EULAR Rheumatology Open 2025 · Peer-reviewed · Open Access
Clinically evaluated.
Nixi AI was used in everyday clinical practice in a peer-reviewed real-world study, in Dr Peer Aries' rheumatology practice (Immunologikum Hamburg), on exactly the kind of consultation that makes documentation hardest: long, complex rheumatology conversations with disease course, medication changes and multiple diagnoses. 88 consultations, 108 patients. The result: in 72 of 88 consultations the treating rheumatologist estimated consultation time as reduced, no major errors, and no patient declined to participate.
Last updated: May 2026
Study at a glance
| Title | Ambient AI scribes in rheumatology: early real-world clinician and patient experience |
|---|---|
| Authors | Dr. med. Johannes Knitza (ORCID 0000-0001-9695-0657), Institute for Digital Medicine, University Hospital Marburg, Philipps-Universität Marburg · Dr. med. Peer Aries, Department of Rheumatology, Immunologikum, Hamburg |
| Journal | EULAR Rheumatology Open |
| Year | 2025 |
| DOI | 10.1016/j.ero.2025.12.002 |
| PubMed Central | PMC13425160 · PMID 42540133 |
| Licence | Open Access · CC BY 4.0 |
| Study design | Real-world study (research letter), single-centre, one treating rheumatologist |
| Sample | 88 consultations, 108 patients |
| Setting | Rheumatology practice, Germany |
| System studied | Nixi AI (ambient AI documentation) |
Primary outcome
Consultation time estimated as reduced in 72 of 88 consultations.
In 72 of 88 consultations the treating rheumatologist estimated consultation time as reduced. In 62 of 88 a saving of two to five minutes was reported, in 10 of 88 more than five minutes. These are the clinician's estimates, not measured times.
72 of 88 consultations where the treating rheumatologist estimated consultation time as reduced
Full breakdown. Perceived time change per consultation:
| Perceived time change per consultation (n=88) | n=88 |
|---|---|
| Consultation time estimated as reduced | 72 of 88 |
| → of which 2–5 minutes | 62 of 88 |
| → of which more than 5 minutes | 10 of 88 |
| No meaningful change | 13 of 88 |
| Consultation time increased (2 by 2–5 minutes, 1 by more than 5 minutes) | 3 of 88 |
Model calculation: how the study values become “about 1 hour per day”
72 of 88 consultations × avg 3.5 min, extrapolated to 20 consultations ≈ 60 min per day
Model calculation for 20 consultations; not directly measured. The study records an estimate per consultation (2–5 minutes, and more than 5 minutes in 10 of 88). A per-day figure only arises by extrapolating to an assumed consultation count, and depends on specialty, caseload and documentation style.
We deliberately don't write "up to 2 hours". The data don't support it.
Accuracy
Zero major errors. 19% minor deviations.
Errors were reported in 17 of 88 consultations (19%): misspelled medication names and incorrect sex detection. All errors were rated as minor. Major errors: zero.
Across the 88 documented consultations: 19% contained minor errors (misspelled medication names, incorrect sex detection). Major errors: zero.
What "minor" means:
The reported errors were misspelled medication names and incorrect sex (female/male) detection. The study rated all of them as minor. The clinician reviews every note before release, which is where such errors are corrected.
What "zero major" means:
Under the framework the study used (Asgari et al.), major errors are those that could alter diagnosis or management if left uncorrected. None occurred: zero out of 88 consultations.
Why we publish the 19%:
An accuracy figure without an error rate is incomplete. We show the 19% so you can judge where the clinician's review remains essential. Our aim is to lower that rate with every release.
Patient acceptance
108 patients surveyed. Zero declined.
Of 108 surveyed patients (64.8% female, mean age 59), no one declined to participate. 56% had a positive attitude toward AI in healthcare, 14% very positive, 30% neutral. 60% agreed or strongly agreed that the clinician spent less time on the screen and more time talking with them.
Patient demographics:
108
Patients
64.8%
Female
59 years
Mean age
0%
Decline rate
Attitude toward AI in healthcare:
| Very positive | 14% |
| Positive | 56% |
| Neutral | 30% |
| Negative or refusing | 0% |
60% of patients agreed (34%) or strongly agreed (26%) that, compared with previous visits, the clinician spent less time on the screen and more time talking with them. 62% would welcome the technology at future appointments.
The most common pushback we hear from clinicians: "My patients won't accept this." In this study none of 108 patients declined to participate, and 60% noticed the clinician was more present with them.
Transparency
What we don't yet know. And why we say it.
Every study has limitations. This one too. We name them because honesty about study limits is a stronger trust signal than a marketing slide without sources.
1/5
Single-centre, one rheumatologist
The study was conducted in a single rheumatology practice, with one treating clinician. The findings are therefore not directly transferable to all specialties, settings, or individual documentation styles.
2/5
Small sample size
88 consultations and 108 patients. That's a robust starting point but not an RCT with thousands of participants. We are planning follow-on studies with larger samples and multiple sites.
3/5
Self-reported time perception
Time savings were not measured by stopwatch but estimated by the clinicians themselves. Self-reported data can be systematically biased, in either direction.
4/5
Short observation window
The study covers a limited time period. Long-term effects, e.g. acclimatisation, workflow changes after 6 or 12 months, have not yet been investigated.
5/5
Copy-paste into the practice system
In the study setting, Nixi AI was not API-connected to the practice management system (PVS). Clinicians transferred the generated documentation manually via copy-paste. Nixi AI Practice Pro removes this step with the optional direct EMR integration. That advantage was not evaluated in this study.
We show the limitations because clinicians see through vendors who only show their best numbers.
Study co-author
What the study co-author says about everyday practice.
“You only realise the true value of Nixi AI when you have to work without it. The other day I had to see patients in a room without Nixi AI and felt genuinely constrained. In just three weeks, Nixi AI has become so fundamental to my practice that I can hardly imagine high-quality patient care without it.”
Disclosure: Dr. Aries is one of the two authors of the EULAR study cited above and is on our Clinical Advisory Board. The study's methodology, data collection and analysis were carried out independently by the study authors and confirmed in peer review; Nixi AI was not involved. His quoted statement here reflects his daily use of Nixi AI in his rheumatology and immunology practice.
For your practice
What this study means for your practice.
If you're a solo clinician:
In 72 of 88 consultations consultation time was estimated as reduced. Extrapolated to a practice day of 20 consultations that is about one hour (model calculation, not directly measured). Basic is free; Practice Pro from €89/month or €828/year.
If you work in a practice team:
The study shows that Nixi AI works in real-world practice, not only under lab conditions. Practice Pro adds the EMR integration that the study had to handle via copy-paste.
If you decide for a clinic or MVZ:
Standardised documentation with measurable time savings, zero major errors, and patients who welcome the use. Those are the three KPIs your leadership wants to see.
Frequently asked
Questions about the EULAR study
Conducted by Dr Johannes Knitza (Institute for Digital Medicine, University Hospital Marburg) and Dr Peer Aries (Rheumatology, Immunologikum Hamburg); the consultations took place in Dr Aries' practice. The Philipps University Marburg ethics committee confirmed that no ethical approval was required (25-215 ANZ). Published in EULAR Rheumatology Open 2025 (DOI 10.1016/j.ero.2025.12.002), externally peer-reviewed, Open Access under CC BY 4.0. Nixi AI provided the system; methodology, data collection and analysis sat with the study authors. Dr Aries is on Nixi AI's clinical advisory board. We state all of this so you can judge independence for yourself.
See for yourself what Nixi AI makes of a consultation.
Ready to get your evenings back?
Join the clinicians using Nixi AI to finish documentation in minutes. Not hours.