An iterative approach to optimization of a behavioral intervention to reduce STIs on college campuses Kari Kugler & David Wyrick OBBI Training May 20th, 2016

Outline Rationale for itMatters Preparation phase Optimization phase Evaluation phase Practical considerations Summary

Outline Rationale for itMatters Preparation phase Optimization phase Evaluation phase Practical considerations Summary

An important health issue College students have high rates of STIs Sexual risk-taking is common Inconsistent condom use Multiple concurrent partners Casual sex (hook-ups) Alcohol use is associated with Increase in number of partners and hook-ups Decrease in condom use

Numerous factors of influence Proximal factors Perceived norms Expectancies Perceived benefits Self-efficacy Contextual factors Gender Relationship status Campus environment

Few interventions target the intersection Combine separate components Personalized normative feedback Most are in online format Limitations Focus on changing norms only Diversity of students (race/ethnicity) Only sexually active students

What is the goal of itMatters study? To engineer an optimized online intervention to prevent sexually transmitted infections among college students Aimed uniquely at the intersection of alcohol use and sexual risk behaviors Every component of itMatters has an empirically detectable effect Effectiveness of itMatters approaches that of other STI preventive interventions

Interdisciplinary team of investigators

Outline Rationale for itMatters Preparation phase Optimization phase Evaluation phase Practical considerations Summary

itMatters applied to MOST Preparation phase Component 1 Component 2 Component 4 Component 3 Component 5

Conceptual model Situational Context Appraisal of sex potential & risk Sexual arousal Cognitive impairment/disinhibition Sexual Risk Behaviors STIs Moderators Gender Other individual factors Relational Institutional context Alcohol Use Protective Behavioral Strategies Descriptive Norms (e.g., perceptions of how many people have hook-ups due to alcohol use) Injunctive Norms (e.g., perceptions of acceptability of sex after drinking alcohol) Proximal Factors Conceptual Pathway Self-efficacy to Prevent Harm (e.g., self-efficacy to negotiate condom use when intoxicated) Outcome Expectancies (e.g., expect alcohol use to enhance or diminish sex experience) Perceived Benefit (e.g., perceived benefits of using PBSs related to alcohol and sexual behaviors) We have had 25 other iterations of this model to try to get it right!

Identification of components Situational Context Appraisal of sex potential & risk Sexual arousal Cognitive impairment/disinhibition Sexual Risk Behaviors STIs Moderators Gender Other individual factors Relational Institutional context Alcohol Use Protective Behavioral Strategies Descriptive Norms (e.g., perceptions of how many people have hook-ups due to alcohol use) Injunctive Norms (e.g., perceptions of acceptability of sex after drinking alcohol) Proximal Factors Conceptual Pathway Self-efficacy to Prevent Harm (e.g., self-efficacy to negotiate condom use when intoxicated) Outcome Expectancies (e.g., expect alcohol use to enhance or diminish sex experience) Perceived Benefit (e.g., perceived benefits of using PBSs related to alcohol and sexual behaviors) Intervention Components Perceived Benefits

Intervention components Develop online modules for five candidate intervention components Descriptive norms Injunctive norms Expectancies Perceived benefits Self-efficacy Component levels: On/Off

Program modules Descriptive norms Injunctive norms Correct misperceptions about alcohol use and sexual behaviors with and without alcohol Strategy: Personalized feedback comparing prevalence to own behaviors Injunctive norms Correct misperceptions about other college students’ approval of alcohol and sexual behaviors (with and without alcohol) Strategy: Personalized feedback perceptions of approval to actual approval

Program modules Outcome expectancies Perceived benefits Increase understanding that expectancies drive behavior and alcohol shouldn’t be given the “credit” Strategy: bar lab simulation and visual representation of how alcohol effects key elements for fun and safe sex Perceived benefits Increase perceptions of perceived benefits of using protective behavioral strategies (PBS) and value of reducing risk Strategy: Valuation of reducing risk of STI and risk assessment of vaious PBS

Program modules Self efficacy to use protective behavioral strategies Increase self-efficacy to use alcohol and sex protective behavioral strategies Strategy: Modeling; Know boundaries, have a plan, practice Knowledge (everybody gets) Increase knowledge about alcohol impairment, condom use skills, STIs Strategy: Decisional balance activity; interactive presentation of important facts

Module development This has been an iterative process working with: eLearning designers External Advisory Panel Students

Example module

Optimization criterion Include only “active” components in the optimized intervention “Active” = d≥.15 How did we decide on this? Meta-analyses suggest alcohol and sexual risk reduction interventions have overall effects of d ~.35 .35/5 components = .07

Outline Rationale for itMatters Preparation phase Optimization phase Evaluation phase Practical considerations Summary

Optimization phase Goal: to have the most effective STI preventive intervention possible Constraints: to include only “active” components Use MOST to accomplish this …taking an iterative approach!

Optimized intervention Optimization phase Revise components showing d<.15 Descriptive norms component Optimized intervention Injunctive norms Expectancies Select best version of each component where d≥.15 Perceived benefits Self efficacy Screening experiment 2 Screening experiment 1 Optimization phase

Experimental Conditions in 25 Factorial Design Screening experiment Experimental Conditions in 25 Factorial Design Experimental Condition Intervention Components KNOW DNORMS INORMS EXPECT BENEFITS SELFEFF 1 Include 2 √ 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 Study design: 25 full factorial Within school randomization 4 schools: public universities, historically black colleges and universities, rural and urban Mode of delivery: online

Intervention implementation Randomize 1st year students to experimental condition Block on gender After last data collection, students will be given access to any lessons that they were not originally assigned

Data collection Students will be asked to provide consent before completing online surveys Survey 1 (pre-test) Survey 2 (immediate post-test) Survey 3 (30-day follow-up)

Measures Proximal factors/mediators will be used as primary outcome of the screening experiment In general, what percentage of students at your college do you believe had penetrative sex during a hookup that also had alcohol before or during the hookup? Measuring intervention effect on mediators is different than effect on behavioral or health outcomes Need to develop our own measures that (a) target intersection and (b) capture intended change

Analytic approach Use a regression approach to ANOVA Assess main effect of each component on hypothesized proximal mediator Hypothesis: The effect of receiving the self-efficacy component on self-efficacy to use protective behavioral strategies was greater than or equal to d≥.15, averaged across the other components. Conduct analyses separately for males and females

Decision making Any component that has effect d<.15 will be revised Informed by focus group data Input from External Advisory Panel Input from eLearning experts Any component that has d≥.15 will not be revised

2nd screening experiment We get a second shot at improving the components to make them more efficient and effective! Use same procedures as for 1st screening experiment Randomize, collect data, and analyze the results

Decision making If component has an effect of d<.15 and does not synergistically interact with other components → do not include in optimized intervention…! This means that we may not bring forward 5 components If component has an effect of d≥.15 and/or synergistically interacts with other components → include in optimized intervention.

Outline Rationale for itMatters Preparation phase Optimization phase Evaluation phase Practical considerations Summary

Optimized intervention Evaluation phase Revise components showing d<.15 Descriptive norms component Optimized intervention Injunctive norms Expectancies Select best version of each component where d≥.15 Perceived benefits Self efficacy Screening experiment 2 Screening experiment 1 Evaluate via RCT Evaluation phase

Evaluation phase Evaluate optimized intervention compared to delayed control condition Use same participating universities Randomize students to experimental condition (optimized intervention v. delayed control) Assess long-term outcomes (i.e., behaviors, STIs)

Outline Rationale for itMatters Preparation phase Optimization phase Evaluation phase Practical considerations Summary

Practical considerations Critical to have a partnership with the participating universities Incentives for screening experiment and RCT needed to be different

Grantsmanship considerations First submission critiques: Include an RCT to evaluate intervention Response: Removed one of the three fully powered screening experiments to free up time and resources to conduct an RCT. Provide rationale for proposed research to target mediators already known to be important Response: Targeting mediators at the intersection of alcohol and sexual risk behaviors is understudied. MOST is being used to engineer an intervention that will target some of the same mediators as previous work but will be more effective.

Outline Rationale for itMatters Preparation phase Optimization phase Evaluation phase Practical considerations Summary

Summary Using an iterative approach to optimization Ensures each component is achieving an effect Maps well on to go/no-go decision making Good approach for new intervention development Collecting data to generate new hypotheses Mediation analyses to better understand process of change Moderation analyses to identify subgroups for whom intervention is working best

Acknowledgments Funders: NIAAA R01 AA022931 Penn State colleagues: UNCG colleagues: Linda Collins Amanda Tanner Jeff Milroy Brittany Chambers Alice Ma X

Thank you! An iterative approach to optimization of a behavioral intervention to reduce STIs on college campuses