Khan 2024 — REACT: Real-time Early Agitation Capture Technology
Citation: Khan, N., Plunk, A., Zheng, Z., Adiani, D., Staubitz, J., et al. (2024). Digital Health, 10, 20552076241287884.
DOI: 10.1177/20552076241287884
Type: Pilot study
Key Contribution
The closest existing prototype to Meltdown Minder. Real-time wearable agitation detection for children with IDD/ASD in naturalistic settings. Pre-trained models and dataset available from Vanderbilt.
Design
- N=3 children with IDD (6-9 yrs)
- Naturalistic (real-world) setting
- Wearable HR + EDA + accelerometry
- Real-time ML classification
Key Results
- F1 = 78.69% at time of agitation
- F1 = 68.20% 20 seconds prior
Limitations
- Very small N (3 participants)
- No published follow-up as of July 2026 (5 citations)
- Standalone pilot — no known active development pipeline
- Detection window (20s) shorter than Imbiriba (3 min) suggesting multiple detection horizons exist
Linked To
- Dataset: react-vanderbilt
- Full study notes: react-study.md