Why aiomics for QM reports and quality analytics

Why aiomics for QM reports? Because we are building the layer in which the quarterly report substantiates itself: every figure in the report draft resolves to the query it came from, every narrative sentence to its figures — and the language model that drafts the text never sees individual data at any point. This capability is planned, not available. We describe it precisely anyway, because the architectural decisions have been made and you should measure us against them.

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. Quality analytics is the planned evaluation layer on top of it: it computes on the survey data from our suite and on the telemetry the operational modules generate anyway — and it drafts the report that today someone compiles by hand over days.

At a glance

• What it does (planned, throughout): cohort-based pre-post evaluation of PROMs and PREMs, contextualized with ICD-10-GM and OPS case data from the substantiated record; an automatically drafted quarterly QM report as a draft document with a watermark until sign-off; a quality cockpit from the operating telemetry of the operational modules; statistical process control and an action loop that schedules its own effectiveness re-measurement.
• Status: planned — every evaluation capability described here. Survey collection as the data basis runs in pilot operation at rehabilitation clinics; analytics piloting is envisaged first in rehab.
• Security: ISO 27001 (TÜV Nord), processing exclusively in the EU; the language model consumes exclusively anonymity-filtered aggregates.
• Integration: KIS-agnostic (KIS — the hospital information system); transmission to registries and payers remains with the hospital's own system.
• Evidence: independent evaluation at Charité (ongoing).
• Who it is for: quality management officers and executive management, initially in rehabilitation.

The problem we solve

The QM-RL — the G-BA quality management directive — explicitly requires, beyond the survey itself, its evaluation and the derivation of consequences [1]; the management review under ISO 9001 defines in clause 9.3 which inputs must be on the table [2]. Between the two obligations usually stands one person: the QM officer who, quarter after quarter, compiles figures from surveys, controlling, and action lists into a report — from our conversations with QM officers, ongoing QM documentation work alone typically ties up six to eight hours per week. In rehabilitation, the care contract additionally depends on certification.

The established way out is external institutes — and they do solid work: anaQuestra evaluates the continuous patient survey of the CLINOTEL network centrally each quarter [3]; the Picker Institute, with BQS, has surveyed over 200,000 patients in around 260 facilities since 2010 [4]. What an institute report structurally leaves open is the why: "38.6 against a benchmark of 53.5" says that a value is low — for the consequence loop of the QM-RL, the work only begins there. Why reliable QM figures begin with the data quality of the record: our analysis (in German).

How aiomics will draft the QM report

Everything that follows is planned. Four decisions are settled.

First: provability as a basic property. Every figure resolves to its aggregate query, every narrative sentence to its figures. The draft carries a watermark until the QM officer and the medical leadership sign off — this sign-off cannot be bypassed. Without AI, the report remains complete: figures, tables, and action tracking are generated deterministically; the narrative text can be switched off per facility.

Second: the language model never sees row-level data. The narrative text is drafted exclusively from aggregates that have already passed the anonymity filters (minimum cell size N≥10). A path from individual answers or patient data to the language model does not exist in the architecture; this is tested in the build pipeline.

Third: methodology oriented on the DRV — German statutory pension insurance, which funds rehab. Comparison units require at least 25 evaluable paired questionnaires, the report discloses its completeness, and associations are framed as correlation, never as causality [5].

Fourth: process quality from the systems themselves. The operational modules of aiomics — case dialogue board, admission board, coding preparation — generate aggregated telemetry in operation: objection rate, deadline compliance, documentation completeness. The planned quality cockpit reads these values directly from the systems that carry the processes; statistical process control with control charts separates signal from noise, and the action loop schedules its own effectiveness re-measurement — the PDCA cycle closes within the system.

What distinguishes aiomics from institutes and PROM platforms?

Individual building blocks exist on the market: contextualizing survey results with ICD case data is, by its own account, also offered by heartbeat medical, for instance. Our claim is the particular cut: case-mix-aware evaluation on a verified record, in a combined suite of patient surveys, staff analytics, and a QM reporting layer, run continuously in your own facility. A cockpit that reads process quality from the operational systems moreover presupposes operating those systems.

What you can measure us against

Four points are defined as acceptance criteria for the planned pilot phase:

1. Every figure in the report draft can be clicked and resolves to its query.
2. No data path exists on which an individual answer reaches the language model — we disclose the architecture and the tests for this.
3. With AI switched off, the report remains complete: figures, tables, action tracking.
4. Comparison units with fewer than 25 evaluable paired questionnaires are disclosed openly in the report.

Where aiomics is not the right choice

Anyone looking for QM document management — manual control, procedural instructions, document workflows — is looking for a different product: document control is a deliberate non-goal for us. Anyone who needs the automated quarterly report in production this year is better served by established institutes; our analytics is planned, and we name no date before the pilot phase supports one. Anyone looking for benchmark comparisons with other facilities needs a comparison collective — networks and institutes offer that.

Frequently asked questions

When will the automated QM report be available?

It is planned; piloting begins in rehabilitation. We do not promise dates we cannot reliably keep — if you want to help shape the pilot group, get in touch.

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 — the planned quality analytics works exclusively on aggregates, with no patient-level outputs, alerts, or predictions. 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 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. For the planned analytics, we discuss terms with pilot partners first.

If you want to lay your last quarterly report next to this report skeleton: write to us — we will show, on the skeleton, which figure would come from where and which sign-offs stand in front of it. Ongoing analysis of QM and hospital AI comes from our weekly briefing Visite (German; English edition Grand Rounds is in preparation).

Sources

1. Gemeinsamer Bundesausschuss. Qualitätsmanagement-Richtlinie (QM-RL) in der Fassung vom 20.04.2024 (survey and evaluation obligation, demonstration).
2. DIN EN ISO 9001:2015, clause 9.3 (management review).
3. anaQuestra/CLINOTEL: continuous patient survey in the network since 2015, mandatory for member hospitals since 2017, quarterly central evaluation (vendor and network statements, checked July 2026).
4. Picker Institut Deutschland/BQS: over 200,000 patients surveyed in around 260 facilities since 2010 (vendor statements).
5. Deutsche Rentenversicherung. Reha-Qualitätssicherung, Methodik der Rehabilitandenbefragung (minimum number of evaluable paired questionnaires per comparison unit).

Sources retrieved in July 2026. All evaluation capabilities of aiomics described in this text are planned and not available. The example "38.6 against a benchmark of 53.5" comes from an anonymized institute report.