Advertisement

School-Based Body Mass Index Screening and Parental Notification in Late Adolescence: Evidence From Arkansas's Act 1220

      Abstract

      Purpose

      In 2003, Arkansas enacted Act 1220, one of the first comprehensive legislative initiatives aimed at addressing childhood obesity. One important provision of Act 1220 mandated that all children attending public schools be screened for their body mass index (BMI) and the information sent home to their parents. Since then, eight other states have adopted similar school-based BMI screening and notification policies. Despite their widespread adoption and implementation, there is a dearth of empirical studies evaluating such policies, particularly for adolescents. The aim of this study was to evaluate whether adolescents, who had been previously screened in early adolescence, experienced changes in their health outcomes if they continued to receive screening and reporting throughout late adolescence (11th and 12th grades).

      Methods

      Secondary data from the Centers for Disease Control's Youth Risk Behavior Survey were analyzed using the method of difference-in-differences. Changes in outcomes between 10th and 12th grade were compared between a group of students who received screenings throughout 11th and 12th grades versus a later comparison group who were exempt from screening and reporting requirements in 11th and 12th grades.

      Results

      BMI screening and parental notification during late adolescence, given prior screening and notification in early adolescence, was not significantly related to BMI-for-age z-scores, the probability of being in a lower weight classification or exercise and dietary intake behaviors.

      Conclusions

      Exposing 11th and 12th graders to BMI screening and reporting, given that they had been exposed in prior grades, was not associated with adolescents' health outcomes.

      Keywords

      Implications and Contribution
      This study demonstrated that screening adolescents for their body mass index and reporting it to their parents in late adolescence, given prior screening and reporting in early adolescence, was not related to adolescents' health outcomes. These results underscored the importance of understanding the timing of adolescents' exposure to body mass indexing screening and reporting.
      Between 1980 and 2010, the adolescent obesity rate in the United States more than tripled [
      • Lavizzo-Mourey R.
      The adolescent obesity epidemic.
      ], reaching an all-time high of 18.4% in 2010 [

      Centers for Disease Control and Prevention (CDC). Centers for Disease Control. Obesity—facts—adolescent and school health. Available at: http://www.cdc.gov/healthyyouth/obesity/facts.htm. Accessed October 10, 2013.

      ]. This increase in adolescent obesity (defined as having a body mass index [BMI] at or above the 95th percentile among children of the same age and sex [

      Centers for Disease Control and Prevention (CDC). Centers for Disease Control. Obesity—facts—adolescent and school health. Available at: http://www.cdc.gov/healthyyouth/obesity/facts.htm. Accessed October 10, 2013.

      ]) has placed a growing number of adolescents at risk for physical and psychosocial problems ranging from Type 2 diabetes and high blood pressure to depression and social isolation [
      • Adair L.S.
      Child and adolescent obesity: Epidemiology and developmental perspectives.
      ,
      • Brownell K.D.
      • Schwartz M.B.
      • Puhl R.M.
      • et al.
      The need for bold action to prevent adolescent obesity.
      ,
      • Wang L.Y.
      • Denniston M.
      • Lee S.
      • et al.
      Long-term health and economic impact of preventing and reducing overweight and obesity in adolescence.
      ,
      • Wile S.
      • Jason S.
      • Schwarz S.W.
      • et al.
      Adolescent obesity in the United States: Facts for policymakers.
      ]. Adolescent obesity also has broader societal consequences, impacting the U.S. economy because of productivity losses brought about by obesity-related diseases [
      • Wang L.Y.
      • Denniston M.
      • Lee S.
      • et al.
      Long-term health and economic impact of preventing and reducing overweight and obesity in adolescence.
      ].
      Schools districts across the United States have played an increasingly prominent role in combating adolescent obesity [
      • Adair L.S.
      Child and adolescent obesity: Epidemiology and developmental perspectives.
      ,
      • Brownell K.D.
      • Schwartz M.B.
      • Puhl R.M.
      • et al.
      The need for bold action to prevent adolescent obesity.
      ,
      • Peterson K.E.
      • Fox M.K.
      Addressing the epidemic of childhood obesity through school-based interventions: What has been done and where do we go from here?.
      ], given their regulatory authority to exert influence over students' eating and exercise behaviors during the school day [
      • Adair L.S.
      Child and adolescent obesity: Epidemiology and developmental perspectives.
      ]. One of the first comprehensive legislative initiatives placing public schools front and center in the effort to address obesity and overweight was Arkansas's Act 1220 [
      • Raczynski J.M.
      University of Arkansas for Medical Sciences. College of Public Health
      Establishing a baseline to evaluate Act 1220 of 2003: An act of the Arkansas General Assembly to combat childhood obesity.
      ]. As part of Act 1220, beginning in 2003, all children attending Arkansas's public schools were required to be screened by a team of trained health professionals for their BMI [

      Conis E. Mandating body mass index reporting in the schools. Available at: http://www.hpm.org/us/a15/2.pdf. Accessed August 12, 2014.

      ], and then confidential letters—known as Child Health Reports—were sent home to parents and guardians [
      • Raczynski J.M.
      • Thompson J.W.
      • Phillips M.M.
      • et al.
      Arkansas Act 1220 of 2003 to reduce childhood obesity: Its implementation and impact on child and adolescent body mass index.
      ]. Letters included information about a child's weight category based on their BMI-for-age (overweight, at risk for overweight, healthy weight or underweight) and the general health consequences of being overweight or at risk for overweight. Also, based on guidelines from the American Academy of Pediatrics, each letter advised parents to ensure their children engaged in frequent exercise, limited their intake of soda, and increased their intake of fruits and vegetables.
      In the first year of implementation, different school districts across the state held screenings at different times of the school year although Child Health Reports were mailed in June [
      • Thompson J.W.
      • Card-Higginson P.
      Arkansas' experience: Statewide surveillance and parental information on the child obesity epidemic.
      ] during which more than 346,000 letters were sent to approximately 450,000 K-12 public school students [
      • Ryan K.W.
      • Card-Higginson P.
      • McCarthy S.G.
      • et al.
      Arkansas fights fat: Translating research into policy to combat childhood and adolescent obesity.
      ]. The reported direct costs for screening and reporting have been estimated at $1.5 million in its initial year and $750,000 annually thereafter (in per student terms, this was $3.00 per student and $1.50 per student, respectively [
      • Thompson J.W.
      • Card-Higginson P.
      Arkansas' experience: Statewide surveillance and parental information on the child obesity epidemic.
      ]). These costs were not definitive as estimates reported in 2009 suggested that the per-school cost of screening and reporting in Arkansas could have been as low as $60 per school or as high as $500 per school [
      • Dietz W.H.
      • Story M.T.
      • Leviton L.C.
      Issues and implications of screening, surveillance, and reporting of children's BMI.
      ]. Since Arkansas implemented its statewide BMI screening and reporting policy in 2003, eight other states have adopted similar BMI screening and parental reporting policies including Alabama, Massachusetts, and Ohio [
      • Ruggieri D.G.
      • Bass S.B.
      A comprehensive review of school-based body mass index screening programs and their implications for school health: Do the controversies accurately reflect the research?.
      ]. However, the justification for such programs and, more broadly, community-based strategies aimed at preventing obesity among adolescents, lacks a comprehensive and rigorous empirical basis [
      • Kaczmarski J.M.
      • DeBate R.D.
      • Marhefka S.L.
      • et al.
      State-mandated school-based BMI screening and parent notification: A descriptive case study.
      ,
      • Nihiser A.J.
      • Lee S.M.
      • Wechsler H.
      • et al.
      BMI measurement in schools.
      ].
      Theoretically, both the health belief model [
      • Glanz K.
      • Rimer B.K.
      • Viswanath K.
      Health behavior and health education: Theory, research, and practice.
      ] and social cognitive theory [
      • Rosenstock I.M.
      • Strecher V.J.
      • Becker M.H.
      Social learning theory and the health belief model.
      ] may help explain why BMI screening and notification may lead to changes in outcomes. The health belief model suggests that notifying parents about their children's BMI—particularly if children classified as overweight or obese—may influence parents' perception of their children's susceptibility and severity [
      • Janz N.K.
      • Becker M.H.
      The health belief model: A decade later.
      ] of being overweight and/or obese, which in turn serves as a cue to action [
      • Kaczmarski J.M.
      • DeBate R.D.
      • Marhefka S.L.
      • et al.
      State-mandated school-based BMI screening and parent notification: A descriptive case study.
      ]. This enhanced perception could induce children (and/or their parents) to change their diet and exercise leading to improved outcomes. Social cognitive theory suggests that informing parents of their children's BMI status will induce parents to undertake changes to improve their children's health if they are convinced that (1) their children's weight status poses a potential health threat; (2) changes in children's behaviors (i.e., exercise and dietary practices) may mitigate any potential threats; and finally, (3) children (and/or parents themselves) possess the capacity to alter their behaviors to influence their well-being [
      • Rosenstock I.M.
      • Strecher V.J.
      • Becker M.H.
      Social learning theory and the health belief model.
      ]. In both models, parents play a critical role since they decide whether to share BMI information with their child. In one prior study on BMI screening and parental notification policies in Minnesota, although about 79% of parents read notification letters in their entirety, 55% did not discuss the information with their child [
      • Kubik M.Y.
      • Story M.
      • Rieland G.
      Developing school-based BMI screening and parent notification programs: Findings from focus groups with parents of elementary school students.
      ]. Furthermore, according to the only peer-reviewed study of BMI screening and parent notification to date from California, notification has not been associated with changes in BMI [
      • Madsen K.A.
      School-based body mass index screening and parent notification: A statewide natural experiment.
      ]. The study, conducted by Madsen [
      • Madsen K.A.
      School-based body mass index screening and parent notification: A statewide natural experiment.
      ], found no differences in the BMI of children in Grades five, seven, and nine across time between one group of districts that reported BMI information to parents versus a comparison group of districts that did not report such information.
      Adherence to BMI screening and reporting in Arkansas has been high with 98.7% of public schools participating in the 2011–2012 academic year [
      Arkansas Center for Health Improvement
      Year nine assessment of childhood and adolescent obesity in Arkansas (Fall 2011–Spring 2012).
      ]. In 2007, because of mounting concerns about a lack of a parental opt out of screening requirements and the administrative burden of screening, the Arkansas General Assembly implemented Act 201, which allowed parental opt out from screenings and exempted children in odd-numbered grades as well as in Grade 12 [
      • Raczynski J.M.
      • Thompson J.W.
      • Phillips M.M.
      • et al.
      Arkansas Act 1220 of 2003 to reduce childhood obesity: Its implementation and impact on child and adolescent body mass index.
      ]. This exemption raised a unique opportunity to compare the health outcomes of adolescents who were subject to screening and reporting in 11th and 12th grades to those who were exempt from screening and reporting to understand the effect of screening and reporting requirements in late adolescence. Accordingly, the aim of this study was to evaluate whether adolescents, who had been previously screened in early adolescence, experienced changes in their health outcomes if they continued to receive screening and reporting throughout late adolescence (11th and 12th grades).

      Methods

      Dataset and measures

      Dataset

      Secondary data were used from the Youth Risk Behavior Survey (YRBS) for Arkansas that included individual-level data on adolescents attending public schools in Arkansas [

      Centers for Disease Control and Prevention. Youth Risk Behavior Survey (YRBS). Washington, D.C.

      ]. The following three repeated cross-sections were used: 2005, 2007, and 2009. Since data were collected in the spring of the academic year (AY) which typically spanned from February to May [
      • Brener N.D.
      • Eaton D.K.
      • Flint K.H.
      • et al.
      Methodology of the youth risk behavior surveillance system-2013.
      ], the YRBS dataset year (e.g., 2009) corresponded to the preceding AY (e.g., AY 2008–2009). The YRBS data contained no individually identifiable information and were publicly available; therefore, the institutional review board at the University of California, Davis determined that the study was not considered research involving human subjects and exempt from review.

      Measures

      Weight status

      The study's primary outcome measure was an adolescent's BMI calculated based on self-reported height and weight. Two versions were used—one continuous, based on their BMI-for-age z-score, whereas the other was categorical corresponding to three age and gender adjusted BMI percentiles: healthy weight (fifth percentile to <85th percentile); overweight (85th percentile to <95th percentile); obese (≥95th percentile) [

      Centers for Disease Control. About BMI for children and teens. Available at: http://www.cdc.gov/healthyweight/assessing/bmi/childrens_bmi/about_childrens_bmi.html. Accessed August 13, 2014.

      ]. BMI percentiles for age and gender have been shown to be highly correlated (.88) with percentage body fat as measured by dual energy X-ray absorptiometry scans [
      • Field A.E.
      • Laird N.
      • Steinberg E.
      • et al.
      Which metric of relative weight best captures body fatness in children?.
      ].
      Prior research demonstrated that among YRBS respondents, discrepancies between self-reported versus measured height and weight were common [
      • Brener N.D.
      • McManus T.
      • Galuska D.A.
      • et al.
      Reliability and validity of self-reported height and weight among high school students.
      ]. On average, students tended to overestimate their height by 2.7 inches and underestimated their weight by 3.5 pounds (p. 284). As a result, the prevalence of adolescents at risk for overweight and obesity was underestimated versus the prevalence using measured height and weight (p. 285). Although the YRBS only included self-reported height and weight, the data were still valuable in understanding adolescents' weight status [
      • Sherry B.
      • Jefferds M.E.
      • Grummer-Strawn L.M.
      Accuracy of adolescent self-report of height and weight in assessing overweight status: A literature review.
      ]; however, readers should be aware of this limitation.

      Exercise behaviors

      Measures of adolescents' exercise behaviors included the frequency of participation in vigorous activities for at least 20 minutes in the past 7 days and the frequency of participation in moderate activities for at least 30 minutes in the past 7 days. Both measures were self-reported on a scale from 0 to 7 days, in whole day increments.

      Dietary intake behaviors

      The study used students' self-reports of whether they consumed fruits at least once a day or more and whether they consumed vegetables at least once a day or more. Both variables were dichotomous, equaling 1 if they responded affirmatively, 0 otherwise.

      Additional measures

      A set of covariates was used including adolescents' self-reported age (in years), gender, and racial/ethnic identification (i.e., non-Hispanic black or African-American; Hispanic or Latino; non-Hispanic white; and other).

      Validity and reliability of exercise and dietary intake behavior measures

      The test–retest reliabilities of the moderate and vigorous activity measures in the YRBS in a sample of middle school students were found to be moderately reliable [
      • Troped P.J.
      • Wiecha J.L.
      • Fragala M.S.
      • et al.
      Reliability and validity of YRBS physical activity items among middle school students.
      ]. As for validity, moderate activity was underestimated versus measures of activity captured by an Actigraph accelerometer, whereas vigorous activity was overestimated [
      • Troped P.J.
      • Wiecha J.L.
      • Fragala M.S.
      • et al.
      Reliability and validity of YRBS physical activity items among middle school students.
      ]. Differences in the self-reported intake of fruits and vegetables in a sample of high school students using the YRBS questionnaire versus a 24-hour recall method showed that there was underreporting of servings of fruit and vegetables consumed [
      • Field A.E.
      • Laird N.
      • Steinberg E.
      • et al.
      Which metric of relative weight best captures body fatness in children?.
      ]. Thus, self-reported measures of exercise and dietary intake were prone to self-reporting bias, a limitation to the study.

      Identification strategy and sample

      Identification strategy

      A straightforward difference-in-differences approach was used to estimate the effect of BMI screening and reporting in late adolescence, given prior screening and reporting in early adolescence. Changes in outcomes between 10th and 12th grade were compared between a group of students who received screenings throughout 11th and 12th grades (Group A, the “treated” group) versus a later comparison group who became exempt from screening and reporting requirements in 11th and 12th grades (Group B, the “comparison” group). Table 1 shows these two groups. Although each group was not randomly assigned to screening and reporting, because of the exemptions, each group's differential exposure to the screening and reporting requirement was plausibly unrelated to characteristics of adolescents themselves which would attenuate any potential self-selection bias.
      Table 1The timing of exposure to BMI screening and reporting requirements under Arkansas's Act 1220 for two groups (A and B)
      YRBS yearAcademic yearEventGroup A (treatment)Group B (comparison)
      GradeSubject to BMI screening and notification?GradeSubject to BMI screening and notification?
      2003–2004Act 1220 enacted9Yes7Yes
      20052004–200510YesGroup A difference8Yes
      2005–200611Yes9Yes
      20072006–200712Yes10YesGroup B difference
      2007–2008Screening exemptions11No
      20092008–200912No
      BMI = body mass index; YRBS = Youth Risk Behavior Survey.
      More formally, in a difference-in-differences framework, the first difference captures, in part, the change in outcomes attributable to BMI screening and reporting in both 11th and 12th grades for Group A (the “treated” group): (Y¯Group A12th gradeY¯Group A10th grade). However, this change also captures other secular trends occurring over time, including exposure to Act 1220's other health-related initiatives. Thus, to isolate the effect of screening and reporting in 11th and 12th grades, it was necessary to know what the change in outcomes would have looked like if these students were not continued to be screened and notified in these grades. To account for this counterfactual trend, Group B (the “comparison” group) students were used as a natural comparison group to form the second difference since they were exempt from screenings in 11th and 12th grades: (Y¯Group B12th gradeY¯Group B10th grade). Subtracting this second difference from the first difference yielded the difference-in-differences estimator which isolated the effect of BMI screening and notification in 11th and 12th grades on outcomes: δ1=(Y¯Group A12th gradeY¯Group A10th grade)(Y¯Group B12th gradeY¯Group B10th grade). Because adolescents experienced screening and reporting before 10th grade, δ1 should be interpreted as the effect, on average, of experiencing additional BMI screening and reporting in 11th and 12th grades, above and beyond the screening and reporting experienced before those grades.
      Regression was used to estimate δ1:
      Y=β0+β1(GroupA)+δ0(12thgrade)+δ1(GroupA×12thgrade)+XZ+ε
      (1)


      where Y is the outcome; Group A is a dummy variable equal to 1 for adolescents in the group which continued to experience screening and reporting in 11th and 12 grades, whereas 12th grade is a dummy variable equal to 1 for adolescents observed in 12th grade (the after time period), 0 otherwise. X′ is a vector of individual-level controls. δ1 is the difference-in-differences estimator. ε is the error term.
      For continuous outcomes (e.g., BMI-for-age z-score), ordinary least squares regression was used. For the categorical weight outcomes, a series of probit regressions were used to estimate the probabilities of adolescents being in a lower versus a higher weight category (e.g., healthy weight vs. overweight). Probit regression was also used for fruit and vegetable consumption outcomes. For all probit models, Norton, Wang and Ai's (2004) method was used to transform the coefficient estimate on the interaction term, the difference-in-differences estimator δ1.

      Analytic sample

      Unweighted sample sizes for each grade-by-year combination used in the analysis were as follows: (1) Group A: 10th graders in 2005 (n = 439) and 12th graders in 2007 (n = 147) and (2) Group B: 10th graders in 2007 (n = 362) and 12th graders in 2009 (n = 133). Based on the total unweighted sample size (n = 1,081), a power analysis was conducted [
      • Faul F.
      • Erdfelder E.
      • Buchner A.
      • et al.
      Statistical power analyses using G* Power 3.1: Tests for correlation and regression analyses.
      ] demonstrating that the study had power of approximately .99 to detect small effects (.04; the average effect size found across prior childhood and adolescent obesity prevention studies [
      • Stice E.
      • Shaw H.
      • Marti C.N.
      A meta-analytic review of obesity prevention programs for children and adolescents: The skinny on interventions that work.
      ]) at a Bonferroni adjusted alpha level of .006 (a more stringent alpha level was adopted to account for multiple hypothesis testing). The samples used were limited given that the two-stage cluster sample design of the YRBS (i.e., schools were sampled first, then students within schools) provided a sample representative of ninth–12th graders as a whole, rather than individual grades. Thus, these subsamples of 10th and 12th graders were unlikely to have come from a representative sample of schools and even by applying sample weights, these samples may not have necessarily reflected the underlying population of 10th and 12th grade adolescents attending Arkansas's public schools.

      Results

      Descriptive statistics

      As shown in Table 2, for each grade-by-year group (e.g., 10th graders in 2005), a majority of adolescents were classified as having a healthy weight, followed by overweight then obese. Also, across each grade-by-year group, approximately a quarter to slightly more than a third of adolescents did not engage in any moderate physical activity, whereas about a fifth of adolescents participated in vigorous exercise for more than 20 minutes in the past 7 days. Regarding dietary intake behaviors, a majority of adolescents did not consume one or more servings of fruits or vegetables in the past 7 days. Finally, the sample was predominately white non-Hispanic, followed by black. Importantly, there were no statistically significant differences in these measures between Groups A and B in 10th grade (before Group B became exempt from screening and reporting); thus, ruling out the possibility that these two groups systematically differed from each other on observable characteristics (results available from author).
      Table 2Weighted sample univariate descriptive statistics (percentages and 95% CIs) for 10th and 12th grade adolescents in Arkansas
      Group A (treatment)Group B (comparison)
      Tenth grade (2005), 95% CITwelfth grade (2007), 95% CITenth grade (2007), 95% CITwelfth grade (2009), 95% CI
      BMI-for-age z-score (mean; n = 1,080).62 (.48–.76).53 (.35–.71).53 (.45–.60).51 (.32–.70)
      Weight status (%)
       Healthy weight (n = 685)61.2 (55.8–66.4)59.2 (51.0–66.8)67 (62.7–71.0)63.5 (54.2–71.9)
       Overweight (n = 252)24 (19.9–28.5)30.3 (22.9–38.9)21.3 (17.7–25.4)21.4 (14.3–30.9)
       Obese (n = 144)14.8 (11.0–19.7)10.6 (7.0–15.5)11.7 (8.6–15.8)15.1 (10.3–21.5)
       Total (n = 1,081)
      Moderate physical activity (number of days in past 7 days for 30 minutes)
       0 days (n = 333)32.8 (29.1–36.8)25 (15.4–38.0)35.3 (30.0–41.0)32.7 (22.2–45.2)
       1 day (n = 132)13 (10.7–15.9)11.5 (6.3–19.9)11.6 (9.1–14.6)14.8 (10.1–21.0)
       2 days (n = 141)13.8 (10.3–18.2)10.5 (5.6–18.7)14.2 (10.6–18.7)17.4 (10.1–28.4)
       3 days (n = 91)10.1 (7.0–14.2)9.4 (4.9–17.3)7.1 (4.3–11.7)5 (2.1–11.4)
       4 days (n = 73)6.8 (4.9–9.2)10.1 (6.6–15.2)7.2 (4.7–10.7)3.5 (1.4–8.4)
       5 days (n = 77)9.1 (6.7–12.3)10.4 (5.6–18.4)6.5 (4.0–10.5)4.7 (2.5–8.5)
       6 days (n = 22)1.0 (.3–2.8)4.1 (1.9–8.5)1.4 (.5–3.9)5.6 (2.1–13.9)
       7 days (n = 167)13.4 (10.0–17.7)19 (12.6–27.6)16.6 (12.0–22.5)16.5 (11.0–24.0)
       Total (n = 1,035)
      Vigorous physical activity (number of days in past 7 days for 20 minutes)
       0 days (n = 215)19.7 (16.0–24.0)19 (11.3–30.0)21.1 (17.0–25.9)23.5 (17.4–30.8)
       1 day (n = 121)11.8 (9.1–15.1)15.5 (9.0–25.4)8.6 (5.5–13.2)9.6 (6.0–14.9)
       2 days (n = 132)14.4 (11.4–18.0)13.9 (8.0–23.1)11.3 (7.9–15.9)15 (8.5–25.1)
       3 days (n = 114)14.1 (10.8–18.3)9.7 (5.0–17.9)8.7 (6.0–12.5)8.8 (5.3–14.0)
       4 days (n = 66)5.3 (3.6–7.7)10.8 (6.2–18.2)6.9 (4.4–10.6)3.4 (1.5–7.8)
       5 days (n = 103)9.7 (6.7–13.7)11.1 (7.8–15.6)10.1 (7.1–14.1)12.3 (6.3–22.6)
       6 days (n = 54)4.9 (3.1–7.4)4.6 (2.0–10.2)5 (2.8–8.6)6.5 (3.7–11.4)
       7 days (n = 229)20.2 (16.5–24.5)15.5 (10.6–22.1)28.4 (23.3–34.1)21 (13.4–31.2)
       Total (n = 1,034)
      Vegetable consumption, once a day or more (%)
       No (n = 962)89.1 (84.8–92.3)90.8 (84.6–94.6)86.9 (77.5–92.8)91.6 (85.2–95.4)
       Yes (n = 119)10.9 (7.7–15.2)9.2 (5.4–15.4)13.1 (7.2–22.5)8.4 (4.6–14.8)
       Total (n = 1,081)
      Fruit consumption, once a day or more (%)
       No (n = 932)85.3 (79.8–89.5)91.4 (83.9–95.5)82.4 (73.8–88.7)88.8 (81.2–93.5)
       Yes (n = 149)14.7 (10.5–20.2)8.6 (4.5–16.1)17.6 (11.3–26.2)11.2 (6.5–18.8)
       Total (n = 1,081)
      Age (mean years; n = 1,081)15.9 (15.8–15.9)17.5 (17.4–17.6)15.8 (15.7–15.9)17.6 (17.5–17.7)
      Race/ethnicity (%)
       Other
      The other race/ethnicity category includes: American Indian/Alaskan Native; Asian; Native Hawaiian or other Pacific Islander; multiple Hispanic/Latino; multiple non-Hispanic/Latino.
      (n = 122)
      3.5 (2.3–5.4)3.6 (2.0–6.5)6.6 (4.6–9.5)6.4 (3.8–10.7)
       White (n = 707)72.1 (61.6–80.7)68 (55.5–78.4)61.0 (45.1–74.9)70.2 (50.0–84.7)
       Black (n = 200)23.1 (14.7–34.3)23.8 (15.1–35.4)29.9 (16.6–47.8)21.6 (9.1–43.2)
       Hispanic (n = 52)1.3 (.7–2.3)4.5 (2.1–9.5)2.5 (1.3–4.7)1.8 (.4–7.6)
       Total (n = 1,081)
      Gender
       Female (n = 545)49.3 (45.1–53.5)49.4 (36.0–62.9)47.5 (40.8–54.2)49.1 (37.4–60.9)
       Male (n = 536)50.7 (46.5–54.9)50.6 (37.1–64.0)52.5 (45.8–59.2)50.9 (39.1–62.6)
       Total (n = 1,081)
      Weighted using YRBS sample weights.
      Unweighted sample sizes provided in parentheses.
      CI = confidence interval; YRBS = Youth Risk Behavior Survey.
      a The other race/ethnicity category includes: American Indian/Alaskan Native; Asian; Native Hawaiian or other Pacific Islander; multiple Hispanic/Latino; multiple non-Hispanic/Latino.

      Examining pre-exemption trends in weight status

      A critical assumption underpinning this study's difference-in-differences strategy was that trends in outcomes between each group would have moved in tandem had exemptions not occurred [
      • Gertler P.J.
      • Martinez S.
      • Premand P.
      • et al.
      Impact evaluation in practice.
      ]. To assess this assumption, pre-trend data were examined under the rationale that if trends in both groups were similar before Group B experienced exemptions in 10th grade, such trends would have continued to be similar after 10th grade. Because pre-trend data were not available in the YRBS dataset, data for the state of Arkansas from the National Survey of Children's Health were used to estimate BMI status for 14–15 year olds in 2003 which roughly corresponded to when 10th graders in 2005 (Group A) were in ninth grade. Also, BMI status was estimated for 12–13 year olds in 2003 corresponding to when 10th graders in 2007 (Group B) were in seventh grade. Although not ideal and a very coarse approximation, relying on data for Arkansas from the National Survey of Children's Health helped determine whether trends were similar in each group before 10th grade.
      As shown in Table 3, the percentage of adolescents in the obese categories declined in both groups, whereas the percentage in the overweight category increased. For the healthy weight category, the two groups exhibited differential trends—the percentage in the healthy weight category decreased slightly in Group A while there was an increase in Group B. However, neither of these changes in the healthy weight status were statistically significant. This descriptive evidence provided some confidence that trends before 10th grade in each group looked relatively similar on observed weight status and therefore, it was plausible to suggest that those trends would have continued to follow a similar pattern after 10th grade in each group.
      Table 3Pre-exemption trends in weight classification outcomes for two groups of 10th graders in Arkansas
      Weight categoryGroup AGroup B
      2003 (NSCH; Arkansas), 14–15 year olds2005 (YRBS; Arkansas), 10th graders2003 (NSCH; Arkansas), 12–13 year olds2007 (YRBS; Arkansas), 10th graders
      % (95% CI)% (95% CI)% (95% CI)% (95% CI)
      Healthy weight63.3 (57.9–68.7)61.2 (55.8–66.4)61.6 (56.2–67.0)67.0 (62.7–71.0)
      Overweight14.8 (10.8–18.8)24.0 (19.9–28.5)12.0 (8.9–15.1)21.3 (17.7–25.4)
      Obese17.5 (13.2–21.7)14.8 (11.0–19.7)20.3 (15.6–25.0)11.7 (8.6–15.8)
      Data for 2003 are for Arkansas from the NSCH. Data for 2005 and 2007 are from the YRBS for Arkansas. Percentages are weighted using sample weights included in each survey.
      CI = confidence interval; NSCH = National Survey of Childhood Health; YRBS = Youth Risk Behavior Survey.

      Main results

      Table 4 reports the difference-in-differences estimators for weight outcomes which are captured by the coefficients on the interaction term (Group A × 12th grade) reported in row three. In terms of the direction of these estimates, these results suggested that screening and reporting in 11th and 12th grades, above and beyond screening and reporting in prior grades, was associated with a lower standardized BMI, a lower probability of being healthy weight versus obese, a lower probability of being overweight versus obese, and a higher probability of being a healthy weight versus overweight. Importantly, although the direction of these associations indicated two potentially beneficial effects of additional screenings in late adolescence (lowered BMI and a higher probability of being a healthy weight), none of these findings were statistically significant and therefore, it was not possible to rule out zero effects. These models were refitted to data with BMIs corrected for the systematic underreporting of weight and overreporting of height, and the results remained consistent to those previously reported (results available from author).
      Table 4Difference-in-differences estimates for the relationship between the exposure to BMI screening and parental reporting in late adolescence (11th and 12th grades) and weight outcomes
      (1) BMI (standardized)(2) Pr(healthy weight vs. obese)(3) Pr(overweight vs. obese)(4) Pr(healthy weight vs. overweight)
      Group A.105 (.087)−.031 (.030).063 (.034)−.025 (.017)
      Twelfth grade.060 (.117).029 (.059)−.016 (.048)−.005 (.027)
      Group A × 12th grade−.087 (.167)−.055 (.071)−.026 (.077).078 (.047)
      Age−.038 (.056)−.019 (.021).033 (.020)−.012 (.010)
      White non-Hispanic.017 (.137).029 (.057)−.043 (.044).010 (.023)
      Black non-Hispanic.149 (.163).023 (.073)−.011 (.059)−.010 (.030)
      Hispanic.157 (.209).055 (.098)−.015 (.088)−.030 (.057)
      Male.195 (.081)−.014 (.031).073 (.038)−.043 (.022)
      N1,0801,0811,0811,081
      Model (1) is based on OLS regression, whereas models (2)–(4) are based on probit regressions. In addition to probit models, linear probability models were refit to the data which yielded consistent results (results available from author). For all probit models (models [2]–[4]), marginal effects are reported. The reported difference-in-differences estimators for each probit model (the estimates on the interaction term Group A × 12th grade) and their associated standard errors were obtained using Stata's inteff command by Norton, Wang and Ai (2004). All models incorporate YRBS survey weights. Standard errors are in parentheses. One observation was dropped from the sample for model (1) due to missing data.
      Models were refitted to adjust BMIs for underreporting of weight and overreporting of height, and the results (available from author) remained consistent to those previously reported.
      BMI = body mass index; OLS = ordinary least squares; Pr = Probability; YRBS = Youth Risk Behavior Survey.
      *p < .05.
      Table 5 reports the effects of screening and reporting in late adolescence on exercise and dietary intake behaviors. As shown, the direction of the estimates indicated that exposure to screening and reporting through late adolescence was related to more frequent moderate and vigorous exercise as well as a higher probability of consuming vegetables once or more per day, while a lower probability of fruit consumption once or more per day. Yet, as with the results for the weight outcomes, none of these effects were statistically significant.
      Table 5Difference-in-differences estimates for the relationship between the exposure to BMI screening and parental reporting in late adolescence (11th and 12th grades) and exercise and dietary intake outcomes
      (1) Moderate exercise(2) Vigorous exercise(3) Pr(vegetable consumption once a day or more)(4) Pr(fruit consumption once a day or more)
      Group A−.063 (.181)−.419 (.196)−.016 (.039)−.012 (.032)
      Twelfth grade−.004 (.370)−.096 (.412)−.074 (.063)−.061 (.065)
      Group A × 12th grade.719 (.369).253 (.463).024 (.057)−.011 (.063)
      Age−.007 (.118)−.171 (.165).019 (.016).011 (.020)
      White non-Hispanic−.333 (.403).390 (.292)−.058 (.041)−.054 (.046)
      Black non-Hispanic−1.136∗∗ (.415)−.169 (.326)−.034 (.030).023 (.032)
      Hispanic−.821 (.601).103 (.540)−.080 (.030).031 (.055)
      Male.262 (.175)1.017∗∗∗ (.177).016 (.018).017 (.019)
      N1,0351,0341,0811,081
      Models (1) and (2) are based on OLS regression, whereas models (3) and (4) are probit regressions. In addition to probit models, linear probability models were refit to the data which yielded consistent results (results available from author). For both probit models (models [3] and [4]), marginal effects are reported and the reported DD estimators for each probit model (the estimates on the interaction term Group A × 12th grade) and their associated standard errors were obtained using Stata's inteff command by Norton, Wang and Ai (2004). All models incorporate YRBS survey weights. Standard errors are in parentheses. Sample sizes for models (1) and (2) reflect missing data.
      BMI = body mass index; DD = difference-in-differences; Pr = Probability; YRBS = Youth Risk Behavior Survey.
      *p < .05, **p < .01, ***p < .001.

      Discussion

      This analysis demonstrated that in Arkansas, BMI screening and parental notification during late adolescence (11th and 12th grades), given prior screening and notification in early adolescence, was not significantly related to changes in either adolescents' weight status or their exercise or dietary intake behaviors. There were several important limitations of this study. First, these results focused exclusively on Arkansas and should not be extrapolated to states that have implemented similar BMI screening and reporting programs. Second, although pre-trends demonstrated that trends were relatively similar before screening exemptions, there was still a possibility that unobservable time-varying factors may differentially affected each group particularly in years in which adolescents were not observed in the dataset (11th grade). Third, as previously addressed, the study's outcome measures were based on adolescents' self-reports which introduced self-reporting bias. Finally, given the parental opt-out provision, adolescents who were exempt from screening may have been systematically different versus the adolescents who had continued to be screened; however, evidence shows that parental opt out was relatively low (e.g., 4.13% in 2007–2008 [

      Q&A with Dr. Thompson surgeon general of Arkansas. Available at: http://mchb.hrsa.gov/researchdata/MCHESP/dataspeak/pastevent/september2008/files/sept2008qanda.pdf. Accessed April 20, 2015.

      ]).
      There are two important implications of this study's results. First, the additional screenings and reportings in 11th and 12th grades did not necessarily lead to any unintended consequences such that the weight status or exercise and dietary intake behaviors of adolescents shifted to higher levels of overweight/obesity or lowered frequency of exercise and consumption of fruits and vegetables. This is particularly relevant in light of the concerns reported in Ikeda et al. (2006) suggesting that the BMI screening process itself and the knowledge of BMI information could have had adverse effects. Additionally, results from supplemental analyses (available from author) examining whether there were changes in behaviors to control weight (i.e., use of diet pills) also demonstrated no significant association. Therefore, the evidence presented here may assuage concerns of potentially adverse effects of experiencing repeated BMI screening and reporting through late adolescence.
      Second, these results have implications for further research on school-based BMI screening and reporting policies. As previously mentioned, eight other states screen for BMI and report the information to parents and guardians; however, although BMI screening and reporting programs are increasingly commonplace, they are often implemented alongside other school-based interventions such as altering school nutrition and physical activity environments [
      • Ruggieri D.G.
      • Bass S.B.
      A comprehensive review of school-based body mass index screening programs and their implications for school health: Do the controversies accurately reflect the research?.
      ]. Given the multipronged nature of school-based obesity prevention efforts, the challenge for researchers is how to isolate and evaluate the effect of BMI screening apart from other interventions. Although a rich body of literature has enhanced our understanding of the process of BMI screening and reporting, including the controversies surrounding BMI screening, this present study alongside one conducted by Madsen [
      • Madsen K.A.
      School-based body mass index screening and parent notification: A statewide natural experiment.
      ] represents one of the first novel attempts to rigorously evaluate BMI screening and reporting. In particular, the results of this study underscore the importance of gaining a more thorough understanding of when and for whom BMI screening and reporting programs might matter the most, especially as adolescents progress from early adolescence into middle and late adolescence. If repeated screenings from middle to late adolescence merely provide information that does not directly alter outcomes, as these findings suggested, then the scarce public resources invested to those screenings could be allocated to better uses, particularly toward interventions that have shown to be effective for adolescents. For example, given that it cost an estimated $1.50 to screen and report each child [
      • Thompson J.W.
      • Card-Higginson P.
      Arkansas' experience: Statewide surveillance and parental information on the child obesity epidemic.
      ] in Arkansas, then not screening and reporting in late adolescence represented a direct annual cost savings (in nominal dollars) of approximately $96,500 in the 2007–2008 school year (based on Arkansas's 11th and 12th grade enrollment of approximately 64,321 students in 2007–2008 [

      Arkansas Department of Education (ADE). Available at: https://adedata.arkansas.gov/statewide/State/EnrollmentByGrade.aspx. Accessed April 20, 2015.

      ]). Finally, these results highlight the need for continued research to find interventions that have promise in impacting obesity and overweight among adolescents, a demographic often overlooked in the obesity research literature [
      • Vash P.
      The complexity of adolescent obesity: Causes, correlates, and consequences.
      ].

      Acknowledgments

      The author would like to thank Eric Cavalli, Michael Hill, and Katherine Fischer for their invaluable research assistance. Also, the author would like to acknowledge helpful feedback from Michal Kurlaender, Paco Martorell, Cassandra Hart, Joseph Price, Kristine West, and four anonymous reviewers. Prior versions of this work were presented at the 2014 Association for Public Policy and Management, the 2015 Association for Education Finance and Policy, and the 2015 Western Social Science Association conferences.

      Funding Sources

      This work was supported by grants from the UC Davis Center for Poverty Research and the UC Davis Small Grant Program.

      References

        • Lavizzo-Mourey R.
        The adolescent obesity epidemic.
        J Adolesc Health. 2009; 45: 6
      1. Centers for Disease Control and Prevention (CDC). Centers for Disease Control. Obesity—facts—adolescent and school health. Available at: http://www.cdc.gov/healthyyouth/obesity/facts.htm. Accessed October 10, 2013.

        • Adair L.S.
        Child and adolescent obesity: Epidemiology and developmental perspectives.
        Physiol Behav. 2008; 94: 8-16
        • Brownell K.D.
        • Schwartz M.B.
        • Puhl R.M.
        • et al.
        The need for bold action to prevent adolescent obesity.
        J Adolesc Health. 2009; 45: 8
        • Wang L.Y.
        • Denniston M.
        • Lee S.
        • et al.
        Long-term health and economic impact of preventing and reducing overweight and obesity in adolescence.
        J Adolesc Health. 2010; 46: 467-473
        • Wile S.
        • Jason S.
        • Schwarz S.W.
        • et al.
        Adolescent obesity in the United States: Facts for policymakers.
        New York, NY: National Center for Children in Poverty, 2010
        • Peterson K.E.
        • Fox M.K.
        Addressing the epidemic of childhood obesity through school-based interventions: What has been done and where do we go from here?.
        J Law Med Ethics. 2007; 35: 113-130
        • Raczynski J.M.
        • University of Arkansas for Medical Sciences. College of Public Health
        Establishing a baseline to evaluate Act 1220 of 2003: An act of the Arkansas General Assembly to combat childhood obesity.
        Little Rock, AR: University of Arkansas for Medical Sciences: College of Public Health, 2005
      2. Conis E. Mandating body mass index reporting in the schools. Available at: http://www.hpm.org/us/a15/2.pdf. Accessed August 12, 2014.

        • Raczynski J.M.
        • Thompson J.W.
        • Phillips M.M.
        • et al.
        Arkansas Act 1220 of 2003 to reduce childhood obesity: Its implementation and impact on child and adolescent body mass index.
        J Public Health Policy. 2009; 30: S124-S140
        • Thompson J.W.
        • Card-Higginson P.
        Arkansas' experience: Statewide surveillance and parental information on the child obesity epidemic.
        Pediatrics. 2009; 124: S73-S82
        • Ryan K.W.
        • Card-Higginson P.
        • McCarthy S.G.
        • et al.
        Arkansas fights fat: Translating research into policy to combat childhood and adolescent obesity.
        Health Aff. 2006; 25: 992-1004
        • Dietz W.H.
        • Story M.T.
        • Leviton L.C.
        Issues and implications of screening, surveillance, and reporting of children's BMI.
        Pediatrics. 2009; 124: S98-S101
        • Ruggieri D.G.
        • Bass S.B.
        A comprehensive review of school-based body mass index screening programs and their implications for school health: Do the controversies accurately reflect the research?.
        J Sch Health. 2015; 85: 61-72
        • Kaczmarski J.M.
        • DeBate R.D.
        • Marhefka S.L.
        • et al.
        State-mandated school-based BMI screening and parent notification: A descriptive case study.
        Health Promot Pract. 2011; 12: 797-801
        • Nihiser A.J.
        • Lee S.M.
        • Wechsler H.
        • et al.
        BMI measurement in schools.
        Pediatrics. 2009; 124: 89
        • Glanz K.
        • Rimer B.K.
        • Viswanath K.
        Health behavior and health education: Theory, research, and practice.
        Jossey-Bass, San Francisco, CA2008
        • Rosenstock I.M.
        • Strecher V.J.
        • Becker M.H.
        Social learning theory and the health belief model.
        Health Education Behav. 1988; 15: 175-183
        • Janz N.K.
        • Becker M.H.
        The health belief model: A decade later.
        Health Education Behav. 1984; 11: 1-47
        • Kubik M.Y.
        • Story M.
        • Rieland G.
        Developing school-based BMI screening and parent notification programs: Findings from focus groups with parents of elementary school students.
        Health Education Behav. 2007; 34: 622-633
        • Madsen K.A.
        School-based body mass index screening and parent notification: A statewide natural experiment.
        Arch Pediatr Adolesc Med. 2011; 165: 987-992
        • Arkansas Center for Health Improvement
        Year nine assessment of childhood and adolescent obesity in Arkansas (Fall 2011–Spring 2012).
        ACHI, Little Rock, AR2012
      3. Centers for Disease Control and Prevention. Youth Risk Behavior Survey (YRBS). Washington, D.C.

        • Brener N.D.
        • Eaton D.K.
        • Flint K.H.
        • et al.
        Methodology of the youth risk behavior surveillance system-2013.
        US Department of Health and Human Services, Centers for Disease Control and Prevention, 2013
      4. Centers for Disease Control. About BMI for children and teens. Available at: http://www.cdc.gov/healthyweight/assessing/bmi/childrens_bmi/about_childrens_bmi.html. Accessed August 13, 2014.

        • Field A.E.
        • Laird N.
        • Steinberg E.
        • et al.
        Which metric of relative weight best captures body fatness in children?.
        Obes Res. 2003; 11: 1345-1352
        • Brener N.D.
        • McManus T.
        • Galuska D.A.
        • et al.
        Reliability and validity of self-reported height and weight among high school students.
        J Adolesc Health. 2003; 32: 281-287
        • Sherry B.
        • Jefferds M.E.
        • Grummer-Strawn L.M.
        Accuracy of adolescent self-report of height and weight in assessing overweight status: A literature review.
        Arch Pediatr Adolesc Med. 2007; 161: 1154-1161
        • Troped P.J.
        • Wiecha J.L.
        • Fragala M.S.
        • et al.
        Reliability and validity of YRBS physical activity items among middle school students.
        Med Sci Sports Exerc. 2007; 39: 416-425
        • Faul F.
        • Erdfelder E.
        • Buchner A.
        • et al.
        Statistical power analyses using G* Power 3.1: Tests for correlation and regression analyses.
        Behav Res Methods. 2009; 41: 1149-1160
        • Stice E.
        • Shaw H.
        • Marti C.N.
        A meta-analytic review of obesity prevention programs for children and adolescents: The skinny on interventions that work.
        Psychol Bull. 2006; 132: 667
        • Gertler P.J.
        • Martinez S.
        • Premand P.
        • et al.
        Impact evaluation in practice.
        World Bank Publications, Washington, DC2011
      5. Q&A with Dr. Thompson surgeon general of Arkansas. Available at: http://mchb.hrsa.gov/researchdata/MCHESP/dataspeak/pastevent/september2008/files/sept2008qanda.pdf. Accessed April 20, 2015.

      6. Arkansas Department of Education (ADE). Available at: https://adedata.arkansas.gov/statewide/State/EnrollmentByGrade.aspx. Accessed April 20, 2015.

        • Vash P.
        The complexity of adolescent obesity: Causes, correlates, and consequences.
        2015