The Goldilocks Conundrum

There is a moment every Monday morning at every home-health branch in America that I came to recognize across the network I worked in, and that I've never seen written about properly. It happens around 6:45 a.m. The branch director opens her schedule, and the schedule looks wrong. Not slightly wrong. Categorically wrong. Either there are three nurses sitting at the office with no patients to see, or there are eleven patients on the visit list and seven nurses on the road, and one of those seven just texted to say her kid has a fever.

The director does what every director does, which is reshuffle the deck. The deck always loses something. A patient who needed a 9 a.m. visit gets pushed to 1 p.m. A nurse who was scheduled for six visits gets four — and at the end of the week, when the math runs, that nurse takes home eighty-two dollars less than she'd planned for, because in this industry we pay clinicians by the visit and not by the hour they spent driving to a town where the patient's family forgot to mention there was a dog.

The next Monday, the schedule looks wrong in a different way.

I spent enough years close to enough of these branches — three hundred and fifty of them, give or take, across one national home-health network — to stop seeing this as a managerial problem and start seeing it as something else. The branches that ran the cleanest weren't the ones with better directors. They were the ones whose directors had stopped expecting to be right. They were planning to be wrong, and choosing carefully which kind of wrong to be that week.

That is the Goldilocks problem.

The fairy tale gets its name from the porridge scene, which I won't recap except to note that the entire economic premise of it is that “just right” exists. Somewhere on the table, there is a bowl of porridge that has been correctly proportioned to the eater. The story does not work without that bowl. The whole architecture of the metaphor — the lesson it teaches children, the way it survives in the operational vocabulary of every staffing meeting I have ever sat in — depends on the assumption that the bowl is findable.

In home-health and hospice, the bowl is not findable. Or, more precisely: the bowl exists for a single moment, on a single Tuesday afternoon, in a single census configuration that will not survive contact with Wednesday morning. The window in which the porridge is the right temperature is shorter than the time it takes to eat it.

Consider the workforce equivalent of that bowl. There is a staffing level on every branch that, if it were perfectly calibrated to the demand of the moment, would produce the right outcomes on every dimension that matters: utilization in a healthy range, patient access at the level the regulator and the patient both expect, operating margin in the band that lets the agency keep its lights on. That level is what every branch is trying to hit. The slider below is the abstraction of every Monday-morning planning meeting I have ever observed.

You can find the spot where all three lights turn green. You can show it to the leadership team. You can put it in the Tuesday deck. And then a week happens, and the green window — which the slider was honest enough to show you isn't fixed — drifts somewhere else.

This is not a planning failure. This is the structural condition of the work.

I keep a mental list of the things a branch has to absorb in any given week. It is not exhaustive and it is not even original — every operations director in this industry knows it, in the way one knows the weather — but I find that writing it down is what clarifies why no system has solved this.

There is census, which moves daily. Patients admit, patients discharge, patients die. There is acuity, which moves per patient — the same name on the schedule can be a twenty-minute visit one day and a ninety-minute visit two days later, depending on what is developing clinically. There is geography. Patients are not where the clinicians live. A nurse who lives twenty minutes east of the branch can absorb the patients in that quadrant; a patient in the north quadrant gets a different nurse, a different drive time, a different effective capacity. Two patients with identical clinical needs can produce wildly different staffing math depending on which side of the river they happen to live on.

There is payer mix, which determines what each visit pays and therefore how much margin loss any single cancellation represents. There is weather. There is illness — clinicians get sick, patients get worse, families call to say not today. There is churn, which reshapes the staffing baseline every month in a way that no quarterly plan can stay ahead of. And there is on-call rotation, which is the part that nothing on a spreadsheet has ever known how to model honestly.

Each of these moves on its own clock. The clocks are not synchronized. And the response time of the workforce — how fast you can redeploy a clinician from one configuration to another — is measured in days, sometimes in weeks. Not in hours.

The director's job, every Monday at 6:45 a.m., is to make a plan for a week that will not stay still long enough for the plan to apply.

There is a number I would like to put a finer point on, but I'm going to gesture at it instead, because the literature is patchy and I'd rather be directionally honest than precisely wrong: home-health margins are thin, and they are pressed thin, in part, by the cost of being unable to perfectly match capacity to demand.

The chart below shows fourteen days of visit demand on a representative branch. Two horizontal lines — one for the FTE capacity the branch carries, one for the additional ceiling available from the per-diem pool — divide the chart into the two ways a week can be wrong. Above the lines: patients who didn't get their visit on the day they expected. Below the lines: clinician hours that were paid for but not productively deployed. A running cost ticker shows the cumulative dollar weight of being wrong in either direction.

You can move the lines. You can experiment with where to draw them. What you will notice is that the cumulative cost never actually goes to zero. There is no FTE line that minimizes both directions of harm. The volatility itself is what is being paid for — and no static staffing decision can absorb a moving target.

The branches I worked closest with all knew this implicitly. The good ones were not better at finding the right line — they were better at choosing, week to week, which kind of wrongness was less harmful. Some weeks they accepted the access miss because the per-diem pool was tapped. Some weeks they accepted the idle time because they couldn't afford the patient experience hit. Every week, every choice was a decision about which harm.

I once watched a branch director hit her census target three months running. The corporate response — and this is not a criticism of corporate, this is a description of how systems behave — was to push her staffing ratios tighter. Which is, of course, exactly what caused the next month to fall apart. The signal that she had absorbed the variance was treated as evidence she had spare capacity. There is no version of that conversation that ends well.

This is the part of the story that almost never makes it into the conversation, because the people most affected aren't the ones writing the strategy decks.

In home health and in hospice, most clinicians are paid per visit. A nurse, a PT, an OT, a speech therapist — most of them are not on a salary. They are paid a rate per completed visit, and their take-home in any given week is the product of two numbers: how many visits they made, and what the rate was on each one. Mileage gets reimbursed at some rate. Sometimes there is a small productivity multiplier. The structure varies. The principle is the same.

The principle is this. When the agency's volume drops, the clinician's pay drops with it. When the schedule reshuffles, the clinician's pay reshuffles with it. When a patient cancels because their daughter took them to urgent care, the PT — who already drove forty-five minutes to the house — gets paid nothing for that visit, plus a partial mileage stipend that does not equal the time. The variance the agency cannot absorb on its spreadsheet shows up, week after week, as an actual reduction in someone's paycheck.

The math of this is worth sitting with.

What the calculator shows is not subtle, but I find that until I see it on a screen I forget how clean it is. The clinician's annual take-home is volatile because the agency's census is volatile, and the agency has chosen — through the structure of its compensation — to pass the volatility through.

There is a name for this, and the name is what has been missing from the conversation.

Pay-per-visit is not really a compensation model. It is a risk-transfer mechanism. It transfers the cost of unpredictable demand from the agency to the clinician. The clinician is, in effect, the float pool the agency could not price into its own model. They absorb the variance. They are paid a small premium for absorbing it — that is the spread between the per-visit rate and what the equivalent salaried rate would have been — but the premium is almost never enough to cover the actual variance, and most clinicians I have talked to don't experience it as a premium anyway. They experience it as a paycheck that is sometimes lower than they planned for, in a job whose other costs have not gone down.

I have heard a phrase for this, recently, in conversations I won't attribute. The volatility absorption tax. I would like to claim it, but I think it was already in the air. It is a useful name. It belongs in the next contract negotiation.

Hospice is structurally harder than home health, not easier. Census is smaller, which means a single admit or a single death moves the daily math more, not less. The interdisciplinary team — the five-role care plan that is the regulatory and ethical signature of the work — means a single staffing gap distorts the whole structure: an aide's visit cadence depends on a nurse's assessment, the chaplain's touchpoint depends on the social worker's read of the family, and every gap propagates. The on-call work, which is the part of hospice that nobody has built a clean staffing motion around, runs by its own logic that no algorithm seems to want to schedule honestly. I have a memory of a hospice on-call coordinator getting paged on a Friday at 11 p.m. because a family who had said don't worry about us this weekend on Wednesday was now actively in the dying process, and the assigned nurse had called out at 8.

The Goldilocks problem is sharper here, not duller. The volatility absorption tax is heavier. And the clinicians shouldering it are the ones whose work is, in every other respect, the most demanding in the field.

I have spent enough time around this problem to be careful about prescribing the solution. I'm not interested in offering a plan I haven't built. But there are observations I can offer about the shape of what would have to be true.

It would have to involve some structure that pooled capacity across more than one agency, because no single agency has the surface area to absorb the variance on its own. I have watched four branches within twenty miles of one another spend the same Tuesday short-staffed and over-staffed simultaneously, in different proportions, with no mechanism to redistribute. The capacity exists. The orchestration does not.

It would have to protect the clinician's income from the agency's census volatility, in some form, because as long as the comp model passes the variance through, the workforce is going to keep churning out of the field at the rate it is churning now. Workforce data on this is unambiguous. The career-tenure curve in home-health PT is shorter than it was ten years ago, and the people leaving are not leaving for reasons that have to do with the patients.

And it would have to be structurally separate from any single agency's P&L — because the agency, as currently constructed, has no incentive to absorb a cost that the clinician is, today, absorbing for it.

I don't know what that structure is going to be called. I have some opinions about what it could look like, and I expect to write about them in the next thing I write. What I am fairly sure of is that it isn't going to come from inside any single existing incumbent — not because the incumbents are bad actors, but because the problem they are optimizing for, given their own constraints, isn't this one.

The platforms that are quietly making real progress on this, when I look closely, are the ones that have stopped trying to solve the Goldilocks problem at the agency level and started solving it one layer up. I find that worth noticing.