The Kardashian index of interventional neuroradiologists: measuring discrepant social media influence

Social media has rendered physicians, scientists, scientific conferences and journals with increased accessibility and wider reach, resulting in increased peer and patient engagement of published content.¹,² The publication of the DAWN and DEFUSE landmark trials that changed the paradigm of mechanical thrombectomy led to an increase in conversations on the popular social media platform, Twitter (Twitter Inc., San Francisco, CA, USA), relating to stroke and endovascular therapies.³ Interventional neuroradiologists (INRs) routinely use Twitter to strengthen scientific connections, foster collaborations and often recruit patients for clinical trials.³ Researchers have also been known to share published articles on their personal Twitter profiles in the form of a microblog called ‘Tweetorials’ or ‘Tweetchats’.⁴,⁵ While social media disseminates and makes information more accessible to a wider audience, these ‘tweets’ may increase the number of followers for the physician/scientist without any correlation to the citation count, a metric of scientific influence. Moreover, while about 10% of PubMed articles are tweeted at least once on Twitter, most tweets are posted by people not involved in academia.⁶ A social metric known as the Kardashian index (k-index) was proposed in 2014 to identify scientists with inflated social media following and adjust one’s expectations of the physician/scientist accordingly.⁷ No previous study has been performed to correlate the social media influence (i.e. number of Twitter followers) with scientific influence (i.e. number of citations) for INRs. The purpose of our study was to calculate the k-index of INRs in North America and attempt to correlate their social media and scientific influence.

Institutional review board approval was not required for this study because all data are available in the public domain. The Society for Neurointerventional Surgery (SNIS) membership list was accessed on 18 May.⁸ Google search (Alphabet Inc., Mountain View, CA, USA) was performed to identify Twitter profiles of the members. All INRs who had a Twitter account were included in the study and the number of followers and number of tweets were recorded. Scopus database (Elsevier, Amsterdam, The Netherlands) was used to record the number of citations and h-index.⁹ The age and gender of each INR were obtained from Healthgrades website (Healthgrades Operating Company Inc., Denver, CO, USA).¹⁰ Online curriculum vitae were accessed to obtain their institution name and type (academic or private practice). The k-index was calculated by using the formula previously described by Hall:⁷ k-index = Ft/F, where Ft is the actual number of followers the INR physician has on Twitter, and F is the number of followers the INR is expected to have based on their number of citations (C). The F factor is calculated using the formula: F = 43.3 (C)^0.32. Data were initially recorded in Microsoft Excel 2016 (Microsoft Inc., Redmond, WA, USA) and analyzed using SPSS version 23 (IBM, Armonk, NY, USA). The non-parametric Mann–Whitney U-test was used to study the differences in k-index between different groups. Pearson correlation was used to study the correlation between k-index and age/h-index. Results were considered statistically significant at a P value of 0.05.

A total of 618 names were identified from the SNIS database, of which 142 INRs had a Twitter profile and were included in study, with an average age of 47.6 years (SD 7.9, median 45, range 34–70). One hundred and thirty (91.5%) were men and 12 (8.5%) were women. Stratified by practice type, 112 (78.9%) INRs were at academic centers, while 30 (21.1%) were in private practice settings. The mean Twitter followers were 561.6 (SD 862.7, median 173.5, range 0–4471), and the mean number of tweets was 508.4 (SD 1504.2, median 62.5, range 0–10100). The mean publications, citations and h-index were 78.3 (SD 104.5, median 44.5, range 0–537), 2091.9 (SD 3757.9, median 685.5, range 0–23342) and 16.1 (SD 14.6, median 13, range 0–71). The calculated mean k-index was 1.74 (SD 3.96, median 0.53, range 0–40.24). Figure 1 shows INRs stratified by k-index. There was no statistical difference in k-index between male and female INRs (median 0.53 vs. 0.41, P = 0.27). There was no statistical difference in k-index between academic and private practice INRs (median 0.53 vs. 0.41, P = 0.07). There was a weak, but statistically significant negative correlation between the age and k-index of INRs (r = –0.231, P = 0.013). There was no statistically significant correlation between the k-index and h-index of INRs (P = 0.43).

Figure 1.

Number of interventional neuroradiologists having a Kardashian index (k-index) in different ranges.

Our study highlighted that the majority of INRs who have an active Twitter account are men with a median age of 45 years and are affiliated to academic centers. In a smaller cohort, Dmytriw et al. reported that about 87% of stroke neurointerventionalists who tweeted about endovascular therapies were from academic institutions.¹¹ The median age of INRs is comparable to that of all physicians using Twitter,¹² and although weak, the negative correlation between the age of INRs and k-index could point to an increased popularity of social media among the younger generation of INRs.¹²

Neil Hall in 2014 reported that some scientists with extensive research contributions were under-recognized or unrecognized, whereas scientists with a heavy Twitter presence are widely recognized on social media, some even in areas beyond their field.⁷ In our study, a majority of INRs (103) had a low k-index (<2), suggesting greater scientific influence when compared to social media influence. This is relevant for both physician and patient ‘followers’ alike. For example, the tweet with the most engagement on Twitter might not come from an expert on that topic. The tweet merely receives attention because of the social media influence of the author, or in other words, ‘famous for being famous’ (the Kardashian effect). Furthermore, there has been a dramatic increase in the hashtag ‘#Stroke’ over the past 4–5 years, implying increasing conversations around this topic related to stroke and stroke interventions.³ INRs who gain social capital on Twitter must be aware that their audience consists of patients and other non-physicians who derive information from ‘influential’ Twitter physician-scientists. At the time of our evaluation, the top three INRs with the highest number of citations (23,342, 18,542, 15,030) had 134, 3 and 303 followers, respectively. Interestingly, the INR with the highest number of followers (4471) had 19 citations. Overall, only a few INRs (12) had a k-index of more than 5. Among this small group, only one out of 12 was a woman. Earlier studies have alluded to a lower proportion of women physicians/scientists on Twitter.⁷,¹³ More recently, in the study by Khan et al., three out of 17 female cardiologists (total of 238 studied) had inflated Twitter influence represented by the k-index.¹³ Whether female INRs are less likely than men to self-promote¹³ or whether it is merely a factor of decreased representation of women in interventional neuroradiology,¹⁴ or a combination of both, is unknown.

Perhaps one of the more important results is the lack of statistically significant association between the h-index and k-index among INRs. A common perception is that re-sharing of the research articles on a social media platforms can escalate the number of views and, consequently, the number of citations for the physician/scientist author.¹⁵ Although tweeting an article has been shown to contribute to the increase in Altmetric (alternative social media metrics) scores for an individual article,¹⁶ there are conflicting data on whether Altmetric scores correlate with a significant increase in citation count and indirectly, the h-index.^17–20 Simply having an active social media presence by ‘twitteratis’ does not necessarily equate to more expertise on the topic.

The present study has some important limitations. First, it is possible that some Twitter accounts have not been identified, despite an extensive search by two investigators. Second, citations and the h-index were obtained from the Scopus database which has its own inherent limitations, particularly for physicians/scientists who may have changed their name during their careers. Third, while the citation count is widely considered as a metric of scientific influence, citations also depend on the field of study, size of the field and relevance of the study. Finally, INRs not actively involved in research (such as private practice) may have lower citation counts due to increased emphasis on clinical care.

In conclusion, the majority of INRs with active Twitter accounts have a low k-index (<2) and are predominantly men. There is no statistically significant association between the k-index and h-index, implying that an inflated Twitter presence does not necessarily equate to impactful work.