PDGM Took 2.9 Visits Off the Average Home Health Stay. Your Phones Absorbed the Difference.

PDGM Took 2.9 Visits Off the Average Home Health Stay. Your Phones Absorbed the Difference.

The Patient-Driven Groupings Model has been in effect since 2020. Long enough that the arguments about it have stopped being about whether visit volume fell and started being about what the reduction means. MedPAC has now put numbers on both questions, and they are not the numbers most operators expect.

Here is what the March 2026 Report to the Congress actually found, and the operational consequence that follows from it — including the part where the data runs out and judgment has to take over.

The visit numbers, stated precisely

MedPAC's mandated report on the impact of recent changes to the home health prospective payment system used interrupted time-series models to separate the effect of PDGM from everything else happening in the same period. From Chapter 14:

"After adjusting for factors such as patient severity, the ITS models estimate an average of 15.9 visits under the PDGM compared with 18.8 visits without the PDGM — a difference of 2.9 visits, or about 15.3 percent fewer visits."

Read the construction carefully, because it is easy to misquote. The 18.8 is not a historical figure. It is a modeled counterfactual — MedPAC's estimate of where visits would have been in 2023 absent PDGM. The honest sentence is "15.9 versus a modeled 18.8," not "fell from 18.8 to 15.9." The unit is visits per stay.

Separately, the observed trend is steeper. Average visits per stay fell from 19.6 to 15.9 between 2016 and 2023 — a decrease of nearly 20 percent. Some of that predates PDGM, which is exactly why MedPAC ran the models.

The home health status report in Chapter 8 measures a different unit and reaches a consistent conclusion. Total in-person visits per full 30-day period declined 18.0 percent since 2019, or 1.8 visits:

Year	In-person visits per full 30-day period
2019	10.2
2020	9.2
2021	8.8
2022	8.6
2023	8.5
2024	8.4

Two caveats on that table. A "full" 30-day period is one where in-person visits meet or exceed the threshold for the assigned payment group — a number ranging from two to five visits. And 2019 predates PDGM, when payment ran on 60-day episodes; MedPAC recalculated 2019 services as 30-day periods to make the comparison possible. It is a reconstructed baseline, not an observed one.

The decline is concentrated. Therapy visits per full 30-day period fell from 4.9 to 3.8, down 21.5 percent, with most of the drop landing in 2020 — almost 20 percent in the first year of PDGM — and staying roughly flat since. Skilled nursing fell from 4.6 to 4.1, down 11.8 percent. Chapter 14 finds the same pattern per stay: 2.4 fewer therapy visits, 21.3 percent fewer.

The finding that complicates the story

If you are building a case that fewer visits means worse care, MedPAC does not give it to you. The Commission found that "despite fewer visits per stay, the quality-of-care outcomes were generally stable or potentially improved with the PDGM," and suggested the therapy decline "may reflect better alignment of home health care services with beneficiary needs."

That is not a small concession to make in a post about visit reductions, and it should change how you frame this internally. The defensible claim is not that patients are being harmed. It is that the same clinical work is being accomplished with fewer scheduled in-home contacts — which is a different statement, and one with a specific operational consequence.

The financial context reinforces that this is not a distress narrative. MedPAC reports the fee-for-service Medicare margin for freestanding home health agencies at 21.2 percent in 2024, projects 19 percent for 2026, and recommends Congress reduce the 2026 base payment rate by 7 percent for calendar year 2027. Roughly 2.7 million FFS beneficiaries received home health care in 2024, from more than 12,000 certified agencies.

Where the data ends and inference begins

Here is the part worth being straight about: no public dataset measures home health call volume. MedPAC does not collect it. CMS does not publish it. There is no verified national figure showing that phone contacts rose as visits fell, and any vendor who gives you one has made it up.

What can be said is structural. A visit is a scheduled, documented, reimbursed contact that also functions as an unscheduled information channel — the clinician in the home notices the new bruise, the medication bottle that is still full, the daughter who has stopped coming by. Remove roughly one visit in six from a stay, and the information that used to arrive incidentally through those visits does not stop existing. It has to arrive some other way, and for most agencies the only other channel is the telephone.

That is an inference, not a measurement. But it is a testable one, and you can test it against your own data this month rather than waiting for someone to publish a national number.

What to check this week

Four things, all answerable from systems you already have:

• Calls per active patient, by month, over the last two years. Not total call volume — that moves with census. The per-patient ratio is the one that tells you whether each patient is generating more phone contact than they used to.
• Inbound call volume against your visit schedule. If calls cluster on days with no scheduled visit, that is the substitution effect showing up in your ACD data.
• Who is calling. A rise in family caregiver calls rather than patient calls is a specific signal: the household is absorbing coordination work that used to happen during a visit.
• What the calls are about. Symptom questions, medication clarification, and "when is someone coming" are three different problems. The third is a scheduling failure. The first two are clinical information arriving through an unclinical channel.

A hypothetical worth working through

Consider an agency running a 400-patient census that has seen no change in complaint volume, no change in outcome scores, and no change in staffing. On paper, nothing is wrong.

Its intake line takes a rising number of calls per patient per month, most of them clinical questions, handled by non-clinical schedulers who route what they recognize and note the rest in a free-text field nobody reads. Nothing in that agency's dashboards is red. The clinical information is arriving — it is simply arriving in a channel that was never designed to receive it, and being triaged by people who were never trained to triage it.

That is the failure mode fewer visits creates. Not worse care delivered in the home, but clinical signal landing on a queue built for logistics.

The channel has to be treated as clinical

If the phone is now carrying information that visits used to carry, it needs the same seriousness as a visit: someone accountable for what is said, a way to find the calls that matter, and escalation that does not depend on whether the person who answered happened to recognize the significance of what they heard.

That is a shift from reviewing calls after something goes wrong to monitoring the channel while it is still in front of you. It is also why the early-warning patterns matter — the phone often knows about a missed visit before the EVV data does, because someone called to ask where the aide was.

Where tooling fits

Reading a phone channel the way you would read a visit note means every call has to be legible, not just the sampled ones. SurfacerIQ operates in the category of call intelligence platforms for healthcare contact centers, where conversations are transcribed, categorized, and made searchable so clinical and operational signal is retrievable rather than remembered. The general capability worth having is the ability to ask a question of your call volume and get an answer from all of it.

If you are trying to work out what your phone channel absorbed as visits came down, talk to us.