The Admission Record Begins Before the Patient Arrives

Tuesday, shortly after seven, internal medicine ward. Ms. K., 78, arrives with a referral from her general practitioner: increasing shortness of breath with known heart failure. She brings what she could find at home — the cardiologist's letter from the spring, a handwritten medication list, plus the remark that her tablets were "changed recently". The GP's current findings are in the admissions fax inbox. Somewhere.

The resident has barely half an hour for the admission. She asks, examines, types up what is legible, and writes the admission record from what could be learned in that half hour. The cardiologist's letter goes into the chart as a copy. It does not appear in the admission record.

Months later, a reviewer from the Medizinischer Dienst (MD — the German payers' medical review service) reads this admission record. He is looking for the answer to a single question: why did this patient need inpatient treatment? What the cardiologist's letter and the GP's findings would have evidenced is not in it. Both were available on the day of admission.

The second reader

An admission record has a second reader whom hardly anyone thinks of while writing it. The first is the treatment team of the coming days; for that team, the record is a working document. The second arrives months later on behalf of a health insurer and reviews whether inpatient treatment was necessary under § 39 SGB V, the German statute governing hospital treatment [1]. His most common line of attack is primary misallocation: could the case have been treated as an outpatient? The answer to that question is created on the day of admission — or it is not created at all. How often it is missing shows in the billing review statistics: across 16 quarters, only around 45 to 52 percent of reviewed hospital bills ended without objection [2]. Why tomorrow's audit quota depends on today's documentation is described in The 2027 audit quota is being created now (in German).

Billing itself also hinges on the day of admission: principal and secondary diagnoses can only be coded if they are documented — and a good share of what is coding-relevant sits in the prior findings that lie in the chart as unread copies.

Add the working reality at the other end: reconstructing a history from brought-along and faxed prior findings costs 20 to 60 minutes per admission, according to internal experience figures from hospital projects [3] — time the resident in the scene above simply does not have. So the record is written from the conversation, and the prior findings remain copies in the chart. The remarkable part: almost everything that makes an admission record audit-proof already exists before the patient arrives — in the referral, the prior findings, the referral documentation. It just never reaches the document. Why the completeness of prior findings determines the quality of the admission is described in Prior-findings completeness and admission quality (in German).

How wide the gap is between customary and defensible admission documentation shows in an example from review practice. "Meets G-AEP A1" — the bare mention of a criterion from the G-AEP catalogue (German Appropriateness Evaluation Protocol, the German admission criteria) [4] — carries little weight in an MD review. The evidenced form is what holds up: "Meets G-AEP A1 (impaired consciousness): GCS 12 at admission on 14.04.2026, 02:15 [U-003]" — where the marker [U-003] points to the source document. Criterion, value, timestamp, source reference: four elements that are within reach on the day of admission and barely reconstructible months later.

From document stack to evidenced record

The path there begins before the first sentence of the record. Four building blocks, in vendor-neutral terms:

Structured extraction before admission. Referral, prior findings and medication lists are read by machine and converted into structured entries — ideally before the patient is in the room. The referral documents often arrive days before the admission date; those days are usable.

Verification against the source. Every extracted entry is checked against the original document before it enters the chart. Extraction errs occasionally; what matters is that the error surfaces while it is cheap to correct.

Provenance per field. Every entry knows its source document and its location within it. Only then can the evidenced form of the admission justification be written at all: timestamp and quotation are one click away. Why this is the foundation for everything else is described in Provenance tracking: a source for every statement (in German).

Completeness check at the time of admission. The customary check happens at discharge — when questions have become expensive. At the time of admission, the referring practice is still in the loop and the patient is present; whoever checks now which entries are missing for justification and coding can still obtain them.

Cutting across all of this is the handling of contradictions. The GP's medication list says 5 mg, the patient says "changed": a system should surface this as an inconsistency between sources and leave the assessment to the physician.

How aiomics prepares the admission record

The foundation is in production: documents are ingested, extracted entries are verified against the source, and every field carries its origin — with a four-level trust taxonomy (Verified, Single-Source, Patient-Reported, AI-Extracted) and a source lookup under 50 milliseconds. Whether an entry comes from the cardiology letter or from the admission interview is no longer a question of memory but an attribute of the field.

On this foundation, document generation produces drafts, including for the admission record. It works with a curated library of around 50 document types and 80 regulatory overlays. For the admission justification, these overlays come in three layers: the basis under § 39 SGB V along the G-AEP criteria, above it the context (emergency, elective, geriatrics, psychiatry, pediatrics), above that procedure-specific admission criteria. The GCS example above comes from this library; it codifies what a justification must look like to withstand a review. The rule behind it applies library-wide: clinical statements without a resolvable source reference do not belong in a draft.

Two building blocks are not yet in production, and we label them as such: the patient overview with timeline — the consolidated view of everything received before admission — is planned. The systematic completeness check at the time of admission, which reports missing entries as soon as the referral documents arrive, is in development and likewise not yet released.

Part of an honest picture is the boundary: aiomics does not assess whether an admission is medically necessary and does not suggest diagnoses. The system structures and evidences the documented medical assessment. Whether Ms. K. stays as an inpatient is the physician's decision; the system makes sure that decision is backed by what was available on the day of admission.

What to measure any vendor against

1. Can every field show its source? A click on an entry should open the location in the original document. Also ask whether the system distinguishes whether an entry comes from a physician letter, from a single source, or from the patient herself.
2. When is completeness checked? At discharge, the referring practice is out of the loop and the patient is at home. Ask what the system can say about missing entries at the time of admission — and whether that function is in production today or announced.
3. What does a generated admission justification look like? Ask for an example and hold it against the standard: criterion, value, timestamp, source reference. A justification without these four elements is an assertion with better formatting.
4. What happens when sources contradict each other? GP list versus patient statement, prior finding versus referral: is the inconsistency displayed and left to the physician's decision — or silently resolved in favor of one source?
5. Who confirms the transfer into the chart? Blanket "accept all" buttons save seconds and cost reliability. Section-by-section confirmation with explicit mandatory fields for medication and allergies is the more defensible standard.

If you want to hold your admission documentation against these five questions — with or without aiomics in the room — write to us. In our weekly briefing Visite (German; English edition Grand Rounds is in preparation), we write regularly about documentation quality and revenue integrity.

Sources

1. § 39 SGB V — hospital treatment. gesetze-im-internet.de
2. GKV-Spitzenverband (the German association of statutory health insurers): reporting on hospital billing reviews; share of reviews without objection stable at around 45–52 percent across 16 quarters.
3. Time required for history reconstruction (20–60 minutes per admission): internal experience figures from hospital projects, to be read as context.
4. G-AEP criteria (German Appropriateness Evaluation Protocol): catalogue of criteria for assessing the necessity of full inpatient hospital treatment, introduced as part of the review procedure under § 17c KHG.