Feasibility of a Web-accessible Game-based Intervention Aimed at Improving Help Seeking and Coping Among Sexual and Gender Minority Youth: Results from a Randomized Controlled Trial

Abstract

Purpose:

To address the gap in interventions for improving sexual and gender minority youth (SGMY; e.g., lesbian, gay, bisexual, and transgender youth) health, we tested the feasibility of a game-based intervention for increasing help-seeking, productive coping skills, resource knowledge/use, and well-being.

Methods:

We conducted a 2-arm randomized controlled trial testing a theory-based, community-informed, Web-accessible computer role playing game intervention. Control condition received a list of resources. Primary hypotheses were high levels of implementation success, game demand, and game acceptability.

Results:

We randomized 240 SGMY aged 14–18 into the intervention (n=120) or control (n=120) conditions. Participants completed baseline (100%), 1-month follow-up (T2; 73.3%), and 2-month follow-up (T3; 64.4%) surveys. Among intervention participants, 55.8% downloaded and played the game. Of those who played, 46.2% reported a desire to play it again and 50.8% would recommend it. Game acceptability exceeded hypothesized benchmarks, wherein participants reported high positive affect (M=2.36; 95% CI:2.13,2.58), low negative affect (M=2.75; 95% CI:2.55,2.95), low tension/annoyance (M=3.18; 95% CI:2.98,3.39), and high competence (M=2.23; 95% CI:2.04,2.43) while playing the game. In multivariable intent-to-treat analyses of 38 secondary/tertiary outcomes, intervention participants reported significantly larger reductions than control participants in cyberbullying victimization (T2 b=−0.28; 95% CI:−0.56,−0.01), binge drinking frequency (T2 b=−0.39; 95% CI:−0.71,−0.06), and marijuana use frequency (T3 b=−2.78; 95% CI:−4.49,−1.08).

Conclusions:

We successfully implemented a Web-accessible game trial with SGMY. The game-based intervention was feasible and acceptable to SGMY, and preliminary results show it improved several health-related behaviors. A larger scale trial is needed to test whether the game-based intervention can reduce health inequities for SGMY.

INTRODUCTION

Sexual and gender minority youth (SGMY; e.g., lesbian, gay, bisexual, transgender youth) are at greater risk than cisgender heterosexual youth for negative health outcomes, including substance use and mental health problems[1–5]. These inequities are driven by general adolescent stressors—such as heightened bullying victimization—and by SGMY-specific factors[6]. When SGMY are bullied, in addition to the typical fears of disclosing their bullying victimization experiences to others, SGMY often fear having to disclose their sexual or gender minority (SGM) identity to adults. This places them at-risk for further discrimination and harassment[7,8] and likely prevents them from reaching out for help. Even when SGMY consider suicide, large proportions of them do not seek help[9], and sexual minority youth have more trouble than heterosexuals with identifying people to talk to about their emotional worries[10]. Sexual minority youth are also more likely to use nonproductive coping strategies (e.g., ignoring problems) to manage life stressors[11,12]. Nevertheless, help-seeking, having support from adults, and engaging in productive coping are associated with reduced substance use and better mental health[13,14], making them promising targets for interventions.

Despite much observational research on SGMY health inequities, few interventions have been tested to improve mental health and reduce substance use and violence exposure[15,16]. One way to improve the health of bullied SGMY is via a web-accessible game intervention. Such interventions are an effective way to reach large numbers of SGMY, including SGMY living in rural and high structural stigma locations (i.e., areas with less SGM-inclusive policies, institutions, and attitudes), who may be insufficiently supported by face-to-face programs[17–20]. Web-accessible gaming programs about mental health[21], alcohol use[22], and smoking[23] have been effective for youth in general. Other advantages are increased fidelity and cost-effectiveness[24,25]. Overall, web-accessible game interventions overcome common impediments of face-to-face interventions and present opportunities to reach SGMY to improve their health. Yet little to no research has examined the feasibility of using web-accessible game interventions with SGMY.

Using a pilot randomized controlled trial (RCT), we sought to test the feasibility of a game-based intervention to increase help-seeking (knowledge, intentions, self-efficacy, and behaviors), productive coping skills use, and coping flexibility, and reduce health risk factors and behaviors among SGMY. As detailed in our protocol paper[26], our primary hypothesis was high RCT implementation fidelity, high game demand, and high game acceptability using a priori benchmarks of success. Our secondary and tertiary hypotheses were, compared to controls, the intervention participants would have greater improvements in short-term outcomes (e.g., help-seeking and knowledge of online resources) and long-term outcomes (e.g., substance use), respectively. We also explored other feasibility issues, such as research implementation procedures, game integration, and adaptation/expansion of the game.

METHODS

Study Design

We conducted a parallel-arm, non-blinded pilot RCT to test the feasibility of a game- based intervention to improve help-seeking behaviors and productive coping strategies to reduce substance use, victimization, and mental health issues among SGMY. Detailed descriptions of study procedures have been published elsewhere[26], and were approved by the Human Research Protection Office at the University of Pittsburgh.

A convenience sample was recruited April-July 2018 using advertisements on social media platforms including Facebook, Instagram, SGM-related Web-based gaming groups, Reddit Gaymer forums and through Pitt+Me, a database that links participants with research studies. Interested individuals clicked on the advertisement and were directed to the Web-based screening questionnaire. To ensure representation of both sexual and gender minorities, enrollment was monitored weekly and specific ads were created and used to target underrepresented groups.

Respondents were eligible if they were English literate, lived in the US, were 14–18 years of age, had experienced bullying/cyberbullying victimization in the past year, had a sexual minority identity (i.e., gay, lesbian, bisexual, or queer) or a gender minority identity (i.e., transgender or nonbinary), had a computer where they could download the game, and had an email address. To determine eligibility, respondents completed a brief Web-based self-reported screening questionnaire before entering in this study. Participants voluntarily consented using a click-to-consent procedure. To protect participants from having to reveal their SGM identities to their caregivers, thereby potentially putting them in harm’s way, we received a waiver of parental consent, allowing participants to assent for themselves.

Once eligible respondents agreed to participate, they were emailed a link to the baseline (T1) survey to complete via REDCap. After completion of the T1 survey, the study team randomized participants to intervention or control conditions. Participants in the intervention condition automatically received an email with a link containing instructions for downloading and installing the game intervention onto their computer. Participants in both the intervention and control conditions were automatically sent a list of resources related to study outcomes.

The first follow-up (T2) survey was activated 4 weeks after T1 survey completion and remained open for 4 weeks. The final follow-up survey (T3) was activated 8 weeks after T1 survey completion and remained open for 8 weeks. All surveys were completed between April and October 2018. The following gift card incentives were given after the completion of each survey: US $10 for T1; $25 for T2; and $50 for T3.

In addition to survey data, we collected game play data. Text files containing milestones achieved, time played, and player choices were automatically sent via a secure file-transfer- protocol (FTP) system to a secure server. Participants’ game play data were tracked using a unique identification number that did not rely on identifiable information.

Randomization

We used permuted block randomization with equal allocation (using block sizes of 2, 4, 6, 8, and 10) to randomize participants. The randomization schema was created using the ralloc package for Stata and implemented in REDCap using the Randomization Module. Participants and researchers were unblinded to intervention assignments.

Intervention Materials: Singularities

Described in detail elsewhere[26], the intervention was a theory-based, community- informed, web-accessible, role-playing game called Singularities, incorporating 3 primary components: encouraging help-seeking behaviors, encouraging use of productive coping, and raising awareness of Web-based resources. This intervention can be primary or secondary/tertiary prevention of mental health and substance use, but only secondary/tertiary prevention for violence victimization. Briefly, Singularities is a role-playing game where the player takes on the role of a Singular, a superhuman individual with special gifts who is located in a school. The player is told that because of their uniqueness, Singulars face prejudice, often driven by fear and misunderstanding. After customizing their character, the player is tasked with finding a team to help them complete their final mission, which is to become a world-class superhero by defeating the robots in the Holochamber Challenge. The player’s mini objectives are to do the following for each of their peers: best identify the nonplayable characters’ (NPC) problems, find the best individuals or resources to help the NPCs, and help the NPC properly communicate or utilize their newfound resources. After the player finishes the Holochamber Challenge, the events that took place are evaluated. For every NPC that was successfully helped, the player was given a positive ending. The players were then encouraged to replay the game and given hints to help them receive positive endings. Each player could replay the game unlimitedly. The intervention group also received a resource list (described below). The intervention was delivered entirely online April-October 2018.

Control Materials

All participants randomized into the control condition received a list of national SGM-inclusive resources for bullying, child abuse, dating violence, suicide, mental health, substance use, homelessness, and crisis services. These materials were delivered via email the day after participants were randomized. After completion of the final T3 survey, control participants were offered a free download of the intervention game. No additional follow-up was initiated.

Measures

All measures/items are described in detail elsewhere[26]. Our primary outcomes (Table 2 for assessment and a priori benchmarks for success, chosen based on prior research[27]) were: enrollment number; randomization; survey completion rates; game download (self-reported); game play (via self-reported items and FTP system); overall impressions of the game (from Gaming Experience Questionnaire[28] with subscales); interest in playing the game again (self-reported); and whether participants would recommend the game to their friends (self-reported).

Table 2.

Primary Outcomes of the Study

Domain and Outcome	Assessment	Hypothesized	Actual
Implementation Procedures
Study population	Total number of participants who were consented and who completed T1 survey	240 participants	240 participants
Participation rate	Total number of people who agreed to participate in the study divided by the total number of people who were eligible to participate in the study	≥ 80%	59%
Number of Randomized Participants	Total number of participants randomized	240 participants	240 participants
Randomization Success	Comparing demographic and potential confounding characteristics at baselinebetween intervention and control groups	No differences between intervention and control condition at baseline	There were no significant differences between intervention and control conditions at baseline. However, gender minorities were slightly higher in the control group.
Retention Rate for T2	Total number of participants who completed T2 survey divided by the total number of participants enrolled in the study	≥ 80%	67.9% (95% CI: 61–74%)
Retention Rate for T3	Total number of participants who completed T3 survey divided by total number of participants enrolled in the study	≥ 80%	64.2% (95% CI: 58–70%)
Retention Rate in T2 and T3	Total number of participants that completed both T2 and T3 surveys divided by the total number of participants enrolled in the study	≥ 75%	58.8% (95% CI: 52–65%)
Game Demand (in Intervention Group)
Game Download	"Did you download the game titled 'Singularities'?" assessed on T2 and T3 surveys	≥ 80% selected "Yes"	55.8% (95% CI: 47–65%)
Any Game Play	"Did you play the game titled 'Singularities'?" assessed on T2 and T3 surveys	≥ 80% selected "Yes"	55.8% (95% CI: 47–65%)
Any Game Play	Total number of people who played the game based on the game-play data from the secure file-transfer-protocol system divided by the total number of participants randomized to intervention condition	≥ 80% played	41.7% (95% CI: 37–47%)
Total Time of Game Play	"In the past month, about how long did you play the game 'Singularities'?" assessed on T2 and T3 surveys	≥ 75% selected 1 hour or greater	68.2% (95% CI: 57–77%)
Total Time of Game Play	The number of hours the game was played based on game-play data from the secure file-transfer-protocol system	≥ 75% played 1 hour or more	63.3% (95% CI: 58–68%)
Game Acceptability (in Intervention Group)
Gamining Positive Affect	Subscale of Gaming Experience Questionnaire assessed on T2 or T3 surveys, measured using 5 items (e.g., "I thought it was fun") and a 5-point Likert scale (ranging from 0 "Not at all" to 4 "Extremely").	Mean score ≥ 2	2.36 (95% CI: 2.13–2.58)
Gaming Negative Affect	Subscale of Gaming Experience Questionnaire assessed on T2 or T3 surveys, measured using 4 items (e.g., "I felt bored") and a 5-point Likert scale (ranging from 0 "Not at all" to 4 "Extremely"). This subscale is reverse coded.	Mean score ≥ 2	2.75 (95% CI: 2.55–2.95)
Gaming Tension and Annoyance	Subscale of Gaming Experience Questionnaire assessed on T2 or T3 surveys, measured using 3 items (e.g., "I felt frustrated") and a 5-point Likert scale (ranging from 0 "Not at all" to 4 "Extremely"). This subscale is reverse coded.	Mean score ≥ 2	3.18 (95% CI: 2.98–3.39)
Gaming Competence	Subscale of Gaming Experience Questionnaire assessed on T2 or T3 surveys, measured using 5 items (e.g., "I felt skillful") and a 5-point Likert scale (ranging from 0 "Not at all" to 4 "Extremely").	Mean score ≥ 2	2.23 (95% CI: 2.04–2.43)
Gaming Sensory and Imaginative Immersion	Subscale of Gaming Experience Questionnaire assessed on T2 or T3 surveys, measured using 6 items (e.g., "It was aesthetically pleasing") and a 5-point Likert scale (ranging from 0 "Not at all" to 4 "Extremely").	Mean score ≥ 2	1.96 (95% CI: 1.71–2.22)
Game Flow	Subscale of Gaming Experience Questionnaire assessed on T2 or T3 surveys, measured using 5 items (e.g., "I was deeply concentrated in the game") and a 5- point Likert scale (ranging from 0 "Not at all" to 4 "Extremely").	Mean score ≥ 2	1.47 (95% CI: 1.22–1.72)
Desire to Play Game Again	"I would like to play this game again" assessed on T2 or T3 surveys	≥ 75% selected "Agree" or "Strongly Agree"	46.2% (95% CI: 34–58%)
Likelihood to Recommend Game to Friends	"How likely would you be to recommend that your friends play this game?" assessed on T2 or T3 surveys	≥ 75% selected "Definitely", "Very Probably", or "Probably"	50.8% (95% CI: 39–63%)

Note. T1: baseline; T2: first follow-up; T3: final follow-up; CI = confidence interval.

Our secondary outcomes were: help-seeking intentions (General Help-Seeking Questionnaire[29] assessing likeliness to seek help for emotional problems and suicidal ideation from different sources); help-seeking self-efficacy (two subscales from Multidimensional Scale of Perceived Self-Efficacy[30]); help-seeking behaviors (adapted from the Help-Seeking Behaviors Scale[31]); coping skill usage (two subscales from Adolescent Coping Scale Second Edition Short Form[32]); coping flexibility (two subscales from Coping Flexibility Scale[33]); and knowledge and use of web-based resources (investigator-created indices[26]).

Our tertiary outcomes were: bullying victimization (adapted version of University of Illinois Victimization scale[34]); cybervictimization (adapted from Cyberbullying Perpetration Scale[26,35]); loneliness (UCLA Loneliness Scale[36]); anxiety (Generalized Anxiety Disorder Child Age 11–17[37]); depressive symptoms (Patient Health Questionnaire-9 for Children Age 11–17[38]); suicidality (3 items from YRBS[39]); substance use (from YRBS[39]); internalized gender minority stigma (Transgender Identity Survey[40]); and internalized sexual minority stigma (adapted from Transgender Identity Survey[40]).

Our exploratory outcomes were focused on a more nuanced understanding of implementation procedures, intervention integration, and intervention adaptation and expansion outcomes.

Sample Size Justification

The sample size for the study was based on point estimation of primary feasibility outcomes rather than hypothesis testing[26]. For success of implementation procedures (e.g., retention rate at T2), assuming 240 participants and 5% type 1 error rate, we had the ability to estimate 95% confidence interval (CI) widths of no more than 0.13. For game demand among 120 intervention condition participants, we had the ability to estimate 95% CI widths of no more than 0.18.

Analyses

For our primary outcomes, we calculated frequencies, percentages, and 95% CIs for binary variables, and means and 95% CIs for continuous variables. All data analyses were conducted in StataSE 15 (College Station, TX).

For our secondary and tertiary outcomes variables, we used generalized linear mixed models, using link functions dependent on the outcome distribution, and a random effect for observations nested within persons. To explore the intervention effects, we used an interaction term of time by condition. We conducted three analyses: (1) an intent-to-treat analysis, wherein all participants were included according to the intervention they were randomized to (i.e., 0=control; 1=intervention) rather than the intervention they ultimately adhered to; (2) a self-reported intensity-adjusted intervention analysis, wherein the intensity-adjusted intervention variable was coded as 1.0 for those who reported completing the game (n_intervention=21), 0.5 for those who played but did not complete the game (n_intervention=46), and 0.0 for those who never played the game, were missing self-reported game play data, or were in the control condition (n_intervention=53; n_control=120); and (3) a game-reported intensity-adjusted intervention analyses, wherein the game-reported intensity adjusted intervention variable was coded as 1.0 for those with game-reported data of 45+ minutes (n_intervention=11), 0.75 for those with game-reported data of 27–44 minutes (n_intervention=15), 0.5 for those with game-reported data of 11–26 minutes (n_intervention=16), 0.25 for those with game-reported data of 1–10 minutes (n_intervention=8), and 0.0 for those who did not play, were missing game-reported data, or were in the control condition (n_intervention=70; n_control=120). Analyses controlled for potential confounding variables that were meaningfully different across the levels of the intervention variables. (We did not conduct multivariable models of suicide attempts because its low prevalence prevented models from converging; we also omitted analyses of perceived stress because the measures were incorrectly assessed in the T2 and T3 surveys.)

For exploratory outcomes, we reported frequencies and proportions for the quantitative outcomes. For the open-ended questions, we conducted qualitative content analysis. First, a group of investigators read participants’ responses and inductively identified the constructs reported within each question. Second, using these constructs, we developed a codebook with code names and definitions. Finally, 2 independent coders coded the qualitative data and summarized the findings in narrative format.

RESULTS

Primary Outcomes

Overall, 2153 individuals clicked the link to the screening questionnaire (Figure 1). In total, 988 individuals completed the screening questionnaire, of which 407 individuals met all eligibility criteria. Overall, 304 individuals consented to participate, and, as hypothesized, 240 participants completed the T1 survey and were randomized. Of those eligible, 59.0% were enrolled into the study (lower than our 80% hypothesis; Table 2). Half of the enrolled participants (n=120) were randomized into the intervention condition and half (n=120) into the control condition. Retention was 67.9% (n=163) at T2 and 64.2% (n=153) at T3, which were both slightly less than hypothesized. T2 retention was higher among control than intervention participants (75% versus 61%, respectively). T3 retention was similar among groups.

Figure 1. Consolidated Standards of Reporting Trials (CONSORT) flow diagram.

Note. T1: baseline; T2: first follow-up; T3: final follow-up.

Participants’ demographic characteristics are in Table 1. Regarding sexual identity, participants identified as gay/lesbian (45.0%), bisexual (18.3%), queer (10.4%), another identity (7.9%), or multiple sexual identities (18.3%). Gender minorities comprised 47.1% of the total sample followed by cisgender boys (36.7%) and cisgender girls (16.25%). The majority of the sample was White (62.1%) with the remaining participants identifying at Latinx (20.8%), Black (3.3%), Asian or Pacific Islander (3.8%), and multiracial (10.0%). There were no significant demographic differences between the intervention and control groups, as hypothesized.

Table 1.

Demographic Characteristics for the Total Sample and By Study Arm

Demographic Characteristic	Total Sample (N=240)		Intervention Group (n=120)		Control Group (n=120)
	n	(%)	n	(%)	n	(%)
Sexual Orientation
Gay/Lesbian	108	(45.0)	52	(43.3)	56	(46.7)
Bisexual	44	(18.3)	22	(18.3)	22	(18.3)
Queer	25	(10.4)	14	(11.7)	11	(9.2)
Another Identity	19	(7.9)	8	(6.7)	11	(9.2)
Multiple	44	(18.3)	24	(20.0)	20	(16.7)
Gender
Cisgender girl	39	(16.3)	18	(15.0)	21	(17.5)
Cisgender boy	88	(36.7)	39	(32.5)	49	(40.8)
Gender minority	113	(47.1)	63	(52.5)	50	(41.7)
Race/ethnicity
White	149	(62.1)	72	(60.0)	77	(64.2)
Latinx	50	(20.8)	28	(23.3)	22	(18.3)
Asian or Pacific Islander	9	(3.8)	2	(1.7)	7	(5.8)
Black	8	(3.3)	7	(5.8)	1	(0.8)
Multiracial	24	(10.0)	11	(9.2)	13	(10.8)
Age (years old)
14	35	(14.6)	16	(13.3)	19	(15.8)
15	66	(27.5)	33	(27.5)	33	(27.5)
16	71	(29.6)	37	(30.8)	34	(28.3)
17	56	(23.3)	28	(23.3)	28	(23.3)
18	12	(5.0)	6	(5.0)	6	(5.0)
Free or reduced-priced lunch
Yes	88	(36.7)	44	(36.7)	44	(36.7)
No	120	(50.0)	59	(49.2)	61	(50.8)
Unsure	32	(13.3)	17	(14.2)	15	(12.5)
Parent highest education level
Did not finish high school	24	(10.0)	11	(9.2)	13	(10.8)
Graduated high school	38	(15.8)	22	(18.3)	16	(13.3)
Attended college but did not complete	43	(17.9)	23	(19.2)	20	(16.7)
Graduated from college	130	(54.2)	62	(51.7)	68	(56.7)
Don’t know	5	(2.1)	2	(1.7)	3	(2.5)
Gender mannerisms/appearance, mean (sd)	3.13	(1.25)	3.10	(1.22)	3.16	(1.28)
Sexual minority structural stigma, mean (sd)	1.00	(2.68)	1.09	(2.74)	0.92	(2.64)
Gender minority structural stigma, mean (sd)	0.36	(2.41)	0.42	(2.45)	0.30	(2.38)
Sexual identity outness, mean (sd)	1.31	(1.00)	1.24	(0.95)	1.39	(1.05)
Gender identity outness, mean (sd)	1.20	(0.93)	1.24	(0.99)	1.16	(0.88)
Social support, mean (sd)	3.80	(1.16)	3.78	(1.19)	3.82	1.13

Note. sd = standard deviation

Regarding game demand, 55.8% of intervention participants self-reported downloading and 55.8% self-reported playing the game (Table 2; lower than our hypothesized thresholds). Of those that played, 68.2% reported playing an hour or greater, which exceeded our hypothesis. Game-playing participants exceeded hypothesized benchmarks and reported high positive affect (M=2.36; 95% CI=2.13–2.58), low negative affect (M=2.75; 95% CI=2.55–2.95), low tension/annoyance (M=3.18; 95% CI=2.98–3.39), and high competence (M=2.23; 95% CI=2.04–2.43) while playing the game (Table 2). The subscales of sensory/imaginative immersion and game flow were lower than hypothesized. Of those who played the game, 46.2% reported a desire to play it again and 50.8% would recommend it to friends.

Secondary and Tertiary Outcomes

Means and standard deviations of the secondary and tertiary outcomes are in Supplemental Table 1. Intent-to-treat models are in Table 3, and intensity-adjusted models are in Supplementary Table 2.

Table 3.

Intent-to-Treat and Intensity-Adjusted Analyses of the Intervention Effects for Secondary and Tertiary Outcomes of the Study

	Intent-to-Treat Intervention Effect
	1-month follow up			2-month follow up
Outcome	b	(95% CI)	P-value	b	(95% CI)	P-value
Secondary Outcomes
Help-seeking intentions for personal or emotional problems	−0.01	(−0.12, 0.09)	0.79	0.02	(−0.08, 0.13)	0.69
Help-seeking intentions for suicidality	0.04	(−0.10, 0.17)	0.59	0.02	(−0.12, 0.16)	0.79
Help-seeking self-efficacy for enlisting social resources	0.07	(−0.09, 0.22)	0.40	0.03	(−0.13, 0.18)	0.74
Help-seeking self-efficacy for enlisting parental and community support	−0.02	(−0.19, 0.15)	0.82	0.09	(−0.08, 0.26)	0.29
Help-seeking behaviors from parent/guardian	0.06	(−0.19, 0.31)	0.65	0.04	(−0.22, 0.30)	0.76
Help-seeking behaviors from relative or family member	0.12	(−0.19, 0.42)	0.46	0.13	(−0.18, 0.44)	0.42
Help-seeking behaviors from teacher	−0.01	(−0.32, 0.31)	0.97	−0.08	(−0.40, 0.23)	0.61
Help-seeking behaviors from friend	−0.04	(−0.36, 0.27)	0.79	−0.17	(−0.48, 0.15)	0.31
Help-seeking behaviors from mental health provider	−0.15	(−0.48, 0.19)	0.39	−0.32	(−0.66, 0.03)	0.07
Help seeking behaviors from help line	−0.13	(−0.37, 0.12)	0.30	−0.19	(−0.44, 0.06)	0.14
Help-seeking behaviors from doctor or nurse	−0.19	(−0.42, 0.03)	0.09	−0.02	(−0.24,0.21)	0.87
Productive (problem-solving) coping skill use	−0.07	(−0.24, 0.09)	0.40	0.01	(−0.16, 0.17)	0.95
Non-productive (passive avoidant) coping skill use	0.11	−(0.09, 0.30)	0.29	−0.11	(−0.31, 0.09)	0.29
Coping flexibility: Evaluation coping	−0.01	(−0.18, 0.16)	0.89	−0.04	(−0.21, 0.13)	0.65
Coping flexibility: Adaptive coping	−0.08	(−0.27, 0.12)	0.45	0.01	(−0.19, 0.21)	0.93
Knowledge of Web-based resources	−0.47	(−1.02, 0.08)	0.10	−0.38	(−0.92, 0.17)	0.18
Use of Web-based resources	−0.08	(−0.43, 0.26)	0.63	−0.23	(−0.57, 0.12)	0.20
Tertiary Outcomes
Bullying victimization	−0.04	(−0.28, 0.20)	0.76	0.13	(−0.12, 0.37)	0.31
Cyberbullying victimization	−0.28	(−0.56, −0.01)	0.05	−0.09	(−0.37, 0.19)	0.52
Loneliness	0.01	(−0.15, 0.18)	0.87	0.02	(−0.15, 0.18)	0.83
Anxiety symptoms	−0.19	(−0.42, 0.04)	0.11	−0.16	(−0.40, 0.08)	0.19
Depression symptoms	−0.03	(−0.22, 0.16)	0.76	−0.06	(−0.26,0.14)	0.55
Suicidal ideation, odds ratio (95% CI)	0.77	(0.22, 2.87)	0.70	0.51	(0.12, 2.20)	0.37
Suicide plan, odds ratio (95% CI)	2.15	(0.47, 9.74)	0.32	0.85	(0.20, 3.70)	0.83
Alcohol use frequency	−0.56	(−1.40, 0.28)	0.19	−0.52	(−1.38, 0.33)	0.23
Binge alcohol use frequency	−0.39	(−0.71, −0.06)	0.02	−0.13	(−0.46,0.20)	0.44
Cigarette smoking frequency	0.20	(−0.57, 0.96)	0.62	0.25	(−0.53, 1.03)	0.53
Cigarettes smoked per day	0.30	(−3.37, 3.97)	0.87	−0.62	(−4.37, 3.12)	0.74
Electronic cigarette use frequency	−0.54	(−1.92, 0.84)	0.44	−0.13	(−1.54, 1.27)	0.85
Marijuana use frequency	−1.37	(−3.04, 0.30)	0.11	−2.78	(−4.49, −1.08)	<0.01
Internalized gender minority stigma: Pride a	0.13	(−0.24, 0.50)	0.49	−0.19	(−0.56, 0.19)	0.34
Internalized gender minority stigma: Passinga	0.17	(−0.37, 0.71)	0.53	−0.05	(−0.60, 0.51)	0.87
Internalized gender minority stigma: Alienationa	−0.29	(−0.79, 0.21)	0.25	−0.12	(−0.63, 0.38)	0.63
Internalized gender minority stigma: Shamea	0.34	(−0.17, 0.85)	0.19	0.04	(−0.48, 0.56)	0.87
Internalized sexual minority stigma: Pride	0.13	(−0.13, 0.38)	0.32	0.17	(−0.09, 0.43)	0.19
Internalized sexual minority stigma: Passing	0.13	(−0.21, 0.47)	0.46	0.07	(−0.28, 0.42)	0.69
Internalized sexual minority stigma: Alienation	−0.07	(−0.48, 0.34)	0.73	−0.03	(−0.45, 0.39)	0.88
Internalized sexual minority stigma: Shame	0.05	Z(−0.29, 0.39)	0.79	0.18	(−0.17, 0.53)	0.31

Note. Boldface indicates p<0.05; CI = confidence interval

^aAmong gender minority youth only

Intent-to-treat Intervention Effect

From T1 to T2, the intervention group reported significantly larger reductions in cyberbullying victimization (b=−0.28; 95% CI:−0.56,−0.01) and binge alcohol use frequency (b=−0.39; 95% CI:−0.71,−0.06) compared with the control group. From T1 to T3, intervention participants reported greater reductions in marijuana use frequency (b=−2.78; 95% CI:−4.49,−1.08) than controls. There were no significant changes in 36 T2 outcomes and 37 T3 outcomes.

Self-Reported Intensity-Adjusted Intervention Effect

Participants who self-reported completing the game reported higher knowledge (b=0.92; 95% CI:0.03,1.81) and greater use of online resources (b=0.60; 95% CI:0.04,1.16) and lower binge alcohol use frequency (b=−0.50; 95% CI:−1.00,−0.01) at T2 compared to participants who reported not playing the game and controls. There were no significant changes for 35 T2 outcomes and all T3 outcomes.

Game-Reported Intensity-Adjusted Intervention Effect

According to the FTP game-play data, participants who played the game for 45 minutes or more (versus controls and those who failed to play the game) reported significantly lower cyberbullying victimization at T2 (b =−0.46; 95% CI:−0.92,−0.002). However, participants who played the game for 45 minutes or more reported lower scores in help-seeking behaviors with doctors (T2 b=−0.37; 95% CI:−0.73,−0.003) and help lines (T2 b=−0.52; 95% CL-0.92,−0.11; T3 b=−0.44; 95% CI:−0.85,−0.03). There were no significant changes in 35 T2 outcomes and 37 T3 outcomes.

Exploratory Outcomes

As shown in Table 4, it took 4 months to enroll all participants with Facebook and Instagram as the primary recruitment sites (n=2,146 clicks). Intervention participants who played the game (versus those who did not) had higher sexual identity outness scores (p=0.03) but did not differ by any other sociodemographics. The majority (81.5%) of intervention participants who self-reported playing the game never thought the intervention interfered with their regular activities, 73.4% were never or rarely concerned with safety, and only a minority thought password protection was very (18.8%) or moderately (14.1%) important. Just over half (55.4%) preferred that future gaming platforms included both computer and phone, 29.2% phone only, and 15.4% computer only. Most (87.5%) thought that other SGMY would enjoy the game. According to FTP data, 21 participants played the game for 45+ minutes. Loss-to-follow-up was similar by study arm (p=0.08) and only associated with one sociodemographic: participants who were ineligible (versus eligible/unsure) for free/reduced-price lunch were less likely to be lost to follow up (p=0.02).

Table 4.

Exploratory Outcomes, Research Questions, Assessments, Results of the Study

Domain and Outcome	Exploratory Question	Assessment	Result
Implementation Procedures
Enrollment Period	How long does it take to get 240 people to enroll in the study?	Measured as number of months between enrollment of first and last participant	4 months
Recruitment Venues	How many people were recruited from which venue?	Assessed using unique links that track the number of clicks on each advertisement	2146 from Facebook and Instagram, 3 from Reddit advertisement, 4 from Pitt+Me
Game Completion	How many intervention condition participants completed the game?	Assessed via game-play data from the secure filed-transfer-protocol system as well as T2 and T3 surveys: "Did you complete the game?"	21 participants
Integration
Email Address Access	How many youth were excluded because they were without an email address?	Assessed via the screening questionnaire: "Do you have an email address?"	32 youth
Ease of Download	Among intervention group: How easily was the game downloaded without contacting our research coordinator?	Assessed via the number of emails to our project email address.	4 emails about game download
Ease of Participation	How many times was the research staff contacted by participants with questions about surveys or intervention materials?	Assessed via the number of emails to our project email address.	28 emails about the project
Game Problems	Among intervention group: How many participants encountered problems with the game?	Assessed via T2 and T3 surveys: "Did you have any problems in the game?"	20.0%
Game Security	Among intervention group: How important was the use of password protection in the game?	Assessed via T2 and T3 surveys: "How important was it to have the game be protected by password?"	18.8% reported "very important" 14.1% reported "important" 12.5% reported "moderately important" 29.7% reported "slightly important" 25.0% reported "not important"
Game Privacy	Among intervention group: How safe did participants feel playing the game?	Assessed via T2 and T3 surveys: "How often were you concerned with other people seeing you play the game?"	73.4% reported "never" or "rarely"
Game Interference	Among intervention group: Did the gaming intervention interfere with participants' regular activities?	Assessed via T2 and T3 surveys: "During the pat 30 days, how many times did playing the game interfere with school, work, or other responsibilities (like being late, missing school, or making it hard to concentrate, etc.)?"	81.5% reported "never"
Adaptation/Expansion
Future Gaming	Among intervention group: Would	Assessed via T2 and T3 surveys: "In the future	55.4% reported "both my computer and phone"
Platforms	participants like to play the game on other platforms, such as phone?	where would you like to play this game?"	29.2% reported "my phone only" 15.4% reported "my computer only"
Future Appeal to Sexual and Gender Minority Youth	Will other sexual and gender minority youth enjoy the game?	Assessed via T2 and T3 surveys: "Do you think other LGBTQ people your age would like to play the game?"	14.1% reported "definitely" 21.9% reported "very probably" 23.4% reported "probably" 28.1% reported "possibly" 7.8% reported "probably not" 4.7% reported "definitely not"

For open-ended questions, intervention participants reported that the game provided valuable learning experiences, including the importance of getting support, where to find SGMY resources online, coping skills, conflict resolution, and how to overcome adversity to reach goals. A minority of intervention participants reported they did not learn anything new.

Several participants commented on how the game increased their capacity for acceptance of themselves and of others. Several participants “felt seen” or related to a character, which increased self-acceptance. Many participants appreciated the diverse team members and felt like the different identities and experiences helped them understand what challenges other people may face. The character development process, including pronoun, attire, and superpower selection, was reported as a highlight of the game but could be improved by having more customization options, including hair and identities.

The most common criticisms were glitches (computer crashes) and unaesthetic graphics. Participants suggested the game include additional features such as key control, multiplayer mode, and the capability to be played on a phone. Some participants reported the game was slow, tedious, or boring and sought more challenge, while others enjoyed the tasks, battles, and game pace.

DISCUSSION

These results suggest our game-based intervention and online trial design was feasible for SGMY. Particularly, game acceptability met our a priori benchmarks for success in several areas: gaming competence, positive affect, negative affect, and tension/annoyance. Qualitatively, participants appreciated the diverse identities and enjoyed the NPCs, tasks, battles, and pace of the game. The participants especially liked the character development process and encouraged more customization options. Additionally, online recruitment and randomization of 240 SGMY was successful and occurred in only four months.

The gaming intervention also showed promise as a method for improving health for SGMY which is important given the lack of evaluated interventions for these populations[16]. Because this pilot was a feasibility study, we were not powered to find significant effects for secondary and tertiary outcomes, so these outcomes must be interpreted cautiously. The intervention significantly reduced binge drinking frequency (intent-to-treat and self-report intensity-adjusted), marijuana use frequency (intent-to-treat), and cyberbullying victimization (intent-to-treat and game-reported intensity-adjusted). Additionally, the intervention increased knowledge and use of online resources for SGMY (self-reported intensity adjusted). Though the gaming intervention was not associated with changes in other secondary and tertiary outcomes, this study’s results suggest game-based interventions can be leveraged to improve certain skills and reduce certain negative health behaviors, thereby warranting future research on game-based interventions for SGMY health promotion.

Despite the many successes of this pilot, there are opportunities for improvement. The biggest concern from this study is that only half of the intervention group reported downloading/playing the game. Prior to implementing a larger trial, we will conduct formative qualitative research with SGMY about ways to increase game downloads and play, which may include clearer expectations of study participation, increasing accessibility through streamlining the processes, and providing encouragement, assistance, and incentive during the game download/play phases. Additionally, retention rates at each timepoint were lower than hoped (75–80%) but we were able to maintain about 60% retention over time, suggesting the need for additional re-engagement measures to improve retention. Participants made suggestions for improvement including fixing glitches, improving the graphics, and adding features. Feedback on game pace and challenge was mixed suggesting we need to be more intentional on what age levels are targeted and providing multiple challenge levels. A hybrid effectiveness- implementation study may be warranted to better understand implementation barriers/facilitators in conjunction with conducting a larger RCT.

Limitations

Despite its many strengths, our study is not without limitations. Our study was a feasibility trial not powered to find differences in secondary/tertiary outcomes, and we tested numerous outcomes which increases potential for Type I error; nevertheless, these feasibility data can be used to design a fully powered trial. Our study enrolled SGMY from across the U.S. but generalizability is limited because we recruited from Facebook and Instagram. Though large proportions of youth use these social media platforms, we may have over-enrolled gender minority youth and under-enrolled Black, Indigenous, Asian, and other youth of color, and youth whose parents did not finish high school. The reasons for under-enrollment of specific groups are unclear, but worthy of future attention. Our study’s retention rates and reported intervention use were less than optimal, increasing the likelihood of attrition bias and exposure bias, respectively, and highlighting the importance of engaging in additional and innovative online retention strategies and intervention uptake strategies. Our study may have social desirability bias. Despite using primarily validated self-report measures, in most instances the scales were not specifically validated among SGMY, thereby potentially leading to measurement bias.

Conclusions

Web-accessible game-based interventions are feasible and show promise for improving SGMY health, a priority research area by the National Academies of Medicine[15]. SGMY-specific game-based intervention trials can overcome barriers of traditional face-to-face SGMY-specific interventions, such as the unintentional outing of SGMY, recruitment of SGMY in rural or stigmatizing environments, and discontinuation of in-person programs due to novel infectious disease pandemics like COVID-19. Our trial found many positive results—such as decreased binge drinking and marijuana use—as well as opportunities for improvement which can inform and be addressed in future game-based interventions aimed at forging SGMY health equity.

IMPLICATIONS AND CONTRIBUTION.

This study found that a game-based intervention and online randomized controlled study design was feasible for sexual and gender minority youth. This gaming intervention showed promise as a method for improving health for sexual and gender minority youth, which is important given the lack of evidence-based interventions for these populations.