The Methodology Center ADVANCING METHODS, IMPROVING HEALTH

Annual Report 2016-17

1 Our mission From the Director is to advance public health by improving experimental design and data analysis in For the past few years, The Methodology In this year’s Annual Report, we highlight a the social, behavioral, and health sciences. Center has been moving into an exciting few of our most productive collaborations, new research area. Mobile health, or and we introduce a few of our stellar trainees. We draw upon and integrate methodological mHealth, is the use of mobile devices As these demonstration projects and our perspectives from a variety of disciplines, such as smartphones to measure former trainees spread out across the including , engineering, health characteristics and deliver health country, so do the reach and impact of interventions. The promise of mHealth is the methods we develop. psychology, and human development, the ability to deliver individually tailored to develop new quantitative methods for interventions to large numbers of people at I hope you enjoy learning about our work. research focusing on vital public health relatively low cost. mHealth studies generate huge amounts of complex data, calling issues, especially drug abuse and HIV. for new methods to maximize the amount of scientific knowledge gained from this research. Linda M. Collins Director, The Methodology Center We are developing methods for the analysis Distinguished Professor, Human Development of data from mHealth studies and for the and Family Studies empirical design of mHealth interventions. Professor, Statistics We collaborate with many of the nation’s top researchers in fields such as smoking cessation, opioid-addiction recovery, autism, cardiovascular health, and HIV/AIDS. These collaborations produce important scientific advances that improve public health and serve as demonstration projects so that other researchers in the relevant fields can see the value of our methods.

TO LEARN MORE ABOUT THE METHODOLOGY CENTER, VISIT methodology.psu.edu

2 3 Our Largest Research Projects

Susan Murphy (right) collaborates with Inbal “Billie” Namum-Shani (left) and Daniel Almirall (center) on research into JITAIs. Almirall and Nahum- Linda Collins (center) collaborates with Kari Kugler (right) and Kate Bethany Bray (right) and Stephanie Lanza (left) lead Center research Stephanie Lanza (left) works to disseminate and expand TVEM, a Runze Li leads the statistical research on variable screening and Shani lead research on SMART, which is based on work by Murphy. Guastaferro (left) on the dissemination and application of MOST. on mixture models, including LCA. statistical method developed by Runze Li (right). selection in massive data sets.

Just-in-Time Adaptive Intervention Sequential, Multiple Assignment, Multiphase Optimization Strategy Latent Class Analysis (LCA) Time-Varying Effect Modeling Variable Screening & Selection (JITAI) Randomized Trial (SMART) (MOST) LCA allows researchers to detect (TVEM) In genetic studies there may be hundreds A JITAI determines when an intervention is SMART experimental designs provide high- MOST is a framework for engineering unobservable (latent) subgroups within a TVEM is a flexible tool that can answer of subjects and hundreds of thousands needed based on data that are collected quality data for constructing empirically effective and efficient behavioral and population. By expanding LCA models, we questions about associations over time. of variables. Nearly all techniques for actively (like smartphone prompts that inquire validated adaptive interventions. Adaptive biobehavioral interventions. MOST is make it possible for intervention scientists It can be used with multiple data types, exploratory data analysis break down when about mood) or passively (like smartphone interventions are treatments that are an iterative process with three phases: to better target the subgroups who will including intensive longitudinal data, panel the number of variables exceeds the sample GPS data that indicate when a person is individually tailored to meet a patient’s preparation, optimization, and evaluation. benefit the most. Over twenty years, we have data, and cross-sectional data. TVEM allows size. High-dimensional variable-screening traveling in a vehicle). Principal Investigator changing needs. A SMART can be used to By following MOST, intervention designers developed longitudinal extensions of LCA, researchers to uncover changes in the procedures allow researchers to narrow Susan Murphy and her team are developing create adaptive interventions that improve can understand which aspects of an methods for LCA with a distal outcome, and relationships between variables. It enables the subset of variables for analysis. We microrandomized trial designs and machine outcomes while decreasing patient burden. intervention are working. By emphasizing other extensions of LCA. Our LCA software researchers to model those relationships developed two SAS procedures that enable learning algorithms to create JITAIs that Dozens of studies around the world are using careful resource management, MOST can for SAS and Stata has allowed thousands of without assuming a parametric relationship researchers to select relevant variables from are responsive to each person’s conditions SMART designs to address a wide of maximize the gain of scientific information. researchers to perform LCA on their data. between the variables. large data sets. and that learn over time what is effective or health problems, including drug dependence, 2015-16 Highlight: Release of Stata functions 2016 Highlight: Sixteen peer-reviewed articles ineffective for each person. ADHD, alcoholism, and autism. 2016 Highlight: Collaboration with Center for 2015-16 Highlight: A year of rapid growth for Drug Use and HIV/HCV Research at NYU on new to perform bootstrap likelihood ratio test and to TVEM, capped by a standing-room-only session on published or accepted for publication 2016 Highlight: Susan Murphy elected to the 2015-16 Highlight: NIDA R01 grant funding five grant to increase treatment engagement among estimate the association between a latent class TVEM at the Society for Prevention Research 2016 National Academy of Sciences more years of research to develop SMART minorities living with HIV variable and a distal outcome Annual Meeting

4 5 The Methodology Center collaborates with top researchers to address a broad variety of health problems. Collaborations Around the Nation Here are a few examples of the depth and breadth of our collaborations.

Mary Larimer, 1. Center-to-Center 5. Center-to-Center Irene Geisner collaboration: University collaboration: Center for University of Washington of Wisconsin Center for Drug Use and HIV/HCV Brittany Rhoades Cooper Tobacco Research and Research at New York University. Dr. Marya Gwadz Alex Rothman Washington State University Pedrag Klasnja University of Minnesota Vedat Verter Intervention (UW-CTRI). and others are working with Linda Collins to build an Group Health McGill University Drs. Michael Fiore, Timothy Baker, Megan Piper, and optimized intervention that will improve treatment Seattle Center for Tobacco others at UW-CTRI became the first research group engagement among African-American and Hispanic Research and Intervention to follow the multiphase optimization strategy in a people living with HIV/AIDS. University of Wisconsin Sam Bacharach, Lisa Marsch large-scale smoking intervention. Their first round 1 Peter Bamberger Dartmouth College of funding resulted in three with 80 6. Dr. Lisa Dierker of Wesleyan Cornell University University different conditions in the field simultaneously. They University collaborates with Stephanie Bonnie Spring Lisa Dierker 4 of Michigan 6 Lanza and Runze Li to apply novel The Methodology Center Wesleyan University are now refining the intervention. Northwestern University analytic methods to uncover the Center for Drug Use and HIV/HCV Research Robin Mermelstein 2. Dr. Hendrée Jones of The University mechanisms that lead young people to 5 New York University University of Illinois, Chicago of North Carolina at Chapel Hill works start and escalate smoking. Penn State Kevin Lynch, Jim McKay with Susan Murphy to implement Michael Dennis, 7 University of Pennsylvania 7. Drs. Roger Meyer and Christy Scott Abigail Norris sequential, multiple assignment, Chestnut Health Turner, Alison Roger Meyer, Scott Bunce randomized trials to develop innovative, Scott Bunce of Penn State Systems Norris, Penn State College of Medicine College of Medicine are John Davis effective, cost-effective ways to prevent drug abuse Ohio State Michael Mason among pregnant women. working with Stephanie Lanza University Virginia Commonwealth University and others at Penn State to 2 Hendrée Jones 3. Dr. Connie Kasari of The University improve treatment outcomes for patients recovering University of North Carolina of California Los Angeles works with from opioid dependence. By following patients Santosh Kumar, David Wyrick, Jeffrey Milroy, Amanda Tanner Daniel Almirall and Susan Murphy for seven months, researchers will track changes 3 MD2K University of North Carolina, Greensboro to implement sequential, multiple in the patients’ brain reward systems from initial Connie Kasari University of Memphis Katie Witkiewitz assignment, randomized trials to detoxification through residential care and a follow- University of California Los Angeles University of New Mexico develop the most appropriate adaptive up period. Eric Heckler Arizona State University interventions for children with autism spectrum disorder who are minimally verbal. 8. Dr. Guillermo “Willy” Prado of the University of Miami School of Medicine 4. Dr. Bonnie Spring of Northwestern and Bethany Bray are using latent

Mildred Maldonado-Molina University collaborates with Inbal “Billie” class analysis to investigate differential University of Florida Nahum-Shani, Linda Collins, and Susan treatment effects of the Familias Unidas Murphy on several projects that combine intervention, a culturally sensitive program to prevent technology with innovative experimental substance abuse and HIV-risk behavior among Guillermo “Willy” Prado designs in order to promote weight loss. Hispanic youth. 6 University of Miami School of Medicine 7 8 mHealth

a smartphone that contains a special app. identified five for a full year. During that time apps will mHealth: Always There When It’s Needed The app prompts participants six times per times throughout deliver more than 300 suggestions to each Smartphones and mobile sensors (like Fitbits) can collect massive amounts of data, and these data can be used to help people change their day to answer questions about a range of the day when participant. This that scientists must behavior to achieve their long-term goals. Information, including location, mood, activity, and stress, can all be obtained unobtrusively or passively. variables, including their level of craving, risk people are most reduce dropout by improving the quality, Methodology Center researchers are working to expand and improve methods for mHealth studies and interventions so that mHealth can provide factors for substance use, level of physical likely to engage in variety, and salience of the HeartSteps maximum benefit to the highest number of people. pain, and exposure to drugs and/or alcohol. physical activity. suggestions, relative to other apps. In initial Using data provided by the participant and The HeartSteps pilot testing, some participants reported Measuring Drinking bracelets that with each participant’s preferences and data passively collected by the phone, the app uses passively feeling like they had a relationship with the From to Moment measure blood characteristics to select an appropriate future app might deliver helpful risk-reducing collected data to app when messages were encouraging alcohol content intervention whenever and wherever it is suggestions to the person whenever high determine how to and appropriate. When, however, they Historically, studies of heavy drinking have (BAC) through needed. Realistically, a clinician can only risk is indicated. The data can also be used tailor and whether received the same message multiple relied on self-reports, but there are several Answering a question in an contact with the check in with participants at appointed times, to improve scientists’ understanding of or not to provide times, they reported feeling betrayed. problems with people’s reports of their own mHealth app skin. By applying but people generally carry their smartphones, how pain affects relapse: Investigators will tailored physical HeartSteps requires a massive number of drinking. First, heavy drinking disrupts the An alcohol monitor used in the innovative methods and those phones contain a lot of information assess whether and how pain moderates the activity suggestions. The suggestions are unique messages, so the investigators are memory. Second, drinking games, which are BAC study to these two data about when each person is or is not at risk. effect of the interventions on substance-use tailored based on the participant’s current “crowdsourcing” early phases of message often part of heavy drinking episodes among types, Russell will investigate the accuracy By leveraging these data, JITAIs can help craving. Investigators will also examine other location, time of day, quality of the weather, authoring using the Amazon Mechanical Turk young adults, can obscure the amount of of using EMA self-reports as a proxy for people exercise more, avoid drug use, stop complicated relationships, such as whether and day of the week. Also, suggestions website. Any improvement to the dropout alcohol ingested. Finally, it can be hard to actual BAC. He will also seek to identify time smoking, and more. recommending craving-reduction exercises are not delivered to participants if they are rate could be a benefit to a broad range of figure out the concentration of alcohol in a points during a drinking episode when an during the day has an impact on substance- already exercising. By rating each prompt, future interventions. particular drink. Only recently has technology intervention delivered via smartphone might Assisting people in recovery who use craving in the evening. The researchers participants help improve the quality of future become available to measure the accuracy of As mHealth becomes a more important curb heavy drinking and the associated are on the verge of lapse will develop methods for building JITAIs, suggestions for all participants. self-reported data on drinking. tool for helping people manage behavior, negative health outcomes. Once this pilot Susan Murphy is collaborating create an adaptive intervention, and increase with Drs. Christy Scott Methodology Center researchers will New data + new methods = new answers project is complete, Russell intends to apply knowledge about the way pain management Solving the mHealth and Michael Dennis of influences participants’ ability to avoid continue to explore and develop new A new study at the for a larger grant to extend the research and Dropout Problem Chestnut Health Systems on substance use lapses. methods to improve the quality of mHealth Methodology Center by develop a JITAI. improving an intervention for One of the barriers to widespread use of research and care. By combining new types Michael Russell combines individuals with a history of Helping cardiac patients get moving mHealth apps is that dropout rates are of data with the appropriate methods, ecological momentary Intervention in Your Pocket Susan Murphy substance use disorders. The In another new study, HeartSteps, famously high: Around 75 percent of mHealth science will help people manage drug use, assessment (EMA) of self- In just-in-time, adaptive interventions (JITAIs), intervention provides individuals recovering researchers are developing a JITAI that apps have been abandoned by the tenth smoking, physical activity, eating, and other reported alcohol use with Michael Russell continuously collected data from a mobile from substance abuse or dependence with encourages physical activity among people use. The HeartSteps project includes two behaviors that influence healthy living. continuous data from ankle device and/or wearable sensor are combined recovering from a cardiac event. Researchers pilot phases, and following those it will run

8 9 Trainees

Jennifer Kam Michael Cleveland Donna Coffman Megan Patrick The Future of Prevention and Methodology PAMT: 2007–2009 PAMT: 2005–2007 PAMT: 2005–2007 PAMT: 2005–2007 The Prevention and Methodology Training (PAMT) program trains doctoral students and postdoctoral researchers in integrating innovative Assistant Professor, Department Associate Professor of Human Assistant Professor of Research Associate Professor, research methods and drug abuse and HIV prevention science. The National Institute on Drug Abuse (NIDA) has funded PAMT continually since of Communication, University of Development, Washington State and , Temple University Institute for Social Research, 2005. In 2016, NIDA awarded five more years of funding to the program. Over its first eleven years, fifty-three young prevention researchers and California–Santa Barbara University Dr. Coffman’s current methodologists were trained through PAMT. PAMT is a collaboration with our sister center at Penn State, the Bennett Pierce Prevention Research Dr. Kam’s research Dr. Cleveland researches work focuses on Dr. Patrick’s research uses the stress-coping- innovative methods advancing and applying focuses on the Center, and is directed by Linda Collins and Denni Fishbein. Below are some of the many excellent scientists we have worked with in PAMT. resilience framework to to study adolescent statistical learning development and investigate experiences and young adult methods for large, consequences of Phylicia Bediako Megan Schuler Daniel "Max" Crowley Melissa Lippold related to ethnic/racial substance use, with an complex data to address adolescent and young PAMT: 2014–2016 PAMT: 2014–2015 PAMT: 2010–2012 PAMT: 2008–2010 identity that may place emphasis on alcohol cutting-edge research adult risk behaviors, Postdoctoral Researcher, University Marshall J. Seidman Fellow in Health Assistant Professor in Human Assistant Professor, School of Social adolescents at risk use and abuse. His questions related to health behaviors. She including alcohol use, of Miami Care Policy, Harvard Medical School Development and Family Studies, Work, University of North Carolina– for adverse psychological and behavioral methodological interests include latent has been the principal investigator of multiple drug use, and risky sexual behaviors. She Dr. Bediako’s research Dr. Schuler’s research Penn State Chapel Hill health outcomes. Her goal is to identify transition analysis, structural equation NIH grants and project director of a scientific works to improve the prevention of health is focused on sexual focuses on the Dr. Crowley investigates Dr. Lippold researches messages that can discourage substance modeling, and multi-level modeling. He component on The Methodology Center’s risk behaviors, statistical methods for and reproductive health application of advanced early education and the the role that parent- use, particularly when adolescents face has served as a co-investigator on many P50. Currently, she has a grant through the modeling behavior and behavior change, issues for underserved statistical methods to prevention of health youth relationships stressors related to ethnic/racial identity grants from the National Institutes of Health. Big Data to Knowledge (BD2K) initiative to and web-based . Dr. populations in global study the effectiveness inequalities and criminal play in the prevention and immigration/acculturation processes. In Dr. Cleveland recently began a National receive training and conduct research in Patrick has been principal investigator on settings. Her most of substance use behavior through of adolescent risky 2016, she received the Early Career Award Institutes of Health-funded project to adapt computational methods for analyzing data eight NIH grants and is a co-investigator current projects are treatment, barriers to evidence-based resource behavior and the from the Interpersonal Communication a parent-based intervention to better serve from wearable sensors. She received the on several others, including “Monitoring centered around sexual decision making and treatment, and comorbidity of substance investments in childhood promotion of adolescent Division at the National Communication African American youth and their families. Early Career Award from the Society for the Future” (funded by NIDA). She was risk taking among low-income youth in South use and mental health disorders. She and adolescence. In 2013, he was invited physical health. She also investigates how Association–the premier association in her He has published forty-three peer-reviewed Prevention Research in 2014, and she has awarded the University of Michigan Research Africa. She has published two peer-reviewed has published ten peer-reviewed papers to become a National Bureau of Economic parent and child characteristics influence field. Dr. Kam has published twenty-eight papers (ten first-authored) in journals such published over fifty peer-reviewed papers Faculty Recognition Award in 2015. She has papers and currently has two papers under (seven first-authored) in journals such as Research Crime Research Fellow. He has variability in parenting behaviors. She has peer-reviewed papers (nineteen first- as Addictive Behaviors, Journal of Studies (seventeen first-authored) in journals such published eighty-eight peer reviewed papers review (both first-authored) at journals American Journal of Epidemiology, Drug published twelve peer-reviewed papers published nineteen peer-reviewed papers authored) in journals such as Communication on Alcohol and Drugs, Alcoholism: Clinical as Statistics in Medicine, Epidemiology, (forty-seven first-authored) in journals such such as Journal of Racial and Ethnic Health and Alcohol Dependence, Health Services (eight first-authored) in journals such as (thirteen first-authored) in journals such as Research, Journal of Communication, and Experimental Research, and Prevention Childhood Obesity, Psychological Methods, as Prevention Science, Drug and Alcohol Disparities and Journal of Immigrant and and Outcomes Research Methodology, and Journal of Adolescent Health, Prevention Journal of Research on Adolescence, Journal Prevention Science, and Journal of Research Science. and Prevention Science. Dependence, JAMA Pediatrics, and Journal of Minority Health. Psychiatric Services. Science, and Criminology & Public Policy. of Family Psychology, Journal of Adolescent on Adolescence. Adolescent Health.

Health, and Prevention Science.

postdoctoral predoctoral % National Research + + fellows fellows Service Awards from completed completed the National Institutes 53 90 of Health 700 former trainees 16 training 37 training of trainees go on to research careers 8 400peer-reviewed publications conference presentations 10 11 Staying Connected

Each month, we gather for a meeting where we share our achievements over the past month. Each individual sets his or her own short and long term goals, and we track everyone’s progress. This lets us share each other’s successes and stay in touch with what is going on around The Methodology Center. And for fun, when we meet our goals, we get a fresh baked cookie!

OUR FUNDING Research at the Methodology Center is funded by grants from the National Institutes of Health, primarily the National Institute on Drug Abuse, the National Cancer Institute, the National Institute of Diabetes and Digestive and Kidney , and the National Institute on Alcohol Abuse and Alcoholism. The Methodology Center also receives significant support from Penn State’s College of Health and Human Development.

TO LEARN MORE ABOUT THE METHODOLOGY CENTER, VISIT methodology.psu.edu

This publication is available in alternative media on request. The University is committed to equal access to programs, facilities, admission, and employment for all persons. It is the policy of the University to maintain an environment free of harassment and free of discrimination against any person because of age, race, color, ancestry, national origin, religion, creed, service in the uniformed services (as defined in state and federal law), veteran status, sex, sexual orientation, marital or family status, pregnancy, pregnancy-related conditions, physical or mental disability, gender, perceived gender, gender identity, genetic information, or political ideas. Discriminatory conduct and harassment, as well as sexual misconduct and relationship violence, violates the dignity of individuals, impedes the realization of the University’s educational mission, and will not be tolerated. Direct all inquiries regarding the nondiscrimination policy to Dr. Kenneth Lehrman III, Vice Provost for Affirmative Action, Affirmative Action Office, The Pennsylvania State University, 328 Boucke Building, University Park, PA 16802-5901; Email: [email protected]; Tel 814-863-0471. U.Ed. HHD 17-018 MPC139350