How We Grade, The ElderGrade Methodology, In Full

The five components

A home starts from 100 points spread across five components. Points are deducted, not awarded, in the two components where the federal record shows something went wrong.

Inspection safety

30 pts

Citations from the last three years of state health inspections, weighted by how much harm CMS says they caused. An immediate-jeopardy finding costs far more than a documentation slip, and total citation volume is compared against the national median rather than a fixed threshold.

Staffing

30 pts

Total nursing hours per resident per day, adjusted for how sick the residents actually are, scored against national percentiles. RN hours and weekend consistency are scored separately, because a good weekday average can hide a thin Sunday.

Staff stability

15 pts

The share of nursing staff who left over the last year, plus administrator departures. Continuity is care: a caregiver who knows a resident notices when something is wrong.

Resident outcomes

15 pts

The CMS quality-measure rating, built from clinical outcomes such as pressure ulcers, falls with injury, infections and antipsychotic use.

Enforcement history

10 pts

Federal fines and denials of payment. A payment denial means CMS stopped paying for new admissions, a serious step regulators do not take lightly.

Findings that cap a grade outright

Some things should not be averaged away by good staffing. These are hard ceilings, applied after the points are totalled:

Special Focus Facility → capped at 54 (grade D)

CMS placed the home in its programme for persistently poor care.
Abuse icon on record → capped at 64 (grade C)

CMS flags the home for citations of abuse that harmed residents, or had potential to.
Immediate jeopardy in the last 3 years → capped at 64 (grade C)

Inspectors found a failure likely to cause serious injury, harm or death.
Immediate-jeopardy fire safety citation in the last 3 years → capped at 64 (grade C)

CMS inspects life safety separately from health, under its own K-tags. Almost every home in the country carries some fire citation, so the count tells you little, what separates homes is severity, and a life safety failure at jeopardy level endangers the same residents as one found on the health survey, so it carries the same ceiling.
Fire safety citation causing actual harm in the last 3 years → capped at 74 (grade B)

A step below jeopardy: harm occurred, but inspectors did not judge it likely to cause death or serious injury.

The letter bands

Grade	Score	Meaning	Homes	Share
A	85+	Excellent	2,616	17.8%
B	70+	Above average	6,700	45.7%
C	55+	Average	4,318	29.5%
D	40+	Below average	1,022	7%
F	0+	Poor	1	0%

Note the shape: most homes land at B or C. A grading system where everyone gets an A tells you nothing, and one where everyone fails is just as useless.

The benchmarks, recomputed each release

Staffing thresholds are not hard-coded. They are percentiles of the current national population, so the curve moves when CMS republishes rather than drifting out of date.

Staffing p25	3.27
Staffing p50	3.77
Staffing p75	4.43
Staffing p90	5.24
Rn p50	0.58
Rn p75	0.81
Deficiency median	0

How we grade hospice providers

Hospice is a different service with a different federal record, so it gets a different formula. CMS publishes a Hospice Care Index, but it is a count of how many of ten indicators a provider is not in the worst group on, 55% of hospices score 9 or 10 out of 10, so on its own it separates almost nobody. We read the ten underlying indicators at the percentile CMS publishes for each, and weight them like this:

Component	What it reads	Points
experience	The CAHPS Hospice Survey, what bereaved family caregivers reported afterwards	30
presence	Visits by a nurse or social worker in the last three days of life	25
nursing	Nurse minutes per routine home care day, and gaps between visits	20
discharge	How often patients are signed off hospice alive, and burdensome transitions	15
process	Whether pain, breathlessness, treatment wishes and beliefs were addressed at admission	10

Three indicators we deliberately left out

Before weighting anything we checked which direction each indicator actually runs, by comparing it against the family survey across every hospice in the country rather than assuming. Three did not survive that check and are shown on provider pages but excluded from the score:

Per-beneficiary spending runs backwards. Hospices in the top quarter for Medicare spending averaged 3.20 stars from families; those in the bottom quarter averaged 3.81. It is a cost measure, not a quality one.
Weekend skilled-nursing minutes showed no separation at all (3.66 stars against 3.65). It sounds like it should matter. In the data it does not.
CHC/GIP days and burdensome transitions type 2 are zero for most providers, so their percentiles cannot tell two hospices apart.

What caps a hospice grade

A Hospice Care Index of 6 or below caps the grade at C, and 4 or below caps it at D, however well the provider scores elsewhere. A low index means CMS flagged it on several indicators at once, and no weighting should be allowed to average that away.

How 5,412 hospices grade out

A 0
B 1,237
C 3,262
D 909
F 4

A further 1,495 providers carry no grade at all: CMS published no index, no family survey and no record of visits in the last days of life for them, usually because they are new or very small.

How we grade home health agencies

CMS already publishes a quality-of-patient-care star rating for home health, and the easy thing would be to reprint it. We tested it first, against what patients said about the same agencies in the federal HHCAHPS survey, across all 4,570 agencies where both are published. It tracks the patient survey at essentially zero. Agencies in the bottom decile of the CMS star average 3.53 patient stars and agencies in the top decile average 4.12, with the eight deciles in between wandering. So we do not grade on it. We grade on the measures that survived the same test:

Component	What it reads	Points
Patient experience	The HHCAHPS survey, what the agency's own patients reported afterwards	30
Staying home	Whether patients went on living in the community, and preventable readmissions to hospital	25
Recovery at home	The discharge function score: how much of the expected improvement patients actually reached	20
Care transitions	Whether the patient and their next doctor were handed a current medicine list	15
Starting on time	Whether care began within the window the doctor ordered	10

The measures we tested and threw out

Every candidate measure was checked for direction and strength against the patient survey across the whole country, and then again inside each agency-volume tertile so a result could not just be an artefact of size. Eight did not survive. They are shown on agency pages and excluded from the score:

Potentially preventable hospitalisation runs backwards. Agencies CMS labels "Worse Than National Rate" average 4.24 stars from patients; agencies it labels "Better" average 3.94. The inversion holds in every volume band.
Falls with major injury runs backwards, and hard: bottom quartile 3.55 patient stars, top quartile 4.12, consistent across all three volume bands. It appears to measure whether an agency records adverse events, not whether they happen.
Flu-shot documentation runs backwards too. It is paperwork, not care.
The OASIS improvement measures, better at walking, bathing, getting in and out of bed, breathing, taking medicines correctly, and timely action on medication issues, show no separation at all. The agency scores its own patients on these.
Medicare spending per episode shows no signal, exactly as it did for hospice. It is a cost measure.
Changes in skin integrity is zero for 37% of agencies, so its distribution cannot tell two agencies apart.

What caps a home health grade

The patient survey is the yardstick every other measure here was validated against, so a bottom-band survey result is the one finding no weighting of the rest should be allowed to average away. Two stars from patients caps the grade at C; one star caps it at D.

Missing inputs score 0.65 of their component rather than half, for the same reason as hospice: CMS suppresses results below a minimum denominator, so a blank tracks how small an agency is, not how bad it is, and half marks quietly demote every small provider.

How 9,085 home health agencies grade out

A 433
B 2,555
C 3,917
D 1,840
F 340

A further 3,369 agencies carry no grade at all: CMS published no patient survey and no quality measure for them, usually because they are new or very small.

When we refuse to grade

40 of the 14,697 facilities on this site carry no grade. Those homes have neither a staffing report nor an inspection rating in the current CMS release. We could put a plausible-looking C on them. We do not, because a family making a decision deserves to know the difference between "average" and "we do not know".

What this grade is not

It is not a substitute for visiting. Data cannot tell you how a place smells at 6pm on a Sunday.
It is not current to the day. CMS republishes on a cycle; a home that changed hands last month may not yet look different here.
It is not medical, legal or financial advice.
It is not for sale. No facility, chain or placement agency pays us anything, and none can change a grade.