3,296 Home Health Agencies Have 'Too Few Cases to Report.' Only 169 Have a Rating That Says Anything.

3,296 Home Health Agencies Have "Too Few Cases to Report." Only 169 Have a Rating That Says Anything.

Open the CMS national home health file and look at the readmission measure distribution. In the July 2026 release, 3,296 agencies have too few cases for public reporting on potentially preventable readmissions. Another 9,397 are rated "no different than the national rate." Fifty-seven are better. One hundred twelve are worse.

So out of 12,862 agencies in that distribution, 169 received a rating that distinguishes them from average. Everyone else got either a shrug or a blank.

This is not a criticism of CMS. Suppressing measures with unreliable denominators is the correct statistical choice. It is a description of what public quality data can and cannot do for the agency you actually run — and if you are a small or mid-size agency, the answer is close to nothing.

The distributions, side by side

All figures below are from the same CMS Home Health Care national data file, July 2026 release.

Measure	Better than national	No different	Worse than national	Too few cases	National observed rate
Potentially preventable readmission (PPR)	57	9,397	112	3,296	4.12%
Discharge to community (DTC)	4,178	4,360	1,640	1,826	77.71%
Potentially preventable hospitalization (PPH)	931	7,928	746	2,143	10.83%

Two patterns are worth sitting with.

First, the suppression rate is not marginal. On PPR, roughly one agency in four cannot be scored at all. On PPH it is closer to one in five. These are not fringe agencies; they are ordinary agencies whose Medicare volume is below the reliability threshold for the measure.

Second, even where agencies are scored, the measure often does not separate them. On PPR, 9,397 of the 9,566 rated agencies land in "no different." A rating that 98% of scored agencies share is not a management signal. Discharge to community behaves differently — it spreads agencies across all four buckets — which is exactly why it is the measure most worth your attention if you are trying to learn something from public data.

How many agencies are we talking about

MedPAC's March 2026 report to the Congress counts 12,234 Medicare-certified home health agencies in 2024, up from 11,356 in 2019. The national trend hides a sharp geographic one: California accounted for 2,914 of the 2024 total, against 1,533 in 2019. Outside California, the count fell from 9,823 to 9,320 over the same period — MedPAC puts nationwide growth at 1.5% annually and the non-California change at a decline of 1.0% annually.

MedPAC also reports that for-profit agencies represented 93% of agencies in 2024 while accounting for 86% of 30-day periods. Read that ratio backward and it tells you the same story the suppression counts tell: a large share of the industry is made up of agencies with small per-agency volume.

Small volume is not a quality problem. It is a measurement problem. A twelve-patient agency can deliver excellent care and will still never generate a statistically stable readmission rate. The public reporting system was built to inform consumers comparing providers, not to give a 40-census agency a feedback loop.

What you lose when you have no rating

Three concrete things, in rough order of how much they hurt.

You cannot see a trend forming

A rated agency that slides from "no different" to "worse" gets a signal, however late. A suppressed agency gets nothing, ever. Whatever is driving your readmissions is happening in a space with no instrument in it.

Referral partners fill the gap with something else

Hospital case managers and ACO network teams compare agencies. When a quality field is blank, they substitute whatever they have — responsiveness, whether the phone gets answered, how fast a referral gets accepted, how the last three discharges went. You are being evaluated; you are just being evaluated on inputs you are not tracking.

Your internal QA has no external anchor

Without a rating, internal quality review becomes self-referential. You are measuring against last quarter's version of yourself, which is a fine way to sustain a plateau and a poor way to find one.

"No different than national" is its own trap

Agencies that clear the case threshold often read a "no different" rating as a passing grade. It is closer to a null result. The rating compares your risk-adjusted rate against the national rate with enough statistical caution that ordinary variation does not get labeled. Landing in the middle bucket tells you that your confidence interval overlaps the national figure. It does not tell you that your performance is stable, that it is not deteriorating, or that the four readmissions you had last month were unavoidable.

The practical consequence is that both large and small agencies end up in the same position on this measure, for different reasons. The suppressed agency has no number. The rated agency has a number that 98% of its peers also have. Neither one can answer the question an operations leader actually needs answered, which is not "how do we compare to the nation" but "is this month worse than last month, and why."

What to use instead

The replacement for a suppressed public measure is not a better public measure. It is a leading indicator you generate yourself, at a volume high enough to be stable. For most agencies that means the interaction record — calls, messages, and scheduling exchanges — because it is the only dataset that produces hundreds of events per week even at a small census.

A forty-patient agency will not produce a reliable annual readmission rate. It will produce several hundred phone calls a month, and those calls contain the antecedents of readmission: symptom reports, medication confusion, caregiver strain, missed visits, and the specific sound of a family losing confidence. That distinction — between retrospective reported outcomes and live operational signal — is the one we have described as the difference between dashboards and surfaces.

What to do this week

1. Look up your own agency's row. Go to Care Compare and note exactly which measures show a rating and which show too few cases. Do it before your next referral partner meeting, not after. Knowing which fields are blank on your public profile is a five-minute exercise most operators have never done.
2. For every suppressed measure, name an internal proxy. Readmission suppressed? Track 72-hour post-discharge contact completion and the rate at which those contacts surface a clinical concern. The proxy does not have to be validated. It has to be countable weekly.
3. Set a floor for review volume, not a percentage. "We review 2% of calls" produces almost nothing at low volume. "We review every call flagged for clinical or complaint language, plus 20 random calls" produces a usable sample. We have argued at length why percentage-based sampling is the wrong frame for small denominators.
4. Write the one-paragraph referral answer. When a hospital asks for your readmission rate and your public row is blank, you need a prepared, honest answer: what your internal number is, over what period, on what denominator, and why the public field is suppressed. Draft it once. Do not improvise it in a meeting.
5. Check the discharge-to-community measure specifically. It has the widest spread of the three and the lowest suppression count, so it is the public measure most likely to actually tell you something about your agency.

The honest framing

Public reporting is not going to become useful for small agencies. Reliability thresholds exist for good reasons and are not going to be relaxed to accommodate low-volume providers. The realistic path is to stop treating Care Compare as a scorecard you are trying to move and start treating it as one sparse, lagging input among several.

SurfacerIQ works in the interaction intelligence category — building operational and compliance signal out of the full volume of patient and family conversations, which for a small agency is usually the only data stream large enough to support week-over-week measurement. That is the category, not a claim about your numbers.

This post is general information about federal public reporting programs and is not legal advice. Measure specifications, reliability thresholds, and reported values change with each CMS refresh; check the current release before relying on any figure here.

If your public quality row is mostly blank and you want to talk about what to measure instead, talk to us.