Do Words Matter? Detecting Social Isolation and Loneliness in Older Adults Using Natural Language Processing

• DOI: 10.3389/fpsyt.2021.728732
• Publication Date: 2021-11-16T05:43:42Z
• Authors: Varsha D. Badal; Camille Nebeker; Kaoru Shinkawa; Yasunori Yamada; Kelly E. Rentscher; Ho-Cheol Kim; Ellen E. Lee
• Keywords: Plural; Social Isolation; Sentiment Analysis

Introduction: Social isolation and loneliness (SI/L) are growing problems with serious health implications for older adults, especially in light of the COVID-19 pandemic. We examined transcripts from semi-structured interviews 97 adults (mean age 83 years) to identify linguistic features SI/L. Methods: Natural Language Processing (NLP) methods were used relevant interview segments (responses specific questions), extract type number social contacts such as sentiment, parts-of-speech, syntactic complexity. examined: (1) associations NLP-derived assessments relationships validated self-report support loneliness; (2) important detecting individuals higher level SI/L by using machine learning (ML) models. Results: associated self-reported loneliness, though these stronger women than men. Usage first-person plural pronouns was negatively positively emotional ML analysis leave-one-out methodology showed good performance (F1 = 0.73, AUC 0.75, specificity 0.76, sensitivity 0.69) binary classification models Comparable also observed when classifying measures. Using models, we identified several (including use pronouns, sentence complexity, similarity) that most strongly predicted scores on scales support. Discussion: Linguistic data can provide unique insights into among beyond scale-based assessments, there consistent gender differences. Future research studies incorporate diverse well other behavioral data-streams may be better able capture complexity functioning identification target subpopulations future interventions. Given novelty, NLP should include prospective consideration bias, fairness, accountability, related ethical implications.