JMIR Publications

ABSTRACT

Background:

Stroke is a worldwide leading cause of long term disabilities. Women experience more activity limitations, worse health-related quality of life and more post-stroke depression than men. Twitter is increasingly used by individuals to broadcast their day-to-day happenings providing unobtrusive access to samples of spontaneously expressed opinions on all types of topics and emotions

Objective:

1) considering the raw frequencies of words in the collection of tweets posted by a stroke survivors’ sample, compare them by gender for the following 8 basic emotions: anger, fear, anticipation, surprise, joy, sadness, trust and disgust. 2) after determining the proportion of each emotion per tweet compare each of them by gender. 3) determine whether any gender significantly uses more or fewer words of a particular emotion, first by identifying over-frequented configurations i.e. pairs that are observed more often than expected (types) and second identifying under-frequented configurations i.e. pairs that are observed less often than expected (antitypes). 4) extract the main topics (represented as sets of words) that occur in the collection of tweets, related to each gender and identify when two topics are likely to co-occur within tweets.

Methods:

Sentiment analysis with state-of-the-art Lexicon with syuzhet R package. The emotion scores for men and women were first subjected to an F-test and then subjected to a Wilcoxon rank sum test. Configural frequency analysis (CFA) with hcfa R package for each combination. Daily hedonometer average scores for all tweets during the period under study. Exploratory clustering with VosViewer software. Structural topic modelling (STM) with stm R package.

Results:

We analyzed 800,424 tweets posted during 01 August 2007 – 01-December 2018, by 479 stroke survivors: women (n=244) posted 396,898 tweets and men (n=235) posted 403,526 tweets. All 479 participants were manually verified in their stroke survivor condition and gender, and with membership in at least 3 stroke specific Twitter lists as active users, their total number of tweets since 2007 is 5,257,433 therefore we analyzed the most recent 15.2% of all their tweets. All positive emotions (anticipation, trust, surprise and joy) are significantly higher (P<.001) in women meanwhile all negative emotions (disgust, anger, fear and sadness) are significantly higher (P<.001) in men in all three analysis we performed: raw frequencies, proportion of emotions and CFA. Also when considering global positive-negative emotions. Similarly, hedonometer mean values all along the considered period show higher levels of happiness in women. Top 20 topics (with percentages and confidence intervals) more likely addressed by gender, are finally presented

Conclusions:

In the selected sample of stroke survivors, positive emotions are much more expressed by women and negative emotions by men in Twitter, in spite of the generally reported worse outcomes, including depression, of women after stroke.