Nursing Shortage Solutions in 2026: Documentation Is the Lever You Control

For four years, the answer to “why is my building short-staffed” has been the same: nobody can hire. That answer expired this summer.

Nursing care facility employment in the U.S. passed its pre-pandemic level in June 2026 — the first time since February 2020, according to the Bureau of Labor Statistics’ own employment series. The workforce came back.

And your building probably still feels short.

Both of those things are true at once, and the space between them is the most useful thing an administrator can understand right now. If you have your people back and the floor still doesn’t work, the binding constraint isn’t headcount any more. It’s what those hours get spent on.

Has the nursing shortage actually ended?Copy link

No. But it has changed shape, and the old shape is the one nearly every article on this topic is still describing.

Here is the BLS nursing care facilities series, seasonally adjusted:

Worth noting what this does to the number everyone has been quoting. The familiar “nursing homes lost around 15%, roughly a quarter of a million workers” claim computes almost exactly off this same series — at the March 2022 trough. It was accurate. It is now four years out of date, and it describes a hole that has since been filled.

The industry’s own association says so, and says it more bluntly than we would. When AHCA/NCAL announced the milestone in July 2026, president Clif Porter made the point himself. He went on to note that demand for caregivers is growing faster than the available labor pool.

That is the loudest incumbent voice in this space conceding the point. The recovery happened. The relief didn’t.

Why does your building still feel short-staffed?Copy link

Three reasons — and none of them is “you’re imagining it.”

The official churn number flatters everyoneCopy link

CMS folded staff turnover into Five-Star in 2022, but its definition excludes any staff member below a 120-hours-in-90-days threshold. Petterson and colleagues rebuilt the measure in Health Affairs Scholar in April 2026, counting all new hires across roughly 24,000 facility observations. By 2022–2023, 45% of nursing hires fell outside CMS’s definition. Counting them, turnover ran about 19 percentage points higher than the published figure for nursing staff overall, and about 16 points higher for RNs. Only 30% of facilities landed in the same decile under both measures.

So the revolving door you experience is real, and part of it is genuinely invisible on your own Care Compare page. That matters for more than morale: a workforce that turns over constantly is a workforce permanently re-learning your charting.

Turnover is improving, and still severeCopy link

The 2025–2026 Nursing Home Salary & Benefits Report, summarized by AHCA from a survey of 917 nursing homes, puts CNA turnover at 42.34%, down from 44.16%. Sign-on bonus usage fell from 65.44% to 55.61% — facilities are buying their way in less often.

Real progress. Also a building still replacing four in ten aides a year, before you account for the churn CMS doesn’t count.

The demand curve is the part that doesn’t resolveCopy link

This is the honest answer to “won’t hiring sort itself out eventually?” For RNs, partly. For the licensed staff skilled nursing leans on hardest, no. HRSA’s federal workforce projections, as reported by the American Organization for Nursing Leadership, have LPN/LVN supply meeting 80% of demand in 2027, falling to 64% by 2037. These are model projections rather than measurements, and the underlying vintage runs on data through roughly 2022 — but the direction is the point. If you can’t buy more LPN hours, the only remaining lever is making the hours you have go further.

One assumption worth dropping. Not every facility with a documentation problem is short of people. Plenty of buildings carry strong staffing hours and low turnover and still take survey findings on care planning, resident assessment and medical records. Whether your bench is thin or fully staffed and still drowning, the hours are going to the same place.

What happened to the federal staffing mandate?Copy link

It’s gone. CMS published the repeal of the minimum staffing standards (90 FR 55687) on December 3, 2025, effective February 2, 2026. The rule follows a change in public law that bars HHS from implementing or enforcing the relevant provisions until September 30, 2034.

What that removes, relative to the 2024 rule: the 3.48 total nursing hours per resident day requirement (including 0.55 RN hours and 2.45 nurse aide hours), and the 24/7 on-site RN requirement — reverting to the prior standard of an RN for at least 8 consecutive hours a day, 7 days a week.

Put those together and the position is unusually clear. No federal floor is going to force anyone’s hand for the better part of a decade. Whatever staffing relief your building gets in 2026, you are going to have to engineer it yourself.

Where do your nursing hours actually go?Copy link

Into the chart. Schwabe and colleagues open their 2026 study by noting that nurses in long-term care spend up to one-third of their working time on documentation. Treat that as a range-topping figure attributed to the literature rather than a measured average for your building — and note the strongest evidence behind it is European.

Even discounted, the arithmetic is hard to ignore. A facility at full headcount is already running something close to a third of its nursing payroll through a keyboard. That is a larger and far more controllable pool of hours than anything recruitment will realistically add in a year — and unlike hiring, it doesn’t depend on a labor market you don’t control.

Does AI documentation actually give those hours back?Copy link

The honest answer: the evidence is real, it is thinner than the marketing implies, and almost all of it was measured somewhere other than a skilled nursing facility. Here is the whole of it, graded.

The one long-term care studyCopy link

Schwabe et al., published in the Journal of Medical Internet Research in April 2026, observed 52 registered nurses across 14 German long-term care facilities for 770 hours, before and after deploying a mobile AI speech assistant. Documentation time per morning shift fell by an adjusted mean of 15 minutes — roughly a 28% reduction (95% CI −21.75 to −8.23, P < .001). Self-reported documentation time and interruptions both fell; satisfaction with the documentation system improved.

The limits belong in the same breath as the number. This was a single-group pre-post study with no control group, in German long-term care rather than U.S. skilled nursing, on one vendor’s product, and the first author is employed by the company that makes it. Workplace satisfaction showed no significant change.

It is worth stating what 15 minutes is: per morning shift. Not per resident, not per hour, not per day.

It’s also worth comparing that peer-reviewed ~28% against the “up to 60%” you’ll see in long-term care documentation marketing — a figure that traces back to a single customer’s own before-and-after numbers. A carefully measured 28% is the more useful planning figure, precisely because someone measured it.

The strongest study design is in the wrong settingCopy link

The best evidence in this field by design quality is a randomized trial — and it isn’t about nurses. Lukac et al. in NEJM AI (November 2025) randomized 238 outpatient physicians at UCLA Health across 14 specialties to Microsoft DAX Copilot, Nabla, or usual care.

• Nabla: −9.5% time-in-note versus control (P = .02).
• DAX Copilot: −1.7% versus control (P = .66 — not statistically significant).

Same trial, same category, opposite results. That is the single most useful thing in this literature for a buyer: “adopt an AI scribe” is not a strategy, because the category does not perform uniformly. Adoption was also far from total — physicians used the tools in roughly a third of eligible visits, and a time saving nobody realizes because staff avoid the tool is the ordinary real-world failure mode.

Burnout moved, in a study that can’t prove whyCopy link

Olson et al. in JAMA Network Open (October 2025) followed 263 ambulatory physicians and advanced practice practitioners across six U.S. health systems for 30 days on an ambient AI scribe. Burnout fell from 51.9% to 38.8%.

Read that as an association, not a cause. It was a quality-improvement study with voluntary participation, no control group, self-reported outcomes, a 30-day window, and a vendor-employed co-author. The reason it belongs here anyway is the mechanism it points at: burnout drives intent to leave, and intent to leave is what shows up later as the turnover in the section above.

The gap nobody has filled. There is still no published study measuring AI documentation time savings or accuracy in a U.S. skilled nursing facility. Not one. The closest evidence is German long-term care; the strongest evidence is U.S. outpatient physicians. Anyone quoting you a SNF-specific number is extrapolating, and should say so.

What separates a tool that helps from one that doesn’t?Copy link

The most useful critique of this technology came from someone with no stake in it. Charlene Ronquillo of the University of British Columbia School of Nursing, writing in JMIR in May 2026, reviewed the German study and made a point that reframes the buying decision: AI scribes redistribute cognitive effort from authoring to verification. Time spent reviewing entries and retrieving information went up even as writing time went down.

If that’s what these tools actually do, then the thing that matters is not how impressive the draft looks. It’s how fast and how safely a nurse can check it. That is a genuinely different buying criterion, and it’s worth insisting on seeing it demonstrated rather than described.

Ronquillo’s second point deserves more attention than it gets in California: speech recognition performance varies across accents and dialects, and the field mostly treats this as a caveat rather than something to measure. A tool that degrades on accented English is a staffing-relief tool that works worst exactly where floors are hardest to cover.

So, a short list of questions worth asking a vendor:

• How long does verification take, measured — not how long generation takes?
• What does the nurse see before they sign, and how much of the note do they have to re-read?
• Has speech recognition accuracy been tested across the accents actually on your floor?
• What’s the realistic adoption rate, and what happens to the promised savings at a third of eligible encounters?
• Is there any evidence from a setting resembling yours — and if not, will they say so plainly?

Where Relic AI fitsCopy link

Relic AI is our platform for long-term care; Charting Assistant is the product aimed squarely at the problem above.

We built it around verification rather than generation, for the reason Ronquillo names: the bottleneck moves to checking, so checking is what has to be fast and legible. Nothing is final until your nursing staff reviews and signs it, and notes land directly in the EHR your facility already uses. Charting Assistant is the detail on how that works in practice.

The documentation that carries survey risk is a related but distinct problem — care plans, assessments, the records that turn into findings. That’s Compliance Assistant.

We’d rather be straight with you about the evidence than quote you a number nobody has measured in a building like yours. Nobody has run that study yet, us included. If you want the wider comparison before you talk to anyone, our buyer’s guide to AI documentation software for skilled nursing covers what to look for across the category — not just here.

The practical next step doesn’t involve us at all: work out what a third of your nursing hours actually costs you, and how many of them you could plausibly get back. That number tells you whether this conversation is worth your time.

DisclaimerCopy link

This article is general information about workforce data, federal and state staffing regulation, and published research on AI documentation tools. It is not legal, regulatory, or clinical advice, and it is not a substitute for advice about your own facility’s obligations. Staffing requirements vary by state and change; confirm your current obligations with qualified counsel or your state licensing agency before acting on anything here.