Opinion for the Court filed by Senior Circuit Judge SILBERMAN.
Appellees are a group of hospitals that serve a significant number of elderly, very low-income patients. Congress assumes that such patients cost more to treat than the average Medicare patients, so these hospitals are entitled to supplemental payments. These are determined by calculating what is called the “disproportionate share percentage” — a formula which is a proxy for the percentage of low-income patients served.
In 2004, the Secretary issued a rule that addressed one aspect of this calculation. Although ostensibly only a detail, the financial impact is apparently substantial, costing the hospitals hundreds of millions of dollars. Not surprisingly, the hospitals sued in district court challenging the rule. The court, holding that the final rule was not a logical outgrowth of the proposed rule and that the Secretary had insufficiently explained a change in policy, granted judgment to the hospitals and vacated the rule. But the court went further, instructing the Secretary to recalculate reimbursement percentages using the alternate methodology. We affirm in part and reverse in part.
I.
Medicare, as is surely well known, is the federal program providing health insurance for all elderly, as well as the disabled. The Medicare statute has three parts relevant in this case: Part A provides direct “fee for service” hospital payments; Part C is an alternative option providing eligible beneficiaries an opportunity to enroll in private health insurance plans; and Part E includes the formula for calculation of the disproportionate share percentage — the added compensation for the treatment of a disproportionate number of low-income patients.1
The size of this adjustment is determined by adding together two fractions. The first fraction, referred to as the Medicare fraction, measures the percentage of all Medicare patients (regardless of means) who are low income, i.e., entitled to supplemental security income benefits. Mathematically, the numerator of this fraction is the number of “patient days” for patients who were “entitled to benefits under Part A and were entitled to supplemental security income benefits.” The denominator is the total number of “patient days for such fiscal year which were made up of patients who (for such days) were entitled to benefits under Part A.” 42 U.S.C. § 1395ww(d)(5)(F)(iv) (emphasis added).
The second fraction accounts for the number of Medicaid patients — who, by definition, are low income — not entitled to Medicare. The numerator is the number of patient days attributable to patients who (for such days) were eligible for Medicaid, but “not entitled to benefits under [Medicare] Part A.” The denominator is the total number of patient days, regardless of whether the patients were enrolled in a federal medical benefits program. Id.
The statutory interpretation question that led to this case is whether enrollees in Part C are “entitled to benefits” under Part A, such that they should be counted in the Medicare fraction, or whether, if not regarded as “entitled to benefits under Part A,” they should instead be included in the Medicaid fraction. As it turns out, if Part C beneficiaries are included in the Medicaid fraction rather than the Medicare fraction, the hospitals receive a great deal more compensation.
*137As we have previously recognized, the phrase “entitled to benefits under Part A” is ambiguous. Northeast Hospital Corp. v. Sebelius, 657 F.3d 1, 13 (D.C.Cir.2011). Because a Part C enrollee must, by definition, have been eligible for Part A, it could mean one was legally entitled to Part A benefits whether or not one chose Part C’s option, or it could mean only those who did not choose Part C, and, therefore, remained legally entitled to Part A benefits. In other words, someone who chose Part C nevertheless could still be “entitled” to Part A within the meaning of the statute.
Prior to 2003, the Secretary treated Part C patients as not entitled to benefits under Part A. Id. at 16-17. They then should have been included in the Medicaid fraction. But there was, apparently, considerable confusion among the hospitals, and since the disproportionate share percentage was calculated by fiscal intermediaries (insurance companies) using privacy protected patient data, the hospitals were unable to confirm that reimbursement rates were correct.
The Secretary, recognizing the confusion, issued a notice of proposed rulemaking, explaining:
We have received questions whether patients enrolled in an M + C Plan[2] should be counted in the Medicare fraction or the Medicaid fraction.... The question stems from whether the M + C plan enrollees are entitled to benefits under Medicare Part A since M + C plans are administered through Part C.
We note that, under 422.50, an individual is eligible to elect an M + C plan if he or she is entitled to Medicare Part A... .However, once a beneficiary has elected to join an M + C plan, that beneficiary’s benefits are no longer administered under Part A. Therefore, we are proposing to clarify that once a beneficiary elects Medicare Part C, those patient days attributable to the beneficiary should not be included in the Medicare fraction of the DSH patient percentage. These patient days should be included in the count of total patient days in the Medicaid fraction (the denominator), and the patient’s days for the M + C beneficiary who is also eligible for Medicaid would be included in the numerator of the Medicaid fraction.
Medicare Program, Proposed Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2004 Rates, 68 Fed.Reg. 27154, 27208 (May 19, 2003).
The Secretary further explained, in estimating the financial impact of the proposal, that “there should not be a major impact associated with this proposed change.” 68 Fed.Reg. at 27416. Only a smattering of hospitals even bothered to comment; their commentary totaled just 26 pages, and a number of them did not understand the proposal.
The next year the Secretary announced a final rule adopting the exact opposite interpretation of the statute. Medicare Part C beneficiaries, according to the rule, were to be counted in the Medicare fraction because “they are still, in some sense, entitled to benefits under Medicare Part A.” Medicare Program: Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2005 Rates, 69 Fed.Reg. 48916, 49099 (Aug. 11, 2004) (emphasis added).3
*138The rule change had enormous financial consequences for the hospitals. Apparently Part C beneficiaries are rarely entitled to SSI payments or eligible for Medicaid. Thus, by including Part C beneficiaries in the Medicare fraction, the denominator (total patient days for Part A eligible patients) is increased, without having a significant impact on the numerator (total patient days for Medicare patients who are also receiving SSI payments). This has the effect of diluting the fraction and significantly reducing reimbursement rates. By contrast, if the Part C patients are counted in the Medicaid fraction, there is no effect on the denominator (total patient days) and a small effect on the numerator (Medicare Part A eligible patients who are also eligible for Medicaid).
In Northeast Hospital Corp. v. Sebelius, a number of hospitals challenged the Secretary’s rule, arguing that it was an impermissible interpretation of the statute and that it could not be retroactively applied to the fiscal years 1999 through 2002. We held that the Medicare statute did not unambiguously foreclose the Secretary’s interpretation of the statute. 657 F.3d 1, 13 (D.C.Cir.2011). In other words, the Secretary’s interpretation passed Chevron step one.4 We did not reach the question whether the Secretary’s interpretation was reasonable under step two. We held that, even if the Secretary’s interpretation was reasonable, that interpretation could not be applied retroactively to the years at issue in the case because, prior to issuing the rule, the Secretary had a settled practice of not counting the Part C days in the Medicare fraction. Id. at 17. Accordingly, after Northeast Hospital, the validity of the Secretary’s rule as applied to future years remained an open question.
When the Secretary, in 2009, published reimbursement calculations for FY 2007 (one of the future years), the petitioners learned that their payments would decrease by tens of millions of dollars per year. The hospitals challenged these calculations before the agency, and ultimately appealed to the district court. The court held that the 2004 rule was invalid on two grounds: It was not a logical outgrowth of the proposed rule, and it did not adequately acknowledge and justify the Secretary’s change in policy. The court vacated the rule and ordered the Secretary to recalculate the hospitals’ reimbursements by counting Part C days under the Medicaid fraction. This appeal followed.
II.
An agency may promulgate a rule that differs from a proposed rule only if the final rule is a “logical outgrowth” of the proposed rule. Ass’n of Private Sector Colleges & Universities v. Duncan, 681 F.3d 427, 442 (D.C.Cir.2012). A final rule is a logical outgrowth if affected parties should have anticipated that the relevant modification was possible. CSX Transp., Inc. v. Surface Transp. Bd., 584 F.3d 1076, 1080 (D.C.Cir.2009) (citations omitted).
The Secretary points out that the 2003 notice proposed to codify one of only two possible interpretations of the statute: Under the Medicare statute, a Part C beneficiary is either entitled to benefits under Part A, or not. Therefore, the Secretary argues, the hospitals should have been on notice that the Secretary might adopt either interpretation. The hospitals counter *139by arguing that the notice did not actually “propose” adopting a rule; rather, the notice proposed merely to “clarify” an existing practice. There is nothing in the text of the notice, the hospitals argue, to suggest that the Secretary was thinking of reconsidering a longstanding practice. Moreover, the notice indicated that “there should not be a major impact associated with this change.” 68 Fed.Reg. at 27416. Although the government’s argument is not insubstantial, we agree with the hospitals in light of the regulatory context that we have just described.
This case is similar to one we decided in 2005. In Environmental Integrity Project v. E.P.A., the EPA issued a notice in which it “proposed to codify” an interpretation of a regulation that the agency had applied in previous adjudications. 425 F.3d 992, 994 (D.C.Cir.2005). In its final rule, however, the agency adopted an interpretation precisely opposite to the one it had proposed codifying. We held that this was unlawful, explaining that there was no indication in the notice that the agency was open to reconsidering the interpretation that it has previously adopted through adjudication. Id. at 998. We said that agencies may not “pull a surprise switcheroo on regulated entities.” Id. at 996.
So, too, here. The hospitals should not be held to have anticipated that the Secretary’s “proposal to clarify” could have meant that the Secretary was open to reconsidering existing policy. The word “clarify” does not suggest that a potential underlying major issue is open for discussion.
The government would distinguish Environmental Integrity Project by arguing that the Secretary did not previously actually include Part C days in the Medicaid fraction, so it cannot be thought that she was merely endorsing a prior policy. But this argument disregards our holding in Northeast Hospital, where we explicitly stated that the Secretary did have a prior practice of excluding Part C days from the Medicare fraction. 657 F.3d at 17. Granted, we did not say the Secretary counted the Part C days in the Medicaid fraction, but the statute unambiguously requires that Part C days be counted in one fraction or the other (a Part C-enrolled individual is either eligible for Medicare Part A, or not), so the necessary implication of our opinion is obvious. Moreover, a party reviewing the Secretary’s notice of proposed rulemaking understandably would have assumed that the Secretary was proposing to “clarify” a then-existing policy, ie., one of excluding Part C days from the Medicare fraction and including them in the Medicaid fraction.
The Secretary’s estimated financial impact of its proposal — that there should not be a major impact associated with this proposed change — supports our conclusion. See 68 Fed.Reg. at 27416. If, as the government contends, the 2003 notice had actually suggested a binary choice, between maintaining a preexisting policy and reversing that policy, then the potential estimated financial impact should have been stated in the hundreds of millions of dollars. That would doubtless have triggered an avalanche of comments, in contrast to the mere 26 pages that were actually submitted.
It should be noted that since the Secretary was disposed to codify an interpretation that was favorable to the hospitals, there was no reason for the hospitals to fear that another party would offer comments opposed to such an interpretation. (There is no obvious constituency opposed to greater compensation for hospitals.) In that regard, this case differs from, for example, environmental regulation cases, where regulated industries can usually anticipate fierce opposition from environmen*140tal groups, and it might be thought prudent to submit comments in support of favorable proposed rules.
We are sympathetic to the view expressed by the Seventh Circuit that proposed rules that might seem obscure to the average reader should alert members of the regulated class to the possible options that an examination of a policy would imply. See Alto Dairy v. Veneman, 336 F.3d 560, 570 (7th Cir.2003); but see Natural Res. Def. Council v. U.S. E.P.A., 279 F.3d 1180, 1188 (9th Cir.2002). But we ask ourselves, would a reasonable member of the regulated class — even a good lawyer— anticipate that such a volte-face with enormous financial implications would follow the Secretary’s proposed rule. Indeed, such a lawyer might well advise a hospital client not to comment opposing such a possible change for fear of giving the Secretary the very idea.
In sum, we agree with the district court that the Secretary’s final rule was not a logical outgrowth of the proposed rule.
The government argues that even if the 2003 notice is inadequate, the hospitals cannot show that they were prejudiced by the procedural defect, so we should find any error harmless. The Administrative Procedure Act does tell us that reviewing courts shall take “due account ... of the rule of prejudicial error.” 5 U.S.C. § 706(2)(F). But, as the hospitals point out, the Medicare statute has no harmless error exception:
If the Secretary publishes a final regulation that includes a provision that is not a logical outgrowth of a previously published notice of proposed rulemaking or interim final rule, such provision shall be treated as a proposed regulation and shall not take effect until there is the further opportunity for public comment and a publication of the provision again as a final regulation.
42 U.S.C. § 1395hh(a)(4).
The government suggests that we ignore this explicit text, citing a few cases in which we have drawn parallels between other aspects of the APA and the Medicare statute. For example, we once stated that the Medicare statute “places notice and comment requirements on the Secretary’s substantive rulemaking similar to those created by the APA.” Monmouth Med. Ctr. v. Thompson, 257 F.3d 807, 814 (D.C.Cir.2001) (emphasis added); cf. Baptist Health v. Thompson, 458 F.3d 768, 776 n. 8 (8th Cir.2006) (holding that § 1395hh preserves the APA’s distinction between legislative and interpretive rules); Warder v. Shalala, 149 F.3d 73, 79 n. 4 (1st Cir. 1998) (same). But that the Medicare statute is similar to the APA hardly means it is identical, and the government has presented no reason to depart from the plain meaning of the text.
We need not decide this question, however, because even if the APA applied, we would reject the government’s harmless error claim. We have not been hospitable to government claims of harmless error in cases in which the government violated § 553 of the APA by failing to provide notice. The most egregious are cases in which a government agency seeks to promulgate a rule by another name— evading altogether the notice and comment requirements. See Sugar Cane Growers Coop. of Florida v. Veneman, 289 F.3d 89, 96 (D.C.Cir.2002). That sort of case can be analogized to an illegal failure to afford a formal hearing under § 554. In that circumstance, we decline to even consider whether a petitioner would be successful if it had had the benefit of a formal hearing. In rulemaking, the comment period performs an analogous function, and the government’s statement of basis and purpose *141equates to the formal decision in an adjudication.
More difficult are the cases in which an agency has relied on data or information that was not disclosed to commenters. They are troublesome because, as the Supreme Court emphasized in the seminal Vermont Yankee opinion, one of the advantages of informal rulemaking is an agency’s ability to rely on internal information in its files. Vermont Yankee Nuclear Power Corp. v. Natural Res. Def. Council, Inc., 435 U.S. 519, 556, 98 S.Ct. 1197, 55 L.Ed.2d 460 (1978). Still, we have held for many years that an agency’s failure to disclose critical material, on which it relies, deprives commenters of a right under § 553 “to participate in rulemaking.” See Air Transp. Ass’n of Am. v. F.A.A., 169 F.3d 1, 7 (D.C.Cir.1999) (citing Association of Data Processing Service Organizations, Inc. v. Board of Governors of the Federal Reserve System, 745 F.2d 677, 684-85 (D.C.Cir.1984)). Perhaps because of the possible tension between Vermont Yankee and our critical material doctrine, we have more carefully examined whether a failure to disclose such material actually harmed a petitioner. But it is sufficient for a petitioner to show that an opportunity to comment regarding an agency’s important information created “enough uncertainty” as to its possible affect on the agency’s disposition. See Chamber of Commerce of U.S. v. S.E.C., 443 F.3d 890, 906 (D.C.Cir.2006).
Our case involves a third category: whether the final rule violates the notice requirement of § 553 because it is not a logical outgrowth of the proposed rule. Even if a final rule were regarded objectively as an abrupt departure from a proposed rule, if parties directed comments to such a denouement, it might well be properly regarded as a harmless error — depending on how pointed were the comments and by who made. If the petitioner itself made such a comment, it would presumably be hoist on its own petard. And even if the comment were made by others, if it were the same point that petitioner would press, it would still presumably be non-prejudicial because all that is necessary in such a situation is that the agency had an opportunity to consider the relevant views. In other words, the concepts of logical outgrowth and harmless error merge if the final rule is, in fact, anticipated, whether or not that anticipation was objectively foreseeable. In this case, there were a few commenters who initially commented in support of the final rule, apparently not understanding its implications, and another commenter who read the proposed rule as if it were the final rule. But the tiny handful of comments mostly revealed hopeless confusion, rather than focused opposition to the final rule, so we cannot conclude that the agency’s error was harmless.
We turn to the question of remedy. The government argues that, under our precedents, vacatur was inappropriate. To be sure, although vacatur is the normal remedy, we sometimes decline to vacate an agency’s action. See Advocates for Highway & Auto Safety v. Fed. Motor Carrier Safety Admin., 429 F.3d 1136, 1151 (D.C.Cir.2005). That decision depends on the “seriousness of the order’s deficiencies” and the likely “disruptive consequences” of vacatur. Allied-Signal, Inc. v. U.S. Nuclear Regulatory Comm’n, 988 F.2d 146, 150-51 (D.C.Cir.1993). Neither factor favors the government.
First, deficient notice is a “fundamental flaw” that almost always requires vacatur. Heartland Reg’l Med. Ctr. v. Sebelius, 566 F.3d 193, 199 (D.C.Cir.2009). Second, there is no indication that vacatur would lead to disruptive consequences. This is not a case in which the “egg has *142been scrambled,” and it is too late to reverse course. Sugar Cane Growers Coop. of Florida v. Veneman, 289 F.3d 89, 97 (D.C.Cir.2002). The district court thus correctly concluded that vacatur was warranted.5
The court went further, however; it ordered the Secretary to recalculate the hospitals’ reimbursements “without using the interpretation set forth in the 2004 Final Rule.” In other words, the district court required the Secretary to affirmatively count Part C days under the Medicaid fraction for 2007. The government complains that, even if the 2004 rule is invalid, the Secretary might achieve the same result through adjudication. The government is right to object. The question whether the Secretary could reach the same result through adjudication was not before the district court; therefore the court erred by directing the Secretary how to calculate the hospitals’ reimbursements, rather than just remanding after identifying the error. Sec. & Exch. Comm’n v. Chenery Corp., 332 U.S. 194, 201, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947) (“After the remand was made, therefore, the Commission was bound to deal with the problem afresh, performing the function delegated to it by Congress.”).6
Because the deficient notice is an adequate basis to vacate the Secretary’s rule, we do not reach the district court’s alternate holding — that the Secretary had failed to adequately explain a major change in policy.
III.
We hold that the Secretary did not provide adequate notice and opportunity to comment before promulgating its 2004 rule, and so affirm the portion of the district court’s opinion vacating the rule. We reverse only the portion of the district court’s opinion directing the Secretary to recalculate the hospitals’ reimbursements using the alternate methodology.
So ordered.