A behavioral health therapy group earns one session at a time with costs that largely move with the sessions. An addiction treatment program — detox, residential, partial hospitalization, intensive outpatient — does not. It is licensed for a number of beds or slots, staffed to ratios that regulation sets, housed in a facility that costs the same whether twelve clients or twenty-four are in it, and paid per client day or per episode. Its economics are closer to a hotel's than to a clinic's, and the operators that understand this run census, length of stay, and cost per client day as the three numbers that govern everything else.
What does a licensed bed cost before the first admission?
Most of a program's cost exists before the first client is admitted. Facility lease or mortgage, utilities, insurance, licensing and accreditation, the medical director, the program director, the nursing coverage the license requires, the minimum clinical and support staffing for the licensed capacity, food service and housekeeping infrastructure, technology, and the utilization-review and admissions functions — these are incurred at licensed capacity, not at actual census. A 24-bed residential program with US$3.1 million of annual fixed cost carries about US$354 of fixed cost per licensed bed per day, whether the bed is occupied or not.
The variable cost per client day is modest by comparison. Food, supplies, medications and laboratory work, transportation, incremental laundry and housekeeping, and the staffing increments that census bands trigger — typically US$90 to US$160 per client day in residential settings, less in intensive outpatient. The gap between the fixed cost per bed and the variable cost per client is why every additional client above break-even is nearly pure contribution, and every client below it is nearly pure loss.
What is the break-even census?
Break-even census is the number of occupied beds at which total daily revenue equals total daily cost. With fixed cost of US$8,490 a day (US$3.1 million over 365), variable cost of US$120 per client day, and net revenue per client day of US$640 after verification realities and denials, break-even is 8,490 divided by (640 − 120), or about 16.3 clients. At 16 the program loses money every day; at 20 it earns roughly US$2,000 a day; at 22 it earns US$3,000. The entire financial story of the program sits in the four or five beds above and below that line.
Net revenue per client day is a measured number, not a rate. Contracted per-diem rates by payer and level of care, reduced by the denial rate on continued-stay reviews, by days delivered beyond authorization, by patient-responsibility amounts never collected, and by the out-of-network reimbursement reality where it applies — computed from remittance data over a trailing period, by payer, and refreshed quarterly. A program that computes break-even on the contracted rate is computing it four or five beds too optimistically.
How should staffing flex with census?
Regulation sets ratio floors; the program sets everything above them. Licensing and accreditation define minimum clinical staff, nursing coverage, and technician ratios by level of care and by shift; those are fixed at licensed capacity. Above the floor, staffing should flex with census in bands — one band at 12 to 16 clients, another at 17 to 21, another at 22 to 24 — with therapist caseloads, group facilitation, case management, and support staff scheduled to the band, and the triggers for moving between bands written down.
The cost of staffing to peak rather than to band. A program that schedules for 24 clients while running 17 carries roughly US$600 to US$900 of unnecessary daily labor — US$220,000 to US$330,000 a year — which is more than the margin on three occupied beds. The cost of staffing below the band is a ratio breach, a survey finding, and a liability, which is why the bands are set from the regulatory floor up, never from the budget down.
Agency and overtime are the cost of not managing census. When admissions arrive in clusters and staffing was set to a lower band, the program fills with agency nurses and overtime; agency rates at 1.5 to 2 times employed cost, applied to a third of nursing hours, add US$150,000 or more a year in a mid-sized program. The admissions pipeline and the staffing bands are the same conversation.
How does length of stay change revenue per episode?
Length of stay converts census into episodes, and episodes into revenue. A 24-bed program running at 20 average daily census with a 28-day average length of stay admits about 21 clients a month; at a 21-day average it admits 28 a month to hold the same census. Shorter stays mean more admissions, more verification and intake labor, more authorization cycles, and more discharge-planning cost per occupied bed — the per-day margin may be unchanged while the per-episode economics and the admissions burden shift materially.
Authorization shapes the curve. Payers authorize in blocks — often seven days at a time for residential, with continued-stay review against medical-necessity criteria — and the revenue per episode is the sum of authorized, delivered, and paid days; days delivered beyond an authorization are unbillable, and a client discharged early because a continued stay was denied produces a shorter episode than clinical judgment would have chosen. The utilization-review function therefore sits inside the length-of-stay economics: its denial rate by payer is a revenue-per-episode variable.
Step-down continuum economics. A client who moves from residential to partial hospitalization to intensive outpatient within the same organization produces a longer total episode at declining per-day rates and lower per-day cost — and keeps the admission's acquisition cost working across three levels of care. Programs without a continuum discharge revenue to someone else at day 28.
How should cost per client day be reported?
The report that runs the program. Cost per client day at the current census — total cost divided by client days — and at each census band, so the leadership sees the curve: US$1,040 at 14 clients, US$815 at 19, US$690 at 23 in the worked example below. Beside it, net revenue per client day by payer, so the margin at each band is visible for each payer; a managed-care per diem that is profitable at 22 clients may be below cost at 16.
The variance that matters is census against plan. A monthly report that shows revenue below budget tells the owners nothing; a report that shows average daily census of 17.4 against a plan of 20, with admissions of 19 against 23 and length of stay of 26 against 28, tells them where to look — and which of the three numbers moved.
Why is this CFO work?
The fixed-cost model requires every contract, lease, license, and staffing floor assembled into a daily number; net revenue per client day requires remittance data by payer and level of care; the census bands require the regulatory floors and the labor model together; the revenue-per-episode curve requires authorization and length-of-stay data joined to billing; and the report by census level is a design someone has to build and defend. Clinical leadership owns the care and the ratios; the CFO owns the economics of the bed.
Worked example
A 24-bed residential program with a 20-slot partial hospitalization step-down, US$3.1 million of fixed cost, US$120 variable cost per residential client day, net revenue per client day of US$640 blended (commercial in-network US$720, out-of-network US$590 after collection realities, managed care US$510).
| Average daily census | Client days / yr | Fixed cost / client day | Variable cost / client day | Total cost / client day | Net revenue / client day | Daily contribution | Annual contribution |
|---|---|---|---|---|---|---|---|
| 14 | 5,110 | US$607 | US$120 | US$727 | US$640 | −US$1,220 | −US$445,000 |
| 16 | 5,840 | US$531 | US$120 | US$651 | US$640 | −US$170 | −US$62,000 |
| 19 | 6,935 | US$447 | US$120 | US$567 | US$640 | +US$1,390 | +US$507,000 |
| 22 | 8,030 | US$386 | US$120 | US$506 | US$640 | +US$2,950 | +US$1,077,000 |
Illustrative figures for a hypothetical organization; not a client's data.
The program had been running at 16.4 average daily census and was staffed for 22 — about US$280,000 of labor above the 16-to-18 band, plus US$140,000 of agency fill during admission clusters. Its continued-stay denial rate with the managed-care payer was 19 percent, which cut the average managed-care episode from 28 to 21 days and forced an extra two admissions a month to hold census. The plan: staffing bands written and implemented (US$280,000 recovered); a utilization-review protocol targeting the managed-care payer's criteria (denials to 9 percent, episodes back toward 26 days); admissions-pipeline targets by referral source tied to the band triggers; and a census target of 20 with the step-down continuum capturing discharges. At 20 average daily census and the lower labor base the program's modeled annual contribution moved from roughly −US$100,000 to +US$780,000 on the same license.