The Referral You Never Called Back: Why Roughly a Third of Home Health Referrals Never Become Care

A referral lands at 4:15 on a Friday. Intake logs it. Someone calls the home number at 5:05, gets voicemail, leaves a message. Monday it is still in the queue, one of nineteen. Tuesday someone reaches a daughter who says her mother is "doing okay for now," and the record is closed as patient declined.

Nobody did anything that would get written up. The referral source never heard back. And the patient never received home health care.

That is not an outlier story. It is close to the base rate.

What the research actually measured

Two studies are worth knowing cold, because they measure the thing most agencies never measure about themselves: whether a referred patient ever actually received care.

The first, published in the Journal of the American Medical Directors Association in 2021, examined 724,700 hospitalizations from October 2015 through December 2016. Its finding: almost 1 in 3 patients discharged from a hospital with a discharge status of home health does not receive home health care — a 29% incomplete referral rate. The rate swung hard by clinical condition, from 17% among joint and musculoskeletal patients to 38% among digestive and endocrine patients. Patients whose referrals were completed had 1.4 percentage points lower mortality and 2.4 percentage points lower readmissions.

The second is more recent and narrower. A 2024 study in the American Journal of Managed Care, drawing on 2021–22 data from a single state, found that 26% of referred Medicare Advantage members — 761 of 2,876 discharges — never received services. Ninety-day readmissions ran 24% in the unfulfilled group versus 16% in the fulfilled group. One-hundred-eighty-day mortality was 14% versus 11%.

Be precise about what these are and are not. Both measure unfulfilled referrals: a patient was referred, and care never happened. Neither is a measure of any single agency's intake conversion rate, and neither should be restated as a conversion statistic. The honest version is the plain one: roughly a quarter to a third of hospital home health referrals never result in care being delivered. Mind the vintage, too — the JAMDA data is now a decade old, and the AJMC data covers one state and one payer type. They point the same direction anyway, five years apart, in different populations.

The gap is nobody's exclusive fault. It spans discharge planning, the patient and family, the payer, and the agency. But the agency is usually the only actor on that list with a phone queue and someone whose actual job is to close the loop.

The industry measures everything after admission and almost nothing before it

Once a patient is on service, the measurement is relentless. CMS publishes a national Discharged to Community observed rate of 77.71% in the July 2026 Care Compare refresh, alongside a wall of outcome and process measures every agency watches monthly. MedPAC's March 2026 report to Congress counts about 2.7 million fee-for-service Medicare beneficiaries who received home health care in 2024, spread across 12,234 Medicare-certified agencies.

There is no comparable public number for the interval between "referred" and "admitted." No star rating covers it. Most agency dashboards report referrals received and admissions completed and treat the difference as a rounding error made of patients who chose otherwise.

It is not a rounding error. It is the largest unexamined pool of lost patients most agencies have, and it is almost entirely made of phone calls.

Five places referrals leak — and why none of them look like losses

1. Phone tag that never escalates

The first call goes to voicemail. So does the second. Nothing in the workflow distinguishes a referral attempted three times without human contact from one attempted once yesterday; both sit in the queue looking equally alive. The first is functionally dead and nobody has said so out loud.

What to look at: attempts per referral before first live human contact, and how many referrals reach four attempts with zero live contact.

2. Voicemails left in both directions

Families call back at 7:40 in the morning, during lunch, or after hours, and land in a general mailbox nobody is staffing. Some of those messages never get returned, and the ones that do often get returned by someone who does not know what the first conversation was about. From the family's side, this reads as an agency that is not especially interested.

What to look at: inbound calls to intake that ended in voicemail, and how many got a return call within the same business day.

3. Capacity declines that never get logged as declines

This one is the quietest and the most expensive. Staffing is tight in a ZIP code. The intake coordinator knows it. So the call with the family is subtly framed — the start of care is described as "probably next week," the aide availability is presented as uncertain, the family is asked to call back and confirm. The family goes elsewhere or does nothing. The record never says declined for capacity, because nobody declined anything. It says patient chose another provider or nothing at all.

An agency that cannot see this pattern cannot staff against it, and cannot tell the hospital which service areas are honestly open. The same blind spot drives attrition after admission, which we covered in The Hidden Cost of Patient Churn in Home Health.

What to look at: declines by ZIP code and by discipline, compared against your actual staffing map. If the two are unrelated, your decline reasons are not being recorded honestly.

4. Insurance verification and authorization stalls

The referral is real, the patient wants care, and the file goes quiet waiting on a benefits check, a signature, or a Medicare Advantage authorization. Nobody tells the family why. Nobody sets a follow-up date. Two weeks pass, and the patient is readmitted, recovers enough to refuse, or goes to a competitor who verified faster. The documentation obligations that sit around intake and referral handling — CMS Conditions of Participation and state visit-documentation rules — are covered in call intelligence for home health agencies.

What to look at: referrals with no outbound contact in the last five days, grouped by the reason they stalled.

5. The referral source never learns why

A discharge planner sends ten referrals and hears back on four. She does not know whether the other six were unreachable, out of area, declined for staffing, or admitted by someone else. What she learns is that your agency is unreliable to track. The next batch goes somewhere else, and the loss shows up months later as a referral volume decline with no obvious cause.

What to look at: what percentage of closed-without-admission referrals resulted in an outbound call or note back to the referring source.

A referral leakage diagnostic you can run this week

You do not need a project for this. You need an afternoon, one person, and the last 60 days of referrals that did not convert.

• Pull every referral from the last 60 days that did not become an admission. All of them, not a sample of the interesting ones.
• Count how many have a recorded reason at all. If more than a small minority are blank, "other," or "patient declined" with no detail, stop here: your reason codes are decoration, and every number built on them is fiction.
• For 20 of them, listen to the actual calls. Every call, in order, start to finish. This is the step people skip and the only step that reliably changes minds.
• Tally what you hear into five buckets: never reached a human; reached a human and the agency went quiet; capacity steered the conversation; verification or authorization stall; genuine refusal.
• Check the clock on the first outbound attempt. Hours from referral receipt to first call, and how many first calls went out after 4 p.m. or on a Friday.
• Check whether anyone called the referral source back. On any of the 20.
• Compare the reason you heard on the call to the reason in the record. The gap between those two columns is your real finding.

If you want a single number out of the exercise, make it this one: of referrals that did not convert, what share never had a live human conversation with the patient or family? That number is usually the one that gets an executive team to move.

Why this is hard to see from inside your systems

That exercise is manual because the systems of record capture the shape of an interaction, not its content. Your CRM knows a call happened, how long it lasted, and which disposition someone picked at the end of a busy shift. It does not know that the coordinator told the family aide coverage was uncertain, or that the return call promised on Tuesday never went out. We wrote about that gap in Your CRM Is Logging Calls, Not Listening to Them.

The disposition field is not lying on purpose. It is one dropdown asked to summarize a five-minute conversation, chosen by the person with the least incentive to record a bad outcome accurately. Any analysis built only on disposition data will confirm whatever the agency already believes.

This is the category of problem SurfacerIQ works on: making what actually happened on intake and referral calls visible to the people accountable for referral outcomes, rather than leaving it in recordings nobody has time to open. The point is not a new dashboard. The point is that the reason a referral died should be recoverable without someone listening to 20 calls by hand.

The part that is not about revenue

Referral leakage gets pitched as a growth problem, and it is one. But both studies above measured mortality and readmissions, not census, and patients whose referrals were completed did better on both. The referral sitting in your queue on day nine is not a lead. It is someone a clinician thought was sick enough to need care at home.

Start with the twenty calls. If what you hear matches your reports, you have lost an afternoon. If it does not, you have found the leak and can point at exactly where it is.

If you want to talk through how to get visibility into what is happening on your intake calls, talk to us.