Accelerometer Bluetooth proximity validation in parents and early years children

Abstract

ABSTRACTThe purpose of this study was to examine the validity of a novel technique determining parent–child proximity using Bluetooth-enabled accelerometers. Nineteen parents and children (18–60 months; 38.4 ± 13.2 months) wore an ActiGraph wGT3X-BT accelerometer for 7 days. Parental accelerometers continuously emitted a Bluetooth signal, while children’s accelerometers recorded signal detection once per minute. Parents reported whether they were “AWAY”/“WITH” their child using time-use diaries in 5-minute intervals for 2 days. Presence (1) or absence (0) of a Bluetooth signal was averaged over the 5-minute time frame. Accelerometer and time-use diary variables produced 8,331 comparative observations. Area under the curve (AUC) via receiver operator characteristic (ROC) curve was 0.84 (95% confidence intervals: 0.84, 0.85), and the optimal cut-point was 0.0. Specifically, sensitivity (0.82) and specificity (0.81) were maximized when classifying all positive proximity tags (0.20–1.00) as “WITH” and all negative proximity tags (0.0) as “AWAY”. Accelerometer-derived parent–child proximity demonstrated good concurrent validity.