Harvesting data from a 20,000-strong study cohort
What are the social, economic, and biological factors during adolescence and young adulthood that lead to chronic disease and poor health outcomes later in life? This is the kind of question that can only be answered by a longitudinal study—that is, a study that tracks a representative cohort of individuals over years, or even decades, assessing their health status at periodic intervals and drawing carefully weighted conclusions from the results.
One of the most influential projects in the social and behavioral sciences, the National Longitudinal Study of Adolescent to Adult Health (Add Health) has been cited in more than 3,000 research papers, and its reams of data have been pored over by tens of thousands of researchers in the United States and abroad. Add Health began in 1994 (when it was known simply as the National Longitudinal Study of Adolescent Health) with a detailed survey administered to a sample of more than 20,000 adolescents in grades 7-12 across the nation. Since then, subsequent data-collection “waves” have tracked this enormous cohort all the way from adolescence to their late 30s and early 40's..
RTI International was involved in Add Health during Wave III, when the cohort was between 18 and 26 years old, and Wave IV, when the cohort was between 24 and 32 years old. Now we are in the middle of collecting data for Wave V, in which the Add Health participants are between 32 and 42 years old. Wave V is focused on the early-life precursors of chronic disease: that is, the social, biological, genetic and behavioral pathways that lead to diabetes, heart disease and kidney failure, among other chronic conditions.
The Fifth Wave: The Challenge of Multimodal Data Collection
Collecting detailed information from nearly 20,000 people across the United States—as well as a few who have moved abroad—is necessarily a labor-intensive process. In previous waves, all or most of the data were collected in-person, but in Wave V the study moved to a multimodal technique, which reduces costs and increases efficiency without affecting data.
We divided the 20,000-strong cohort into three samples. In 2016, Sample 1, consisting of 7,931 respondents, either filled out a web survey or a paper survey (either of which took about 50 minutes to complete). In 2017, Sample 2a (2,716 respondents) also had the option of choosing between web and paper surveys, while the 1,550 respondents in Sample 2b were interviewed in-person, in their homes.
Why the split in Sample 2? Add Health researchers want to assess “mode effects”—the possibility that the medium used to ask a question can affect the way it is answered. Since Waves I, II, III, and IV used in-person interviews, we wanted to assess and correct for any identified effects of switching to web and paper surveys. After successfully fielding Samples 1 and 2, we progressed to the 7,631 respondents in Sample 3, who will be surveyed via web and paper over the course of 2018.
Supplementing Surveys with Solid Medical Data
The Add Health questionnaire is extremely detailed. For example, here is question #121 from the paper survey:
In the past 30 days, which of the following types of prescription drugs have you taken that were not prescribed for you, taken in larger amounts than prescribed, more often than prescribed, for longer periods than prescribed, or that you took only for the feeling or experience they caused?
a. Sedatives or downers, such as barbiturates, sleeping pills, Quaalude, or Seconal
b. Tranquilizers, such as Librium, Valium, or Xanax
c. Stimulants or uppers, such as amphetamines, prescription diet pills, Ritalin, Preludin, or speed
d. Pain killers or opioids, such as Vicodin, OxyContin, Percocet, Demerol, Percodan, or Tylenol with codeine
There are comparable in-depth questions not only about medical and health issues, but also about the respondents’ finances, educational level, social and family life, and early childhood experiences.
As exhaustive as these questions are, though, they can only provide a partial window into the overall health of the Add Health cohort. For this reason, the data collected by RTI are supplemented by in-home visits (conducted by another contractor) to all consenting respondents, during which blood samples are collected, blood pressure is measured, and various other biomeasures are taken. These tests are largely the same as those conducted for Wave IV, with the addition of whole blood collection and a renal disease test (since this is the age at which chronic kidney problems can begin to manifest).
One innovation of the Wave V data-collection effort is the collection of birth records from respondents, and from their children, in six states. While the cohort members have always been asked detailed questions about their children, for the first time Add Health is also asking a subset of respondents for permission to access data from their and their children’s birth records, which will provide valuable information about birth weight, gestational age, and other measures of the circumstances of birth.
Finally, we are self-funding our own ancillary study on the Add Health cohort. We are asking a subset of a few hundred respondents about wearable technology (e.g., Fitbits, Apple watches): whether and how long they have owned these devices, how often they use them, and what kinds of data they are tracking (e.g., weight, number of steps taken, blood pressure, etc.). Our ultimate goal is to be granted access to the data these devices generate, which will open a new window into respondents’ health and provide a useful check against what they claim in surveys.
The Future of Add Health: Waves VI and Beyond
Add Health has already answered—or begun to answer—major questions about the links between adolescent experiences and adult health outcomes. For one example, there is now solid evidence, from earlier waves, that middle and high-schoolers who have problems with homework are at increased risk to take up drinking and smoking, and that adolescents who were highly embedded in social networks with friends, family, and in the school and local community had better cardiovascular and metabolic health 15 years later than adolescents with fewer network connections. Furthermore, DNA collected from the full Add Health sample during Wave IV is currently being mined to uncover the ways in which the environment affects genetic influence (i.e., gene-by-environment interactions) on health and health behaviors in adulthood.
Given our role in Waves III, IV and V of the Add Health study, we are excited for the opportunity to continue to work with the Eunice Kennedy Shriver National Institute of Child Health and Human Development and the UNC Carolina Population Center on Waves VI and beyond. In a sense, we are only halfway through this project’s lifetime. Assuming that most of the participants reach the age of 75 or 80, there is plenty of data still to be collected that will shed invaluable light on health outcomes across the entire human life cycle.
