2.2. Measures

Independent variables (i.e., pubertal timing measures), dependent variables (i.e., psychosocial well-being, health behaviors, physical health in adolescence and adulthood), and baseline covariates are described in detail below.

Two self-reported measures of pubertal timing were assessed: intra-individual development (i.e., maturational timing) and inter-individual development (i.e., peer-relative pubertal timing). Self-reported age at menarche served as a proxy for maturational timing for girls. Although this event occurs later in the pubertal process, it is commonly used as a marker of pubertal timing for girls and facilitates the comparison of findings across studies. Age of menarche was measured in whole years at Wave I, Wave II, and Wave III. Of the female respondents, 89.63% reported their age at menarche at Wave I; if a respondent had not had menarche or did not report it at Wave I, we used the next available report at Wave II (5.38%) or Wave III (3.42%). Boys’ maturational timing was derived from selected questions about bodily changes that become evident in mid-to-late puberty (i.e., facial hair, body hair, and voice change) on a continuous scale that ranged from “least developed” (1) to “most developed” (5). Responses on the three items were averaged and then standardized within each whole number age group so that a higher value represents more advanced development among same-aged boys (i.e., earlier maturation). For both sexes, early/late maturational timing was defined as one standard deviation below/above the mean. These cut-offs have been used in previous research with the Add Health sample and others (e.g., (Foster, Hagan, & Brooks-Gunn, 2008)Ge et al., 2001).

Peer-relative pubertal timing was assessed by the question, “How advanced is your physical development compared to other boys/girls your age?” Responses could range from “I look younger than most” (1) to “I look older than most” (5). We defined early peer-relative timing as physical development older than most, and late as younger than most. The distributions of early/on-time/late peer-relative timing roughly match that of maturational timing in both females and males (see Table 1).

Descriptive statistics for the analytic sample (n = 14, 545).

Notes. All descriptive statistics are drawn from the final imputed dataset and weighted using Add Health survey weights.

In Waves I and IV, participants completed the Center for Epidemiological Studies Depression Scale (CES-D), which is a short, self-report scale designed to measure depressive symptomatology in the general population, and is also valid for use in adolescent populations (Perreira, Deeb-Sossa, Harris, & Bollen, 2005, Radloff, 1991). Sample items include: felt blue; bothered by things that don't usually bother you; did not enjoy life (0 = never; 1 = sometimes; 2 = a lot of the time; 3 = most/all of the time). The full version was administered at Wave I; a 10-item abbreviated version was administered at Wave IV. To keep measurement consistent across waves, we used the abbreviated version. Participants' responses were summed to form the depressive symptoms scale (ranging from 0 – 30) at Waves I (α=.80) and IV (α=.84).

Antisocial behaviors were assessed by self-reported frequency of antisocial behaviors over the past 12 months (0 = never; 1 = 1 or 2 times; 2 = 3 or 4 times; 3 = 5 or more times). Five items were asked at both Waves I and IV, including property damage, stealing something worth <$50, stealing something worth >$50, selling drugs, and breaking into a building. Wave I included five additional items: running away from home, lying to parents, driving a car without the owner's permission, shoplifting, and being loud and rowdy in public. Wave IV included three additional items: deliberately writing a bad check, using others' debit card without permission, and buying or selling stolen property. Participants' responses were summed at Waves I (α=.78) and IV (α=.62).

We assessed five key health behaviors: sexual risk-taking, drug use, physical activity, screen time, and sleep. Sexual risk-taking was conceptualized as the total number of sexual partners. Responses of 50 partners or more were winsorized (i.e., top coded) at 50 (0.23% of all responses in Wave I and 1.48% in Wave IV) to reduce skewness.

At Waves I and IV, participants reported the number of days they smoked cigarettes, the number of times they used marijuana and illicit drugs (e.g., cocaine, inhalants) in the last 30 days, and the number of days they drank five or more alcoholic drinks in a row over the past 12 months. These four variables were dichotomized as whether a participant used cigarettes, marijuana, or illicit drugs in the past 30 days or binge drank 3–5 days or more in the past 12 months (approximately once per month;1 = yes; 0 = no). A drug use index was created as the sum of these items (ranging from 0 to 4), based on previous research (McDade et al., 2011).

Physical activity was assessed in Wave I by three questions about the frequency of skating/biking, playing an active sport, and exercising in the past week (0 = not at all; 1 = 1 – 2 times; 2 = 3 – 4 times; 3 = 5 or more times). In Wave IV, physical activity was assessed by the frequency of participation in seven activities in the past week: skating/biking, snowboard/racquet/aerobics, team sports, individual sports, gymnastics/weights/strength training, golfing/fishing/baseball, and walking. Responses to the activities were summed (Simpkins, Schaefer, Price, & Vest, 2013). To measure screen time, we summed the number of hours per week spent in three behaviors in Wave I: TV hours, video hours, and screen games (Gordon-Larsen, McMurray, & Popkin, 2000) and three behaviors in Wave IV: TV hours, internet hours, and screen games.

In Wave I, participants were asked about how many hours of sleep they usually get with responses in whole hours. In Wave IV, participants were asked about weekday/weekend sleep via four items: “on days when you go to work, school, or similar activities, what time do you usually wake up?”; “what time do you usually go to bed the night before?”; “on days when you don't have to get up at a certain time, what time do you usually wake up?”; “on these days, what time do you usually go to sleep the night before?”. We calculated the hours of sleep per night on weekdays and weekends, and computed a weighted average (i.e., weighing the weekday items by 5/7 and the weekend items by 2/7) to assess sleep duration in a typical day during a week (Maslowsky & Ozer, 2014).

Self-reported good health was assessed by the same, single question in Waves I and IV: “In general, how is your health?” (0 = poor; 1 = fair; 2 = good; 3 = very good; 4 = excellent). Self-reported height and weight were used to calculate baseline adolescent BMI because objective measures were not available at Wave I. Height and weight, measured by trained interviewers at Wave IV, were used to calculate adult BMI (i.e., the ratio of weight in kilograms over height in meters squared).

All baseline covariates were measured at Wave I. Demographic data included age in years and race/ethnicity (coded as Non-Hispanic White, Black/African American, Hispanic/Latino, Asian, or other). Measures of the family environment included family socioeconomic status (SES) – measured by approximate years of parents’ education and whether or not their family received public aid – as well as parental marital status (married, single parent, widowed, divorced, separated) and father absence (father present, father left when participant was 0–5 years, father left when aged 6–13 years). The environment was addressed by urbanicity (urban, suburban, or rural.