Examining spoken words and acoustic features of therapy sessions to understand family caregivers' anxiety and quality of life.

BACKGROUND: Speech and language cues are considered significant data sources that can reveal insights into one's behavior and well-being. The goal of this study is to evaluate how different machine learning (ML) classifiers trained both on the spoken word and acoustic features during live conversations between family caregivers and a therapist, correlate to anxiety and quality of life (QoL) as assessed by validated instruments. METHODS: The dataset comprised of 124 audio-recorded and professionally transcribed discussions between family caregivers of hospice patients and a therapist, of challenges they faced in their caregiving role, and standardized assessments of self-reported QoL and anxiety. We custom-built and trained an Automated Speech Recognition (ASR) system on older adult voices and created a logistic regression-based classifier that incorporated audio-based features. The classification process automated the QoL scoring and display of the score in real time, replacing hand-coding for self-reported assessments with a machine learning identified classifier. FINDINGS: Of the 124 audio files and their transcripts, 87 of these transcripts (70%) were selected to serve as the training set, holding the remaining 30% of the data for evaluation. For anxiety, the results of adding the dimension of sound and an automated speech-to-text transcription outperformed the prior classifier trained only on human-rendered transcriptions. Specifically, precision improved from 86% to 92%, accuracy from 81% to 89%, and recall from 78% to 88%. INTERPRETATION: Classifiers can be developed through ML techniques which can indicate improvements in QoL measures with a reasonable degree of accuracy. Examining the content, sound of the voice and context of the conversation provides insights into additional factors affecting anxiety and QoL that could be addressed in tailored therapy and the design of conversational agents serving as therapy chatbots.

Spoken words as biomarkers: using machine learning to gain insight into communication as a predictor of anxiety.

OBJECTIVE: The goal of this study was to explore whether features of recorded and transcribed audio communication data extracted by machine learning algorithms can be used to train a classifier for anxiety. MATERIALS AND METHODS: We used a secondary data set generated by a clinical trial examining problem-solving therapy for hospice caregivers consisting of 140 transcripts of multiple, sequential conversations between an interviewer and a family caregiver along with standardized assessments of anxiety prior to each session; 98 of these transcripts (70%) served as the training set, holding the remaining 30% of the data for evaluation. RESULTS: A classifier for anxiety was developed relying on language-based features. An 86% precision, 78% recall, 81% accuracy, and 84% specificity were achieved with the use of the trained classifiers. High anxiety inflections were found among recently bereaved caregivers and were usually connected to issues related to transitioning out of the caregiving role. This analysis highlighted the impact of lowering anxiety by increasing reciprocity between interviewers and caregivers. CONCLUSION: Verbal communication can provide a platform for machine learning tools to highlight and predict behavioral health indicators and trends.