Telemedicine in Nursing Homes Did Not Cut Hospital Admissions, German Trial Finds
Key Takeaways:
- A large cluster-randomised trial across 24 nursing homes in western Germany found no statistically significant reduction in hospital admissions or total days spent in care following an intersectoral, telemedicine-based intervention.
- Researchers tested 1,260 separate model specifications; around two-thirds pointed to a numerical trend towards fewer admissions, but none reached statistical significance (P < .05).
- The authors attribute the null result partly to external shocks, staffing pressures, limited GP engagement and short adaptation windows, and recommend narrower outcome measures and simpler, automated technology in future.
What the study set out to test
A prospective, multicentre cluster-randomised trial has concluded that introducing an intersectoral, telemedicine-based model of care into nursing homes did not produce a statistically significant reduction in hospital admissions among residents.
The research was led by John Grosser, Sophie Pauge, Birthe Aufenberg and Prof. Dr. Wolfgang Greiner of the Department of Health Economics and Health Care Management at Bielefeld University. They worked alongside Miriam Hertwig, Dr. med. Jenny Unterkofler, Dr. med. Christian Hübel and Prof. Dr. med. Jörg Christian Brokmann from the Department for Acute and Emergency Medicine at University Hospital RWTH Aachen, in collaboration with Dr. med. David Brücken of Rhine-Meuse Hospital Würselen and the Optimal@NRW Research Group.
The findings, published in JMIR Aging, evaluate the Optimal@NRW project as it was rolled out across 24 nursing homes in western Germany between May 2021 and April 2023.
The problem the intervention was designed to solve
Optimal@NRW was built around two well-documented pressures in long-term care: non-emergency hospital admissions that could plausibly have been managed in place, and persistent resource shortages across the sector. Transfers to hospital are disruptive for people living in nursing homes, costly for the wider system, and in a meaningful proportion of cases potentially avoidable if clinical assessment and decision support can be brought to the resident rather than the resident to the hospital.
Inside the model of care
The intervention combined four components intended to work as a single, joined-up pathway:
- A 24/7 telemedical consultation centre, giving nursing staff round-the-clock access to remote clinical input.
- Mobile non-physician medical assistants, able to attend the home and carry out assessments in person.
- A virtual emergency hub, coordinating escalation decisions across settings.
- A software-based early warning system for vital signs, designed to flag deterioration before it became acute.
Crucially, the model was intersectoral by design: it was meant to link the nursing home, emergency medicine and primary care rather than sit within any one of them.
A deliberately exhaustive statistical approach
Rather than relying on a single headline model, the research team used specification curve analysis to test how robust any effect was to analytical choices. They evaluated 1,260 distinct mixed-effects regression model specifications, drawing on both primary data collected during the trial and statutory health insurance claims data, with hospitalisation rates and total days spent in care as the outcomes of interest.
The result was consistent across that curve. Roughly two-thirds of the model variations indicated a numerical trend towards reduced hospital admissions, but not one of the 1,260 specifications demonstrated a statistically significant intervention effect at the conventional threshold (P < .05). In other words, the direction of travel was mildly encouraging, but the evidence did not support a claim that the intervention worked.
Why the intervention may not have delivered
The authors are candid that several practical and contextual factors are likely to have compromised the intervention’s chances of success.
External disruptions. The trial period coincided with severe regional flooding in western Germany and with ongoing COVID-19 pandemic restrictions. Both disrupted day-to-day operations in participating facilities and, importantly, distorted baseline hospitalisation rates against which any effect would have been measured.
Staff workload and usability. The model asked a great deal of nursing teams already contending with severe labour shortages. Complex, multi-component tasks – daily manual recording of vital signs among them – added meaningful operational stress rather than relieving it. A digital system that increases the documentation burden on staff is, in practice, competing with the very work it is meant to support.
Physician integration. Engaging general practitioners directly within the telemedical framework proved difficult. Because GP involvement was central to the intersectoral logic of the intervention, that gap limited how far the programme could genuinely connect primary care with the nursing home and the emergency pathway.
Short adaptation windows. Intervention phases ran for between 6 and 15 months per group. The authors suggest this may simply have been too short for staff to embed unfamiliar digital workflows into routine practice, particularly given the competing pressures above.
What the researchers recommend next
The team’s conclusions are constructive rather than dismissive of telemedicine in long-term care. They recommend that future digital health interventions in this setting should:
- Target specific, avoidable admission metrics rather than overall hospitalisations, which are influenced by too many factors outside the intervention’s reach to serve as a sensitive outcome measure.
- Allow extended implementation phases, giving teams realistic time to adapt to new digital workflows.
- Favour simplified, automated solutions such as wearable monitoring devices, reducing the manual data entry burden on nursing staff.
- Pursue deeper structural integration with general practitioners, so that primary care is built into the pathway rather than invited to join it.
What this means for practice
The wider lesson here is one that recurs across digital health evaluation: the technology itself is rarely the binding constraint. Implementation conditions – staffing capacity, workflow design, clinical buy-in and the length of time teams are given to adapt – tend to determine whether a well-conceived tool produces measurable benefit. Practitioners who are asked to assess, adopt or lead on digital tools increasingly need a framework for judging fit and feasibility, not just functionality, which is the ground covered by CPD-accredited training such as CCH’s AI Essentials for Primary Care: Tools, Ethics and Everyday Applications.
For services considering remote monitoring or telemedical support in care homes, the Optimal@NRW findings are a useful corrective. They suggest that ambition should be matched by realism about what frontline teams can absorb, and that outcome measures should be chosen precisely enough to detect an effect if one exists.
CCH insight
Digital tools are arriving in primary and community care faster than most teams can evaluate them. AI Essentials for Primary Care: Tools, Ethics and Everyday Applications is a CPD-accredited short course covering how to appraise digital and AI-enabled tools, work within governance requirements and get reliable results in everyday practice. It carries 3.5 CPD points and counts towards appraisal and revalidation.
Explore AI Essentials for Primary Care →
Source: JMIR Aging
