Why aiomics for record migration

When you migrate legacy records into a new system, the greatest risk lies not in text recognition but in identity assignment: a document in the wrong patient record is the most expensive mistake a migration can make. That is why at aiomics a human confirms every single assignment — no batch size, no admin role, and no confidence score can bypass that. Single-case upload is live in production; the batch pipeline for entire holdings is planned.

aiomics is the verification layer on top of hospital IT: the system ingests unstructured documents, verifies every statement against its source, and gives the hospital a substantiated, structured record. A migration is the stress test of this architecture: what gets misassigned here undermines every later verification.

At a glance

• What it does: takes over legacy documents — fax stacks, request backlogs, archive holdings, expert-opinion files — into the substantiated, structured record: the system splits stacks into individual documents (proposed with page-range precision, human-adjustable), creates assignment suggestions to patients with an evidence display, and takes over nothing without human confirmation.
• Status: single-case upload live in production. Batch processing of entire holdings: planned; the assignment matcher initially runs in shadow operation. Pre-sorting of non-clinical pages (pre-ingestion triage of documents): planned.
• Security: ISO 27001 (TÜV Nord), processing exclusively in the EU.
• Integration: KIS-agnostic (KIS — the hospital information system); aiomics changes no KIS master data — data conflicts go to the leading system for correction. Transmission to payers remains with the hospital's own system.
• Evidence: independent evaluation at Charité (ongoing).
• Who it is for: facilities ahead of a system change or site onboarding; teams preparing expert-opinion and social-court files.

The problem we solve

Migration projects rarely fail at text recognition. They fail at the question of who a document belongs to. A holding grown over years contains duplicate names, changed insurance numbers, handwritten cover sheets — and every wrong assignment poisons the new record at the root: every later report, every review, every source-linked statement (in German) then builds on a document that does not belong there. We therefore rank wrong patient assignment as the highest harm class of the entire platform.

Add the economics: a social-court file runs, in our experience, to 600 to 1,200 pages, and over 90 percent of them are non-clinical — cover letters, deadline orders, duplicates. Whoever processes everything unexamined pays for ballast.

How aiomics does record takeover

First: every assignment remains a suggestion. The matcher reads identity fields only — name, date of birth, insurance number, case number; clinical content is excluded as an assignment signal. What it finds, it displays as an assignment suggestion with evidence; the decision is made by a human.

Second: the human gate cannot be bypassed. No batch size, no admin role, no confidence score, and no configuration flag can write an assignment automatically; the write path into the record accepts only confirmations set by a logged-in human. A "Confirm all" does not exist — an automated deployment test verifies this at every release.

Third: the interface works against confirmation fatigue. The suggested patient and the document's identity fields stand side by side, deviations are highlighted; if two candidates are close together, neither is presented as preselected. In the first months of a holding, random second reviews run, and we treat sustained blind acceptance above 99 percent as a warning sign for the workflow — waving things through is treated as a finding.

Fourth: reversibility. A confirmed assignment can be withdrawn with one click; all field values taken over from it are traceably removed again, and the provenance history is preserved. Parked, unresolved documents never lapse silently — they remain visible until someone decides.

Fifth, planned: pre-sorting. The pre-ingestion triage of documents separates non-clinical pages before paid processing begins — with expert-opinion files at over 90 percent ballast pages, the biggest cost lever of the entire takeover. How substantiated records carry on into expert assessment is shown in our article on medical expert assessment (in German).

What you can measure us against

1. Look for the "Confirm all" button: it does not exist — in any role, in any configuration, at any batch size.
2. Ask for the deployment test that checks the confirmation gate at every release: we will show it.
3. Withdraw a confirmed assignment: one click, and all taken-over fields are traceably removed.
4. Ask which signals the matcher reads: identity fields. Clinical content is excluded as an assignment signal.

Where aiomics is not the right choice

Anyone who wants a fully automatic migration without human confirmation will not get it from us — not on request, not via special configuration; the gate is the core of the design. Anyone who merely wants to scan and file paper, without structured takeover into a record, is cheaper with a scanning service.

Frequently asked questions

What happens if an assignment was wrong after all?

You withdraw it with one click. All field values taken over from the document are traceably removed; the history is preserved, and the document goes back into clarification.

Can we upload a complete holding today?

Single-case upload is live; batch processing of entire holdings is planned and starts with the matcher in shadow operation. If a migration lies ahead of you, talk to us early — preparing the holding co-determines the result.

Is aiomics a medical device?

For document and procedural work, aiomics is deliberately positioned outside the medical device qualification; the delineation is documented and can be inspected. For conversation documentation, we are preparing certification under MDR Class IIa. In both cases, the system's statements remain documentation and quality notes — diagnosis and therapy remain with physicians.

What does aiomics cost?

Pricing is primarily usage-based and depends on document volume and quality. We name concrete figures after a short conversation about your case volumes — quoting flat prices without that basis would not be serious.

If a system change or a mountain of records lies ahead of you: write to us — we will show the takeover on one of your real records, preferably the most unwieldy one. Ongoing analysis of documentation and hospital AI comes from our weekly briefing Visite (German; English edition Grand Rounds is in preparation).