A machine learning language model approach to evaluating mental health awareness content across Spanish- and English-language social media posts on Twitter.

Purpose: Mental health information appears on social media in varying levels of quality and may or may not be productive information to users, particularly in relation to healthcare decision-making and community living among diverse populations coping with mental health problems. To better understand the mental health landscape on social media, this study validated a language model approach to evaluating the availability and sentiment of mental health awareness content across Spanish- and English-language social media posts on Twitter (currently X) to inform future mental health communication guidelines.

Methods: A comprehensive list of mental health awareness hashtags in Spanish and English was developed by bilingual investigators to download tweets containing these hashtags in both languages from the Twitter Academic API from 09/19/22 - 10/10/22. Data extraction and cleaning of duplicate tweets resulted in a final sample of 28,268 Spanish and 205,774 English tweets for sentiment and structural topic analysis across the two languages.

Results: Fifteen unique topics emerged for both Spanish and English tweets including overlapping themes of awareness, self-care, lived experience, and service providers. Topics in Spanish tweets were more often significantly associated with negative emotions compared to English tweets. Yet English tweets also included misappropriation of mental health labels to make political statements and market products.

Conclusions: Mental health awareness content on Twitter appears not to be consistently available or aligned with clinical values, disadvantaging Spanish-language social media users, possibly leading to divergent priorities concerning population mental health. Nevertheless, natural language processing techniques offers a viable method to further understand unequal mental health awareness content across various language and cultural social media.