Psychiatry on Twitter : Content Analysis of the Use of Psychiatric Terms in French

Psychiatry on Twitter : Content Analysis of the Use of Psychiatric Terms in French

Auteurs : S. Dela­nys, F. Bena­ma­ra, V. Mori­ceau, F. Oli­vier, J. Mothe

Réfé­rence : Jour­nal of Medi­cal Inter­net Research (v.6(2) ; 2022 Feb.)

Abstract

Back­ground
With the advent of digi­tal tech­no­lo­gy and spe­ci­fi­cal­ly user-gene­ra­ted contents in social media, new ways emer­ged for stu­dying pos­sible stig­ma of people in rela­tion with men­tal health. Seve­ral pieces of work stu­died the dis­course conveyed about psy­chia­tric patho­lo­gies on Twit­ter consi­de­ring most­ly tweets in English and a limi­ted num­ber of psy­chia­tric disor­ders terms. This paper pro­poses the first stu­dy to ana­lyze the use of a wide range of psy­chia­tric terms in tweets in French.

Objec­tive
Our aim is to stu­dy how gene­ric, noso­gra­phic, and the­ra­peu­tic psy­chia­tric terms are used on Twit­ter in French. More spe­ci­fi­cal­ly, our stu­dy has 3 com­ple­men­ta­ry goals : (1) to ana­lyze the types of psy­chia­tric word use (medi­cal, misuse, or irre­le­vant), (2) to ana­lyze the pola­ri­ty conveyed in the tweets that use these terms (posi­tive, nega­tive, or neu­ral), and (3) to com­pare the fre­quen­cy of these terms to those obser­ved in rela­ted work (main­ly in English).

Methods
Our stu­dy was conduc­ted on a cor­pus of tweets in French pos­ted from Janua­ry 1, 2016, to Decem­ber 31, 2018, and col­lec­ted using dedi­ca­ted key­words. The cor­pus was manual­ly anno­ta­ted by cli­ni­cal psy­chia­trists fol­lo­wing a mul­ti­layer anno­ta­tion scheme that includes the type of word use and the opi­nion orien­ta­tion of the tweet. A qua­li­ta­tive ana­ly­sis was per­for­med to mea­sure the relia­bi­li­ty of the pro­du­ced manual anno­ta­tion, and then a quan­ti­ta­tive ana­ly­sis was per­for­med consi­de­ring main­ly term fre­quen­cy in each layer and explo­ring the inter­ac­tions bet­ween them.

Results
One of the first results is a resource as an anno­ta­ted data­set. The ini­tial data­set is com­po­sed of 22,579 tweets in French contai­ning at least one of the selec­ted psy­chia­tric terms. From this set, experts in psy­chia­try ran­dom­ly anno­ta­ted 3040 tweets that cor­res­pon­ded to the resource resul­ting from our work. The second result is the ana­ly­sis of the anno­ta­tions sho­wing that terms are misu­sed in 45.33% (1378/3040) of the tweets and that their asso­cia­ted pola­ri­ty is nega­tive in 86.21% (1188/1378) of the cases. When consi­de­ring the 3 types of term use, 52.14% (1585/3040) of the tweets are asso­cia­ted with a nega­tive pola­ri­ty. Misu­sed terms rela­ted to psy­cho­tic disor­ders (721/1300, 55.46%) were more frequent to those rela­ted to depres­sion (15/280, 5.4%).

Conclu­sions
Some psy­chia­tric terms are misu­sed in the cor­po­ra we stu­died, which is consistent with the results repor­ted in rela­ted work in other lan­guages. Thanks to the great diver­si­ty of stu­died terms, this work high­ligh­ted a dis­pa­ri­ty in the repre­sen­ta­tions and ways of using psy­chia­tric terms. Moreo­ver, our stu­dy is impor­tant to help psy­chia­trists to be aware of the term use in new com­mu­ni­ca­tion media such as social net­works that are wide­ly used. This stu­dy has the huge advan­tage to be repro­du­cible thanks to the fra­me­work and gui­de­lines we pro­du­ced so that the stu­dy could be rene­wed in order to ana­lyze the evo­lu­tion of term usage. While the new­ly build data­set is a valuable resource for other ana­ly­ti­cal stu­dies, it could also serve to train machine lear­ning algo­rithms to auto­ma­ti­cal­ly iden­ti­fy stig­ma in social media.