This repository contains the code associated with the study:

An interpretable machine learning tool for in-home monitoring of agitation episodes in people living with dementia: a proof-of-concept study The Lancet eClinicalMedicine, 2025 https://www.thelancet.com/journals/eclinm/article/PIIS2589-5370(24)00611-4/fulltext

Summary

We present an interpretable machine learning model for the weekly monitoring of agitation episodes in people living with dementia, using remotely and passively collected in-home sensor data.

Using explainability techniques, we identified actionable environmental and behavioral factors—such as sleep patterns, ambient light, and temperature—that are associated with agitation episodes.

To support the development of personalised, non-pharmacological interventions, we developed an interactive interface that allows users to:

Simulate changes in modifiable environmental factors (e.g. lighting, temperature)

Observe how these changes influence the model’s predicted agitation probability

Explore and plan tailored care strategies with transparency and control

This tool aims to empower caregivers and clinicians with data-driven, interpretable, and actionable insights into agitation in dementia care.