Variations in Facebook Posting Patterns Across Validated Patient Health Conditions: A Prospective Cohort Study

Robert J. Smith, Patrick Crutchley, H. Andrew Schwartz, Lyle Ungar, Frances S. Shofer, Kevin Padrez, Raina M. MerchantJournal of Medical Internet Research 2017 · DOI

TL;DR

This study explored how Facebook posting patterns vary across different patient health conditions, finding that individuals with depression tend to post more frequently and about health-related topics. This suggests that social media platforms could be valuable tools for targeted health interventions, particularly for those experiencing depression.

What Problem Does This Paper Address?

The increasing prevalence of social media has opened up new avenues for understanding and potentially intervening in public health. However, there's a significant gap in understanding which patient groups are most engaged with social media and how their posting behavior might relate to their health conditions. This paper aims to bridge this gap by investigating if social media posting frequency and content differ based on validated patient health conditions, thereby identifying potential targets for social media-based health interventions. The authors highlight that while social media offers a vast platform for engagement, its utility for health interventions depends on understanding user behavior and its correlation with health status.

Methods

The researchers employed a prospective cohort study design, recruiting participants from an urban emergency department. Key methods include:

  • Data Collection: Participants consented to share data from their Facebook accounts and electronic medical records (EMRs).
  • Natural Language Processing (NLP): Latent Dirichlet Allocation (LDA), a topic modeling technique, was used to analyze the content of Facebook posts. LDA identifies underlying "topics" within a corpus of text by assuming each document is a mixture of topics and each topic is a mixture of words. This allowed the researchers to distill language into a smaller feature space and identify themes.
  • Statistical Analysis:
    • Language and Posting Frequency: Standardized regression coefficients (β) were used to correlate language variables (topics) with posting frequency, controlling for demographics (age, sex, race). Multiple comparison corrections (Benjamini-Hochberg) were applied.
  • Posting Frequency and Diagnoses: Log transformations were applied to the number of posts due to non-normal distribution. Two-sample t-tests and one-way ANOVA were used for comparisons. Analysis of covariance (ANCOVA) was employed to adjust for covariates like sex, age, and race when comparing posting frequencies across disease diagnoses.
  • Perception of Social Media Use: Participants' self-reported posting frequency was compared to their actual posting behavior.

Datasets

  • Facebook Data: Status updates, photos, likes, and demographics were extracted from participants' Facebook accounts.
  • Electronic Medical Records (EMRs): Demographic information, International Classification of Diseases, Ninth Revision (ICD-9) codes for prevalent health conditions, and Patient Health Questionnaire-2 (PHQ-2) depression screen results were obtained.

No new benchmarks were introduced; the study utilized existing data sources and standard NLP and statistical methods.

Key Insights and Ideas

  • Health-Related Content from High-Frequency Posters: A significant finding is that the language topics most positively correlated with high Facebook posting frequency were related to health and illness (e.g., "sleep," "insomnia," "cough," "headache"). This suggests that individuals who are more active on social media may be more inclined to discuss their health.
  • Depression as a Key Correlate: The study found a strong association between depression and higher Facebook posting frequency. Individuals who screened positive for depression or had a history of depression posted significantly more than those without. This suggests depression might be a key indicator for increased social media engagement.
  • Social Media as a Potential Intervention Platform: Given that high-frequency posters discuss health and that depression is linked to higher posting rates, the authors propose that patients with depression could be ideal targets for health-related interventions delivered via social media.
  • Patient Self-Prediction is Reliable: The study indicated that patients are relatively accurate predictors of their own social media posting frequency, which is a valuable insight if social media usage becomes a recognized risk factor for certain conditions.

Key Results

  • Posting Frequency Variation: Significant variation in Facebook posting frequency was observed within the patient sample.
  • Demographic Differences: Women posted more than men, younger participants (18-29) posted more than older participants (>49), and individuals identifying as African American had the highest posting means.
  • Depression Impact: Individuals with a positive depression screen or a history of depression posted significantly more (mean 38 posts) compared to those without (mean 22 posts) (P=.001).
  • Other Conditions: No significant posting differences were found for hypertension, diabetes, asthma, headaches, back pain, anemia, or cancer, after controlling for demographics.
  • Language Topics: High-frequency posters discussed health-related topics, while low-frequency posters' language was more associated with social events like birthdays and celebrations.

Limitations and Future Work

  • Sample Demographics: The study's sample was drawn from a single, urban, academic medical ED and was predominantly young and black women. This limits the generalizability of the findings to the broader population.
  • Facebook Specificity: The analysis was limited to Facebook, while social media encompasses a wider range of platforms.
  • "Posting" as a Proxy: Posting frequency is only one measure of social media engagement; passive observation ("lurking") is not captured.
  • Causality: The direction of causality between depression and social media use remains unclear. It's unknown if depression leads to increased posting or if increased posting contributes to depression. Future work should explore this bidirectional relationship.
  • Intervention Development: The authors suggest further research into developing and evaluating social media-based interventions, particularly for individuals with depression.

Summary updated

Abstract

High-frequency posters in our sample were more likely to post about health and to have a diagnosis of depression. The direction of causality between depression and social media use requires further evaluation. Our findings suggest that patients with depression may be appropriate targets for health-related interventions on social media.
BibTeX
@article{smith2017variations,
  author = {Robert J. Smith and Patrick Crutchley and H. Andrew Schwartz and Lyle Ungar and Frances S. Shofer and Kevin Padrez and Raina M. Merchant},
  title = {{Variations in Facebook Posting Patterns Across Validated Patient Health Conditions: A Prospective Cohort Study}},
  journal = {Journal of Medical Internet Research},
  year = {2017},
  doi = {10.2196/jmir.6486},
}
Figure