ACT Algorithm Calibration: Synthetic Data Generation and Validation

Supplementary data repository for: Brill, M., & Whalley, J. H. (2025). Development of a Digital Anticipatory Care Tool for People Living with Dementia: A Co-Design and Technical Validation Study.This repository contains the synthetic dataset (100 persons, 2,400 observations), generation code, validation outputs, and supplementary figures for the algorithm calibration component of the Anticipatory Care Tool (ACT). ACT is a digital application supporting non-clinical caregivers in observing and reporting wellbeing changes in people living with dementia.The synthetic data was generated using a parametric bootstrap with per-person AR(1) temporal modelling, fitted to observational data from 12 participants over 24 weeks at the Lifecare Centre, Edinburgh. Four adapted QoL-AD dimensions (1-7 Likert scale) are included: Sad/Happy, Inactive/Active, Dependent/Independent, andSolitary/Social. Known decline scenarios (gradual, sudden, correlation decoupling, recovery) were injected into 20% of trajectories for algorithm sensitivity testing.Validation confirmed no significant distributional differences (all KS p > 0.30), preserved temporal and cross-variable correlation structure, and no privacy memorisation risk.Contents: synthetic cohort CSV, data dictionary, Python generation code, Jupyter notebook, supplementary information document, 10 validation figures, and validation summary.