AudreyElliottArticle2

Understanding Side Effects of Antidepressants

Authors and Affiliation

• Koustuv Saha (BTech), John Torous (MD), Emre Kiciman (PhD), Munmun De Choudhury (PhD)

  • 1Georgia Institute of Technology, Atlanta, GA, United States

  • 2Department of Psychiatry, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, United States

  • 3Microsoft Research, Redmond, WA, United States

• Corresponding Author: Koustuv Saha, BTech

  • Email: koustuv.saha@gatech.edu

  • Phone: 1 4046929496

Abstract

• Background:

  • Antidepressants have heterogeneous effects on individuals, which complicates understanding their efficacy.

  • Social media serves as a large-scale, unobtrusive data source to observe individuals' behavior while taking antidepressants.

• Objective:

  • This study aims to analyze side effects of antidepressants through social media expressions.

• Methods:

  • Utilized Twitter data from over 300,000 users between Jan 2014 and Feb 2018, conducting a quasi-experimental study to explore keywords linked to improvements or lack thereof in individuals taking antidepressants.

• Results:

  • Identified five major side effects: sleep, weight, eating changes, pain, and sexual issues.

  • Keywords associated with these side effects were extracted, showing a spectrum ranging from decrease to increase across effects.

• Conclusions:

  • Findings offer insight into the side effects of antidepressants and suggest the potential use of social media for digital pharmacovigilance, informing clinicians and patients about expected aftereffects.

Keywords

• antidepressants, symptoms, side effects, digital pharmacovigilance, social media, mental health, linguistic markers, digital health

Introduction

• Context:

  • Rising mental health issues, worsened by the COVID-19 pandemic, necessitate clearer understanding of antidepressant impacts.

  • Traditional efficacy assessment via randomized controlled trials can be biased; individual subjective experiences are essential but challenging to incorporate in evaluations.

• Common Side Effects:

  • Include gastrointestinal issues, weight changes, sleep disturbances, and sexual dysfunction.

  • Side effects contribute to medication discontinuation, underscoring the importance of accurate prediction.

• Research Gap:

  • Despite existing evaluations, understanding individual-specific effects requires larger sample sizes. Social media can reveal hidden side effects not previously observed in clinical settings.

• Methodological Approach:

  • Leveraging social media data for unobtrusive, real-time insights into psychosocial concerns and behaviors enables broader understanding of mental health.

Data and Methods

• Data Source:

  • Utilized Twitter API to gather posts mentioning FDA-approved antidepressants from Jan 2015 to Dec 2016, eventually analyzing a dataset of 112,052,496 posts from 23,191 users who self-reported drug intake.

  • Established a control group from users without reported medication intake.

• Analysis Techniques:

  • Employed quasi-experimental design and propensity score matching to assess treatment effects on users' symptomatic expression.

  • Utilized the Sparse Additive Generative (SAGE) model for keyword extraction to identify linguistic markers related to side effects.

Results

Overview

• Comparison of post-treatment linguistic markers reveals unique keywords associated with symptomatic outcomes. Keywords were categorized into five significant side effects based on their prevalence among users showing varying levels of improvement.

Key Findings

• Linguistic Markers:

  • Identified notable keywords for increased/decreased experiences related to each side effect, giving insights into how individuals articulate their responses on social media.

Side Effects Analysis

1. Sleep Issues:

   • Common keywords for improved users included phrases indicating ease of sleeping, whereas those less improved expressed frustration with sleeplessness.

2. Weight Changes:

   • Keywords indicating weight loss were common among improved users, while others reported gains.

3. Eating Behavior:

   • Contrasting phrases reflected improvements or struggles with eating habits.

4. Pain Management:

   • Keywords relating to pain were prominent in responses to medications like Duloxetine, showcasing individualized pain experiences.

5. Sexual Function:

   • Variations in keywords revealed differing experiences of sexual side effects related to antidepressant use.

Discussion

• Implications:

  • This study highlights how social media's vast data can offer insights into underreported side effects of antidepressants, which may help inform clinical practices.

• Future Directions:

  • Encourages development of approaches combining social media with traditional sources to enhance understanding of patient experiences and drug side effects.

• Note on Limitations:

  • While providing robust findings, the quasi-experimental design does not establish true causality, thus further research is warranted.

Acknowledgments

• Supported by National Institutes of Health grant #R01MH117172.

• No conflicts of interest declared.

References

• A selection from various studies on antidepressants, social media analysis, and their interplay with mental health outcomes is provided.