Gender Differences in Twitter Use and Influence Among Health Policy and Health Services Researchers

Abstract

This study describes gender differences in Twitter use and influence among health services researchers who attended AcademyHealth’s 2018 Annual Research Meeting.

Ample research has documented the lower visibility and success of women compared with men in academic medicine. Against this setting, social media platforms such as Twitter offer academics opportunities to promote their research, network professionally, gain visibility, and, in turn, foster opportunities for career advancement.¹ These opportunities are particularly critical in health policy and health services research, in which dissemination of policy-relevant research and engagement with health care decision-makers impacts academic influence, recognition, and promotion. Herein, we describe gender differences in Twitter use and influence among health services researchers.

Methods

Using publicly available data, we identified names and affiliations of all 6442 speakers and coauthors of research presented at AcademyHealth’s 2018 Annual Research Meeting, the largest academic meeting for health services research. Using this sampling group, we conducted online searches to identify each attendee’s degree(s), title/position, and gender, and identified Twitter users by searching for each individual’s Twitter profile and handle. We included individuals who had an MD, PhD, or equivalent and worked as independent non-trainee–level researchers in the United States. We used Twitter’s Application Program Interface to extract metrics on Twitter use for each individual and all accounts in their social networks, including the most recent 3200 tweets. Review of this study was waived by the University of Pennsylvania’s institutional review board, which determined that informed consent was not applicable to this study because it is based on publicly available data.

We used an established natural-language-processing model to predict the gender of users’ followers and those they followed.² We compared unadjusted average metrics for Twitter use between genders. P values for comparisons of proportions are calculated using χ² tests; all other P values are calculated using a Wilcoxon signed rank test. Tests were 2 sided, and P = .05 was considered statistically significant. We also tested the sensitivity of our results to time on Twitter by standardizing each metric for years on Twitter (eg, dividing each metric by number of years), which did not affect the results and is thus not reported.

Results

Of 3148 health services researchers included in our sample (1668 [53%] women and 1480 [47%] men), approximately one-third used Twitter (Table; n = 919). Across all conference participants, 29.5% of women (n = 492) and 28.9% of men (n = 427) were on Twitter. Among those in faculty positions, 33.3% of women (n = 370) and 32.0% of men (n = 341) were on Twitter.

Table. Twitter Use Among Health Policy and Health Services Researchers, by Gender and by Academic Rank.

Variable	Female	Male	P Valuea
Conference participants, No.	1668	1480
On Twitter, No. (%)b	492 (29.5)	427 (28.9)	.69
Faculty, No. (%)	1112 (66.7)	1065 (72.0)
Faculty who are on Twitter	370 (33.3)	341 (32.0)	.53
Assistant professors (478 women, 320 men)	168 (35.1)	119 (37.2)	.56
Associate professors (299 women, 264 men)	99 (33.1)	75 (28.4)	.23
Professors (355 women, 481 men)	103 (30.8)	147 (30.6)	.96
Engagement on Twitter
Years on Twitter, mean (SD)
All conference participants	4.5 (2.5)	5.1 (2.6)	<.001
Assistant professors	4.5 (2.5)	4.7 (2.7)	.51
Associate professors	4.5 (2.6)	5.4 (2.6)	.02
Professors	4.6 (2.4)	5.2 (2.5)	.04
No. of original tweets per year, mean
All conference participants	70.8 (112.7)	98.1 (208.9)	.06
Assistant professors	70.6 (94.7)	73.7 (103.5)	.88
Associate professors	86.6 (136.2)	146.9 (393.8)	.55
Professors	56.1 (88.0)	92.6 (146.9)	.04
No. following, meanc
All conference participants	332.4 (480.6)	375.3 (655.9)	.68
Assistant professors	363.2 (439.6)	381.1 (477.1)	.86
Associate professors	288.1 (357.2)	364.6 (544.1)	.75
Professors	337.8 (669.0)	377.0 (838.1)	.68
% following who are women, mean (SD)	54.8 (14.6)	42.6 (14.8)	<.001
Influence on Twitter
No. of followers, meand
All conference participants	567.5 (1819.7)	1162.3 (3056.2)	<.001
Assistant professors	402.3 (685.0)	627.5 (1512.0)	.29
Associate professors	629.6 (1689.3)	1117.7 (2796.1)	.03
Professors	985.1 (3396.0)	1770.1 (4341.0)	.02
% followers who are women, mean (SD)	58.0 (14.5)	48.1 (13.8)	<.001
Follower-to-following ratio, mean (SD)
All conference participants	1.7 (3.1)	4.8 (11.9)	<.001
Assistant professors	1.4 (4.0)	2.2 (5.5)	.34
Associate professors	1.9 (2.4)	5.0 (14.3)	.01
Professors	2.4 (2.3)	7.6 (15.9)	<.001
Total No. of likes of original tweets per year, mean (SD)e
All conference participants	315.6 (659.8)	577.6 (1281.8)	.01
Assistant professors	307.2 (702.6)	455.4 (1029.2)	.07
Associate professors	439.7 (859.1)	636.7 (1551.8)	.70
Professors	289.2 (484.3)	691.3 (1401.0)	.03
Total No. of retweets of original tweets per year, mean (SD)e
All conference participants	207.4 (403.6)	399.8 (876.6)	<.001
Assistant professors	181.3 (289.7)	305.1 (634.0)	.008
Associate professors	298.0 (573.7)	464.1 (1146.7)	.58
Professors	189.2 (355.3)	486.2 (981.5)	.02
Average No. of likes per original tweet, mean (SD)e
All conference participants	3.8 (4.8)	4.5 (4.8)	.02
Assistant professors	4.0 (7.1)	4.7 (6.0)	.48
Associate professors	4.1 (2.8)	3.7 (2.3)	.23
Professors	4.0 (3.4)	5.2 (5.3)	.06
Average No. of retweets per original tweet, mean (SD)e
All conference participants	2.4 (2.2)	3.1 (3.4)	.01
Assistant professors	2.2 (2.8)	3.1 (4.2)	.41
Associate professors	2.7 (1.7)	2.6 (1.6)	.56
Professors	2.5 (2.0)	3.6 (3.9)	.01

^aP values for comparisons of proportions are calculated using χ² tests; all other P values are calculated using a Wilcoxon signed rank test.

^bTwitter is an online microblogging platform that allows users to post up to 280 characters at a time (“tweet”).

^cNumber following on Twitter refers to the number of Twitter accounts that each researcher is following. Results are similar when standardizing by these results by the length of time on Twitter.

^dNumber of followers on Twitter refers to the number of Twitter users that follow each researcher. Results are similar when standardizing by these results by the length of time on Twitter.

^eA user’s posts can be “liked” or “retweeted” (reposted) by others. Likes and retweets highlight or share the original tweet with a user’s followers, who in turn may share it with others, helping to promote the content and the tweet’s original writer.

Women had used Twitter for fewer mean (SD) years than men (4.5 [2.5] vs 5.1 [2.6], P < .001) but had a similar mean (SD) number of original tweets per year (70.8 [112.7] vs 98.1 [208.9], P = .06) and followed a similar mean (SD) number of people (332.4 [480.6] vs 375.3 [655.9], P = .68). Women were more likely to follow other women than were men (54.8% of users followed by women were women, whereas 42.6% of users followed by men were women).

Women also had substantially less influence on Twitter than men. Women had half the mean (SD) number of followers as men (567.5 [1819.7] vs 1162.3 [3056.2], P < .001) and were more likely to be followed by women compared with men (58.0% of users following women were women, whereas 48.1% of users following men were women). Women’s tweets generated fewer mean (SD) likes and retweets per year (315.6 [659.8] likes vs 577.6 [1281.8] likes and 207.4 retweets [403.6] vs 399.8 [876.6] retweets) and per tweet (3.8 [4.8] likes vs 4.5 [4.8] likes and 2.4 [2.2] retweets vs 3.1 [3.4] retweets) compared with men’s. Most gender differences were largest among full professors.

Gender differences in influence held across the distribution of number of tweets, with men having more followers than women at every level of activity (Figure).

Figure. Association Between the Number of Followers and the Number of Original Tweets for Women and Men.

Discussion

Twitter is used frequently among health policy and health services researchers. Although it may be an effective way to gain professional visibility and career advancement opportunities, in this study, men had a greater Twitter audience compared with their female peers. These gender differences were observed among both academic and nonacademic researchers and, within academics, were most pronounced among full professors.

Our findings offer some bright spots. Similar rates of Twitter use between genders suggest that social media offers women opportunities for engagement, perhaps with fewer barriers than may be present in day-to-day academic interactions. Moreover, the differences in influence on Twitter were less pronounced among junior researchers, suggesting greater gender parity among younger cohorts.

Our sample is limited to health policy and health services researchers, which is a group for whom Twitter may be particularly important but which may reduce generalizability. Additionally, our measures of Twitter use and influence are not comprehensive, and thus some of the gender differences in these measures may reflect differences in gender norms but not necessarily disparities in influence.

Some have hoped that social media would help level the playing field in academic medicine by giving women an accessible and equitable platform on which to present themselves.¹ However, our findings—that women’s voices on Twitter appeared to be less influential and have less reach than men’s—suggest that these forums may do little to improve gender parity and may instead reinforce disparities.