upon the preliminary order for injunction and we may go no further.”).¹

* * * * *

Normally, I do not think it is a good idea for an appellate court to decide in the first instance whether or not to grant preliminary injunctive relief. We are not, for example, in the business of fact-finding, and often facts will drive some of the preliminary injunction factors (i.e., likelihood of success, irreparable harm, and the balancing of the equities). But because we are only addressing the substantial likelihood of success prong on a legal issue, and ruling adversely to Florida on that score, doing so here does not seem too problematic.

As the majority opinion notes, we are not conclusively deciding the merits of Florida's APA claims. In the words of the Supreme Court, “it is generally inappropriate for a federal court at the preliminary-injunction stage to give a final judgment on the merits.” Univ. of Tex. v. Camenisch, 451 U.S. 390, 395 (1981). Compare Munaf v. Geren, 553 U.S. 674, 691 (2008) (“Adjudication of the merits is most appropriate if the injunction rests on a question of law and it is plain that the plaintiff cannot prevail.”) (emphasis added). Although there is some support for Florida's interpretation of 42 U.S.C. § 1396b(w)(4)(C)(i), see Texas v. Brooks-LaSure, 680 F. Supp. 3d 791, 808–09 (E.D. Tex. 2023), the statutory analysis conducted by the majority—whether ultimately correct or not—is sufficient to demonstrate that Florida has not shown a substantial likelihood of success on the merits. And that is enough to deny Florida the preliminary injunction it seeks.

* * * * *

I close with one final observation. Even if we wanted to, we could not definitively decide the merits of Florida's APA claims in favor of CMS in this appeal. The district court dismissed the action without prejudice because it concluded that the Bulletin did not constitute final agency action. See D.E. 53 at 2. Because CMS has not cross-appealed, we cannot issue a merits ruling that expands its victory in the district court by converting the dismissal into one with prejudice. See, e.g., Trustees of Atlanta Iron Workers, Local 387 Pension Fund v. S. Stress Wire Corp., 724 F.2d 1458, 1459 (11th Cir. 1983) (“Absent a cross-appeal, an appellee may not attempt to enlarge his own rights or decrease the rights of his adversary[.]”).²

FOOTNOTES

1.   A few different standards of review apply here. We review de novo a district court's dismissal of a complaint for lack of subject matter jurisdiction. Tufts v. Hay, 977 F.3d 1204, 1208 (11th Cir. 2020). We review for abuse of discretion the denial of a preliminary injunction, “but the underlying legal conclusions are reviewed de novo.” Cambridge Christian Sch., Inc. v. Fla. High Sch. Athletic Ass'n, Inc., 942 F.3d 1215, 1229 (11th Cir. 2019) (emphasis omitted). And “[a]s a general matter, when an agency interprets a statute, judicial review of the agency's interpretation is de novo.” Seven Cnty. Infrastructure Coal. v. Eagle Cnty., 145 S. Ct. 1497, 1511 (2025) (emphasis omitted). But arbitrary-and-capricious review is “deferential.” Id.In his separate opinion, Judge Jordan contends that legal issues embedded within a district court's decision to grant or deny a preliminary injunction shouldn't necessarily be subject to de novo review. See Jordan Concurring Op. at 1. His argument is characteristically thoughtful and rigorous, and not without some force. But it runs headlong into a wall of our precedent expressly holding that although we review the grant or denial of a preliminary injunction for abuse of discretion, we review constituent legal determinations de novo. E.g., Wood v. Fla. Dep't of Educ., 142 F.4th 1286, 1289 (11th Cir. 2025); Gonzalez v. Governor of Georgia, 978 F.3d 1266, 1270 (11th Cir. 2020); Bloedorn v. Grube, 631 F.3d 1218, 1229 (11th Cir. 2011); LSSi Data Corp. v. Comcast Phone, LLC, 696 F.3d 1114, 1119 (11th Cir. 2012).

2.   Nor could they. CMS's conclusions in the Bulletin are neither tentative nor poised for revision; no further “administrative steps [will] necessarily [ ] be taken” with respect to CMS's interpretation of the hold-harmless rule. Norton, 324 F.3d at 1238 (citing City of San Diego v. Whitman, 242 F.3d 1097, 1101 (9th Cir. 2001)). To the contrary, the Bulletin's explication of CMS's views is decisive and unambiguous, as it purports to “reiterate[ ]” the agency's “longstanding position.” Bulletin, supra, at 1.

3.   We reject CMS's suggestion that the Bulletin isn't a final agency action because Florida could raise its arguments in defense to an enforcement action. The Supreme Court long ago held that an agency order that threatened penalties for noncompliance and “specif[ied] which commodities the [agency] believed were exempt by statute from regulation, and which it believed were not,” was reviewable, even though the order “would have effect only if and when a particular action was brought against a particular carrier.” Hawkes, 578 U.S. at 599–600 (citation modified) (describing Frozen Food Express v. United States, 351 U.S. 40 (1956)). So too here.

4.   A brief word about our decision to reach the merits: The district court never assessed the merits because it dismissed Florida's suit and denied its preliminary-injunction motion on jurisdictional grounds. Although we usually “go no further into the merits than is necessary to decide the interlocutory appeal,” that “is a rule of orderly judicial administration only”—we may go further if it's appropriate. Callaway v. Block, 763 F.2d 1283, 1287 n.6 (11th Cir. 1985). Deciding the merits of Florida's preliminary-injunction request is appropriate because “both sides’ arguments go to the merits, no facts are at issue and the questions raised are purely legal ones.” Id. at 1287. Were we to remand without addressing the request, it “would unnecessarily delay the disposition of th[e] case.” Boyes v. Shell Oil Prods. Co., 199 F.3d 1260, 1266 (11th Cir. 2000).

5.   Of course, at this preliminary stage, we do not conclusively decide the merits of Florida's challenges; rather, we hold only that Florida is unlikely to succeed on the merits and, therefore, isn't entitled to a preliminary injunction. See Univ. of Tex. v. Camenisch, 451 U.S. 390, 395 (1981) (“[I]t is generally inappropriate for a federal court at the preliminary-injunction stage to give a final judgment on the merits.”).

6.   Explaining exactly why this scheme would constitute an indirect guarantee requires a detour into the regulatory weeds. According to (C)(ii), a decisionmaker seeking to determine whether a funding arrangement amounts to an indirect guarantee should look to 42 C.F.R. § 433.68(f)(3)(i). That regulation, in turn, outlines a two-part test. First, the decisionmaker should ask whether “the health care-related tax or taxes on each health care class are applied at a rate that produces revenues less than or equal to 6 percent of the revenues received by the taxpayer.” 42 C.F.R. § 433.68(f)(3)(i)(A). If so, there is no indirect guarantee. Id. If, however, the tax generates revenue above the 6 percent threshold, the decisionmaker should proceed to ask whether “75 percent or more of the taxpayers in the class receive 75 percent or more of their total tax costs back in enhanced Medicaid payments or other State payments.” Id. § 433.68(f)(3)(i)(B). If the answer is yes, then an impermissible indirect guarantee exists.

7.   It's worth noting that even under Florida's reading of (C)(i), the clause's test overlaps substantially with other provisions of § 1396b(w)(4). Consider a scheme that functions like the private transfers at issue here but that the state operates directly. Under such a program, the state would impose a broad-based provider tax and then remit extra payments to hospitals that pay the tax but receive little or no Medicaid funding. Subparagraph (A) would cover this arrangement, because the state—which imposed the tax—would “provide[ ] ․ for a payment ․ to taxpayers and the amount of such payment [would be] positively correlated ․ to the difference between the amount of the tax and the amount of payment under the State [Medicaid] plan.” 42 U.S.C. § 1396b(w)(4)(A). The scheme would also satisfy (C)(i) under either Florida or CMS's interpretation, because the state would be “provid[ing] ․ directly ․ for [a] payment ․ that guarantees to hold taxpayers harmless”—regardless of whether it's the state or the payment doing the guaranteeing. Id. § 1396b(w)(4)(C)(i).

1.   The district court here did not address the merits. Because there is no underlying merits decision to give deference to, the majority's statement about plenary review is dicta as to the merits.

2.   On remand, therefore, the district court will need to address the merits of Florida's APA claim.

Newsom, Circuit Judge: