Social media as a measurement tool of depression in populations

TLDR

Depression is a serious public health challenge, and social‑media data can complement traditional surveys by providing finer‑grained, large‑scale measurements over time. We examine whether social‑media postings can serve as a new lens for understanding depression in populations. We crowdsourced a large corpus of Twitter posts from clinically depressed users and trained a probabilistic model that uses social‑activity, emotional, and linguistic signals to detect depression. The resulting depression index captures geographic, demographic, and seasonal patterns that align with psychiatric findings and correlate strongly with CDC depression statistics.

Abstract

Depression is a serious and widespread public health challenge. We examine the potential for leveraging social media postings as a new type of lens in understanding depression in populations. Information gleaned from social media bears potential to complement traditional survey techniques in its ability to provide finer grained measurements over time while radically expanding population sample sizes. We present work on using a crowdsourcing methodology to build a large corpus of postings on Twitter that have been shared by individuals diagnosed with clinical depression. Next, we develop a probabilistic model trained on this corpus to determine if posts could indicate depression. The model leverages signals of social activity, emotion, and language manifested on Twitter. Using the model, we introduce a social media depression index that may serve to characterize levels of depression in populations. Geographical, demographic and seasonal patterns of depression given by the measure confirm psychiatric findings and correlate highly with depression statistics reported by the Centers for Disease Control and Prevention (CDC).