IRMAA Surcharge Planning (Medicare Part B/D)

How it works

IRMAA is not a tax. It is a reduction in the roughly 75 percent subsidy the federal government normally pays toward a Medicare beneficiary's Part B premium, recovered instead as a higher premium from beneficiaries whose income clears a set threshold. Congress created the Part B version in 2003, effective for premiums starting in 2007, and added a parallel Part D version in 2010. The statutory home for each is Social Security Act section 1839(i), codified at 42 U.S.C. 1395r(i), for Part B, and section 1860D-13(a)(7), codified at 42 U.S.C. 1395w-113(a)(7), for Part D.

The mechanism is a sliding scale: instead of the standard applicable percentage of about 25 percent of the full Part B cost, a beneficiary above the threshold pays 35, 50, 65, 80, or 85 percent of it, depending on how far above the threshold income falls. Part D runs a related formula off the same income tiers, against the Part D base beneficiary premium instead. What makes both expensive to miss is that this is a cliff, not a phase-in: the table is a step function of income bands, and one dollar into the next band applies that entire tier's add-on for the full year.

It is also assessed per beneficiary, not per return. A married couple filing jointly uses one joint MAGI figure to look up the tier, but if both spouses are enrolled in Medicare, each one owes that tier's full add-on separately, doubling the household cost. And the premium a beneficiary pays this year is not priced from this year's income; it is priced from a return filed two calendar years earlier, so a decision made today has no visible cost until a notice arrives roughly two years from now, long after that return is closed.

The MAGI this uses, and why it is not the MAGI from other tests

IRMAA MAGI is adjusted gross income under Internal Revenue Code section 62, with specific items added back: the foreign earned income exclusion under section 911, certain U.S.-possession income under section 931, Puerto Rico source income under section 933, and the savings-bond education interest exclusion under section 135, plus tax-exempt interest. That is broader than the MAGI used for the net investment income tax, which carries none of those addbacks, and broader than the Affordable Care Act premium-tax-credit MAGI, which adds back only foreign earned income and tax-exempt interest. A model built for either of those can understate exposure here.

What this means in Florida

Florida has no individual income tax and imposes no state-level overlay on Medicare premiums, so IRMAA is a purely federal problem here, identical to what a retiree in any other state faces. There is no state-side planning angle to add to the federal one.

Who this applies to

Every Medicare Part B or Part D enrollee is technically within reach of this rule, but the planning conversation matters most for someone near or over 63 with a controllable, lumpy income year ahead. That covers more situations than it first sounds like.

• A business sale. Any transaction large enough to move adjusted gross income for the year it closes creates a look-back problem two years later, regardless of how the deal is structured.
• A Roth conversion, including one built through a backdoor Roth or mega backdoor Roth step. Whatever portion turns out taxable lands in adjusted gross income in the year of conversion, the same as any other conversion.
• A large capital gain, whether realized outright or spread through an installment note.
• A qualified charitable distribution, for an IRA owner already 70½ or older, which runs the other direction: a common lever that lowers this year's MAGI rather than raising it.

That last item matters for a specific reason: a qualified charitable distribution is an exclusion under Internal Revenue Code section 408(d)(8)(A), so it never reaches the adjusted-gross-income line at all. A gift written as a personal check and itemized on Schedule A instead is subtracted after AGI is already computed, and does nothing for MAGI.

None of this requires the beneficiary to already be on Medicare when the income event happens. The look-back tests the return filed two years before the premium year, not concurrent enrollment. A client who sells a business at 64 and enrolls in Medicare at 66 can find a surcharge baked into the very first Part B bill.

Filing status changes who this reaches hardest. A taxpayer married but filing separately, having lived with the spouse at any point in the year, faces a substantially harsher version of the same structure, covered next. A household whose baseline income already sits comfortably under the lowest threshold every year, with no lumpy event ahead, has nothing to plan around here.

How the income is earned, or which retirement vehicle funds it, does not change any of this; IRMAA MAGI starts from adjusted gross income the way the return already reports it. Whether an S corporation suits a business heading toward a sale, or a solo 401(k) suits a self-employed client better than a SEP IRA, are separate questions I cover in my Florida S-corp guide and my solo 401(k) versus SEP IRA guide.

What it requires

The look-back period is fixed by statute: the individual's last taxable year beginning in the second calendar year before the year involved. A 2026 premium, for example, is set from the 2024 return. If the IRS has not supplied usable data for that year by October 15 of the preceding year, the statute allows a fallback to the next-older year's return instead.

For calendar year 2026, priced from 2024 MAGI, the standard Part B premium is $202.90 per month. The table below shows the monthly add-on for a single or head-of-household filer and for a joint filer, with the joint amounts at their own breakpoints:

Every row is a full-year, all-or-nothing add-on. A dollar into the next row applies that row's figures for the entire year, for each Medicare-enrolled spouse separately on a joint return.

A taxpayer married but filing separately, who lived with the spouse at any point in the year, gets no joint-return doubling at all. The statute shrinks every single-filer breakpoint, including the top one, by the same base amount, which collapses most of the structure into two live tiers: nothing at or under the base single threshold, then the second-highest tier for nearly everything above it, until the top tier takes over at its own reduced breakpoint, $391,000 for 2026 rather than the $500,000 everyone else faces. Living apart from the spouse all year avoids this compressed version entirely.

The lower breakpoints move most years, indexed to the CPI-U, though frozen for stretches before, most recently 2011 through 2017. The top-tier dollar threshold is carved out of that indexing and fixed through 2027, so a client's top-tier exposure is knowable years ahead in a way the lower bands are not.

What you need to document

The return that actually set the number

The specific taxable year the look-back points to, and each of the section 911, 931, 933, and 135 addbacks applied to it, if any apply.

Which spouse was enrolled, and from when

Because the surcharge attaches to each Medicare-enrolled beneficiary separately, the file needs enrollment dates for each spouse. A spouse not yet enrolled during the relevant premium year owes nothing for that year no matter what the joint MAGI was.

Evidence for a life-changing-event request

Form SSA-44 plus evidence tied to the specific qualifying category: a signed statement from an employer, proof of a business transfer, or a signed statement under penalty of perjury for work stoppage or reduced hours.

The amended return, if the MAGI figure changed

A retained copy of the amended return and the IRS acknowledgment or transcript confirming it, requested within three calendar years following the close of the amended tax year. Miss that window and the correction cannot be made this way.

Which dispute path applies

A disagreement with how the rules were applied calls for a reconsideration request. A disagreement that the IRS-supplied MAGI figure is itself wrong does not: that gets dismissed until the return is corrected with the IRS first.

Where it goes wrong

This is not an audit-risk item the way many strategies in this library are. It runs through the Social Security Administration's own process, not an IRS exam, and the mistakes here are structural and timing mistakes.

One of them is reusing a net investment income tax or premium-tax-credit MAGI model for this purpose. Both use a narrower MAGI than IRMAA does, and whenever the foreign earned income exclusion or one of the other addbacks above is present, that model understates the real number.

Mistaking a high-income year for a life-changing event

The seven categories that let a beneficiary substitute a more recent year's MAGI for the statutory look-back year are exhaustive: a spouse's death, marriage, divorce or annulment, stopping work or reducing hours, a non-self-directed loss of income-producing property, a pension plan's cessation, or a closure-related employer settlement. Every one reaches a decrease in income tied to a specific event, never a year that was simply, and correctly, high. A business sale, a large conversion, or a big capital gain is never itself a life-changing event, and there is no path to undo the surcharge it causes. Retirement can qualify, under the stopped-working category, but only for what it does to MAGI going forward, not to undo a surcharge the sale-year income already triggered.

Filing the wrong kind of dispute

If the disagreement is with how the rules were applied to a correct MAGI figure, a reconsideration request is the right document. If the disagreement is that the IRS-supplied MAGI figure is itself incorrect, a reconsideration filed on that basis gets dismissed; the return has to be corrected with the IRS first, and proven, before a new determination can be requested.

• Forgetting the per-spouse doubling. Modeling this for the household instead of for each Medicare-enrolled spouse understates real exposure by half whenever both are enrolled.
• Assuming a spread always helps. A gain spread thin enough to land under every threshold every year is a genuine improvement. Spread instead into a middle tier across several consecutive years, and the same gain can cost several times what a single top-tier year would have, because this prices the number of elevated years, not the total dollars recognized.
• Letting the amended-return window close. It runs three years from the close of the amended tax year and requires both the retained return and the IRS confirmation.
• Treating every threshold as if it moves the same way. The top-tier figure is fixed through 2027. The bands below it are not.

A situation where this comes up

The situation I see most often starts as an entirely different conversation: a client in their early-to-mid sixties deciding when to sell a business or how aggressively to convert a traditional IRA before required distributions begin. Medicare is not what brought them in. In an ordinary year their income sits nowhere near any threshold, built from Social Security, a pension, and ordinary portfolio distributions, so nothing here flags a problem.

What changes the conversation is the lag. The year a sale or a conversion closes is not the year the bill arrives. That year's MAGI sets a premium a notice will price roughly two years later, often after the client has stopped thinking about the return that caused it. Left unmodeled, the first the client hears of this is a letter from Social Security showing a materially higher Part B premium.

The version that worries me is the one where the number could have been managed and nobody looked. A gain that could have been sized under a threshold instead clears it by a small margin and locks in a full year of the next tier over very little. Or a gain gets spread across an installment note without anyone checking what that spread does here, and it turns out more expensive for this one purpose.

None of this changes whether the underlying transaction should happen. A sale large enough to move this needle is usually still the right decision for the reasons that made it right in the first place. What belongs in the file before the closing date, not after, is the number this creates two years out, so it lands on the client's desk as something already expected, not a letter that reads like a mistake.