IRMAA cliffs and the two-year lookback no one budgets for.

The conversation that exposes IRMAA as a planning blind spot tends to go something like this. A retired client calls — irritated, sometimes alarmed — because their Medicare premium has jumped. They don't know why. They went looking and found out the new premium tier traces back to a single tax return from two years prior, and that the year in question was the year of a Roth conversion their advisor ran for them. They want to know how something they did in 2024 can be costing them four thousand dollars in 2026.

The honest answer is that the conversion was modeled against tax rates and not against the IRMAA bracket two years forward. That's the kind of mistake the structure of IRMAA quietly invites, and the kind of mistake that becomes invisible once you've internalized the shape of it.

This is the post I'd hand to an advisor before that call ever happens.

What IRMAA actually is

The Income-Related Monthly Adjustment Amount is a means-tested surcharge added to Medicare Part B and Part D premiums for higher-income beneficiaries. It is not a tax. It does not show up on a tax return. It shows up on a Social Security benefit deduction or a direct billing notice from CMS, and it lands quietly enough that clients often don't realize what they're paying until they look at their year-end Social Security 1099.

The structure is bracketed, with a base premium for incomes below the first threshold and progressively higher surcharges in each tier above it. There are five surcharge tiers above the base, plus a separate IRMAA on Part D that scales independently. The brackets are indexed annually, but the structure is stable: same number of tiers, same MAGI thresholds applied to the same definition of MAGI.

MAGI for IRMAA purposes is straightforward — it's the AGI from the relevant tax return plus tax-exempt interest. No add-back gymnastics, no schedule-by-schedule rebuild. The thing that is not straightforward is which tax return SSA pulls.

The two-year lookback

Here is the part that breaks plans:

The IRMAA bracket applied to your premium in any given year is determined by your MAGI from two years prior.

A client paying their 2026 Medicare premium is paying based on the AGI on their 2024 Form 1040. A client paying their 2027 premium is paying on 2025 AGI. The tax return SSA has access to is the most recent one available — and the most recent one available, on the cycle SSA pulls, is two years old.

This means every dollar of MAGI you add in a given year is a dollar against the IRMAA bracket two years forward. Not next year. Two years.

For an advisor, this is a planning shape with two unfamiliar features:

1. The expensive consequence is delayed and decoupled. The conversion you run in March of 2026 doesn't affect the premium until January of 2028. By the time the client sees the bill, the decision is twenty months in the rearview.

2. It is a cliff, not a gradient. A dollar above the threshold puts the entire premium into the higher tier. The IRMAA tier doesn't phase in. There is no marginal IRMAA rate. There is the threshold, and there is the next threshold. [TODO: insert current-year bracket structure with the specific MAGI thresholds — these adjust annually, so cite the current SSA notice rather than hardcoding.]

The combination is unforgiving. Most planning conversations about Roth conversions are framed around marginal income tax rates, which are continuous functions. IRMAA is a step function with two-year delay. They do not compose into a single "marginal rate" the way you'd want them to.

Where this bites in practice

A few patterns I've seen often enough that I now check for them by default:

The conversion-fill blind spot

We model a Roth conversion that fills the 24% federal bracket and feels efficient on a spreadsheet. Two years later, the same household has crossed an IRMAA threshold by a few hundred dollars and is paying the next-tier surcharge on both spouses' Part B and Part D premiums. The marginal cost on the dollar that crossed the threshold is enormous — the entire surcharge differential, divided by the dollars over the threshold. I have seen this work out to effective marginal rates north of 100% on the last few hundred dollars of the conversion.

The fix is simple: model the IRMAA bracket two years forward as part of the conversion-sizing decision, and back the conversion off the threshold by enough to handle inflation in the bracket between now and then. Not all the way back to the threshold — far enough that a normal year of indexing won't pull the threshold below your MAGI.

The widow trap

A surviving spouse files single starting the year after the spouse's death. The MFJ IRMAA brackets are roughly twice the single brackets, so the same income that was comfortably mid-bracket joint becomes top-bracket single. If the surviving spouse's income doesn't drop materially — pension survivor benefits, Social Security survivor benefit, RMDs continuing — the IRMAA jump can be substantial and is structurally hard to plan around.

The two-year lookback makes this worse. If the spouse dies in early 2026, the surviving spouse's 2026 single-filer return drives the 2028 IRMAA bracket. Form SSA-44 is the appeal mechanism (more on that below), and "death of spouse" is a qualifying life-changing event — but the appeal has to be filed and supported, not assumed.

The capital-gain year

A client sells a property, recognizes a one-time large gain, and the IRMAA tier jumps for two years following. This is one of the cases where the appeal mechanism actually works well — but only if you know to file it. Most clients don't.

The Social Security step

Social Security benefits become taxable above provisional income thresholds that haven't been indexed for inflation since they were created in the 1980s. The taxable portion of Social Security is part of MAGI. So the year a client claims SS, their MAGI typically jumps — and the IRMAA bracket two years forward is going up with it. This isn't bad planning; it's just a thing to model rather than discover.

The appeal mechanism

Form SSA-44 ("Medicare Income-Related Monthly Adjustment Amount — Life-Changing Event") is the formal appeal. The qualifying events SSA accepts are narrower than the cases where the bracket math feels unfair:

• Marriage, divorce, or death of spouse
• Work stoppage or work reduction
• Loss of income-producing property (involuntary, e.g., natural disaster)
• Loss of pension income
• Settlement payment from a former employer

Notably not on the list: a one-time Roth conversion. A one-time capital gain. A bracket-fill year. SSA's position is that those are voluntary income decisions, and you don't get to appeal a voluntary decision after the fact.

The appeal pathway closes most of the doors that planning errors open. It catches genuine life events; it doesn't bail out a conversion that overshot.

Why most planning tools handle this badly

The tooling problem is straightforward to describe and harder to solve:

1. Premium projections require knowing the bracket structure two years forward. Bracket thresholds are indexed annually using the same inflation factor used for the base premium. A planning tool needs to project both the brackets and the MAGI two years forward to land the premium correctly.
2. MAGI projections have to round-trip through the tax engine. AGI and tax-exempt interest are downstream outputs of the tax engine. If your tool computes IRMAA off a separate income figure (gross income, taxable income, take-home), it's modeling the wrong number.
3. The bracket cliff has to actually be a cliff in the projection. Smoothed projections — say, a planner that quietly averages the surcharge across MAGI — give the wrong answer at the threshold. The point of modeling IRMAA is to see the cliff, not to round it off.

We built Foundry Planning so that IRMAA bracket projections are tied to the tax engine's MAGI output, projected to the right two-year-forward year, and computed against the cliff structure rather than a smoothed approximation. That's the bar. It is not the bar most planning tools clear.

What to tell clients

When the planning topic comes up, I tend to land on a few specific lines:

Medicare premiums are means-tested. The surcharge depends on what your tax return looked like two years ago, and it doesn't show up gradually — it steps up at brackets. We watch this every year alongside everything else.

If we run a Roth conversion this year, it will affect your premium two years from now. We model that bracket before sizing the conversion, not after.

If something happens that drops your income — you stop working, you lose a pension, your spouse dies — there is an appeal path. It is a real form, and we will help you file it. It is not something to assume will happen automatically.

The point of the conversation is not to scare anyone off MAGI-affecting decisions. It is to make sure the decision is being priced with the IRMAA bracket included in the price tag.

Closing

IRMAA is a small line in the tax stack until it isn't. The two-year lookback makes it the kind of mistake that doesn't show up until the decision is too far back to undo, which is exactly the kind of mistake you want a planning tool to catch on the way in.

If you're an advisor and your current tool handles this with a pre-baked premium estimate that doesn't sit downstream of the tax engine, the cash-flow pillar is the larger pattern worth getting right. The IRMAA cliff is the most visible symptom of a tool that approximates instead of computes — the tooling pillar walks through how I'd evaluate any candidate replacement.

If you'd like to compare notes on how we model this — or what the bracket-fill conversation should sound like when IRMAA is in the room — we should talk.

---

Primary references: Social Security Administration, "Medicare Premiums: Rules for Higher-Income Beneficiaries" (Publication 05-10536); Form SSA-44 instructions; CMS Annual Premium Notice. Bracket thresholds and base premium amounts adjust annually — verify against the current year's SSA publications before quoting figures to clients.