Machine learning prediction of blood alcohol concentration: a digital signature of smart-breathalyzer behavior

Longitudinal Patterns of Engagement and Clinical Outcomes: Results From a Therapist-Supported Digital Mental Health Intervention

OBJECTIVE

Digital mental health interventions (DMHIs) are an effective treatment modality for common mental disorders like depression and anxiety; however, the role of intervention engagement as a longitudinal "dosing" factor is poorly understood in relation to clinical outcomes.

METHODS

We studied 4978 participants in a 12-week therapist-supported DMHI (June 2020-December 2021), applying a longitudinal agglomerative hierarchical cluster analysis to the number of days per week of intervention engagement. The proportion of people demonstrating remission in depression and anxiety symptoms during the intervention was calculated for each cluster. Multivariable logistic regression models were fit to examine associations between the engagement clusters and symptom remission, adjusting for demographic and clinical characteristics.

RESULTS

Based on clinical interpretability and stopping rules, four clusters were derived from the hierarchical cluster analysis (in descending order): a) sustained high engagers (45.0%), b) late disengagers (24.1%), c) early disengagers (22.5%), and d) immediate disengagers (8.4%). Bivariate and multivariate analyses supported a dose-response relationship between engagement and depression symptom remission, whereas the pattern was partially evident for anxiety symptom remission. In multivariable logistic regression models, older age groups, male participants, and Asians had increased odds of achieving depression and anxiety symptom remission, whereas higher odds of anxiety symptom remission were observed among gender-expansive individuals.

CONCLUSIONS

Segmentation based on the frequency of engagement performs well in discerning timing of intervention disengagement and a dose-response relationship with clinical outcomes. The findings among the demographic subpopulations indicate that therapist-supported DMHIs may be effective in addressing mental health problems among patients who disproportionately experience stigma and structural barriers to care. Machine learning models can enable precision care by delineating how heterogeneous patterns of engagement over time relate to clinical outcomes. This empirical identification may help clinicians personalize and optimize interventions to prevent premature disengagement.

SpeCamX: mobile app that turns unmodified smartphones into multispectral imagers

We present the development of SpeCamX, a mobile application that enables an unmodified smartphone into a multispectral imager. Multispectral imaging provides detailed spectral information about objects or scenes, but its accessibility has been limited due to its specialized requirements for the device. SpeCamX overcomes this limitation by utilizing the RGB photographs captured by smartphones and converting them into multispectral images spanning a range of 420 to 680 nm without a need for internal modifications or external attachments. The app also includes plugin functions for extracting medical information from the resulting multispectral data cube. In a clinical study, SpeCamX was used to implement an augmented smartphone bilirubinometer, predicting blood bilirubin levels (BBL) with superior performance in accuracy, efficiency and stability compared to default smartphone cameras. This innovative technology democratizes multispectral imaging, making it accessible to a wider audience and opening new possibilities for both medical and non-medical applications.

Development and Performance Evaluation of an IoT-Integrated Breath Analyzer

Although alcohol consumption may produce effects that can be beneficial or harmful, alcohol consumption prevails among communities around the globe. Additionally, alcohol consumption patterns may be associated with several factors among communities and individuals. Numerous technologies and methods are implemented to enhance the detection and tracking of alcohol consumption, such as vehicle-integrated and wearable devices. In this paper, we present a cellular-based Internet of Things (IoT) implementation in a breath analyzer to enable data collection from multiple users via a single device. Cellular technology using hypertext transfer protocol (HTTP) was implemented as an IoT gateway. IoT integration enabled the direct retrieval of information from a database relative to the device and direct upload of data from the device onto the database. A manually developed threshold algorithm was implemented to quantify alcohol concentrations within a range from 0 to 200 mcg/100 mL breath alcohol content using electrochemical reactions in a fuel-cell sensor. Two data collections were performed: one was used for the development of the model and was split into two sets for model development and on-machine validation, and another was used as an experimental verification test. An overall accuracy of 98.16% was achieved, and relative standard deviations within the range from 1.41% to 2.69% were achieved, indicating the reliable repeatability of the results. The implication of this paper is that the developed device (an IoT-integrated breath analyzer) may provide practical assistance for healthcare representatives and researchers when conducting studies involving the detection and data collection of alcohol consumption patterns.

The Clinical Course of Alcohol Use Disorder Depicted by Digital Biomarkers

Aims: This study introduces new digital biomarkers to be used as precise, objective tools to measure and describe the clinical course of patients with alcohol use disorder (AUD).

Methods: An algorithm is outlined for the calculation of a new digital biomarker, the recovery and exacerbation index (REI), which describes the current trend in a patient's clinical course of AUD. A threshold applied to the REI identifies the starting point and the length of an exacerbation event (EE). The disease patterns and periodicity are described by the number, length, and distance between EEs. The algorithms were tested on data from patients from previous clinical trials (n = 51) and clinical practice (n = 1,717).

Results: Our study indicates that the digital biomarker-based description of the clinical course of AUD might be superior to the traditional self-reported relapse/remission concept and conventional biomarkers due to higher data quality (alcohol measured) and time resolution. We found that EEs and the REI introduce distinct tools to identify qualitative and quantitative differences in drinking patterns (drinks per drinking day, phosphatidyl ethanol levels, weekday and holiday patterns) and effect of treatment time.

Conclusions: This study indicates that the disease state—level, trend and periodicity—can be mathematically described and visualized with digital biomarkers, thereby improving knowledge about the clinical course of AUD and enabling clinical decision-making and adaptive care. The algorithms provide a basis for machine-learning-driven research that might also be applied for other disorders where daily data are available from digital health systems.