Social media use and anxiety as measured by the PROMIS Anxiety Short Form

Colditz J.B., Shensa A., Jones J.R., Lin L., Pilkonis P.A., Rollman B.L., & Primack B.A. (May 2015). Social media use and anxiety as measured by the PROMIS Anxiety Short Form. Oral presentation in Philadelphia, PA: Inaugural Conference of the Patient-Reported Outcomes Measurement Information System (PROMIS) Health Organization.


Objective: To determine associations between social media use and Patient-Reported Outcomes Measurement Information System (PROMIS) 4-item Anxiety Short Form scores in a representative sample of US young adults.

Methods: We surveyed a nationally representative sample of US young adults between the ages of 18 to 30. Data collection was assisted by a survey research company that recruited participants via random-digit dialing and address-based sampling frames, representing over 97% of the US population. Participants responded to online surveys over a 1-month period in October/November of 2014. We assessed our independent variable, social media use, in three different ways, including (1) self-reported time spent with online social networks per day, (2) average number of social network site (SNS) visits per week, and (3) responses on a global frequency scale that was adopted from the Pew Internet Research Questionnaire. Our primary dependent variable was self-report anxiety as measured by the 4-item PROMIS short form. We used ordered logistic regression with sample weights to assess associations between SNS use and anxiety while controlling for relevant socio-demographic, personal, and environmental factors. All logistic regression models satisfied the proportional odds assumption.

Results: Our sample of 1,781 participants was 61.8% female, 64.1% Caucasian, 10.0% African American, 16.6% Hispanic and 9.3% Mixed or other race. Our fully adjusted multivariate models included all covariates, including age, sex, race, ethnicity, relationship status, living situation, household income, and education level. T-Scores for anxiety (M=51.43, SD=9.73) were similar to established PROMIS norms, although they demonstrated a strong floor effect, with roughly 1/3 of participants having the lowest possible T-Score of 40.3. Therefore, anxiety T-Scores were converted to tertiles prior to analysis to compensate for the skewed distribution and non-normality (Shapiro-Wilk W=0.99, p<.001). Multivariate analyses demonstrated that, compared to those in the lowest quartile for total SNS use per day (0-30 minutes of use), participants in the highest quartile (2 or more hours of use) had increased odds of having greater anxiety (AOR=1.86, 95% CI=1.28-2.69). In addition, compared with those in the lowest quartiles, those in the highest quartiles of SNS checks per week (AOR=2.49, 95% CI=1.70-3.67) and the global frequency scale (AOR=2.34, 95% CI=1.80-3.03) reported greater anxiety. All associations between independent variables and anxiety demonstrated strong, linear, dose-response trends (p≤.001) that was robust to sensitivity analyses.

Conclusion: We identified a strong, positive association between anxiety and SNS use among a representative sample of US young adults. This association persisted across three different methods of operationalizing SNS use and after adjusting for socio-demographic covariates. This study highlights an important association, given the increasing prevalence of SNS use and the substantial morbidity associated with anxiety. It also suggests that it may be valuable to use SNS to identify individuals at risk for anxiety or mood disorders.