P1571 Echocardiograms on Twitter: the good, the bad and the ugly

N Irenji,C Mcaloon
DOI: https://doi.org/10.1093/ehjci/jez319.991
2020-01-01
Abstract:Abstract Background - Case reports are a well-established part of the medical literature presenting interesting cases which require patient consent and peer review as a prerequisite. The advent of social media platforms has changed the way knowledge is shared. Twitter is a micro-blogging site utilised by the medical profession including cardiologists and allied health care professionals (HCP), allowing sharing of interesting cases. Echocardiography is particularly complementary to the medium as images and short films can be included in Tweets. However, its immediacy means traditional safeguards are absent, meaning patients might be identifiable. The study aim was to sample the prevalence of echocardiography cases being shared on Twitter and analyse the degree of identifiable information present. Methods - A retrospective observational study of videos of echocardiograms uploaded to Twitter under the hashtag #echofirst over 4 months (Nov 18-Feb 19) was performed in May 2019. Only English language Tweets were included. All Tweeter Demographics and Followers were manually collected by reviewing biographies. Tweets were screened for ‘direct’ (name, address, date-of-birth) and ‘indirect’ (age, gender, ethnicity, diagnosis, clinical diagnosis, date study performed) patient identifiers. Numbers of interactions (replies, likes, retweets, views) were recorded at the time of viewing tweets. Results - 324 of screened tweets (120 unique users; 114 individuals, 6 organisations) met study criteria. 81 (67%) stated their city. 35 (29%) stated their hospital. One Tweet contained directly identifiable information. Patient consent was not mentioned in any Tweet. Table 1 demonstrates Tweets with the number of identifiers cross-referenced with Tweeters" location. Discussion - Social media is a highly accessible method for cardiologists and allied HCP to instantly share and discuss echocardiography cases. However, our analysis potentially demonstrates that specific patients could be identified based location and identifiers. Caution must be applied when posting cases as Tweets. Specific guidelines on this practice should be considered. Table 1 Identifiers (n) Worldwide (n = 324) Country (n = 284) City (n = 221) Hospital (n = 112) 0 98 (30%) 88 (31%) 64 (29%) 27 (24%) 1 114 (35%) 97 (34%) 77 (35%) 39 (35%) 2 50 (15%) 45 (16%) 33 (15%) 19 (17%) 3 24 (7%) 20 (7%) 15 (7%) 5 (4%) 4 36 (11%) 32 (11%) 31 (14%) 21 (19%) 5 2 (1%) 2 (1%) 1 (<1%) 1 (1%)
What problem does this paper attempt to address?