This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
In the recent years, social networking sites (SNSs, also called social media) have been adopted in organ donation campaigns, and recruiting opinion leaders for such campaigns has been found effective in promoting behavioral changes.
The aim of this paper was to focus on the dissemination of organ donation tweets on Weibo, the Chinese equivalent of Twitter, and to examine the opinion leadership in the retweet network of popular organ donation messages using social network analysis. It also aimed to investigate how personal and social attributes contribute to a user’s opinion leadership on the topic of organ donation.
All messages about organ donation posted on Weibo from January 1, 2015 to December 31, 2015 were extracted using Python Web crawler. A retweet network with 505,047 nodes and 545,312 edges of the popular messages (n=206) was constructed and analyzed. The local and global opinion leaderships were measured using network metrics, and the roles of personal attributes, professional knowledge, and social positions in obtaining the opinion leadership were examined using general linear model.
The findings revealed that personal attributes, professional knowledge, and social positions predicted individual’s local opinion leadership in the retweet network of popular organ donation messages. Alternatively, personal attributes and social positions, but not professional knowledge, were significantly associated with global opinion leadership.
The findings of this study indicate that health campaign designers may recruit peer leaders in SNS organ donation promotions to facilitate information sharing among the target audience. Users who are unverified, active, well connected, and experienced with information and communications technology (ICT) will accelerate the sharing of organ donation messages in the global environment. Medical professionals such as organ transplant surgeons who can wield a great amount of influence on their direct connections could also effectively participate in promoting organ donation on social media.
Since the 1960s, many countries have initiated organ donation programs, and at least eighty countries are now known to have a national organ donation program [
Social networking sites (SNSs, also called social media) are a popular platform for promoting organ donation in the United States [
To provide the background for the study, this paper will first review the literature on health communication campaigns and opinion leadership on social media. Following that literature review, this paper’s hypotheses are set forth. Finally, the Methods, Results, and Discussion are presented.
Recruiting opinion leaders has proven to be an effective strategy in Web-based organ donation promotions. For example, Stefanone et al [
The classic two-step flow of communication hypothesis suggests that opinion leaders are individuals who directly receive information from mass media and, in turn, pass on what they know to their everyday associates through interpersonal communication [
Moreover, besides the “two step flow” process [
In terms of the operationalization of opinion leadership, scholars have employed various methods to identify opinion leaders who are able to assist in the implementation of behavior modification efforts [
Closeness is the sum of shortest distances from a member to all other members in the network [
In additional to the indicators of opinion leadership, scholars have also long investigated the factors associated with opinion leadership. In 1957, Katz’s [
Sociability is the first factor related to leadership in offline and online contexts [
On social media, some users’ accounts include a verified badge on their profile that shows the authenticity of their identities as key individuals or organizations. To obtain the verified badge, a verification request is usually submitted by the user and then confirmed by the SNS platform. Previous research has stated that the verification badge indicates a user’s credibility [
In addition to being associated with one’s activeness and identification on the SNS platform, opinion leadership in cyberspace may be subject to certain external environmental factors such as opportunities to access information and communications technologies (ICTs), including the Internet, cell phones, and personal digital assistants [
An individual’s expertise or knowledge about a social issue has long been regarded as a critical contributor to his or her stature as an opinion leader on the topic [
SNS users not only integrate their offline social relationships into a cyber network but also develop new online social ties on social media [
Weibo is a Twitter-like microblogging service site that was launched in 2009 and that has become one of the most popular social media platform in China, with 600 million registered users [
The popular organ donation messages were defined as the messages whose number of retweets ranked in the top three percentiles out of all 6701 messages (n=206). The retweet network of the 206 popular messages was extracted using Python Web crawler in April 2016, resulting in a retweet network with 505,047 unique Weibo users. Next, Python Web crawler was employed to extract the profile information of the Weibo users who received at least one retweet from others in the retweet network (n=44,074). The Python Web crawler extracted all existing accounts as of April 2016, which included 43,510 users. The information about these users’ profiles included the account’s username, verification status, self-introduction, self-reported location and gender, as well as his or her number of followers, followings, and posts on Weibo.
The retweet network of the popular organ donation messages was constructed such that if user
An individual’s opinion leadership was measured via three network metrics from the retweet network, including his or her indegree for local opinion leadership, as well as in-closeness and betweenness for global opinion leadership. The indegree ranged from 1 to 59,061 with a mean of 12.51 (standard deviation [SD] 515.32). The in-closeness ranged from 3.92e-12 to 6.20e-12 with a mean of 3.95e-12 (SD 1.17e-13). The betweenness ranged from 0 to 58,442,335.20 with a mean of 80,485.50 (SD 939,107.11).
The measurement of a user’s activeness on Weibo was adapted from Zhang et al [
If a user had a verified badge in his or her Weibo profile, this account was regarded as a verified account. It was a dichotomous variable:
The correlation matrix among all continuous variables.
Variable | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
1. Indegree | _ | ||||||
2. In-closeness | 0.266a | _ | |||||
3. Betweenness | 0.117a | 0.567a | _ | ||||
4. The number of followers | 0.587a | 0.163a | 0.043a | _ | |||
5. The number of followings | 0.068a | 0.115a | 0.076a | 0.133a | _ | ||
6. Activeness on Weibo | 0.117a | 0.163a | 0.200a | 0.171a | 0.361a | _ | |
7. Location | 0.037a | 0.043a | 0.041a | 0.045a | 0.027a | 0.111a | _ |
a
The user’s location was recoded as a continuous variable according to its degree of ICT development. According to China’s ICT development index [
If a user has a medical-focused profile which lists the user’s professional position in a clinic, hospital, or university, this variable was coded as 1 (n=84).
This information was listed in users’ profiles. The number of followers of all the users ranged from 0 to 50,563,948 with a mean of 38,801.45 (SD 827,423.30). The number of followings ranged from 1 to 5832 with a mean of 543.84 (SD 545.83).
The general linear model (GLM) was employed for analysis. Due to highly skewed distributions, indegree, in-closeness, betweenness, activeness, and the numbers of followers and followings were square root transformed to meet the assumptions of GLM. The correlation matrix among continuous variables is presented in
The retweet network is a connected network including 505,047 nodes and 545,312 edges. The length of the maximum distance between nodes (ie, diameter) in this network is 21. The network has a low density: only 0.0002% of possible edges between all the nodes are connected. The reciprocity values and clustering coefficient of this network are extremely low at 0.0016 and 0.000003, respectively. The low values of density, reciprocity, and clustering coefficient indicate a sparse network. In addition, indegree centralization is 0.1169, and outdegree centralization is 0.00017 for this network. This indicates that, in terms of indegree, links are retweeted disproportionately to a small group of users. The distribution of indegree within this retweet network is highly skewed (see
The network-level characteristics of the retweeting network.
Social network metric | Definition | Possible range | Value |
Size | The number of nodes (eg, users) in the network | N/Aa | 505,047 |
Diameter | The largest geodesic distance, which is the shortest distance from one node to another in the network | N/A | 21 |
Density | The proportion of all possible dyadic connections that are presented in the network | 0-1 | 0.000002 |
Reciprocity | The proportion of all pairs in the network that have a reciprocated tie between them | 0-1 | 0.0016 |
Clustering coefficient | The degree to which nodes in the network tend to cluster together | 0-1 | 0.000003 |
Indegree centralization | The extent to which the distribution of indegree centrality in the network deviates from a perfectly equal distribution | 0-1 | 0.1169 |
Outdegree centralization | The extent to what the distribution of outdegree centrality in the network deviates from a perfectly equal distribution | 0-1 | 0.00017 |
aN/A: not applicable.
The log transformed distribution of indegree.
Visualization of the sharing network of popular organ donation messages. The figure includes only nodes whose indegree equal or larger than 50. Nodes represent Weibo users (n=362). The size of node depends on its indegree. The larger the node, the greater amount of retweets the user received. Lines represent retweet relationship between Weibo users. The weight of line depends on the number of retweets.
The GLM effects of predictors on individuals’ opinion leadership.
Predictors | Local opinion leadership | Global opinion leadership | |||||
Indegree | In-closeness | Betweenness | |||||
Coefficients | Standard error | Coefficients | Standard error | Coefficients | Standard error | ||
Intercept | 1.57a | 0.31 | 0.000002a | 0.000000003 | −53.76 | 29.67 | |
Activeness on Weibo | 0.001a | 0.0003 | 0.00000000006a | 0.000000000003 | 0.89a | 0.03 | |
Verification | −0.33a | 0.05 | −0.0000000007 | 0.0000000005 | −17.36a | 4.98 | |
Location | 0.02 | 0.01 | 0.0000000005a | 0.0000000001 | 3.72a | 1.05 | |
Medical-focused | 1.14a | 0.30 | 0.000000001 | 0.000000003 | −8.02 | 29.35 | |
The number of followers | 0.01a | 0.00008 | 0.00000000002a | 0.0000000000008 | 0.02b | 0.01 | |
The number of followings | −0.003 | 0.002 | 0.0000000002a | 0.00000000002 | 0.34c | 0.16 | |
∆ |
34.4% | 5.2% | 4.2% | ||||
Gender (female=1, male=0) | −0.14a | 0.03 | −0.000000002a | 0.0000000003 | 7.60c | 3.16 | |
∆ |
0.0% | 0.0% | 0.0% | ||||
Total |
34.4% | 5.2% | 4.2% |
a
b
c
The GLM results are reported in
H1 predicted a positive effect of users’ activeness on Weibo on (1) local and (2) global opinion leadership on the organ donation topic. The analysis revealed that the number of messages one posted on Weibo was significantly and positively associated with one’s local opinion leadership,
H2 proposed that, compared with an unverified user, a verified user exhibits more (1) local and (2) global opinion leadership within the retweet network about organ donation. However, the results showed an opposite direction of effect. The unverified users exhibited significantly more local opinion leadership than verified users within the retweet network,
H3 predicted positive effects of ICT development level in users’ location on his or her (1) local and (2) global opinion leadership in the retweet network of organ donation message. The results showed that level of ICT development was not significantly associated with local opinion leadership. However, it was positively associated with two global opinion leadership indicators: in-closeness,
H4 made predictions about the effects of professional, medical knowledge on (1) local and (2) global opinion leadership in the retweet network of organ donation messages. The results showed that medical-focused users significantly exhibited more local opinion leadership than other users,
H5 and H6 considered the effects of social position on Weibo on one’s opinion leadership on the organ donation topic. H5 predicted a positive effect of the number of followers on (1) local and (2) global opinion leadership. The results showed that users with a higher number of followers were more likely to exhibit more local as well as global opinion leadership in retweet network about organ donation:
This study investigates organ donation information on Weibo by mapping its sharing (ie, retweet) network and examining the local as well as global opinion leadership in the network. This work explores the role of personal attributes, professional knowledge, and social position in obtaining influence according to Katz’s [
The sharing network of popular organ donation messages on Weibo is extremely sparse and centralized, resembling a star-like network structure. Only a very small portion of users in this network receives retweets from others, whereas more than 90% of users do not receive any retweets from others and occupy peripheral positions in the network. This result indicates that few central users control the flow of organ donation information and could act as critical peer leaders in organ donation promotions on Weibo. After mapping the network, subsequent analysis explores how individual and social factors affect these users’ ability to influence the information flow (ie, opinion leadership). In addition, the opinion leadership on social media is conceptualized as a two-dimensional construct, including a direct influence in neighborhood (ie, local opinion leadership), as well as an indirect impact in the whole environment (ie, global opinion leadership).
The findings show that two personal attributes are significant predictors of both local and global opinion leadership on organ donation: activeness on Weibo and verification status. In detail, compared with inactive users, active users are more likely to show greater local and global opinion leadership in the organ donation information diffusion network on Weibo. This finding is consistent with theories on developing influence [
Nevertheless, the other personal attribute, verification status, negatively impacts opinion leadership, which is the opposite of H2’s prediction. This study found that, compared with verified users, unverified users are more likely to show greater local as well as global opinion leadership about organ donation on Weibo. One possible explanation is that a user’s influence on social media is topic-sensitive [
The other possible explanation could be that, in general, verified users may enjoy less rather than more credibility than unverified ones. Indeed, the Chinese government has hired a large number of people to fabricate posts on popular websites and social media, and the number of pseudonymous and deceptive social media posts could reach 488 million a year [
The third personal attribute examined in this study is the level of ICT development in one’s location. Unlike the abovementioned two attributes, which are relevant to a user’s activities and identity on Weibo, this one is an environmental factor. The results show that a person’s direct impact on the neighborhood (ie, local opinion leadership) is highly associated with his or her characteristics and identity on Weibo but not with the ICT development level in his or her area. A user’s indirect influence on other users (ie, global opinion leadership) depends on that user’s characteristics on Weibo as well as this environmental factor. Indeed, a previous study found that ICT development was highly associated with users’ influence in friendship networks on social media [
The second type of opinion leadership predictor suggested by Katz [
The last predictor of opinion leadership included in this study is a user’s social position on Weibo. The results reveal that compared with obtaining local opinion leadership, securing global influence requires a well-connected social location in the network. For local opinion leadership, the number of followers, but not followings, is a significant predictor. The number of a user’s followers is the number of users on Weibo who will be directly exposed to his or her posts (ie, the user’s direct audience). As documented in these results, the larger the size of a user’s direct audience, the greater level of that user’s local influence. On the other hand, global opinion leadership depends on not only the size of one’s direct audience but also the size of the user’s information sources (ie, followings) on Weibo. Users with large audiences and many information sources occupy well-connected positions in the network and have updated information on the topic, thus exhibiting more global opinion leadership than others who occupy peripheral social positions. For public health professionals, they may recruit peer leaders according to campaign objectives. A user with a large direct audience will be competent to impact his or her neighbors, but only users with a large audience and many information sources will be capable of controlling the dissemination of organ donation information on Weibo.
Although previous organ donation campaigns have employed social media, the campaign advertisements and strategies were specifically designed for college students, and the information dissemination was mainly controlled by the researchers [
There are several areas worthy of further research in opinion leadership in the topic of organ donation on social media. First, examining the retweet paths of all the organ donation messages (n=6701) would yield an extremely large dataset and be computationally intensive, so this study focuses on only the most popular messages. In fact, this study initiates an exploration of opinion leadership of organ donation promotion on social media with an innovative and advanced method. Future research may replicate this research on other SNS platforms or with a larger dataset. Second, this study analyzes a snapshot of the retweet network instead of a dynamic diffusion network that evolves over time. Subsequent work may employ more sophisticated data mining and data analyzing techniques to detect how organ donation messages go viral on social media and who facilitates the dissemination, which would offer valuable information for future SNS organ donation promotion. Third, apart from medical knowledge, some other factors such as experience with organ donation may contribute to a user’s opinion leadership on the topic of organ donation. Future studies could explore other measures or indicators of users’ competence on the topic of organ donation. Fourth, this study examines the retweet network of all organ donation tweets regardless of their content. However, the structure of the retweet network may vary by how organ donation is covered or framed in the tweets. Future research should investigate whether content shapes the retweet paths and opinion leadership. For example, are myths about organ donation disseminated the same way as stories about an organ recipient? Findings from this research will greatly enhance the design and implementation of organ donation campaigns using social media.
general linear model
information and communications technology
standard deviation
social networking site
The authors thank Dr Tai-Quan Peng and Mr Xiaohui Wang for their helpful comments on an earlier draft of this paper.
None declared.