Rambam Health Care Campus
INSPIRE, an LLM tool that drafts nerve and muscle test reports (randomized trial)
A randomized trial at Israel's Rambam hospital found AI-drafted nerve test reports did not raise report quality
Problem
Neurologists at Rambam write a report for each electrodiagnostic (EMG and nerve conduction) exam. The hospital's team wanted to know whether an AI-drafted first version would improve the reports.
Approach
The team built INSPIRE, a multi-agent system on a GPT-4o-class model that retrieves from an EMG textbook. In January and February 2025, 200 patients at the outpatient EMG clinic were randomized to physician-only reports or physician plus AI draft. Three neurologists rotated between the two arms.
Results
Takeaway
Report quality did not differ significantly between the groups. Neurologists rated efficiency and ease of use poorly.
Categories
- Industry
- Healthcare and life sciences
- Business function
- Internal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- Generative AIAI agents
- Adoption
- Built in-house
- Evidence
- Independent study
Sources
Source summaries
A single-center trial at Rambam in Haifa randomized 200 patients to physician-only or physician-plus-AI reporting. Report quality was 0.94 in both arms (p = 0.32). The AI flagged abnormal studies 93.3% of the time and normal ones 92.0%, but physicians rated efficiency, ease of use and workload reduction low.
Full details are in the original sources.