- Email: [email protected]
Twitter Activity Is Associated With a Higher Research Citation Index for Academic Thoracic Surgeons
Journal Pre-proof Twitter Activity Is Associated with a Higher Research Citation Index for Academic Thoracic Surgeons Michal Coret, BHSc, Matthew Rok, BSc, MSc, Jane Newman, BSc, Deven Deonarain, BSc, John Agzarian, MD, MPH, Christian Finley, MD, MPH, Yaron Shargall, MD, Peter RA. Malik, BSc, Yogita Patel, BSc, Waël C. Hanna, MDCM, MBA PII:
S0003-4975(19)31704-7
DOI:
https://doi.org/10.1016/j.athoracsur.2019.09.075
Reference:
ATS 33219
To appear in:
The Annals of Thoracic Surgery
Received Date: 23 January 2019 Revised Date:
8 September 2019
Accepted Date: 23 September 2019
Please cite this article as: Coret M, Rok M, Newman J, Deonarain D, Agzarian J, Finley C, Shargall Y, Malik PR, Patel Y, Hanna WC, Twitter Activity Is Associated with a Higher Research Citation Index for Academic Thoracic Surgeons, The Annals of Thoracic Surgery (2019), doi: https://doi.org/10.1016/ j.athoracsur.2019.09.075. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 by The Society of Thoracic Surgeons
Twitter Activity Is Associated with a Higher Research Citation Index for Academic Thoracic Surgeons Running Head: Twitter activity is associated with a higher h-index in academic thoracic surgery
Michal Coret (BHSc), Matthew Rok (BSc, MSc), Jane Newman (BSc), Deven Deonarain (BSc), John Agzarian (MD, MPH), Christian Finley (MD, MPH), Yaron Shargall (MD), Peter RA Malik (BSc), Yogita Patel (BSc), Waël C. Hanna (MDCM, MBA)
Division of Thoracic Surgery, McMaster University, Hamilton, ON, Canada
Meeting Presentation: Society of Thoracic Surgeons 55th Annual Meeting, San Diego, California, January 27-29, 2019.
Word Count: 3023
Corresponding Author: Waël C. Hanna, MDCM, MBA McMaster University St. Joseph’s Healthcare Hamilton 50 Charlton Ave E suite T2105, Hamilton, ON L8N 4A6 [email protected]
1
ABSTRACT Background: Academic surgeons are encouraged to promote their work on social media. We hypothesized that thoracic surgeons who are active on Twitter have a higher research citation index (h-index) than their counterparts who are not.
Methods: Thoracic surgeons on CTSNet.org in Canada and the United States were queried for profiles with an h-index on Google Scholar (GS) and/or Research Gate (RG) in July 2018. Surgeons were categorized by whether they possessed a Twitter account (T+) or not (T-), and hindex values were compared. Within the T+ cohort, a multivariate regression model was used to identify independent predictors of increased h-index among variables related to Twitter activity.
Results: Of 3,741 surgeons queried, 19.3% (722) had a known h-index. The mean (SD) h-index for the entire cohort was 14.54 (15.73). The median (range) h-index was 10 (0-121), and the 75th percentile h-index was 20. T+ surgeons had a median (range) h-index of 10 (0-66), and Tsurgeons had a median (range) h-index of 10 (0-72, p=0.25). The 75th percentile h-index for T+ surgeons was 23, compared to 20 for T- surgeons (p=0.24). For T+ surgeons, the regression model identified the number of followers (p=0.029), the number of people followed (p=0.048), and the frequency of tweeting (p=0.046) as independent predictors of a higher h-index.
Conclusions: The median h-index for an academic thoracic surgeon in Canada and the United States is 10. Surgeons who engage in Twitter activity are more likely to have their research cited by others.
2
The dissemination of research findings is increasingly occurring through social media and online networks, such as Twitter, ResearchGate (RG), and Google Scholar (GS). Thoracic surgeons may choose to have a public Twitter account through which they can promote their work, their name, and their research to reach wider audiences.
A common metric to evaluate a researcher’s productivity and influence in a field is their total citations count, which simply sums all the citations attributed to a certain researcher. This metric has been criticized because it can be heavily skewed to reflect one or two highly cited influential papers by the researcher, regardless of their productivity. To address this problem, the American physicist Jorge Hirsch proposed the Hirsch-index (or h-index) in 2005 as an alternative to the total citation count (1). A researcher’s h-index is defined as the number of papers, h, that have at least h citations. Thus, the h-index is reflective of the impact of a particular researcher’s collective work. Consequently, the h-index has become one of the preferred unified metric for evaluating productivity and impact. “High” h-index values vary widely across the different fields of science. For example, in physics, 84% of Nobel prize winners had an h-index > 30 (1). Average h-indices also vary across disciplines ranging from 2.8 (in law), through 3.4 (in political science), 3.7 (in sociology), 6.5 (in geography) and 7.6 (in economics) (2).
The benchmark h-index in general thoracic surgery is unknown. We hypothesized that thoracic surgeons who actively promote their research on Twitter have a higher h-index than surgeons who do not. Through this study, we set out to test this hypothesis and to determine the benchmark h-index value in General Thoracic Surgery.
3
Material and Methods Data Collection The online directory of the Cardiothoracic Surgery Network (www.CTSnet.org), a voluntary database of contact information for cardiothoracic surgeons around the world, was used. The search was restricted to surgeons practicing in General Thoracic surgery in Canada and the United States. In the advanced directory search option of CTSNet.org, “General Thoracic Surgery” was selected in the Subspecialty field drop-down menu, and the desired country (Canada or the United States) in the Country field. All other fields were left blank. The results included any surgeons who registered identified themselves as practising “General Thoracic Surgery” in those two countries.
Google Scholar The first and last names of each of those surgeons were then queried in the profile finder bar in Google Scholar (scholar.google.com). Using the available information within the profile such as interests, past research, publication titles, and current occupation, each profile was validated or rejected as belonging to the surgeon name being queried. Any discrepancies in identity were validated by cross-matching a respective surgeon’s education posted in the GS profile with that provided by CTSNet. As a last option, a visual match would be attempted between the surgeon’s photo on the CTSNet (if present) and the GS profile (if present). Once a GS profile was identified and validated, data was collected from the table on the right side of the screen, from the headings “citations,” “h-index,” and “i10 index”.
Research Gate
4
The first and last names of each surgeon were then queried in the Author Search Bar on Research Gate (RG) (ww.researchgate.net). Author identity was validated using a variety of available profile data. Discrepancies or uncertainties of identity were validated in a similar method to that used to validate GS accounts. Once the profile was identified and validated on RG, data was collected from the “Research items,” “Reads,” and “Citations” numbers on the first page of the profile. Using the “Scores” tab, the RG score and h-index was attained for each participating surgeon.
Scopus Consideration for utilizing Scopus.com as a data source was given. The advantage of Scopus.com over RG and GS is that it automatically creates online profiles for published surgeons, and does not require individuals to create and maintain profiles of their own. As such, the population of published surgeons that could have been studied in this work would have been much larger. However, after close examination of the data from Scopus, it was determined that multiple conflicting online profiles could be present for one individual. This is likely based on the fact that names on publications can be spelled slightly differently, or on the fact that many authors change institutions over their careers, prompting the algorithm to assume that they are two different people. As such, it was elected that Scopus.com could not be used as a source for clean and reliable data for this project.
Twitter The first and last names of each surgeon were finally queried into the Twitter search bar (www.twitter.com) using the subheading “People”. If there were no results, the first and last
5
names were pasted into Google with the keywords “twitter,” “thoracic,” “surgeon,” and “MD” and searched. All Twitter accounts with the same name were checked to verify if any one of them was the corresponding surgeon's account. This was done by first observing if their biography mentioned their position as a thoracic or cardiac surgeon, or simply as a doctor or an “MD.” If not, the content of their tweets/retweets, accounts they follow, and accounts they are being followed by would be assessed for their relevance to the field of thoracic surgery. If this failed, a final visual attempt was made to match the profile photos, if present, on Twitter and on CTSNet. Once the Twitter account was validated as belonging to the surgeon in question, the number of tweets, number of accounts following, number of followers, and number of likes were collected. No distinction was made between tweets and retweets, since these are not reported separately on the Twitter website, and a manual count was not deemed feasible. In the biography, the month and year when the surgeon created their account (joined Twitter) was also noted.
Geography Each surgeon’s location was noted from CTSNet. Geographical divisions were based on longitudinal meridians on the American continent. “East” location was categorized as < 90º. “Central” location was between 90º-110º. Finally, “West” was any region >110º.
Informed Consent Due to the public nature of the information on the queried websites (CTSnet, Twitter, RG and GS), the requirement for informed consent for participation in this study was waived by the research ethics board at our institution.
6
Statistical Analysis The unit of analysis was the individual surgeon. Surgeons who were found on CTSNet.org, but who did not have any profiles on RG or GS were excluded from the analysis because their hindex could not be calculated. Surgeons who had a public profile on RG and/or GS, allowing for a valid h-index to be collected, were included in the analysis. Descriptive statistics such as mean +/- standard deviation, median and range, and percentages, were then generated for the entire cohort. The cohort was then categorized into two subgroups, by whether surgeons possessed a Twitter account (T+) or not (T-). The median h-index values between the T+ and T- subgroups were compared using the Wilcoxon Rank-Sum test and quantile regression equations. Within the T+ cohort, a step-down multivariate linear regression model was used to identify independent predictors of increased h-index among variables including geographical location, time on Twitter, number and frequency of tweets, number of followers, number of people followed, and number of liked posts. Statistical significance was set at a p<0.05. Analyses were undertaken with STATA 14.0 (Statacorp, College Station, TX, USA).
Results A total of the 3,741 General Thoracic Surgeons practising in Canada and the United States were identified on CTSNet.org. The majority of surgeons were located in the United states (95.2%) and 4.8% were located in Canada. Of the Canadian surgeons, 76% were located in the East, 17% in Central Canada, and 7% in the West. In the United States, the majority of surgeons were also located in the East (57%), whereas 17% were located in the Central United States and 27% in the West.
7
Of all the surgeons queried, 375 (10.02%) were found to have a public Twitter account, and 722 (19.30%) were found to have a known h-index value from either an RG or GS profile. Only surgeons with a known h-index value were included in the analysis (Table 1). The majority of surgeons (84.62%) had a RG profile only, whereas a small minority (4.99%) had a GS profile only, and the remainder had active profiles on both (10.39%). For surgeons who had active profiles on both RG and GS, the higher of the two h-index values was used. The mean (SD) hindex for the entire cohort was 14.54 (15.73). The median (range) h-index was 10 (0-121), and the 75th percentile h-index was 20.
Surgeons included in the analysis were then assigned to one of two subgroups, according to whether they had a public active Twitter profile (T+) or whether they did not (T-) (Table 1). The majority of surgeons were in T- subgroup (n= 534/722, 73.96%). Both cohorts were equally likely to have a RG profile with an available h-index. However, T+ surgeons had a significantly higher proportion of GS profiles with an available h-index. Twitter analytics for the T+ cohort are also summarized in Table 1. T+ surgeons had an active Twitter account for a mean (SD) 5.23 (2.88) years, and tweeted a median (range) 1 (0-1,939) tweets/month T+ surgeons had a median (range) h-index of 10 (0-66), and T- surgeons had a median (range) h-index of 10 (0-72, p=0.25). The 75th percentile h-index for T+ surgeons was 23, compared to 20 for T- surgeons (p=0.24).
For T+ surgeons, the multivariate linear regression model identified the number of followers (p=0.029), the number of people followed (p=0.048), and the frequency of tweeting (p=0.046) as independent predictors of a higher h-index. Examination of the dataset reveals that engagement on Twitter is heavily skewed to the to 25th percentile of the T+ cohort. Surgeons in the top 25th
8
percentile had significantly higher times on Twitter, total tweets, tweets per month, likes, followers, and people following. However, the h-index did not differ significantly between the top 25th percentile and the rest of the cohort.
Comment This work demonstrates that the average h-index for an academic general thoracic surgeon in North America (excluding Mexico) is 14, the median h-index is 10, and the 75th percentile is 20. These numbers can be valuable benchmarks to gauge the productivity and impact of a thoracic surgeon. The mere fact of having a Twitter account does not seem to influence the h-index of academic thoracic surgeons, as the median h-index was similar between the T+ and T- subgroups in our study. However, for surgeons who have a Twitter account, it appears that a higher level of engagement and activity is related to a higher h-index. We have shown that the number of followers and people followed, as well as the frequency of tweeting, are positively predictive of a higher h-index, based on the regression analysis. However, when the top 25th percentile of Twitter users is compared to the rest of the T+ cohort, there did not appear to be significant differences in the h-index values. This discordant finding is surprising, but could likely be explained by certain hypotheses. One hypothesis for the lack of detected difference in h-index between the top 25th percentile of Twitter users and the rest of the cohort is that this is related to the small number of users in the top 25th percentile. The fact that the mean h-index difference is very close to significance (p=0.07), suggests that a larger sample size may have in fact detected a difference in h-index between avid Twitter users and occasional ones. An alternative hypothesis for this lack of difference is that there are a multitude of unmeasured confounding variablessuch as age, years in practice, and papers published by association- which could not be
9
controlled for in the regression model. Age for example, is likely a major contributor for this discordance. As the top 25th percentile of Twitter users are expected to be younger surgeons, their h-indexes may reflect their early-career stage, and thus may not be as high as one would expect based on their academic productivity. Conversely, older and more prolific surgeons, who tend to have higher h-indexes based on long careers in science, may not be as Twitter savvy, and are likely located in the bottom 75th percentile of Twitter users. As such, it remains unknown whether the association between the frequency of Twitter usage and higher h-index is causal, or not. A surgeon who is highly engaged on Twitter could potentially bring attention to their work, raising the number of citations, and in consequence raising their h-index. Conversely, it is possible that a surgeon whose work is already widely cited, will have their work discussed on Twitter by their peers, prompting involvement of this surgeon in Twitter exchanges.
Online presence for thoracic surgeon is becoming increasingly important in the digital age. The two major North American societies, the Society of Thoracic Surgeons (STS) and the American Association for Thoracic Surgery (AATS), both have Twitter accounts through which they communicate with their audiences and patients. The STS also holds a regular session on social media at its annual meeting. Surgeons wishing to increase their online presence in a professional manner should consider starting a Twitter account and following those two associations. Academic surgeons who wish to promote their research should consider starting an account on CTSnet.org or Research Gate. Keeping those accounts active and regularly updated will facilitate communication with peers and patients alike. There are multiple other social media platforms that surgeons can use to promote their work, such as Facebook, Instagram, LinkedIn and
10
Doximity. We have chosen to study only one platform, in an effort to maintain data integrity and better understand associations that may be observed.
Moreover, the internet and social media platforms are both global phenomena. Consequently, associations between Twitter activity and h-index may exist in markets other than North America. While, we have chosen to restrict our study to North America, in an effort to collect data that was primarily available in the English language, it is possible that the results of this study may apply to other countries, or even be independent of geographical distributions.
The limitations of this work are inherent to the limitations of the h-index as a measure, and of Twitter as a social media platform. The h-index does not take into account the author order on publications, or whether the research is original scientific work as opposed to a review article or secondary data analysis. As such, the h-index does not reflect the amount of work and involvement a researcher might have had in each of the papers cited. It is also possible that surgeons who are tweeting about topics other than science (politics, sports, etc…), are drawing attention to themselves and their work without directly engaging in scientific activity. The hindex also does not take into account the age, or number of years in practice of the researcher, which would impact how many papers they have had time to publish and promote. Hirsch has developed an alternative citation measure, called the m-index, which adjusts for age, enabling comparison despite different career stages (3). Unfortunately, this number is less popular and therefore hard, if not impossible, to find online for the majority of surgeons. Notwithstanding the disadvantages of the h-index, it has been shown to have more predictive power than total citation counts, citations per paper, and total paper counts (4). Age is also reflective in Twitter usage. The
11
2018 statistics from Pew Research Centre indicate that the distribution of adults using Twitter decreases with age: 40% of Twitter users are between the ages of 18-29, 27% between 30-49, 19% between 50-64, and 8% are over 65 years (5). In contrast, a researcher’s h-index is expected to increase with age, for they would have had more time to publish papers and get citations. As such, surgeons in T+ subgroup are expected to be younger than those in the T- subgroup. Consequently, the fact that the h-indices are similar in the two groups could actually be demonstrative of the effect of Twitter on raising the median h-index for this younger population of surgeons. Unfortunately, this bias could not be adjusted for, because the age of surgeons in our cohort is a data point that could not be collected reliably and consistently.
Conclusion We have shown that the median h-index for an academic thoracic surgeon in Canada and the United States is 10. Having a Twitter account, in and of itself, does not seem to increase a researcher’s h-index. However, for surgeons who have a Twitter account, being engaged in more Twitter activity appears to be associated with a higher h-index.
12
References 1. Hirsch JE. An index to quantify an individual’s scientific research output. Proc Natl Acad Sci U S A. 2005;102(46):16569-16572. 2. The London School of Economics and Political Science. Key measures of academic influence. LSE Impact Blog. Available at http://blogs.lse.ac.uk/impactofsocialsciences/the-handbook/chapter-3-key-measures-ofacademic-influence/. Accessed on January 2, 2019. 3. Hirsch JE, Buela-Casal G. The meaning of the h-index. Int J Clin Health Psychol. 2014;14:161-164. 4. Hirsch JE. Does the h index have predictive power? Proc Natl Acad Sci U S A. 2007;104(49):19193-19198. 5. Pew Research Centre. Social Media Use in 2018. Available at http://www.pewinternet.org/2018/03/01/social-media-use-2018-appendix-a-detailedtable/. Accessed on September 4, 2018.
13
Table 1. Descriptive data on surgeons included in the analysis. Variable
T+ Cohort
T- Cohort
p-value
n=188
N=534
RG Profile
179 (95.2%)
507 (95.1%)
0.96
GS Profile
43 (22.8%)
68 (12.73%)
0.001
Time on Twitter
5.23 (2.88)
N/A
Tweets
41 (0-125,000)
N/A
Tweets per month
1 (0-1,911)
N/A
Followers
60 (0-16,400)
N/A
Following
77 (0-7,939)
N/A
Likes
12 (0- 24,000)
h-index (mean)
15.34 (16.06)
14.25 (15.62)
0.41
h-index (median)
10 (0-66)
10 (0-72)
0.25
75th percentile of h-
23
20
0.24
(years)
index
Proportions are presented as number (percentage). Summary statistics are presented as Mean (Standard Deviation) or Median (range)
14
Table 2. Comparison between the top 25% percentile and the rest of T+ cohort
Top 25th Percentile
Bottom 75th Percentile p-value
n=44
n=144
6.5 (2.55)
4.8 (0.77)
0.002
Tweets
278 (252-125,000)
12 (0-247)
0.001
Followers
491 (49-16,400)
22 (0- 3043)
0.001
Following
279 (35-7,939)
50 (0- 671)
0.001
Likes
606 (0- 24,000)
4 (0- 838)
0.001
h-index (mean)
17.7 (15.18)
14.31 (15.18)
0.07
h-index (median)
9 (0-121)
10 (0-117)
0.4772
Variable
Time on Twitter (years)
Proportions are presented as number (percentage). Summary statistics are presented as Mean (Standard Deviation) or Median (range)
15
S0003-4975(19)31704-7
DOI:
https://doi.org/10.1016/j.athoracsur.2019.09.075
Reference:
ATS 33219
To appear in:
The Annals of Thoracic Surgery
Received Date: 23 January 2019 Revised Date:
8 September 2019
Accepted Date: 23 September 2019
Please cite this article as: Coret M, Rok M, Newman J, Deonarain D, Agzarian J, Finley C, Shargall Y, Malik PR, Patel Y, Hanna WC, Twitter Activity Is Associated with a Higher Research Citation Index for Academic Thoracic Surgeons, The Annals of Thoracic Surgery (2019), doi: https://doi.org/10.1016/ j.athoracsur.2019.09.075. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 by The Society of Thoracic Surgeons
Twitter Activity Is Associated with a Higher Research Citation Index for Academic Thoracic Surgeons Running Head: Twitter activity is associated with a higher h-index in academic thoracic surgery
Michal Coret (BHSc), Matthew Rok (BSc, MSc), Jane Newman (BSc), Deven Deonarain (BSc), John Agzarian (MD, MPH), Christian Finley (MD, MPH), Yaron Shargall (MD), Peter RA Malik (BSc), Yogita Patel (BSc), Waël C. Hanna (MDCM, MBA)
Division of Thoracic Surgery, McMaster University, Hamilton, ON, Canada
Meeting Presentation: Society of Thoracic Surgeons 55th Annual Meeting, San Diego, California, January 27-29, 2019.
Word Count: 3023
Corresponding Author: Waël C. Hanna, MDCM, MBA McMaster University St. Joseph’s Healthcare Hamilton 50 Charlton Ave E suite T2105, Hamilton, ON L8N 4A6 [email protected]
1
ABSTRACT Background: Academic surgeons are encouraged to promote their work on social media. We hypothesized that thoracic surgeons who are active on Twitter have a higher research citation index (h-index) than their counterparts who are not.
Methods: Thoracic surgeons on CTSNet.org in Canada and the United States were queried for profiles with an h-index on Google Scholar (GS) and/or Research Gate (RG) in July 2018. Surgeons were categorized by whether they possessed a Twitter account (T+) or not (T-), and hindex values were compared. Within the T+ cohort, a multivariate regression model was used to identify independent predictors of increased h-index among variables related to Twitter activity.
Results: Of 3,741 surgeons queried, 19.3% (722) had a known h-index. The mean (SD) h-index for the entire cohort was 14.54 (15.73). The median (range) h-index was 10 (0-121), and the 75th percentile h-index was 20. T+ surgeons had a median (range) h-index of 10 (0-66), and Tsurgeons had a median (range) h-index of 10 (0-72, p=0.25). The 75th percentile h-index for T+ surgeons was 23, compared to 20 for T- surgeons (p=0.24). For T+ surgeons, the regression model identified the number of followers (p=0.029), the number of people followed (p=0.048), and the frequency of tweeting (p=0.046) as independent predictors of a higher h-index.
Conclusions: The median h-index for an academic thoracic surgeon in Canada and the United States is 10. Surgeons who engage in Twitter activity are more likely to have their research cited by others.
2
The dissemination of research findings is increasingly occurring through social media and online networks, such as Twitter, ResearchGate (RG), and Google Scholar (GS). Thoracic surgeons may choose to have a public Twitter account through which they can promote their work, their name, and their research to reach wider audiences.
A common metric to evaluate a researcher’s productivity and influence in a field is their total citations count, which simply sums all the citations attributed to a certain researcher. This metric has been criticized because it can be heavily skewed to reflect one or two highly cited influential papers by the researcher, regardless of their productivity. To address this problem, the American physicist Jorge Hirsch proposed the Hirsch-index (or h-index) in 2005 as an alternative to the total citation count (1). A researcher’s h-index is defined as the number of papers, h, that have at least h citations. Thus, the h-index is reflective of the impact of a particular researcher’s collective work. Consequently, the h-index has become one of the preferred unified metric for evaluating productivity and impact. “High” h-index values vary widely across the different fields of science. For example, in physics, 84% of Nobel prize winners had an h-index > 30 (1). Average h-indices also vary across disciplines ranging from 2.8 (in law), through 3.4 (in political science), 3.7 (in sociology), 6.5 (in geography) and 7.6 (in economics) (2).
The benchmark h-index in general thoracic surgery is unknown. We hypothesized that thoracic surgeons who actively promote their research on Twitter have a higher h-index than surgeons who do not. Through this study, we set out to test this hypothesis and to determine the benchmark h-index value in General Thoracic Surgery.
3
Material and Methods Data Collection The online directory of the Cardiothoracic Surgery Network (www.CTSnet.org), a voluntary database of contact information for cardiothoracic surgeons around the world, was used. The search was restricted to surgeons practicing in General Thoracic surgery in Canada and the United States. In the advanced directory search option of CTSNet.org, “General Thoracic Surgery” was selected in the Subspecialty field drop-down menu, and the desired country (Canada or the United States) in the Country field. All other fields were left blank. The results included any surgeons who registered identified themselves as practising “General Thoracic Surgery” in those two countries.
Google Scholar The first and last names of each of those surgeons were then queried in the profile finder bar in Google Scholar (scholar.google.com). Using the available information within the profile such as interests, past research, publication titles, and current occupation, each profile was validated or rejected as belonging to the surgeon name being queried. Any discrepancies in identity were validated by cross-matching a respective surgeon’s education posted in the GS profile with that provided by CTSNet. As a last option, a visual match would be attempted between the surgeon’s photo on the CTSNet (if present) and the GS profile (if present). Once a GS profile was identified and validated, data was collected from the table on the right side of the screen, from the headings “citations,” “h-index,” and “i10 index”.
Research Gate
4
The first and last names of each surgeon were then queried in the Author Search Bar on Research Gate (RG) (ww.researchgate.net). Author identity was validated using a variety of available profile data. Discrepancies or uncertainties of identity were validated in a similar method to that used to validate GS accounts. Once the profile was identified and validated on RG, data was collected from the “Research items,” “Reads,” and “Citations” numbers on the first page of the profile. Using the “Scores” tab, the RG score and h-index was attained for each participating surgeon.
Scopus Consideration for utilizing Scopus.com as a data source was given. The advantage of Scopus.com over RG and GS is that it automatically creates online profiles for published surgeons, and does not require individuals to create and maintain profiles of their own. As such, the population of published surgeons that could have been studied in this work would have been much larger. However, after close examination of the data from Scopus, it was determined that multiple conflicting online profiles could be present for one individual. This is likely based on the fact that names on publications can be spelled slightly differently, or on the fact that many authors change institutions over their careers, prompting the algorithm to assume that they are two different people. As such, it was elected that Scopus.com could not be used as a source for clean and reliable data for this project.
Twitter The first and last names of each surgeon were finally queried into the Twitter search bar (www.twitter.com) using the subheading “People”. If there were no results, the first and last
5
names were pasted into Google with the keywords “twitter,” “thoracic,” “surgeon,” and “MD” and searched. All Twitter accounts with the same name were checked to verify if any one of them was the corresponding surgeon's account. This was done by first observing if their biography mentioned their position as a thoracic or cardiac surgeon, or simply as a doctor or an “MD.” If not, the content of their tweets/retweets, accounts they follow, and accounts they are being followed by would be assessed for their relevance to the field of thoracic surgery. If this failed, a final visual attempt was made to match the profile photos, if present, on Twitter and on CTSNet. Once the Twitter account was validated as belonging to the surgeon in question, the number of tweets, number of accounts following, number of followers, and number of likes were collected. No distinction was made between tweets and retweets, since these are not reported separately on the Twitter website, and a manual count was not deemed feasible. In the biography, the month and year when the surgeon created their account (joined Twitter) was also noted.
Geography Each surgeon’s location was noted from CTSNet. Geographical divisions were based on longitudinal meridians on the American continent. “East” location was categorized as < 90º. “Central” location was between 90º-110º. Finally, “West” was any region >110º.
Informed Consent Due to the public nature of the information on the queried websites (CTSnet, Twitter, RG and GS), the requirement for informed consent for participation in this study was waived by the research ethics board at our institution.
6
Statistical Analysis The unit of analysis was the individual surgeon. Surgeons who were found on CTSNet.org, but who did not have any profiles on RG or GS were excluded from the analysis because their hindex could not be calculated. Surgeons who had a public profile on RG and/or GS, allowing for a valid h-index to be collected, were included in the analysis. Descriptive statistics such as mean +/- standard deviation, median and range, and percentages, were then generated for the entire cohort. The cohort was then categorized into two subgroups, by whether surgeons possessed a Twitter account (T+) or not (T-). The median h-index values between the T+ and T- subgroups were compared using the Wilcoxon Rank-Sum test and quantile regression equations. Within the T+ cohort, a step-down multivariate linear regression model was used to identify independent predictors of increased h-index among variables including geographical location, time on Twitter, number and frequency of tweets, number of followers, number of people followed, and number of liked posts. Statistical significance was set at a p<0.05. Analyses were undertaken with STATA 14.0 (Statacorp, College Station, TX, USA).
Results A total of the 3,741 General Thoracic Surgeons practising in Canada and the United States were identified on CTSNet.org. The majority of surgeons were located in the United states (95.2%) and 4.8% were located in Canada. Of the Canadian surgeons, 76% were located in the East, 17% in Central Canada, and 7% in the West. In the United States, the majority of surgeons were also located in the East (57%), whereas 17% were located in the Central United States and 27% in the West.
7
Of all the surgeons queried, 375 (10.02%) were found to have a public Twitter account, and 722 (19.30%) were found to have a known h-index value from either an RG or GS profile. Only surgeons with a known h-index value were included in the analysis (Table 1). The majority of surgeons (84.62%) had a RG profile only, whereas a small minority (4.99%) had a GS profile only, and the remainder had active profiles on both (10.39%). For surgeons who had active profiles on both RG and GS, the higher of the two h-index values was used. The mean (SD) hindex for the entire cohort was 14.54 (15.73). The median (range) h-index was 10 (0-121), and the 75th percentile h-index was 20.
Surgeons included in the analysis were then assigned to one of two subgroups, according to whether they had a public active Twitter profile (T+) or whether they did not (T-) (Table 1). The majority of surgeons were in T- subgroup (n= 534/722, 73.96%). Both cohorts were equally likely to have a RG profile with an available h-index. However, T+ surgeons had a significantly higher proportion of GS profiles with an available h-index. Twitter analytics for the T+ cohort are also summarized in Table 1. T+ surgeons had an active Twitter account for a mean (SD) 5.23 (2.88) years, and tweeted a median (range) 1 (0-1,939) tweets/month T+ surgeons had a median (range) h-index of 10 (0-66), and T- surgeons had a median (range) h-index of 10 (0-72, p=0.25). The 75th percentile h-index for T+ surgeons was 23, compared to 20 for T- surgeons (p=0.24).
For T+ surgeons, the multivariate linear regression model identified the number of followers (p=0.029), the number of people followed (p=0.048), and the frequency of tweeting (p=0.046) as independent predictors of a higher h-index. Examination of the dataset reveals that engagement on Twitter is heavily skewed to the to 25th percentile of the T+ cohort. Surgeons in the top 25th
8
percentile had significantly higher times on Twitter, total tweets, tweets per month, likes, followers, and people following. However, the h-index did not differ significantly between the top 25th percentile and the rest of the cohort.
Comment This work demonstrates that the average h-index for an academic general thoracic surgeon in North America (excluding Mexico) is 14, the median h-index is 10, and the 75th percentile is 20. These numbers can be valuable benchmarks to gauge the productivity and impact of a thoracic surgeon. The mere fact of having a Twitter account does not seem to influence the h-index of academic thoracic surgeons, as the median h-index was similar between the T+ and T- subgroups in our study. However, for surgeons who have a Twitter account, it appears that a higher level of engagement and activity is related to a higher h-index. We have shown that the number of followers and people followed, as well as the frequency of tweeting, are positively predictive of a higher h-index, based on the regression analysis. However, when the top 25th percentile of Twitter users is compared to the rest of the T+ cohort, there did not appear to be significant differences in the h-index values. This discordant finding is surprising, but could likely be explained by certain hypotheses. One hypothesis for the lack of detected difference in h-index between the top 25th percentile of Twitter users and the rest of the cohort is that this is related to the small number of users in the top 25th percentile. The fact that the mean h-index difference is very close to significance (p=0.07), suggests that a larger sample size may have in fact detected a difference in h-index between avid Twitter users and occasional ones. An alternative hypothesis for this lack of difference is that there are a multitude of unmeasured confounding variablessuch as age, years in practice, and papers published by association- which could not be
9
controlled for in the regression model. Age for example, is likely a major contributor for this discordance. As the top 25th percentile of Twitter users are expected to be younger surgeons, their h-indexes may reflect their early-career stage, and thus may not be as high as one would expect based on their academic productivity. Conversely, older and more prolific surgeons, who tend to have higher h-indexes based on long careers in science, may not be as Twitter savvy, and are likely located in the bottom 75th percentile of Twitter users. As such, it remains unknown whether the association between the frequency of Twitter usage and higher h-index is causal, or not. A surgeon who is highly engaged on Twitter could potentially bring attention to their work, raising the number of citations, and in consequence raising their h-index. Conversely, it is possible that a surgeon whose work is already widely cited, will have their work discussed on Twitter by their peers, prompting involvement of this surgeon in Twitter exchanges.
Online presence for thoracic surgeon is becoming increasingly important in the digital age. The two major North American societies, the Society of Thoracic Surgeons (STS) and the American Association for Thoracic Surgery (AATS), both have Twitter accounts through which they communicate with their audiences and patients. The STS also holds a regular session on social media at its annual meeting. Surgeons wishing to increase their online presence in a professional manner should consider starting a Twitter account and following those two associations. Academic surgeons who wish to promote their research should consider starting an account on CTSnet.org or Research Gate. Keeping those accounts active and regularly updated will facilitate communication with peers and patients alike. There are multiple other social media platforms that surgeons can use to promote their work, such as Facebook, Instagram, LinkedIn and
10
Doximity. We have chosen to study only one platform, in an effort to maintain data integrity and better understand associations that may be observed.
Moreover, the internet and social media platforms are both global phenomena. Consequently, associations between Twitter activity and h-index may exist in markets other than North America. While, we have chosen to restrict our study to North America, in an effort to collect data that was primarily available in the English language, it is possible that the results of this study may apply to other countries, or even be independent of geographical distributions.
The limitations of this work are inherent to the limitations of the h-index as a measure, and of Twitter as a social media platform. The h-index does not take into account the author order on publications, or whether the research is original scientific work as opposed to a review article or secondary data analysis. As such, the h-index does not reflect the amount of work and involvement a researcher might have had in each of the papers cited. It is also possible that surgeons who are tweeting about topics other than science (politics, sports, etc…), are drawing attention to themselves and their work without directly engaging in scientific activity. The hindex also does not take into account the age, or number of years in practice of the researcher, which would impact how many papers they have had time to publish and promote. Hirsch has developed an alternative citation measure, called the m-index, which adjusts for age, enabling comparison despite different career stages (3). Unfortunately, this number is less popular and therefore hard, if not impossible, to find online for the majority of surgeons. Notwithstanding the disadvantages of the h-index, it has been shown to have more predictive power than total citation counts, citations per paper, and total paper counts (4). Age is also reflective in Twitter usage. The
11
2018 statistics from Pew Research Centre indicate that the distribution of adults using Twitter decreases with age: 40% of Twitter users are between the ages of 18-29, 27% between 30-49, 19% between 50-64, and 8% are over 65 years (5). In contrast, a researcher’s h-index is expected to increase with age, for they would have had more time to publish papers and get citations. As such, surgeons in T+ subgroup are expected to be younger than those in the T- subgroup. Consequently, the fact that the h-indices are similar in the two groups could actually be demonstrative of the effect of Twitter on raising the median h-index for this younger population of surgeons. Unfortunately, this bias could not be adjusted for, because the age of surgeons in our cohort is a data point that could not be collected reliably and consistently.
Conclusion We have shown that the median h-index for an academic thoracic surgeon in Canada and the United States is 10. Having a Twitter account, in and of itself, does not seem to increase a researcher’s h-index. However, for surgeons who have a Twitter account, being engaged in more Twitter activity appears to be associated with a higher h-index.
12
References 1. Hirsch JE. An index to quantify an individual’s scientific research output. Proc Natl Acad Sci U S A. 2005;102(46):16569-16572. 2. The London School of Economics and Political Science. Key measures of academic influence. LSE Impact Blog. Available at http://blogs.lse.ac.uk/impactofsocialsciences/the-handbook/chapter-3-key-measures-ofacademic-influence/. Accessed on January 2, 2019. 3. Hirsch JE, Buela-Casal G. The meaning of the h-index. Int J Clin Health Psychol. 2014;14:161-164. 4. Hirsch JE. Does the h index have predictive power? Proc Natl Acad Sci U S A. 2007;104(49):19193-19198. 5. Pew Research Centre. Social Media Use in 2018. Available at http://www.pewinternet.org/2018/03/01/social-media-use-2018-appendix-a-detailedtable/. Accessed on September 4, 2018.
13
Table 1. Descriptive data on surgeons included in the analysis. Variable
T+ Cohort
T- Cohort
p-value
n=188
N=534
RG Profile
179 (95.2%)
507 (95.1%)
0.96
GS Profile
43 (22.8%)
68 (12.73%)
0.001
Time on Twitter
5.23 (2.88)
N/A
Tweets
41 (0-125,000)
N/A
Tweets per month
1 (0-1,911)
N/A
Followers
60 (0-16,400)
N/A
Following
77 (0-7,939)
N/A
Likes
12 (0- 24,000)
h-index (mean)
15.34 (16.06)
14.25 (15.62)
0.41
h-index (median)
10 (0-66)
10 (0-72)
0.25
75th percentile of h-
23
20
0.24
(years)
index
Proportions are presented as number (percentage). Summary statistics are presented as Mean (Standard Deviation) or Median (range)
14
Table 2. Comparison between the top 25% percentile and the rest of T+ cohort
Top 25th Percentile
Bottom 75th Percentile p-value
n=44
n=144
6.5 (2.55)
4.8 (0.77)
0.002
Tweets
278 (252-125,000)
12 (0-247)
0.001
Followers
491 (49-16,400)
22 (0- 3043)
0.001
Following
279 (35-7,939)
50 (0- 671)
0.001
Likes
606 (0- 24,000)
4 (0- 838)
0.001
h-index (mean)
17.7 (15.18)
14.31 (15.18)
0.07
h-index (median)
9 (0-121)
10 (0-117)
0.4772
Variable
Time on Twitter (years)
Proportions are presented as number (percentage). Summary statistics are presented as Mean (Standard Deviation) or Median (range)
15