PROMs and PREMs: The Survey Obligation Hardly Any Facility Evaluates

In the corridor outside the dining hall of a rehab clinic stands a plexiglass box, next to it a stack of questionnaires: four pages, checkboxes, at the end two blank lines for comments. Departing patients are asked to fill one in. Many do — paper surveys have a better response rate than their reputation suggests. At the end of the quarter, a staff member packs the forms into a box and sends them to a survey institute.

Months later, a PDF comes back: means, benchmarks, bar charts. Satisfaction with catering is below the reference value. Why, the report does not say. Which ward, which period, which group of patients — no longer reconstructable from the aggregated figures. The QM officer files the report away. For the audit, that is enough.

This is how surveying works in many German facilities, in acute care as in rehab. Nothing about it is negligent; the obligation is fulfilled. It is just that almost nothing happens with the answers.

The obligation is growing faster than the practice

The quality management directive of the Federal Joint Committee (QM-RL — the G-BA quality management directive) lists patient surveys and staff surveys in Section 4 as instruments that must be applied; under Section 6, implementation must be demonstrated on request [1]. That means more than filing an institute's PDF: whoever surveys must be able to show what became of the answers. In rehabilitation, the new QM agreement of the Federal Rehabilitation Council (BAR) has additionally been in force since July 1, 2025 — around 1,200 inpatient rehab facilities are subject to mandatory certification [2].

Alongside this exists the rehab patient survey of the DRV (German statutory pension insurance, which funds rehab): around 120,000 questionnaires per year, collected and evaluated centrally [3]. For the individual facility, the results arrive late, aggregated, and only for DRV cases. What a facility can learn from them for its own management is limited — the in-house, timely view across all payers is missing.

Those who buy in the obligation externally pay for it: survey institutes cost, in our market experience, 10,000 to 30,000 euros per year and facility — for quarterly waves on paper with central evaluation and a benchmark report that arrives months after the last answer.

Yet collection itself is not the bottleneck. In a pilot at a rehab clinic, the response rate of the paper-based survey was 88.2 percent. The problem is not the response rate; it is the time between answer and insight — and the evaluation, which often never happens at all.

That leaves the second half of the obligation, the one less often talked about: the staff survey. It rarely fails on technology and frequently on trust. Employees who are not sure their answers will remain anonymous answer diplomatically or not at all. And the Betriebsrat (works council) has good reasons to be skeptical: "the data is anonymous" is a promise it cannot verify with the means at its disposal.

What a survey solution must be able to do

Vendor-neutrally, software for patient and staff surveys must deliver four things at once.

First, longitudinal measurement: PROMs — patient-reported outcome measures — measure change, meaning the same person at admission, at discharge, and at follow-up. A solution that can only run single waves measures momentary mood. PREMs and satisfaction surveys, by contrast, need proximity to the experience: at the bedside, before departure, while memory is fresh.

Second, reach: whoever only sends out a web link surveys the digitally inclined and calls the result representative. Accessibility, plain language, additional language versions, and a paper fallback determine whether the sample reflects the facility.

Third, validated instruments with cleared licenses: PHQ-9 and WHO-5 are free to use; EQ-5D-5L, SF-36v2, and COPSOQ require licenses. A solution that lets you create arbitrary questionnaires makes this license risk invisible — until a rights holder makes it visible.

Fourth, anonymity that can be demonstrated: for the staff survey, a promise is not enough. Anonymity only becomes verifiable when the protective mechanisms live in the architecture and can be demonstrated to a works council.

How aiomics builds it

The aiomics survey suite is in pilot operation at rehab clinics; the expansion to the full suite — including the staff package — is planned. The construction decisions behind it are fixed; they apply to the pilot operation as to every planned expansion stage.

The minimum cell size lives in the code. Analyses below ten answers per cell are blocked at the query level; the industry norm would be a floor of five. Complementary cells are suppressed along with them so that small groups cannot be reconstructed by subtraction; queries below the threshold are refused and logged. This can be demonstrated to a works council, with real queries.

The works agreement is part of the planned staff package. aiomics ships a works agreement template and a technical anonymity whitepaper as product artifacts — and activation of the staff survey is technically blocked as long as no works council review is documented. The staff channel stores no personal identifiers and no IP addresses and coarsens timestamps.

Instruments come from a registry with license metadata. Free instruments such as PHQ-9, WHO-5, WAI, or OLBI can be activated immediately; license-bound ones such as EQ-5D-5L, SF-36v2, or COPSOQ can only be activated once the facility's license is on file — and instruments in a prohibited class, such as the Gallup Q12, cannot be created at all. aiomics ships no instrument content; the license relationship remains between the facility and the rights holder.

Collection and evaluation are separated, for a regulatory reason. The suite computes no scores, no severity grades, no red-flag markers at the individual level. Sensitive individual answers — item 9 of the PHQ-9, for instance — are stored verbatim, without automated detection or alerting: exactly as on paper, and exactly so it is communicated. A check in the build pipeline fails any software version in which score computation appears in the collection path. Behind this detail lies the boundary between a survey tool and a medical device.

Waves follow fixed rules. Whether a follow-up questionnaire is triggered depends on the wave's schedule — answer content is architecturally not addressable from the wave logic. Instruments are also version-pinned: a running wave computes with the version it started with until it ends, otherwise the year-over-year view silently compares two different questionnaires. For facilities with strict requirements, there is a postal, reminder-free operating mode.

In everyday use it stays fast: handing over an iPad with a survey in progress at the bedside takes under 30 seconds; for paper patients there is a printable blank PDF whose answers are entered afterwards. The interfaces are designed for accessibility to WCAG 2.2 AA and for plain language; Turkish and Kurdish are prioritized as the first language versions. Wave setup in under 30 minutes and results in under 24 hours are target values of the expansion — targets, not yet measurements.

The suite is also intended to cover the smaller obligations: surveys of Zuweiser (referring physicians) with a maximum of three questions via token link on correspondence that is sent anyway, and an anonymous incident inbox modeled on CIRS systems — both planned as part of the suite expansion.

The analytics layer above it — cohort analyses, case context, automatically drafted QM reports — is likewise planned. Why usable evaluation is ultimately a question of data quality is described in our article on data quality as the underrated lever in hospital quality management (in German).

What to measure any vendor against

1. Where does the minimum cell size live — in a policy or in the code? Have them show you live what happens when an analysis falls below the threshold.
2. What does your works council get to hold? An anonymity promise — or a technical whitepaper, a works agreement template, and a demonstration of the protective mechanisms?
3. How are instrument licenses managed? And who carries the risk when a license-bound questionnaire runs without a license on file?
4. What happens with sensitive individual answers? A vendor promising automated alerts on individual answers is promising a medical device — ask for the certification.
5. Does the survey reach all patients? Ask about accessibility, plain language, language versions, paper fallback — and about the response rate the vendor measures in its own deployments.

If you want to sort out your survey obligations — QM-RL, BAR, DRV, works council —, write to us; the conversation is worthwhile even without an aiomics evaluation. Or subscribe to Visite, our weekly briefing on documentation and data quality in German healthcare — an English edition, Grand Rounds, has a waitlist.

Sources

1. Gemeinsamer Bundesausschuss. Qualitätsmanagement-Richtlinie (QM-RL) in der Fassung vom 20. April 2024, §4 und §6. https://www.g-ba.de/richtlinien/87/
2. Bundesarbeitsgemeinschaft für Rehabilitation (BAR). Vereinbarung zum internen Qualitätsmanagement nach §37 Abs. 2 SGB IX, gültig ab 01.07.2025. https://www.bar-frankfurt.de/
3. Deutsche Rentenversicherung. Reha-Qualitätssicherung: Rehabilitandenbefragung. https://www.deutsche-rentenversicherung.de/