• radar-base@kcl.ac.uk | radar-base@thehyve.nl
  • KCL +44 (0)20 7848 0924 | The Hyve +31 (0)61 169 6842

Blog

đź””SPOTLIGHT: Using Local Bluetooth Device Count to Predict Depressive Symptom Severity
News, Spotlight

đź””SPOTLIGHT: Using Local Bluetooth Device Count to Predict Depressive Symptom Severity

PAPER Spotlight: “Predicting Depressive Symptom Severity Through Individuals’ Nearby Bluetooth Device Count Data Collected by Mobile Phones: Preliminary Longitudinal Study” 10.2196/29840.

Bluetooth-enabled devices are pervasive in the modern environment, most commonly in the form of smartphones. This study used the RADAR-base Passive app to probe this environment, reporting the count of nearby devices detected. Nearby Bluetooth Device Counts (NBDC) were used to investigate the relationship between this metric and mood/depression. Studies in the field of mental health have discovered correlations between depression and various behaviours and states of individuals, including their social connections, employment status, physical mobility, and experiences of isolation and loneliness. These behaviours and circumstances can be approximated using the nearby Bluetooth device count (NBDC) captured by Bluetooth sensors on mobile phones.

The present study utilized a dataset comprising 2886 bi-weekly PHQ-8 assessments obtained from 316 participants sourced from three research locations in the Netherlands, Spain, and the United Kingdom, as part of the EU RADAR-CNS study. From the NBDC data available for two weeks prior to each PHQ-8 score, 49 Bluetooth features, including statistical and non-linear features, were extracted to measure the individuals’ rhythm periodicity and regularity. Associations between Bluetooth features and PHQ-8 scores were explored using linear mixed-effects models. Subsequently, hierarchical Bayesian linear regression models were employed to predict PHQ-8 scores based on the extracted Bluetooth features.

CategoryAbbreviationDescriptionNumber of features (N=49)
Statistical features[Second-order feature]_[Daily feature], eg, Max_MeanSecond-order features (max, min, mean, and standard deviation) calculated in the PHQ-8a interval based on daily statistical Bluetooth features (max, min, mean, and standard deviation).16
Multiscale entropy (MSE)MSE_1, MSE_2, …, MSE_24Multiscale entropy of the NBDCb sequences from scale 1 to scale 24.24
Frequency domaincLF_sum, MF_sum, HF_sumThe sums of spectrum power in LF, MF, and HF.3
Frequency domainLF_pct, MF_pct, HF_pctThe percentages of spectrum power in LF, MF, and HF to the total spectrum power.3
Frequency domainLF_se, MF_se, HF_seSpectral entropy in LF, MF, and HF.3
Summary of 49 Bluetooth features used in this paper and their short descriptions.

aPHQ-8: 8-item Patient Health Questionnaire.

bNBDC: nearby Bluetooth device count.

cLF: low frequency (0-0.75 cycles/day); MF: middle frequency (0.75-1.25 cycles/day); HF: high frequency (>1.25 cycles/day).

The analysis of the data revealed a number of statistically significant associations between Bluetooth features and the severity of depressive symptoms. In general, it was found that as the severity of depressive symptoms worsened, one or more of the following changes were observed in the two weeks preceding the PHQ-8 assessment: (1) a decrease in the amount of data, (2) a decrease in the variance, (3) a decrease in the periodicity, particularly in the circadian rhythm, and (4) a heightened irregularity in the NBDC sequence. The proposed hierarchical Bayesian linear regression model demonstrated improved prediction performance compared to commonly used machine learning models, with a coefficient of determination (R2) of 0.526 and a root mean squared error (RMSE) of 3.891. The inclusion of Bluetooth features in the model accounted for an additional 18.8% of the variance in the PHQ-8 scores as compared to a baseline model without Bluetooth features (R2=0.338, RMSE=4.547).

Boxplots of the nearby Bluetooth devices count (NBDC) for every hour in the whole population. Boxes extend between the 25th and 75th percentiles, and green solid lines inside the boxes are medians. Note the relative stationary NBDC during the night-time hours

The results of our statistical analysis suggest that the NBDC data can serve as an indicator of changes in an individual’s behaviour and well-being concurrent with changes in their depressive state. The prediction results indicate a significant value in using NBDC data for the prediction of depressive symptom severity. These findings hold potential applications in the field of mental health monitoring in real-world settings.

The project is led by Dr Yuezhou, Dr Amos Folarin and Prof. Richard Dobson

Read more: