Long-term care stress test: the cost, the deduction, the house, and the year it ends.
Ask a couple in their late sixties what worries them about the next twenty years and long-term care makes the list, usually near the top. Ask what their financial plan says about it, and in most planning software the honest answer is nothing.
Until this week the workaround in Foundry was the same one advisors use everywhere. You keyed in an expense for the care years, then edited the life expectancy by hand to match. It did not switch on and off. It did not know the care cost was deductible. And taking it back out meant remembering to undo two edits in two different places.
The Solver now has a Long-term care stress test. This post covers what it models, the four design decisions inside it, and what it deliberately leaves out.
How likely this is, and what it costs
The Department of Health and Human Services' planning office estimates that 56% of Americans turning 65 will develop a disability serious enough to need long-term services and supports. Many will need help for less than three years. About one in five, 22%, will need it for more than five. Families pay 37% of the cost out of pocket.
And Medicare, which most clients quietly assume will cover it, does not. In Medicare's own words: "Since most long-term care is non-medical, Medicare and most health insurance, including Medicare Supplement Insurance (Medigap), don't pay for long-term care services, including care in a nursing home or in the community."
The same brief puts the average future cost at $120,900 per person, in today's dollars. That is the wrong number to plan with. It averages in the 44% who never develop a serious need. A plan has to hold up for the household it was written for, not for the average of everyone turning 65. That is why this is built as a stress test you switch on, not an expected cost spread across every plan.
What the stress test does
- One or both clients go into care at an age you choose, for a number of years you choose. Each person gets their own start age, length of care, care setting, annual cost, and cost growth.
- Each person in care dies in their last year of care.
- Living expenses can be cut during the care years.
- The home can be sold to help pay for it.
- Add as change saves the whole event into a scenario as a single row you can switch on and off.
A fresh test starts from the client, in a private nursing room, for three years, starting at 85, or at their current age if they are already past 85. Every one of those is a starting point, not a recommendation.
The cost starts from a published number
Pick a care setting and the yearly cost fills in from CareScout's 2025 Cost of Care Survey, released in March. These are national medians, in today's dollars.
| Care setting | 2025 national median | Basis |
|---|---|---|
| In-home care | $80,080 a year | Non-medical caregiver, $35 an hour, 44 hours a week |
| Assisted living | $74,400 a year | $6,200 a month |
| Nursing home, semi-private room | $114,975 a year | $315 a day |
| Nursing home, private room | $129,575 a year | $355 a day |
Costs vary widely by state and by metro, so the amount is an editable field. If you know what the facility down the road charges, type that in.
Care costs get their own growth rate, separate from the plan's general inflation, and it defaults to 5%. In the 2025 survey, assisted living rose 5% and a private nursing room rose 1%. A single year is noise in either direction, and a stress test should lean conservative. The rate is editable per person.
Here is what that compounding does. A client who is 66 today goes into a private room at 85, for three years:
| Year | Age | Care cost |
|---|---|---|
| 2045 | 85 | $327,430 |
| 2046 | 86 | $343,801 |
| 2047 | 87 | $360,991 |
| Total | $1,032,222 |
In today's dollars, three years of a private room is $388,725. In the dollars the household will actually spend, it is $1,032,222. Both numbers are true. The second one is what the portfolio has to produce.
In the Cash Flow report's drill-downs the care cost is its own line, named for the person in care, so it does not disappear into a general other expenses bucket.
The plan ends when care does
Care starting at 85 for three years means care at 85, 86, and 87, and death in the year the person turns 87. That holds even when it is later than the plan's own life expectancy. If the plan had them dying at 86, the plan extends a year to reach the end of care.
This is the decision we spent the most time on, and there are three reasons for it.
It is the stress version. A spell of care that ends in recovery costs less and tests less. The useful question is whether the plan survives the expensive case, and ASPE's 22% tail is exactly that case.
It keeps every report in agreement. Life expectancy is read in dozens of places across the engine: the death year, survivor income, Social Security, the estate. Rather than teach each of them about care, the stress test overrides life expectancy once, before the projection starts. Every consumer then agrees on the same year by construction. The Cash Flow report's years and ages, the life expectancy markers on the charts, and the estate's death years all follow the scenario's lifespan.
It switches off cleanly. A shortened lifespan entered as a separate change would stay behind when you turned the care off. Because the death is part of the care event, turning off the event brings the original lifespan back with it.
For a couple, the plan carries on for the survivor after the first death. Their spending picks back up wherever the base plan has it.
The deduction most spreadsheets skip
Long-term care is a medical expense. Under IRS rules, care for a chronically ill person under a plan of care from a licensed practitioner counts as medical care, and so does the full cost of a nursing home when medical care is the main reason for being there (Publication 502). Medical expenses above 7.5% of AGI are deductible, as an itemized deduction.
Foundry now applies that deduction. It shows in the tax detail under Other Itemized as Medical expenses above 7.5% of AGI, and it counts only in years where itemizing beats the standard deduction.
The deduction has a circularity, and handling it is most of the work. To pay the care bill, the plan withdraws money. If that money comes out of an IRA, it raises AGI. A higher AGI raises the 7.5% floor, which changes the deduction, which changes the tax, which changes how much has to be withdrawn. Foundry resolves this inside the same convergence loop that already sizes every year's withdrawals against that year's tax, so the withdrawal and the deduction settle together. There is a test whose job is to rerun an IRA funded care year and fail if anything moves.
With real numbers: in 2045 the care costs $327,430. Say the year's AGI comes to $400,000, most of it the IRA withdrawal that paid the bill. The floor is 7.5% of that, $30,000. The deduction is $297,430.
Which leads to a planning point that is easy to miss. The care years can be the cheapest years in the whole plan to draw down pre-tax money. Most of what comes out of the IRA to pay for care is offset by the deduction. That is worth knowing before recommending a large Roth conversion: a dollar converted today at 24% may be a dollar that would have come out nearly tax free in a care year. Whether that matters for a particular household is exactly what running the plan with the care test on and off will show you.
Two caveats, both about AGI. The medical deduction is itemized, so it lowers taxable income and does not lower AGI. Anything keyed to AGI still sees the full withdrawal. That includes the taxable share of Social Security, and Medicare premiums two years later through the IRMAA lookback. Foundry models both.
And one caveat about the law. Foundry treats the full care cost as deductible. For a nursing home that is usually right. For assisted living, how much is deductible can depend on the facts of the case, so read the deduction as a planning estimate, not as a tax return.
Spending during care
Tick Cut living expenses during care and the household's living expenses are cut by 100% in the care years. The percentage is editable. The cut applies only to living expenses, so planned items like education and insurance premiums keep running, and it can never touch the care cost itself. It applies in any year that at least one person is in care, and spending resumes afterward.
For a single client moving into a nursing home, 100% is about right. The facility is the housing and the food. For a couple where one spouse goes into care, 100% is wrong. The spouse at home still has a house, groceries, and a car. Set the cut to what actually stops.
Where the cut is applied matters as much as its size. It sits at the one point every household expense calculation passes through, so the lower spending shows up in the cash flow, the withdrawals, and the deductions all at once. A cut applied anywhere else would show less spending while the plan went on withdrawing the full amount.
Selling the house
A lot of families pay for care with the house. Tick Sell the home and choose:
- Home. Defaults to the primary residence.
- Sale year. Defaults to the year care starts.
- Price. The projected value in that year, or a custom amount.
- Selling costs. Default 6%.
A preview shows the projected value in the sale year, the mortgage left, the selling costs, and the estimated cash to the household before tax. When the projection cannot supply one of those figures, the preview shows a dash. It never prints a guessed $0.
The sale then runs through the same machinery as any other sale in Foundry. The home sells at the start of the year. A primary residence gets the home sale exclusion, and a rental sold to pay for care does not. The gain lands in that year's tax, property tax stops, the proceeds go to cash, and every loan secured by the property is paid off at closing.
That last point is a fix that applies to every sale, not just this one. Until this week a full sale paid off one linked loan. A house carrying both a mortgage and a HELOC kept paying the HELOC after the house was gone. A full sale now pays off every loan secured by the property. If one of your plans sells a property with two loans against it, its projection changed this week, and it is now right.
On and off, side by side
The stress test lives in a scenario, never in the base plan. In the base case the Add as change button is disabled, with a hint to pick or create a scenario first.
Inside a scenario, Add as change saves the whole event as one row on the Changes tab. Switch that row off and everything comes back together: the lifespan, the spending, and the house. Click the row to edit the event in a dialog. The presentation's assumptions page lists the event too, so a client deck says plainly what was assumed.
The comparisons worth running are simple. The base plan against the care scenario. Care paid from the portfolio against care paid partly by the house. One spouse in care against both. Each is a scenario, and each is one switch.
What it does not do yet
- Insurance. Long-term care policies are the next piece: traditional standalone coverage and LTC riders on a life policy, paying against the care cost during the event, with a rider drawing down the life policy's death benefit. Until then the stress test is the uninsured case. That is also the number that tells you whether a policy is worth having.
- Medicaid. Spend-down and partnership asset protection are not modeled.
- Partial care and recovery. Care starts on January 1 of the start year and runs to the end of the person's life. There is no mid-year start and no recovery.
- Local costs. The presets are national medians. Type in the local number.
What to say at the table
The version that lands is mostly arithmetic.
"We ran your plan with three years in a private nursing room starting at 85. In today's money that is about $389,000. By 2045, at the pace care costs have been rising, those three years come to just over a million dollars. Here is your plan with that in it, and here it is without. Here is what selling the house covers, and here is what is left for the kids either way."
That is a conversation about a decision: whether to self insure, whether to buy coverage, whether the house is the reserve. It is a much better conversation than "don't worry, you'll probably be fine."
The long-term care stress test is live now in the Solver for every Foundry Planning firm. If you want to see it on a real household, pricing is here.
Related reading: the IRMAA cliffs and the two year lookback, which the care years tend to trip, and what a Plan Confidence number actually counts.
Working notes on planning, written for the people who do the work. If there's a planning conversation you'd like covered, support@foundryplanning.com.