- Residential and detox billing is built from bed-days, not encounters. The census is the source document.
- Facility claims usually go on a UB-04 rather than a CMS-1500, with different fields and different failure modes.
- The two most expensive errors are billing a day the patient was not present, and billing beyond the authorized length of stay.
- If the census board and the billing system are separate, someone reconciles them by hand every month — and the errors survive that process.
Outpatient billing is a question about encounters: what happened, who delivered it, how long did it take. Residential billing is a question about occupancy: was this person in this bed, at this level of care, on this date, with authorization covering it.
Systems built for the first question handle the second badly, which is why so many facility programs run their census on a whiteboard and rebuild the claim from it at month end.
Three structures, three different risks
Fee for service
Familiar, and mostly used for the outpatient services a facility also provides. Each encounter is a line, priced at a contracted rate. The risk is coding and modifiers.
Per diem
A bundled daily rate covering everything delivered that day. Common for PHP and IOP, and for some residential contracts. The risk shifts from coding to attendance: whether the day meets the payer's minimum service requirement, and whether the patient was actually present.
Case rate
A single price for an episode, or a daily rate against an authorized length of stay. The risk is the authorization: care delivered past the approved days is usually unrecoverable, and continued-stay reviews are the mechanism that prevents it.
In per-diem and case-rate billing, the census is not an operational convenience. It is the primary billing document.
What the census has to be able to answer
For any given date, for any given patient, a facility needs to be able to state — and evidence — the following:
- Was the patient admitted and physically present at the census-taking time?
- At what level of care, and did that change during the stay?
- Which bed and unit, and were there any transfers?
- Was the day covered by an active authorization?
- Was the minimum service requirement for that day met and documented?
- Were there leave days, and how does the contract treat them?
That last one catches people. Therapeutic leave, hospitalisation mid-stay, and against-medical-advice departures are all handled differently by different contracts, and billing a leave day the contract excludes is a clean overpayment finding.
Where UB-04 differs from what outpatient teams know
Facility claims carry information the professional claim does not: revenue codes describing the accommodation, occurrence and value codes carrying dates and amounts, condition codes, and admission and discharge status. The discharge status field alone is a frequent denial source, because it has to match the clinical record and often does not when the two are maintained separately.
The practical consequence is that a billing team fluent in CMS-1500 does not automatically produce clean UB-04s, and a platform that only knows how to build professional claims will make them do it manually.
The reconciliation problem
Here is the pattern in almost every facility running separate systems:
- The unit maintains the census — often a board, sometimes a spreadsheet.
- Admissions and discharges are recorded in the EHR, sometimes hours or days later.
- At month end, billing builds bed-day counts from one of those two sources.
- The two disagree. Someone reconciles them from memory and shift notes.
- The claim reflects whichever source won that argument.
Every step of that is an opportunity to bill a day that cannot be evidenced, or to fail to bill a day that could have been. Both happen, and the second one is invisible — programs almost never discover the days they forgot to bill.
Take one month. Count bed-days from the census, count billed days from the claims, and compare. If those two numbers differ by more than a rounding error, that gap is either an overpayment waiting to be found or revenue you already gave away.
What removes the problem
Not better reconciliation. One source.
When admission, transfer, level-of-care change, leave, and discharge are recorded as events on the patient record, the census is a view of those events rather than a parallel document. The bed-day count is then derived rather than assembled, the authorization draw-down happens against the same events, and there is nothing to reconcile because there was never a second version.
That is a structural fix rather than a process fix, which is why it is hard to retrofit and worth insisting on when choosing a platform.