AI wearables can now detect the physiological signs of dementia agitation before any visible crisis appears. A 2026 review of 13 studies found machine learning models predicting these episodes with striking accuracy, and early trials show real reductions in distress scores for people wearing the devices.
How the Technology Actually Works
The hardware is simpler than you might expect. Wearable devices, some resembling fitness bands and others designed as specialized socks, track heart rate variability, skin conductance, and movement patterns continuously. When readings deviate from a person’s individual baseline, the system alerts a caregiver in time to intervene calmly.
The real power sits in the AI layer. Deep learning models called LSTM networks are trained on each person’s own data rather than population averages, which keeps false alarms low. In one study, an LSTM model achieved up to 98.6% accuracy and 84.8% recall in predicting agitation episodes. Alerts flow to caregiver smartphones or facility dashboards, enabling de-escalation before a crisis fully develops.
Real World Trials and Results
Early results are genuinely encouraging. A feasibility study of SmartSocks found that participants showed reduced agitation scores on both the Abbey Pain Scale and the Neuropsychiatric Inventory after two weeks of wearing the devices. Over 70% of dementia patients are cared for at home, making community-ready wearables a critical development priority. Most trials remain small and controlled, so broader community studies are still needed before these tools reach widespread use.
