Two-Thirds of Direct Care Workers Earn Under $35,000. That's Your Retention Problem.

Two-Thirds of Direct Care Workers Earn Under $35,000. That's Your Retention Problem.

In 2024 there were 2.3 million direct care workers providing long-term care to adults 65 and older and to people under 65 with disabilities. 66% of them work in home care settings, 22% in nursing facilities, and 12% in residential care.

And 66% earn less than $35,000 a year, against 31% of all other adult workers. In home care specifically the figure is 72%, compared with 56% in nursing facilities.

Those are KFF's numbers, published July 9, 2026. They describe a market, not a management failure. But the operational consequences land squarely on the people running schedules and phones.

The wage floor is close enough to the median to explain a lot

BLS puts the median annual wage for home health and personal care aides at $35,800 as of May 2025 — $17.21 per hour.

Set that against the demand side. BLS projects employment in the occupation to grow 18 percent from 2025 to 2035, much faster than average, with about 760,500 openings per year over the decade.

Three-quarters of a million openings a year, at a median just under $36,000, in a labor market where warehouse and retail employers compete for the same applicants without requiring travel between homes. Any retention strategy that assumes you can out-wait the market is planning against that arithmetic.

The composition of the workforce changes what retention means

The aggregate wage figure gets quoted constantly. The composition data gets quoted rarely, and it is more actionable.

Characteristic	Direct care workers	All other workers
Age 50 or older	41%	33%
Female	85%	47%
Work part-time	39%	18%
Immigrants	30%	18%
Enrolled in Medicaid	32%	9%
Uninsured	13%	9%

All figures: KFF, July 2026. The workforce is also more racially diverse than the broader labor force — 30% Black compared with 11% of other workers, and 23% Hispanic compared with 19%.

Read the rows together and a different retention picture emerges.

39% part-time means a large share of your workforce is assembling income from more than one source. When you cancel a shift, you are not trimming hours at the margin. You are competing with the other employer who did not cancel.

41% age 50 or older means a meaningful part of the workforce has caregiving obligations of its own, and less tolerance for schedules that change with two hours of notice.

32% on Medicaid and 13% uninsured means a raise can cost a worker money. Benefit cliffs are real, and a caregiver who declines extra hours is often making a rational calculation rather than showing low engagement.

None of this is solved by a referral bonus.

The resignation shows up on the phone before it shows up in the HRIS

Here is the operational gap. Agencies measure turnover after it happens — a termination code, a quarterly report, a number in a board deck. The evidence that predicts it is sitting in the scheduling and staffing calls nobody reviews.

Consider a hypothetical illustrative case: a scheduling coordinator takes forty calls in a day. Six are caregivers calling about the same thing — a client whose hours were cut, a visit window that keeps moving, a mileage question that has now been asked three times. In a disposition-code system, those six calls are logged as "Scheduling — Outbound" and disappear. Nothing in that record tells a manager that three of the six callers are within a month of leaving.

The signals that tend to precede an exit are specific and, importantly, they are spoken:

• Repeated calls about the same unresolved pay or mileage discrepancy — the same caregiver, the same issue, a third or fourth contact
• Requests to drop a specific client, particularly when the reason given shifts between calls
• Questions about hours guarantees, minimums, or whether a case is ending
• Declining offered shifts while continuing to ask about availability elsewhere in the territory
• Calls that end without a stated resolution, which the coordinator closes anyway
• Escalating travel-time and commute complaints after a territory or assignment change

Every one of those is retrievable from the conversation. None of them is retrievable from a dropdown.

What you can do this week

Pick a two-week window of caregiver-facing calls — scheduling, staffing, payroll questions — and do four things.

1. Count repeat callers. Not call volume. Distinct caregivers who called about the same issue more than twice in fourteen days. That list is your intervention list, and it is usually short enough to work through in a morning.
2. Find the calls that ended without a resolution. "I'll check and get back to you" is a promise. Sample twenty of them and see how many produced a follow-up. The gap between promised and delivered follow-ups is a retention metric, whatever your HR system calls it.
3. Separate schedule churn from schedule volume. Track how many assignments changed after the caregiver was notified, by coordinator. Part-time workers absorb cancellations worse than full-time workers, and 39% of this workforce is part-time.
4. Fix one thing that generates repeat calls. Usually it is mileage reimbursement, a pay-period cutoff, or an unclear cancellation policy. Repeat calls about a policy are usually evidence the policy is not written down anywhere the caregiver can reach.

The staffing side of this has a companion piece: we looked at how schedule volatility loads the phone queue in Caregiver Turnover Fell to 34% and That Still Means Rebuilding a Third of Your Workforce Every Year.

Then make it a cadence, not an audit

A one-time review finds problems. A weekly rhythm changes behavior. The coordinators talking to caregivers are the largest single lever you have on whether someone stays through the next schedule change, and they are usually coached less than the intake team that talks to referral sources.

Our weekly coaching cadence for contact-center agents lays out a structure that works for scheduling and staffing teams with almost no modification: a small number of reviewed calls, one behavior at a time, same day each week.

Why exit interviews arrive too late to help

Most agencies learn why someone left from an exit interview, which has two structural problems. The person has already gone, so the information cannot change the outcome. And they are answering while thinking about a reference, so the answer skews toward the polite version — "scheduling," "commute," "found something closer to home."

The version told three weeks earlier, to a coordinator, in the middle of a frustrated call about a shift change, is more specific and more useful. It names the client, the coordinator, the pay period, the policy. That is the version you can act on, and it is the one nobody writes down.

The same asymmetry applies to the caregivers who stay. A worker who calls twice about mileage and gets it resolved both times is being retained by an operational process, not by a benefit. You will not find that in any engagement survey either.

The honest limit

You cannot fix a $35,800 median from a scheduling office. Wage levels in this workforce are driven substantially by Medicaid payment rates and state policy, and those are decided somewhere other than your building.

What is in your control is whether a caregiver who is already deciding between two marginal jobs experiences yours as the one that answers the phone, resolves the mileage question, and does not move the visit window twice in a week. At these wage levels, that difference is not a soft benefit. It is most of the differentiation available to you.

SurfacerIQ works on the underlying visibility problem: turning the volume of caregiver and patient conversations an agency already handles into something reviewable at scale, rather than sampled and forgotten.

This post is general information for operators, not legal or HR advice; wage, benefit, and classification questions are fact-specific and vary by state.

If you want to look at what your caregiver-facing calls are actually surfacing, talk to us.