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.
Social networks such as Twitter offer the clinical research community a novel opportunity for engaging potential study participants based on user activity data. However, the availability of public social media data has led to new ethical challenges about respecting user privacy and the appropriateness of monitoring social media for clinical trial recruitment. Researchers have voiced the need for involving users’ perspectives in the development of ethical norms and regulations.
This study examined the attitudes and level of concern among Twitter users and nonusers about using Twitter for monitoring social media users and their conversations to recruit potential clinical trial participants.
We used two online methods for recruiting study participants: the open survey was (1) advertised on Twitter between May 23 and June 8, 2017, and (2) deployed on TurkPrime, a crowdsourcing data acquisition platform, between May 23 and June 8, 2017. Eligible participants were adults, 18 years of age or older, who lived in the United States. People with and without Twitter accounts were included in the study.
While nearly half the respondents—on Twitter (94/603, 15.6%) and on TurkPrime (509/603, 84.4%)—indicated agreement that social media monitoring constitutes a form of eavesdropping that invades their privacy, over one-third disagreed and nearly 1 in 5 had no opinion. A chi-square test revealed a positive relationship between respondents’ general privacy concern and their average concern about Internet research (
The data and findings from this study contribute to the critical dialogue with the public about the use of social media in clinical research. The findings suggest that most users do not think that monitoring Twitter for clinical trial recruitment constitutes inappropriate surveillance or a violation of privacy. However, researchers should remain mindful that some participants might find social media monitoring problematic when connected with certain conditions or health topics. Further research should isolate factors that influence the level of concern among social media users across platforms and populations and inform the development of more clear and consistent guidelines.
The success of clinical trials depends on the enrollment of study participants, also referred to as research participant recruitment. Recruitment involves attracting and selecting suitable study participants. It can be conducted through different communication channels (eg, newspapers, radio, television, posters, brochures, email, and social media). Without their involvement, medical and scientific progress that benefits patients would be impossible [
In the context of the Internet and social media, user privacy is commonly considered a process of boundary management where individuals regulate disclosures in their social relationships through adjustments to the transmission and sharing of personal information online. In her theory of communication privacy management, Petronio argues that individuals are regularly engaged in decisions about disclosing or concealing private information within any given context [
Nearly one-quarter of American adults (22%) use Twitter [
To test the feasibility of Twitter monitoring for recruiting clinical trial participants, we decided to develop a use case for a multisite cancer study on acute myeloid leukemia (AML) with patients in remission. These patients present a uniquely challenging population to recruit for clinical studies. AML, when active, typically leads to severe symptoms and hospitalization. Hospitalized patients are more accessible to screen, identify, and recruit for clinical trials. Once AML patients have completed their consolidation chemotherapy, they only visit their doctor every 3-4 months. The clinical trial we chose for this case study was designed to recruit patients in the first 3 months after they complete their consolidation chemotherapy, precisely the time when these patients have only sporadic contact with the health care system. Traditional techniques employed during routine patient contact would not be possible for this population. Furthermore, since postremission maintenance therapy is not a routine part of clinical practice for AML, we were unlikely to receive referrals from community physicians for this clinical trial. Therefore, we sought to examine the feasibility of a social media monitoring-enabled solution.
However, in their review of the study protocol, the Central Institutional Review Board (CIRB) of the National Institute of Cancer raised concerns about the potential breach of privacy using monitoring techniques on Twitter. The CIRB committee noted the following:
Those who openly share their information via social media platforms may still have an expectation of privacy and/or be unaware of the platform’s privacy policies. To contact people after utilizing the approach of “active listening” may be perceived by some potential participants as eavesdropping on their conversations about their health... This may produce distrust and potential participants may interpret this as an invasion of their privacy even though social media is understood by many to be a public sphere. Privacy risks specific to [a Twitter user’s] diagnosis may be increased by taking part in the study. The study team, by echoing the information about an individual’s diagnosis, may amplify this information, so it’s more likely to come to the attention of the public or an employer.
We used this feedback as guidance and motivation for designing the following research study to ascertain people's attitudes and level of concern about the use of social media monitoring on Twitter for targeted clinical trial recruitment.
Scientists have pointed out a lack of the inclusion of public views to inform future practices in social media research and social media-enabled recruitment [
This study tested four primary and three additional hypotheses related to potential privacy concerns with the use of Twitter monitoring for clinical trial recruitment (see
Our results are based on the views of the public and they support the formulation of evidence-based guidelines to assist researchers and Institutional Review Board (IRB) professionals using social media in clinical research recruitment. The data contribute to the critical dialogue with the public to understand the ethical issues involved in social media-enabled research and recruitment as well as the procedural solutions that are required to protect the rights and safety of research participants.
Hypothesis 1: People perceive social media monitoring on Twitter for clinical trial recruitment as eavesdropping on their conversations about their health and as an invasion of their privacy.
Hypothesis 2: Twitter users’ expectations of privacy relate to their level of concern about the use of social media monitoring for clinical trial recruitment.
Hypothesis 3: General literacy and knowledge about the Twitter platform are associated with the level of concern about the use of social media monitoring on Twitter for clinical trial recruitment.
Hypothesis 4: People’s concerns over Twitter monitoring for clinical trial recruitment are similar to more traditional, offline scenarios (eg, discretely being approached in person as the patient leaves a medical facility).
Hypothesis 5: The type of information monitored to identify and recruit individuals for clinical trials is associated with the level of concern over the use of social media monitoring on Twitter for clinical trial recruitment.
Hypothesis 6: The type of disease recruited for is associated with the level of concern over the use of social media monitoring on Twitter for clinical trial recruitment.
Hypothesis 7: The type of entity performing the monitoring is associated with the level of concern over the use of social media monitoring on Twitter for clinical trial recruitment.
We developed an open 39-item survey (see
Using the survey, we collected the following types of information: previous use and knowledge of Twitter, general concern about Internet privacy, specific concerns about privacy related to the monitoring of Twitter activity for clinical trial recruitment, and demographic data. Clinical trials were defined for respondents in accordance with the National Institutes of Health definition for nonspecialist audiences [
The goal of clinical trials is to determine if a new drug, device, or procedure works and is safe, or they can look at other aspects of care, such as improving the quality of life for people with chronic illnesses. People participate in clinical trials for a variety of reasons, for example, to help others [and] to contribute to moving science forward.
Finally, we used a set of vignettes to assess the association between the level of concern and different variables, such as the disease or health topic of the clinical trial and the entity that monitors social media user activity on Twitter.
Eligible participants were adults, 18 years of age or older, who lived in the United States. People with and without Twitter accounts were included in the study.
We used two online methods for recruiting study participants, who made up our convenience sample: the open survey was (1) advertised on Twitter between May 23 and June 8, 2017, and (2) deployed on TurkPrime, a crowdsourcing data acquisition platform, between May 23 and June 8, 2017 [
The Twitter ads appeared as promoted tweets in users’ Twitter feeds. Twitter ads provide a number of targeting options for reaching a specific target audience. Targeting features used for ads in this study included (1)
The second sample used in this study was recruited through TurkPrime [
TurkPrime... uses transport layer security encryption (also known as HTTPS) for all transmitted data. All data access is blocked except for explicitly whitelisted IP addresses, in addition to being secured with user passwords. Furthermore, [the] data, including Access Key ID and Secret Access Key, are encrypted with AES-256 encryption, the standard adopted by the National Institute of Standards and Technology.
Study data were collected and managed using Research Electronic Data Capture (REDCap), an electronic data capture tool, hosted at the University of Southern California. REDCap is a secure, Web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry, (2) audit trails for tracking data manipulation and export procedures, (3) automated export procedures for seamless data downloads to common statistical packages, and (4) procedures for importing data from external sources [
The paid ads posted on Twitter included a link to the survey hosted on REDCap. Respondents filled out a multipage survey online on either a mobile device or desktop. On TurkPrime, each respondent was provided with a unique link to a separate survey hosted on REDCap. The datasets used for analysis were generated directly from REDCap using the platform’s reporting tools. Please see the Twitter and TurkPrime recruitment sections for further details.
A total of 603 participants completed the survey and passed the attention-check questions in this study: 94 (15.6%) on Twitter and 509 (84.4%) on TurkPrime. Among the initial 704 respondents on Twitter alone, we used Excel filters to identify and remove 70 respondents (9.9%) who did not show correct completion of the attention-check questions and 540 respondents (76.7%) who gave fraudulent responses with unique characteristics. Regarding the fraudulent responses, all of them (1) showed the same age (ie, 22 years old); (2) were submitted about 5-10 minutes apart from each other over a period of 5 days; (3) used email addresses with a consistent pattern, namely, a random English word followed by three to six random letters (eg, upgradeyhujer@gmail.com and imageiunmed@gmail.com); and (4) were confirmed to be fraudulent when respondents were asked to verify the information provided through the survey about their first name, age, and highest degree or level of school they had completed. For each filtered entry, we manually reviewed the email address to identify fraudulent emails (ie, email addresses that included a random English word followed by three to six random letter patterns). Finally, we manually sent a message to each email address and asked the users to verify the information they provided in their survey responses. Among the initial 738 responses on TurkPrime, we removed 229 responses (31.0%) that did not show correct completion of the attention-check questions.
We did not use any methods to adjust the sample, such as weighting of items or propensity scores. We analyzed the data on two levels: (1) at the respondent level to test control variables (individual factors: level 2) and (2) at the vignette level to test independent variables (contextual factors: level 1). Survey responses were first analyzed through descriptive statistical methods to assess the distribution of participants across our dependent and independent variables, such as the degree of privacy concern and demographic factors. Next, data regarding the different levels of concern for each vignette were further analyzed using pivot tables to identify any relationships between the levels of general privacy concern and the participants’ attitudes regarding the vignettes. We defined a high level of concern as responses that indicated
The study was reviewed and approved by the IRB at the University of Southern California (HS-17-00348).
Overall, the sample of 603 participants showed the following distribution (see
We further assessed Twitter usage among the 603 survey participants (see
We attempted to assess the level of Twitter literacy and knowledge among study participants (see
We sought to learn more about general privacy concerns associated with the use of the Internet (see
We also assessed respondents’ concerns about Internet research activities, which pertain to the use of their Twitter data for research purposes (see
Hypothesis 1 states that social media monitoring on Twitter for clinical trial recruitment is perceived as eavesdropping and an invasion of privacy.
To gauge respondents’ overall perception of Twitter monitoring for clinical trial recruitment, we tested the language as stated by the CIRB that
We isolated responses for only those respondents (409/603, 67.8%) who expressed some level of general concern about their privacy while using the Internet; we combined
We also examined the responses of those participants (178/603, 29.5%) who expressed little or no general concern about their overall privacy while using the Internet; this allowed us to assess whether those with little general privacy concern might still have elevated privacy concern about Twitter monitoring. Those with lower general Internet privacy concern indicated lower concern in response to the questions about eavesdropping, privacy, and confidentiality (see
Hypothesis 2 states that the expectation of Internet privacy relates to the level of concern about Internet research and Twitter monitoring for clinical trial recruitment.
We wanted to gauge whether the presence of general Internet privacy concern is related to increased concern about Internet research (see
Isolating for only those respondents (178/603, 29.5%) who expressed little or no general privacy concern, we found that this population generally had lower levels of Internet research privacy concern (see
Finally, we stratified the responses to the Twitter-monitoring vignettes (see
We also isolated responses for those respondents (178/603, 29.5%) who expressed little or no general concern about their overall privacy while using the Internet; this allowed us to assess whether those with little general privacy concern might still have elevated privacy concern about the types of Twitter monitoring described in the vignettes. As reported in
Finally, we performed chi-square tests to explore whether there was a relationship between general Internet privacy concern and levels of concern expressed with each vignette. The tests revealed a statistically significant relationship in all cases (
Stratified analysis of vignette scenarios for respondents who indicated that they were Very concerned or Somewhat concerned about Twitter monitoring.
Vignette | Respondents (N=603), n (%) | Respondents with high general privacy concern (n=409), n (%) | Respondents with low general privacy concern (n=178), n (%) | Respondents who were active Twitter users (n=199), n (%) |
Cancer vignette | 300 (49.8) | 244 (59.7) | 51 (28.7) | 75 (37.7) |
Obesity vignette | 299 (49.6) | 241 (58.9) | 52 (29.2) | 76 (38.2) |
HPVa vignette | 298 (49.4) | 243 (59.4) | 51 (28.7) | 75 (37.7) |
HIV/AIDS vignette | 349 (57.9) | 269 (65.8) | 73 (41.0) | 106 (53.3) |
Smoking vignette | 255 (42.3) | 207 (50.6) | 45 (25.3) | 66 (33.2) |
aHPV: human papilloma virus.
Chi-square analysis of concern expressed by respondents for each vignette based on their general privacy concern.
Vignette | Number of valid cases, N | Pearson chi-square |
|
|
Cancer vignette | 603 | 175.9 | 16 | <.001 |
Obesity vignette | 603 | 126.7 | 16 | <.001 |
HPVa vignette | 603 | 124.4 | 16 | <.001 |
HIV/AIDS vignette | 603 | 79.6 | 16 | <.001 |
Smoking vignette | 603 | 102.5 | 16 | <.001 |
aHPV: human papilloma virus.
Hypothesis 3 states that general Twitter literacy is associated with the level of concern about the use of social media monitoring on Twitter for clinical trial recruitment.
There was a significant association (
Related to the CIRB’s concerns, we also found a significant association between respondents’ Twitter literacy and whether they considered Twitter monitoring for clinical trial recruitment as eavesdropping (
Chi-square analysis of concerns expressed by respondents based on their Twitter literacy.
Respondents’ concerns | Number of valid cases, N | Pearson chi-square |
|
|
Concern about the ability for researchers to monitor their Twitter activity, generally | 536 | 22.7 | 6 | .001 |
Concern about researchers monitoring particular information types on Twitter (eg, hashtags, public tweets, and profile description) | 556 | 19.3 | 6 | .004 |
Overall concern with researchers monitoring Twitter activity | 513 | 7.2 | 2 | .03 |
Consider Twitter monitoring for clinical trial recruitment as eavesdropping | 602 | 38.1 | 4 | <.001 |
Consider Twitter monitoring for clinical trial recruitment as an invasion of privacy | 603 | 15.8 | 4 | .003 |
Felt Twitter monitoring jeopardizes confidentiality | 603 | 3.9 | 4 | .43 |
Hypothesis 4 states that there are differences in attitudes toward Twitter monitoring for clinical trial recruitment compared to a more traditional, offline scenario.
We also used the vignettes to assess the attitudes toward a more traditional, offline scenario (see
Hypothesis 5 states that the level of concern is associated with the type of information monitored for the purpose of identifying individuals to recruit for clinical trials.
We assessed the level of concern about the type of information medical researchers or research institutions might monitor and review in order to identify individuals for recruiting them into clinical trials (see
Hypotheses 6 and 7 state that there is a level of concern associated with the type of disease recruited for and the type of entity performing the monitoring.
We used the set of vignettes (see
We further stratified responses for each vignette’s subquestions, isolating responses for those who expressed some concern—indicated
All relevant data that support the findings of this study are available in the data repository figshare:
Responses from Twitter users: Monitoring Twitter for clinical trial recruitment [
Responses from TurkPrime workers: Monitoring Twitter for clinical trial recruitment [
Public social networks such as Twitter provide access to user information, including personal and sensitive data, without necessarily requiring an individual's knowledge or consent. While previous studies explored the unique ethical challenges of social media as a health research tool and research data source [
The goal of this study was to contribute data that reflect public views of Twitter users and nonusers and to inform the scientific discourse about the use of Twitter user data for clinical trial recruitment. We discuss our findings in relation to our hypotheses (see
Summary of study findings by study hypothesis.
Hypotheses | Overall findings (nonstratified) | |
|
||
|
Hypothesis 1: Social media monitoring on Twitter for clinical trial recruitment is perceived as eavesdropping and as an invasion of privacy. | Not supported. While nearly half the respondents indicated agreement that social media monitoring constitutes a form of eavesdropping that invades their privacy, over one-third disagreed and nearly 1 in 5 had no opinion. Fewer respondents felt that social media monitoring jeopardizes confidentiality. |
|
Hypothesis 2: Twitter users’ expectations of privacy relate to their level of concern about the use of social media monitoring for clinical trial recruitment. | Supported. Chi-square tests revealed a positive relationship between respondents’ general privacy concerns and their average concerns about Internet research (N=603): χ216=143.0, |
|
Hypothesis 3: General literacy about the Twitter platform is associated with the level of concern about the use of social media monitoring on Twitter for clinical trial recruitment. | Supported. There was a statistically significant association ( |
|
||
|
Hypothesis 4: People’s concerns over Twitter monitoring for clinical trial recruitment are similar to those of more traditional, offline scenarios (eg, discretely approaching a patient in person as they leave a medical facility). | Supported. Most people were either indifferent, did not know, or were less comfortable with an in-person approach, regardless of previous Twitter usage and across all disease types. They did not find Twitter monitoring any more concerning than the more traditional means of clinical trial subject recruitment. Overall, the data presented here support the use of the |
|
||
|
Hypothesis 5: The type of information monitored for the purpose of identifying individuals to recruit for clinical trials is associated with the level of concern over the use of social media monitoring on Twitter for clinical trial recruitment. | Partially supported. While not a majority, nearly half the respondents did indicate general concern about researchers actively monitoring users’ Twitter activity to identify and contact potential participants for clinical trials. The greatest concern was related to reviewing the text of their profile description, with less concern expressed related to monitoring hashtags or the text of individual tweets. |
|
Hypothesis 6: The type of disease recruited for is associated with the level of concern over the use of social media monitoring on Twitter for clinical trial recruitment. | Supported. Nearly 6 out of 10 respondents expressed concern about monitoring for an HIV/AIDS trial compared to other disease topics that raised less concern, such as cancer, obesity, HPVb vaccination, and smoking. |
|
Hypothesis 7: The nature of the entity performing social media monitoring on Twitter is associated with the level of concern over this monitoring for clinical trial recruitment. | Supported. The factor that most impacted the level of concern was the entity or person who conducted the Twitter monitoring and research. The exception was the HIV/AIDS scenario, where respondents who expressed overall concern noted that |
aCIRB: Central Institutional Review Board.
bHPV: human papilloma virus.
When we tested the concerns raised by the CIRB that
Furthermore, the CIRB committee noted that “those who openly share their information via social media platforms may still be unaware of the platforms’ privacy policies.” We found that there is a significant association between respondents’ Twitter literacy and their concerns about the ability for researchers to monitor their Twitter activity, generally, for the purpose of clinical trial recruitment. We further found a significant association between respondents’ Twitter literacy and their concerns about researchers monitoring particular information types on Twitter (eg, hashtags, public tweets, and profile description) for the purpose of clinical trial recruitment. We cannot state, however, that these concerns necessarily increase as Twitter literacy increases. Related to the CIRB’s concerns, we also found a significant association between respondents’ Twitter literacy and whether they consider Twitter monitoring for clinical trial recruitment as eavesdropping and an invasion of privacy; however, there was no significant association between Twitter literacy and whether respondents felt Twitter monitoring jeopardizes confidentiality. Overall, there is a significant association between respondents’ Twitter literacy and their overall concern with researchers monitoring Twitter activity, suggesting that the more that users understood about Twitter as a platform, the greater they were concerned about researchers monitoring their Twitter activity. This presents a challenge seen in many areas of online literacy, as confirmed in studies of Internet users, in general [
Gelinas et al suggested employing a
Following Marwick and Boyd [
Our findings further support the point previously made by Bender et al [
HIV a very serious and private disease... it is something that needs to be discussed in person.
On Twitter, users are using the specific language. These users have already disclosed their opinions or diagnosis. I feel like it's similar to outing someone on accident if a company were to just randomly ask people.
However, respondents also argued in favor of using Twitter monitoring for clinical trial recruitment:
If you talk about HIV/AIDS on Twitter or any social media, you have to know it's not private.
As long as the person or researcher making contact with the target is being very transparent about the source of the research and is happy to give information to verify their identity and intent, I wouldn't be alarmed or put off.
We identified additional factors that influenced the level of concern about monitoring Twitter user data for clinical trial recruitment. With the exception of the HIV/AIDS scenario as stated above, the factor that most impacted the level of concern was the type of entity or the person who conducted the research. Researchers who may use this approach should ensure investigator transparency; for example, investigators should refrain from fabricating online identities and clearly disclose the goal and design of the research [
Finally, the form of contact on Twitter (ie, public replies versus private messages) played a more important role for the HIV/AIDS, obesity, and HPV scenarios, where a noticeably larger portion of the respondents expressed some concern. Respondents argued as follows:
This condition definitely need[s] to be addressed privately and not through a public reply.
I think the public reply instead of a dm [direct message] could be embarrassing.
The nature of this can be very embarrassing and a public reply could be damaging.
This may be due to the stigma [
This study was limited to two populations: Twitter users and TurkPrime workers. The range of ages, education levels, and socioeconomic statuses of these populations could be more limited than those found in the general public. A total of 22% of US adults use Twitter, nearly equally among white, black, and Hispanic adults across all ages but with the highest usage among those 18-29 years of age [
Additionally, this was an exploratory study prompted by the feedback from a national research organization (ie, CIRB) and the sample size of this study was limited. More robust studies with a larger sample could yield additional insights. Finally, we acknowledge that while we chose seven hypotheses for this initial study, there are certainly other issues and variables that deserve further attention related to the subject in future studies.
The data we presented here contribute to the critical dialogue with the public about the use of social media in clinical research. Public social networks such as Twitter offer the clinical research community a novel opportunity for identifying and engaging potential study participants based on user activity data. However, the availability of public social media data has led to new ethical challenges about respecting user privacy and the appropriateness of monitoring social media for clinical trial recruitment. The results of this study suggest that most users do not think monitoring Twitter for the purpose of clinical trial recruitment constitutes inappropriate surveillance or a violation of privacy. Our data further support the previously suggested use of the
Open 39-item survey used in this study.
Respondents’ demographics.
Respondents’ Twitter usage.
Respondents’ Twitter literacy.
Respondents’ general Internet privacy concerns.
Respondents’ Internet research privacy concerns.
Respondents’ overall opinion of social media listening on Twitter for clinical trial recruitment.
Stratified analysis of the overall opinion of social media listening on Twitter for clinical trial recruitment for those responses that indicated agreement with the concerning issues of eavesdropping, invasion of privacy, and jeopardized confidentiality.
Stratified analysis of respondents who indicated “Very concerned” or “Somewhat concerned” about their privacy while using the Internet.
Respondents’ levels of concern about scenarios described in vignettes with a focus on study disease and the entity monitoring the social media user activity.
Responses to vignette subquestions, stratified based on level of overall Internet privacy concern with vignette scenario.
A select sample of respondents’ comments in response to the vignettes.
acute myeloid leukemia
Checklist for Reporting Results of Internet E-Surveys
Central Institutional Review Board
human papilloma virus
Institutional Review Board
Research Electronic Data Capture
None declared.