Your nursing home's staffing rating is calculated on a Wednesday

Nurse staffing in American nursing homes falls 18% between Wednesday and Sunday. The federal rating families use to compare facilities is a quarterly average, so it reports neither day.

When a family needs to choose a nursing home, most reach for the same number: hours per resident day. It sits on Nursing Home Compare, the federal government’s comparison tool. It looks like a precise measure of how much nursing care a facility provides.

It is a quarterly average. Ninety-one days, one number.

That averaging treats all seven days of the week as interchangeable. A facility that staffs five nurses on Wednesday and three on Sunday reports a number somewhere in between. The Wednesday disappears. The Sunday disappears. What remains is a figure that doesn’t accurately describe either day.

CMS collects daily staffing data through the Payroll-Based Journal system and publishes it. This paper uses ten quarters of it: every Medicare-certified nursing home in the country, every day, for three and a half years, 10.8 million facility-days across 15,829 facilities. I have not found a published analysis at this resolution.

Coverage falls 18% from Wednesday to Sunday

Coverage peaks on Tuesday and Wednesday, at a national median of 3.96 nurse hours per resident. It stays near that level through Thursday. Then it falls.

Friday drops 3.4% from the Wednesday peak. Saturday drops another 14%. Sunday is the low point: 3.24 hours per resident. An 18.2% decline from what the facility provided on Wednesday.

Three days carry full coverage. Four days do not. The quarterly average blends them into one number.

This matters for a specific reason: that average is the number used in the federal star rating system, cited by hospital discharge planners choosing post-acute partners, and consulted by families making one of the highest-stakes decisions they will make for an aging parent. None of those readers know they are looking at a Wednesday number.

What it means in a real facility

Take a nursing home with 78 residents, which is the national median. On Wednesday, that facility provides roughly 309 nurse hours. On Sunday, it provides about 253. That is 56 fewer hours. Six eight-hour shifts gone from the building.

Every Sunday. Every week. At nearly every facility in the country.

Over a year, weekends alone account for approximately 5,200 fewer nurse hours than the midweek schedule would suggest. That is the equivalent of 2.5 full-time nurses’ worth of care that appears in the quarterly average and not in the actual weekend.

The gap is not concentrated in a particular type of facility. It appears in for-profit and nonprofit facilities, in 1-star and 5-star facilities, in large chains and independent operators. Across all 10 quarters in the dataset, the median weekend drop held within a 1.5 percentage point band. It did not narrow.

43% of compliant facilities fail the standard every weekend

In April 2024, the federal government finalized a minimum staffing standard for nursing homes: 3.48 total nurse hours per resident day. Measured against daily data rather than the quarterly average:

• 77% of facilities meet 3.48 HPRD on weekdays
• 37% of facilities meet 3.48 HPRD on weekends
• Of every facility that clears 3.48 on its quarterly average, 43% fall below that same standard every weekend

That is 4,421 facilities. Approximately 391,000 residents live in them.

They are not in violation of anything. The measurement says they comply. The measurement is a weekly average.

Congress subsequently blocked the standard from taking effect until 2034. The measurement question this finding raises is independent of that outcome: a floor measured as a weekly average is not a floor. Whatever standard is eventually enforced, the compliance picture it produces depends entirely on which days the measurement covers.

California drops half as much as every other state

California’s nursing homes drop 10.9% on weekends. Every other state averages 16.7%. The gap of five and a half percentage points held in every quarter in the dataset without exception.

California is the only large state with mandatory minimum staffing ratios.

This is not proof that minimums caused the smaller drop. But it is the closest available approximation of a controlled experiment, running for three years, in real facilities with real residents. The pattern is large enough to take seriously. The largest drops in the dataset belong to Alabama and Mississippi, both at 20.2%, which is nearly double California’s. For a facility with 78 residents, the difference between a California weekend and an Alabama weekend is approximately 22 fewer nurse hours every Saturday and Sunday.

The COVID artifact

In spring 2020, nursing homes were in crisis. Residents were dying. Families were pulling loved ones out. Admissions froze.

The staffing metric improved. National average HPRD rose 4% in 2020 Q2.

It did not improve because facilities hired more nurses. It improved because residents left. The same nurses covering fewer people produced a better-looking ratio during the worst quarter nursing homes had faced in decades. A metric that improves when residents flee during a public health crisis is not measuring what people think it measures.

What this does and does not support

This analysis does not prove that lower weekend staffing causes harm. Connecting staffing patterns to falls, infections, or hospitalizations requires joining this data to inspection deficiency and Medicare claims records, work that follows from this foundation rather than being part of it.

What it shows is that the number families use to evaluate a nursing home reflects three days of the week accurately and smooths over four. The days with the least coverage are the days when families visit least, when administrative oversight thins, and when the gap between a 1-star and a 5-star facility quietly widens.

The full technical paper, including methodology, limitations, and code for reproduction, is available as Working Paper No. 01.

If you are choosing a nursing home, ask about the weekend. The rating on the website was calculated on a Wednesday.

Data: CMS Payroll-Based Journal Public Use Files, 2017Q1–2020Q2. 10,832,137 facility-days, 15,829 facilities. All code and data snapshots: zatara-research-data/01-weekend-staffing-cliff. Full technical paper: Working Paper No. 01.

@techreport{zatara2024paper1,
  title       = {Weekend Staffing Shortfalls Hidden by Quarterly Averages: Evidence from 10.8 Million Nursing Home Facility-Days},
  author      = {Zatara, David},
  year        = {2024},
  institution = {Independent Researcher},
  type        = {Working Paper},
  number      = {1}
}