870: Posts to an anonymous digital health platform may help assess depression risk among pregnant women

Abstract

The Edinburgh Postnatal Depression Scale (EPDS) is the most widely used tool for depression screening among perinatal women. Recent literature leveraging machine learning and natural language processing has found associations between the content and structure of social media posts and the likelihood of screening positive for depression. The objective of this study was to determine which one- and two-word combinations used in the Community feature of Ovia Pregnancy — a platform where users can post and respond to questions anonymously — are associated with a score of “probable depression” (13+) on the EPDS. A retrospective observational study was performed based on data from 148,347 Ovia Pregnancy app users who completed the EPDS between May 2017 and April 2018. A statistical modeling software used regression methodologies to assess the association between posts in Ovia Pregnancy’s Community and EPDS scores. Women who scored either ≤7 (n=86,235) or 13+ (n=62,112) on the EPDS were included (n=148,347). Questions (608,330) and comments (4,520,889) from Ovia’s Community were analyzed. The words ‘depression’, ‘depressed’, ‘anxiety’, and ‘boyfriend’ were most strongly associated with high EPDS scores, while ‘nursing’, ‘read’, ‘husband’, and ‘hubby’ were the four words most strongly associated with low EPDS scores. Women who scored 13+ on the EPDS were more likely to post words related to mental health conditions. Additionally, there seems to be an association between the type of committed relationship a woman is in during her pregnancy and her EPDS score. The way app users present themselves and their thoughts in an anonymous digital platform may help assess risk of depression.