Menu
Ibe C, Ranney M, Pateña J, Wray T. The impact of remote recruitment and intervention on adolescents at-risk for cybervictimization. HPHR. 2021;40. DOI: 10.54111/0001/NN2
Although online recruitment is increasingly used for studies, the representativeness of populations recruited this way is largely unknown. This study, completed prior to COVID-19, compares the reach and representativeness of recruitment for an adolescent cybervictimization intervention using in-person versus online methods.
In Study A, English-speaking adolescents ages 13-17 were recruited in-person at a hospital-based pediatric clinic to participate in a hybrid cybervictimization intervention program (N=38). In a separate, subsequent study (Study B), adolescents of the same age were recruited remotely via Instagram advertisements to participate in a fully online cybervictimization program (N=79). Screening, 8-week post-enrollment, and 16-week post-enrollment surveys consisted of validated measures divided into four sub-groups (Demographics, At-Risk Factors, Support Factors, and User Experience). Statistical analyses were performed using R
Chi-square tests revealed participants were not significantly different for six of 10 demographic measures. Study B participants were underrepresented in low socioeconomic status, Hispanic identity, and non-white race. However, significantly more remote participants identified as non-heterosexual, an identity associated with notably increased risk for cyberbullying. T-tests revealed participants in Study B demonstrated significantly more risky behavior for eight of 12 measures. By 16-week follow up, all risky behavior measures were not significantly different between the two groups, largely due to decreasing reported risk in Study B participants following intervention. Participants in both studies were not significantly different for two of three Support measures. However, Study A subjects reported significantly less perceived social support, the only non-demographic measure in which those recruited in-person exhibited more risk for cybervictimization. User experience was not significantly different between cohorts.
Results of this study demonstrate remote methodology enhances access to at-risk adolescents without compromising the success of the subsequent intervention.
Findings should serve as evidence of the acceptability and feasibility of remote recruitment and a call to action for further cyberbullying intervention research.
The ongoing digital revolution has brought computer-mediated communication (CMC) to the forefront.1 Ninety-five percent of U.S. teens report having a smartphone or access to one, and 89% describe their Internet usage as occurring several times a day or nearly constantly.2 Adolescents have a wide array of communication tools at their disposal, and social media is currently the dominating preference, with 70% of teens reporting at least two social media platforms comprising a majority of their time spent online.2 However, only 31% of teens describe their social media experience as positive.2
With a powerful medium comes great opportunity for abuse. Up to 72% of U.S. adolescents ages 10 to 19 report being the victim of cyber harm,3 which most commonly occurs as offensive name-calling and spreading of false rumors.4 A Pew Research Center survey found that 90% of U.S. teens believe online harassment is a problem and are frustrated with the lack of intervention by authoritative figures.4 Adolescents feel teachers, social media companies, and politicians especially have failed to address their concerns.4
Harm perpetuated online holds distinct power.5,6 Without the face-to-face interaction, a cyberbully is able to hide their identity with ease. This anonymity reduces feelings of accountability and fear of consequence.7,8 Adolescents also enjoy less supervision, if any, on the Internet, making the bullying easier and intervention harder.9 Moreover, the threat of permanence haunts victims, with release of any private images or offensive remarks on social networks guaranteed to have a lasting digital presence discoverable by anyone.10
Adolescents facing cyberbullying are burdened with lasting mental distress. The harm can lead to psychosomatic symptoms including headaches, abdominal pain, and difficulty sleeping.10 Victims also report higher rates of depressive symptoms, suicide planning, and weapon carrying.11 Even beyond adolescence, cyberbullying of undergraduates is associated with career problems, difficulty in interpersonal relations, depression, and suicidal ideation.12
The COVID-19 pandemic has made effective cyberbullying intervention both more urgent, and more difficult.13 Digital communication surged amidst lockdown orders,14 but researchers discovered an association between the increased screen time and worsened mental health.15 According to a study from the L1ght organization, a creator of AI solutions to detect and remove toxic web behaviors, there has been a 70% increase in hate speech amongst young teens since the onset of the pandemic.16 Racial and sexual minorities are disproportionally impacted by both cyberbullying17,18 and COVID-19,19–22 meaning the most in need became the hardest to reach.
The harm of cyberbullying can be mitigated by developing emotional social skills and learning how to regulate emotions effectively.23,24 Traditional preventive measures commonly involve curriculum-based programs, yet teachers feel deficient in their knowledge of handling cyberbullying.25.26 The Brown-Lifespan Center for Digital Health conducted an innovative study testing a text-message based cyberbullying intervention for high-risk youth identified in an outpatient clinic.27 However, the recruitment was overly laborious, as it required great dedication of resources and staff. Moreover, this process would be especially difficult to replicate given COVID-19 distancing guidelines.
Online intervention has proven efficacious in several research areas such as treating anxiety and depression,28–30 promoting physical activity,31 and aiding smoking cessation.32 Notably, online recruitment, commonly in the form of a social media campaign, has been used in an advantageous manner. Online recruitment has allowed access to more diverse populations,33,34 led to higher enrollment rates,35 required fewer study resources,35 and proven to be more cost effective overall.36 Enhanced accessibility to sexual and ethnic minorities, or the groups most at-risk for cyberbullying, is especially relevant given the longstanding problems with diversity in research.37
How can we deliver an easily disseminated cyberbullying intervention proven to be efficacious in a diverse cohort of at-risk youth? Given the dire need for intervention and the limitations of in-person research, remote recruitment and intervention can prove useful. To better understand the influence of remote methodology in finding and engaging at-risk youth, this analysis utilizes data from the “Intervention to Prevent Adolescent Cyber-victimization through Text messaging (iPACT)” and “Intervention Media to Prevent Adolescent Cyber-conflict through Technology (IMPACT)” studies. The aim of this inquiry is to characterize who is reached and identify the efficacy of intervention when remote methods are utilized. As the research community seeks to progress past the novice stage of cyberbullying interventions, our findings can help guide effective trial and intervention design for the future.
iPACT evaluated the acceptability and feasibility of a two-part (clinic session and tailored eight-week text program) hybrid intervention for youth reporting cybervictimization. Participants were recruited from March 2018 to June 2018 at an outpatient pediatric clinic located in a Northeastern urban city. Participants were offered a small value incentive of 1 pack of gum. Research Assistants (RAs) pre-screened for English-speaking patients 13-17 years of age who were not in police or state custody, diagnosed with a developmental disability, or pregnant. Following verbal consent and assent from the patient and their legal guardian (s), eligible adolescents completed a screening survey on REDCap.38
Similarly, the IMPACT study evaluated the acceptability and feasibility of a two-part (video session and tailored eight-week app program) remote intervention for youth reporting cybervictimization. Participants were recruited from January 2020 to May 2020 through targeted Instagram advertisements, which relied on varying ad campaigns and placements. Advertisements were promoted through Reach campaigns, showing ads to users within a target audience (e.g., all genders between 13-17 years old), and Traffic campaigns, showing ads to users that click on ads frequently. Advertisements were featured in Story format, a post that disappears after 24 hours; Feed format, a lasting post within a user’s feed that one can engage with easily; and Explore format, a lasting post shown to the user based on similar content they have engaged with previously.39
Users who engaged with the advertisement by pressing the “Learn More” button were prompted to complete an anonymous screening survey on REDCap.38 Akin to iPACT, the survey pre-screened for adolescents who were 13-17 years of age, U.S. residents, and English speaking. Users who completed the survey were invited to provide their email and enter a raffle to receive a $50.00 Amazon gift card.
Following completion of the screening surveys for both studies, participants who reported at least one incident of past year cybervictimization and access to a phone with texting abilities were deemed eligible for the pilot randomized controlled trial (RCT). Eligible participants who assented with written informed consent and passed verification of their and their legal guardian(s)’ contact information were enrolled into the study and randomly assigned to the control or intervention group.
Participants were presented with an RA-led information session in the pediatric clinic (iPACT) or via video conferencing (IMPACT) discussing bullying-prevention skills and Cognitive Behavioral Therapy (CBT) techniques, including the thoughts-feelings-action triangle that defines the relationship between what we think, feel, and do. Following the first phase of intervention, participants were enrolled in a text message program developed by Reify Health (iPACT) or an app program developed by JourneyLabs (IMPACT), delivering daily curriculum on how to respond to harassment, manage emotions, and reduce likelihood of future victimization. Participants also received a daily questionnaire assessing their wellbeing (i.e., How are you feeling in general today? 1=really bad, 5=great) and cybervictimization experience (i.e., Any drama or conflict online today? “Yes” or “No”). Participants in the control group were provided with a standard sheet (iPACT) or link to a website (IMPACT) on bullying and mental health resources.
Following completion of the intervention, participants completed 8-week post-enrollment and 16-week post-enrollment surveys and participated in semi-structured exit interviews.
Across the surveys, iPACT and IMPACT evaluate seven constructs: Cybervictimization, Other Violence, Mental Health, Mechanisms of Behavior, Other Risky Behavior, Technology Usage, and Demographics. Accounting for questions repeated in post-enrollment surveys, the iPACT surveys consist of 307 questions, and the IMPACT surveys consist of 258 questions. For the purpose of this analysis, participant responses from 193 questions were sourced from each study. Selection of measures was made at the discretion of the authors based on identical questions in both intervention surveys, relevance to the objective of this analysis, and completion of questions from participants in both groups. All measures in this analysis have been validated,40-53 and they can be re-categorized into four categories: Demographics, At-Risk Factors, Support Factors, and User Experience. The qualitative Demographics measures and quantitative At-Risk, Support, and User Experience measures are detailed in Tables 1 and 2 respectively.
Table 1. Qualitative Demographics Measures
Measure | Response options* | Response categorization for χ2 |
Grade level | 6th grade 7th grade 8th grade | Middle school |
9th grade 10th grade 11th grade 12th grade | High school | |
Academic performance | Mostly As Mostly As and Bs Mostly Bs Mostly Bs and Cs | Strong academic performance |
Mostly Cs Mostly Cs and Ds Mostly Ds Mostly Ds and Fs Mostly Fs | Weak academic performance | |
Socioeconomic status | Qualifies for reduced lunch OR guardian receives public assistance | Low SES |
Does not quality for reduced lunch AND guardian does not receive public assistance | High SES | |
Parental status | Has a child | Parent |
Does not have a child | Not a parent | |
Hispanic identity | Hispanic | Hispanic |
Not Hispanic | Not Hispanic | |
Race | White | White |
American Indian / Alaska Native Asian Native Hawaiian or Other Pacific Islander Black or African American Other | Non-white | |
Sex at birth | Male | Male |
Female | Female | |
Gender | Male Female | Cis gender |
Transgender Other gender not listed | Non-cis gender | |
Sexuality | Straight | Heterosexual |
Gay or lesbian Bisexual Not listed above | Non-heterosexual | |
Counseling outpatient status | Has seen a counselor as an outpatient for help with emotional problems | Counseling outpatient |
Has not seen a counselor as an outpatient for help with emotional problems | Not a counseling outpatient |
*Prefer not to answer included as an option for all measures.
Table 2. Quantitative At-Risk, Support, and UEX Measures
Measure | Number of questions* | Question type / scoring | Total scoring |
Youth Internet Safety Score (YISS) | 1 | Binary: 0-1 | Sum: 0-1 |
Past 12mo. Cybervictimization | 5 | Likert scale: 0-4 | Sum: 0-20 |
Past 2mo. Cybervictimization | 5 | Likert scale: 0-4 | Sum: 0-20 |
Impairment in Daily Life | 3 | Likert scale: 0-3 | Sum: 0-9 |
Emotional Toll | 7 | Likert scale: 0-4 | Sum: 0-28 |
Physical Fights | 1 | Likert scale: 0-4 | Sum: 0-4 |
Past 12mo. Dating Violence | 5 | 1Binary: 0-1 4Likert scale: 0-4 | Sum: 0-17 |
Past 2mo. Dating Violence | 5 | 1Binary: 0-1 4Likert scale: 0-4 | Sum: 0-17 |
Past 12mo. In-Person Bullying | 7 | Likert scale: 0-4 | Sum: 0-28 |
Past 2mo. In-Person Bullying | 7 | Likert scale: 0-4 | Sum: 0-28 |
Illicit Drug and Alcohol Usage | 3 | Likert scale: 0-4 | Sum: 0-12 |
Witnessing Bullying | 1 | Likert scale: 0-4 | Sum: 0-4 |
Cybervictimization Solutions | 9 | Binary: 0-1 | Sum: 0-9 |
Feelings of Bystander Accountability | 13 | Likert scale: 1-5 | Average: 1-5 |
Social Support | 12 | Likert scale: 1-7 | Average: 1-7 |
User Experience | 4 | Likert scale: 0-2 | Sum: 0-8 |
*Refers to number of questions in a given survey. Does not account for number of times a question is repeated across surveys.
All statistical analyses were conducted using R and RStudio.54 A series of Chi-square tests were used to determine statistical significance for the Demographics measures. Subsequently, F tests were used to examine equality of variances for responses in the At-Risk Factors, Support Factors, and User Experience categories. For data sets with unequal variances, a Welch’s t-test was used to compare means, whereas those with equal variances were examined with a Student’s t-test.
Table 3. Significance of Demographics in Adolescent Study Populations
Classification | iPACT (N=38) | IMPACT (N=79) | χ2 | df | Sig. |
Middle School | 39.5% | 20.3% | 0.490 | 1 | 0.484 |
High School Prefer not to answer | 61.5% 0% | 77.2% 2.5% | |||
Strong Academic Performance | 84.2% | 96.2% | 0.0495 | 1 | 0.824 |
Weak Academic Performance Prefer not to answer | 13.2% 2.6% | 3.8% 0% | |||
Low SES | 84.2% | 32.9% | 7.51** | 1 | 0.00613 |
Not Low SES | 10.5% | 63.3% | |||
Prefer not to answer | 5.3% | 3.8% | |||
Parent | 7.9% | 0% | 0.0233 | 1 | 0.879 |
Not a Parent | 89.5% | 100% | |||
Prefer not to answer | 2.6% | 0% | |||
Hispanic | 60.5% | 12.7% | 16.89** | 1 | 3.96E-05 |
Not Hispanic | 21.1% | 87.3% | |||
Prefer not to answer | 18.4% | 0% | |||
White | 2.6% | 68.4% | 10.83** | 1 | 0.0001 |
Non-white | 81.6% | 29.1% | |||
Prefer not to answer | 15.8% | 2.5% | |||
Male | 31.6% | 22.8% | 0.386 | 1 | 0.535 |
Female | 65.8% | 77.2% | |||
Prefer not to answer | 2.6% | 0% | |||
Cis Gender | 92.1% | 84.8% | 0.00279 | 1 | 0.958 |
Non-Cis Gender | 2.6% | 10.1% | |||
Prefer not to answer | 5.3% | 5.1% | |||
Heterosexual | 81.6% | 40.5% | 4.09** | 1 | 0.0431 |
Non-Heterosexual | 10.5% | 54.4% | |||
Prefer not to answer | 7.9% | 5.1% | |||
Counseling outpatient Not a counseling outpatient Prefer not to answer | 32.0% 68.0% 0% | 46.8% 53.2% 0% | 0.702 | 1 | 0.402 |
** refers to p<.05.
As depicted in Table 3, Chi-square analysis revealed that the iPACT and IMPACT cohorts were not significantly different for majority of Demographics measures, including grade level (p=0.484), academic performance (p=0.824), parental status (p=0.879), biological sex (p=0.535), gender choice (p=0.958), and receiving outpatient emotional care (p=0.702). The studies were successful in reaching groups with a primarily similar demographic makeup. However, the IMPACT cohort had significantly more representation for sexuality and less for low SES, Hispanic identity, and non-white race.
Table 4. Screening and Baseline Descriptive Statistics
Factor | N | Mean | Standard Deviation | Standard Error Mean |
iPACT Youth Internet Safety Score (YISS) | 38 | 0.237 | 0.431 | 0.0699 |
IMPACT Youth Internet Safety Score (YISS) | 79 | 0.722** | 0.451 | 0.0508 |
iPACT Past 12mo. Cybervictimization | 38 | 2.08 | 1.28 | 0.208 |
IMPACT Past 12mo. Cybervictimization | 79 | 5.32** | 4.73 | 0.532 |
iPACT Past 2mo. Cybervictimization | 38 | 0.869 | 1.02 | 0.165 |
IMPACT Past 2mo. Cybervictimization | 79 | 3.34** | 3.71 | 0.418 |
iPACT Impairment in Daily Life | 38 | 2.21 | 2.42 | 0.392 |
IMPACT Impairment in Daily Life | 79 | 2.37 | 2.53 | 0.284 |
iPACT Emotional Toll | 38 | 4.21 | 4.91 | 0.797 |
IMPACT Emotional Toll | 79 | 9.33** | 7.49 | 0.842 |
iPACT Physical Fights | 38 | 0.211 | 0.474 | 0.0769 |
IMPACT Physical Fights | 79 | 0.0886 | 0.328 | 0.0369 |
iPACT Past 12mo. Dating Violence | 38 | 0.579 | 0.721 | 0.117 |
IMPACT Past 12mo. Dating Violence | 79 | 0.646 | 0.906 | 0.102 |
iPACT Past 2mo. Dating Violence | 38 | 0.395 | 0.595 | 0.0964 |
IMPACT Past 2mo. Dating Violence | 79 | 0.468 | 0.731 | 0.0822 |
iPACT Past 12mo. In-Person Bullying | 38 | 2.37 | 2.08 | 0.338 |
IMPACT Past 12mo. In-Person Bullying | 79 | 7.13** | 4.76 | 0.535 |
iPACT Past 2mo. In-Person Bullying | 38 | 1.53 | 1.61 | 0.261 |
IMPACT Past 2mo. In-Person Bullying | 79 | 4.56** | 3.85 | 0.433 |
iPACT Illicit Drug and Alcohol Usage | 38 | 0.395 | 0.974 | 0.158 |
IMPACT Illicit Drug and Alcohol Usage | 79 | 1** | 1.78 | 0.200 |
iPACT Witnessing Bullying | 38 | 0.579 | 0.826 | 0.134 |
IMPACT Witnessing Bullying | 79 | 1.53** | 1.26 | 0.142 |
IPACT Cybervictimization Solutions | 38 | 4.45 | 2.60 | 0.421 |
IMPACT Cybervictimization Solutions | 79 | 4.39 | 2.16 | 0.243 |
iPACT Feelings of Bystander Accountability | 38 | 3.48 | 0.669 | 0.109 |
IMPACT Feelings of Bystander Accountability | 79 | 3.40 | 0.624 | 0.0702 |
iPACT Social Support | 38 | 3.58** | 0.675 | 0.109 |
IMPACT Social Support | 79 | 3.93 | 0.921 | 0.104 |
** refers to p<.05.
Table 5. 8-Weeks Post-Enrollment Descriptive Statistics
Factor | N | Mean | Standard Deviation | Standard Error Mean |
iPACT Past 2mo. Cybervictimization | 18 | 1.28 | 1.36 | 0.321 |
IMPACT Past 2mo. Cybervictimization | 36 | 3.92** | 4.12 | 0.686 |
iPACT Impairment in Daily Life | 18 | 0.833 | 1.65 | 0.390 |
IMPACT Impairment in Daily Life | 36 | 1.67 | 1.97 | 0.329 |
IPACT Emotional Toll | 18 | 3.56 | 5.03 | 1.19 |
IMPACT Emotional Toll | 36 | 6.44 | 5.47 | 0.912 |
iPACT Physical Fights | 18 | 0.278 | 0.575 | 0.135 |
IMPACT Physical Fights | 36 | 0.0278 | 0.167 | 0.0278 |
iPACT Dating Violence | 18 | 0.611 | 0.850 | 0.200 |
IMPACT Dating Violence | 36 | 0.333 | 0.478 | 0.0797 |
iPACT Past 2mo. In-Person Bullying | 18 | 2.78 | 2.29 | 0.540 |
IMPACT Past 2mo. In-Person Bullying | 36 | 2.89 | 3.03 | 0.505 |
iPACT Cybervictimization Solutions | 18 | 2.94 | 2.80 | 0.659 |
IMPACT Cybervictimization Solutions | 36 | 3.94 | 2.66 | 0.444 |
iPACT Feelings of Bystander Accountability | 18 | 3.36 | 0.861 | 0.203 |
IMPACT Feelings of Bystander Accountability | 36 | 3.63 | 0.475 | 0.0792 |
iPACT Social Support | 18 | 3.58 | 0.922 | 0.217 |
IMPACT Social Support | 36 | 3.72 | 0.692 | 0.115 |
iPACT User Experience | 18 | 5.56 | 1.58 | 0.372 |
IMPACT User Experience | 36 | 6.17 | 1.40 | 0.234 |
** refers to p<.05.
Table 6. 16-Weeks Post-Enrollment Descriptive Statistics
Factor | N | Mean | Standard Deviation | Standard Error Mean |
iPACT Past 2mo. Cybervictimization | 18 | 2.56 | 4.98 | 1.17 |
IMPACT Past 2mo. Cybervictimization | 35 | 3.14 | 2.71 | 0.459 |
iPACT Physical Fights | 18 | 0.167 | 0.514 | 0.121 |
IMPACT Physical Fights | 35 | 0 | 0 | 0 |
IPACT Past 2mo. Dating Violence | 18 | 0.889 | 1.60 | 0.378 |
IMPACT Past 2mo. Dating Violence | 35 | 0.800 | 2.53 | 0.428 |
iPACT Past 2mo. In-Person Bullying | 18 | 2.39 | 2.64 | 0.622 |
IMPACT Past 2mo. In-Person Bullying | 35 | 2.69 | 2.21 | 0.373 |
iPACT Feelings of Bystander Accountability | 18 | 3.33 | 0.875 | 0.206 |
IMPACT Feelings of Bystander Accountability | 35 | 3.70 | 0.495 | 0.0836 |
iPACT Social Support | 18 | 3.56 | 0.837 | 0.197 |
IMPACT Social Support | 35 | 3.94 | 0.717 | 0.121 |
** refers to p<.05.
Table 7. Significance of At-Risk and Support Measures at Screening and Baseline
F Test for Equality of Variances | t-test for Equality of Means |
| ||||||
Factor | F | Sig. | t | df | Sig. | 95% CI Lower | 95% CI Upper | Cohen’s d |
Youth Internet Safety Score (YISS) | 1.10 | 0.773 | 5.52** | 115 | 2.12E-07 | 0.311 | 0.659 | 1.09 |
Past 12mo. Cybervictimization | 13.6 | 1.21E-13 | 5.67** | 98.8 | 1.44E-07 | 2.10 | 4.37 | 0.817 |
Past 2mo. Cybervictimization | 13.3 | 1.79E-13 | 5.51** | 99.2 | 2.90E-07 | 1.58 | 3.36 | 0.795 |
Impairment in Daily Life | 1.09 | 0.781 | 0.318 | 115 | 0.751 | -0.818 | 1.13 | 0.0628 |
Emotional Toll | 2.32 | 0.00562 | 4.41** | 104.2 | 2.48E-05 | 2.82 | 7.42 | 0.756 |
Physical Fights | 0.478 | 0.00641 | -1.43 | 54.6 | 0.159 | -0.293 | 0.0490 | 0.320 |
Past 12mo. Dating Violence | 1.58 | 0.127 | 0.396 | 115 | 0.693 | -0.267 | 0.400 | 0.0783 |
Past 2mo. Dating Violence | 1.51 | 0.166 | 0.540 | 115 | 0.590 | -0.196 | 0.343 | 0.107 |
Past 12mo. In-Person Bullying | 5.20 | 3.69E-07 | 7.52** | 114.3 | 1.37E-11 | 3.50 | 6.01 | 1.16 |
Past 2mo. In-Person Bullying | 5.73 | 9.40E-08 | 6.00** | 113.4 | 2.41E-08 | 2.03 | 4.03 | 0.920 |
Illicit Drug and Alcohol Usage | 3.33 | 0.000122 | 2.38** | 113.0 | 0.0192 | 0.101 | 1.11 | 0.387 |
Witnessing Bullying | 2.32 | 0.00569 | 4.88** | 104.2 | 3.76E-06 | 0.566 | 1.34 | 0.837 |
Cybervictimization Solutions | 0.694 | 0.178 | -0.120 | 115 | 0.904 | -0.959 | 0.849 | 0.0238 |
Feelings of Bystander Accountability | 0.870 | 0.598 | -0.630 | 115 | 0.530 | -0.329 | 0.170 | 0.124 |
Social Support | 1.86 | 0.0385 | -2.35** | 96.3 | 0.0207 | 0.0555 | 0.654 | 0.417 |
** refers to p<.05.
Tables 4, 5, and 6 illustrate the mean scoring of participants for measures at the three points of the study respectively, i.e., screening, 8-weeks post-enrollment (close of intervention), and 16-weeks post-enrollment. As detailed in Table 7, baseline scoring for online participants revealed they were of equal or greater risk of cyberbullying. Of the At-Risk measures, impairment in daily life (p=0.751), number of physical fights (p=0.159), and past-year (p=0.693) and past 2-month (p=0.590) dating violence were not significantly different, which remained consistent at follow-ups. However, the remaining eight measures demonstrate participants recruited online were significantly more at-risk than their in-person counterparts.
Table 8. Significance of At-Risk, Support, and UEX Measures at 8-Weeks Post-Enrollment
F Test for Equality of Variances | t-test for Equality of Means | |||||||
Factor | F | Sig. | t | df | Sig. | 95% CI Lower | 95% CI Upper | Cohen’s d |
Past 2mo. Cybervictimization | 9.11 | 1.38E-05 | 2.64** | 52 | 0.0110 | 0.632 | 4.65 | 0.762 |
Impairment in Daily Life | 1.42 | 0.445 | 1.54 | 52 | 0.129 | -0.252 | 1.92 | 0.445 |
Emotional Toll | 1.18 | 0.732 | 1.88 | 52 | 0.0661 | -0.199 | 5.98 | 0.542 |
Physical Fights | 0.0842 | 1.39E-09 | -1.81 | 18.4 | 0.0869 | -0.540 | 0.0399 | 0.703 |
Past 2mo. Dating Violence | 0.316 | 0.00390 | -1.29 | 22.5 | 0.211 | -0.724 | 0.169 | 0.445 |
Past 2mo. In-Person Bullying | 1.75 | 0.218 | 0.137 | 52 | 0.892 | -1.52 | 1.74 | 0.0395 |
Cybervictimization Solutions | 0.906 | 0.776 | 1.28 | 52 | 0.206 | -0.568 | 2.57 | 0.370 |
Feelings of Bystander Accountability | 0.305 | 0.00291 | 1.25 | 22.3 | 0.226 | -0.180 | 0.723 | 0.432 |
Social Support | 0.563 | 0.148 | 0.609 | 52 | 0.545 | -0.313 | 0.585 | 0.178 |
User Experience (UEX) | 0.790 | 0.539 | 1.45 | 52 | 0.154 | -0.237 | 1.46 | 0.417 |
** refers to p<.05
Table 9. Significance of At-Risk and Support Measures at 16-Weeks Post-Enrollment
F Test for Equality of Variances | t-test for Equality of Means | |||||||
Factor | F | Sig. | t | df | Sig. | 95% CI Lower | 95% CI Upper | Cohen’s d |
Past 2mo. Cybervictimization | 0.296 | 0.00247 | 0.466 | 22.3 | 0.646 | -2.03 | 3.21 | 0.162 |
Physical Fights | 0 | < 2.2E-16 | -1.37 | 17 | 0.187 | -0.426 | 0.0892 | 0.562 |
Dating Violence | 2.49 | 0.0490 | -0.156 | 48.6 | 0.877 | -1.24 | 1.06 | 0.0393 |
Past 2mo. In-Person Bullying | 0.699 | 0.367 | 0.434 | 51 | 0.667 | -1.08 | 1.67 | 0.126 |
Feelings of Bystander Accountability | 0.320 | 0.00457 | 1.66 | 22.7 | 0.110 | -0.0908 | 0.830 | 0.572 |
Social Support | 0.734 | 0.432 | 1.73 | 51 | 0.0894 | -0.0607 | 0.823 | 0.502 |
** refers to p<.05.
The Instagram cohort scored significantly higher for number of stranger interactions or YISS (p<0.0001) , past-year (p<0.0001) and past 2-month (p<0.0001) cyberbullying incidences, emotional toll (p=<0.0001), past-year (p<0.0001) and past 2-month (p<0.0001) in-person bullying incidences, substance use (p<0.05), and number of bystander incidences (p<0.0001). As depicted in tables 8 and 9, the aforementioned measures repeated at follow-ups were not significantly different after the intervention, as risk was brought to a comparably lower level for both populations. Thus, results indicate remote methodology was successful in enhancing access to high-risk populations while maintaining the efficacy of the subsequent intervention.
Of the Support measures, participants in respective cohorts scored similarly in feelings of bystander responsibility (p=0.530) and number of solutions used by participants to stop cybervictimization (p=0.904) at baseline, which sustained through follow-ups. However, in-person participants reported significantly lower perceived support in their network. Social support is the only non-Demographic measure in which those recruited in-person exhibited more risk for cyberbullying. By 8-weeks post-enrollment, feelings of social support were comparably high between participants.
Finally, there was no significant difference in user experience between participants, as both groups scored comparably high.
As signified by the results of our analysis, online recruitment increased access to hard-to-reach populations for cyberbullying. The IMPACT methodology succeeded in enrolling significantly more sexual minorities, with 54.4% of the cohort identifying as non-heterosexual. Similarly, 10.1% of online participants self-identified as transgender, nonbinary, or another non-cis gender, in comparison to just 2.6% of in-person subjects. Social media ultimately served as a tool for reaching these particularly underrepresented cohorts in research.37
LGBTQ+ youth are harassed online notably more than their heterosexual or cis-gender counterparts.17 Moreover, victims of homophobic bullying in childhood express more severe depression, anxiety, and physical pain in early adulthood, indicating the importance of early mitigation.55 Despite their heightened vulnerability to cyberbullying, sexual and gender minority youth have historically not been targeted for interventions.17 All people deserve a right to participate in research that affects them, and our findings on online recruitment highlight a path to increasing connection to adolescents in the LGBTQ+ community.
As cited in previous studies, a contributing factor to the lack of access to research for LGBTQ+ individuals is the fear of being outed if data is mishandled.37,56 Online mediums have successfully been utilized to help overcome this reluctance to self-reporting orientation.34 Guillory et al. found that use of audience tailored Facebook and Instagram advertising was crucial in recruiting one of the largest LGBTQ+ young adult (18-24 years old) samples for tobacco cessation intervention.34 Similar to cyberbullying, LGBTQ+ individuals are at an elevated risk for tobacco use but are difficult to target for intervention, signifying the need for novel methodology.57 The added layer of anonymity afforded with online recruitment can enhance participant comfortability in disclosing information that may be sensitive or non-public, such as one’s orientation.
Although we found the online-recruited cohort was significantly more diverse in gender and sexuality, they were underrepresented in low socioeconomic status, Hispanic identity, and non-white race. However, in-person participants were selected from a pediatric outpatient clinic dedicated nearly exclusively to serving underprivileged communities. Eighty-four percent of the patients at this facility are Medicaid enrollees, majority of whom are racial and ethnic minorities.58 Researchers in organizations that do not serve a similarly diverse population may still benefit from using online recruitment as a tool to enhance access to their studies.
The argument for the acceptability of online research recruitment is further supported by IMPACT participants demonstrating equal or heightened risk for cyberbullying for all At-Risk measures and two out of three Support measures. For example, we discovered that significantly more online participants habitually spoke to people they did not know offline. The average teenage social media user spends one hour and 11 minutes on their preferred networks every day, granting them ample time to connect with a stranger.59 A friend one only knows online benefits more from the Internet’s anonymity cloak, thus increasing the likelihood of cyberbullying.60 Comparison of the baseline surveys confirms the IMPACT group had experienced significantly more cyberbullying incidences at baseline.
The use of Instagram specifically may have also aided in connecting with teens who face cyber-harm more frequently. Instagram is the second most commonly used social media site for teens in the United States, and it has the highest occurrence of cyberbullying across all platforms.61 Historically, the preferred tool for social media recruitment has been Facebook,62 but those interested in cyberbullying research should examine Instagram as a resource.
Those recruited online also reported a significantly greater emotional toll due to cyberbullying, indicating increased relevance and need for intervention. This finding is well explained by the higher incidence of cyberbullying in this group and literature confirmation that those who struggle with their emotions are already more prone to victimization.63 In other words, the teens that are least prepared to deal with the emotional fallout of cyberbullying are most frequently targeted.
Online recruitment improves accessibility to those who need the guidance the most, granting them the opportunity to be involved in the improvement of the intervention. Participants recruited online reported significantly more occurrences of witnessing the bullying of another peer, and a study conducted by Panumaporn et al., found that cyberbullying victims are significantly more likely to intervene as a bystander.64 By equipping the most at-risk participants with the tools to respond to their own bullying, we concurrently create better bystanders. Creating and refining an intervention around those who stand to benefit most from its usage is crucial for widespread success upon dissemination. The comparably high scoring in user satisfaction confirms researchers can utilize online methods without compromising results or the participant experience.
Though the evidence is compelling, there are limitations to our data analysis. To begin, the sample size of both cohorts is relatively small. Analysis of screening data compared in-person samples of N=38 to online samples of N=79, which may not be fully representative of the population of adolescents online.
Moreover, all analysis was reliant upon self-reported measures. Given the sensitive nature of the questions asked, youth may not have felt motivated to answer with absolute honesty. This may have resulted in underreporting. However, the limitations of self-reporting may not have affected both groups equally, as more research is needed to determine how online contexts contribute to trust and disclosure.
Finally, the degree of generalizability is limited. Although our results imply online recruitment may be useful in reaching at-risk populations, this may be exclusive to youth. Our data was sourced from studies that restricted participation to individuals aged 13-17, a group that self-describes their time spent online as almost constant.2 More research would be needed to determine if online recruitment increases accessibility to various adult subpopulations.
Our analysis reveals online recruitment may be a promising tool for enhancing access to potential study participants. Social media advertising was successful in gathering participants demonstrating comparable or increased risk for majority of demographic factors and all but one behavioral factor. The findings from this paper should serve as evidence of the acceptability and feasibility of remote recruitment and a call to action for further cyberbullying intervention research, especially targeting sexual and gender minorities. Interventions should be designed around the communities most in need, and the first step towards a more representative study population is expanding reach in recruitment. Given the act of cyberbullying occurs online, it serves as a relevant issue to test innovative remote methods. Moreover, our growing understanding of the lasting consequences of cyberbullying points to urgency in evidence-based tools for early mitigation. Utilization of novel digital methods can act as a vital aid in overcoming challenges of accessibility in research and healthcare at wide.
Chiaka Ibe is a recent alumna of the Brown-Lifespan Center for Digital Health and Brown University, graduating magna cum laude with a degree in Health and Human Biology.
Her studies centered on analyzing health crises around the world and their exacerbation by common societal failures, such as poor administrative efforts, growing refusal of vaccinations, and ever-complicated drug pricing policies. She translated her passion for public health to the lab, working under the guidance of her co-authors to engage in an array of innovative digital health studies. Her research contributions have been featured by the Journal of Maternal-Fetal & Neonatal Medicine and the American Public Health Association.
Chiaka is now an Associate Consultant at Bain & Company, working primarily with the Private Equity and Healthcare practices.
Megan Ranney MD, MPH is a practicing emergency physician, researcher, and advocate for innovative approaches to health. Her work focuses on the intersection between digital health, violence prevention, and population health.
She is the founding Director of the Brown-Lifespan Center for Digital Health, as well as the Associate Dean of Strategy and Innovation for the Brown University School of Public Health. She graduated from Harvard University summa cum laude with a Bachelor of Arts in History of Science in 1997. She served as a Peace Corps Volunteer in Cote d’Ivoire prior to attending medical school at Columbia University College of Physicians & Surgeons in NYC. She graduated with AOA status and received the Leonard Tow Humanism in Medicine award from the Gold Humanism Society on graduation. She completed internship, residency, and chief residency in Emergency Medicine, as well as a fellowship in Injury Prevention Research and a Master of Public Health, at Brown University.
She is currently the Warren Alpert Endowed Associate Professor in the Department of Emergency Medicine at Rhode Island Hospital/Alpert Medical School of Brown University. Her work has been featured by hundreds of media outlets, including CNN, MSNBC, the BBC, the New York Times, the Washington Post, and Fox News.
John Pateña, MPH, MA is a public health professional with a passion for promoting mental wellness. He is currently the Program Director at the Brown-Lifespan Center for Digital Health.
John graduated from the Brown University School of Public Health (MPH ’14) focusing in behavioral and social sciences interventions. He also has a degree in counseling psychology and practiced as a mental health counselor working with youth in hospitals, schools, and community centers. John is currently a Doctor of Public Health (DrPH) student at New York University School of Global Public Health focusing in implementation science: translating research into practice, disseminating evidence-based programs, and scaling up programs/policies that support population mental health.
Dr. Tyler Wray, PhD is the Edens Family Assistant Professor of Healthcare Communication & Technology at the Brown University School of Public Health (SPH).
Dr. Wray is a clinical psychologist who studies how technology can be used to encourage health behavior change, particularly in the areas of addiction and HIV prevention. Since joining the faculty at Brown, his work has been continuously funded by the National Institutes of Health (NIH) and has received over $9 million in total support to date. Dr. Wray’s research involves designing and testing software/devices that use the unique strengths of various computing tools to deliver behavior change techniques in ways that maximize their potency and relevance.
BCPHR.org was designed by ComputerAlly.com.
Visit BCPHR‘s publisher, the Boston Congress of Public Health (BCPH).
Email [email protected] for more information.
Click below to make a tax-deductible donation supporting the educational initiatives of the Boston Congress of Public Health, publisher of BCPHR.
© 2024 BCPHR: An Academic, Peer-Reviewed Journal