- Email: [email protected]
The Effects of Requiring Parental Consent for Research on Adolescents' Risk Behaviors: A Meta-analysis
Journal of Adolescent Health xxx (2017) 1e8
www.jahonline.org Original article
The Effects of Requiring Parental Consent for Research on Adolescents’ Risk Behaviors: A Meta-analysis Chao Liu, M.A. a, *, Ronald B. Cox Jr., Ph.D. a, Isaac J. Washburn, Ph.D. a, Julie M. Croff, Ph.D. b, and Hugh C. Crethar, Ph.D. b a b
Department of Human Development and Family Science, Oklahoma State University, Stillwater, Oklahoma School of Applied Health and Educational Psychology, Oklahoma State University, Stillwater, Oklahoma
Article history: Received August 17, 2016; Accepted January 31, 2017 Keywords: Parental consent; Active consent; Passive consent; Risk behaviors; Adolescents; Meta-analysis
A B S T R A C T
Purpose: Requiring parental consent may result in sampling biases that confound scientific conclusions and stifle the representation of children most at risk for adverse outcomes. This study aims to investigate whether active parental consent, compared with passive parental consent, creates a bias in response rate, demographic makeup, and adverse outcomes in adolescent samples. Methods: A meta-analysis was performed on peer-reviewed articles and unpublished dissertations from 1975 to 2016 in five computerized databases ERIC, PsycINFO, MEDLINE, PubMed and ProQuest. Quantitative studies were retained if they included the following keywords: active consent (or informed consent or parental consent), passive consent (or waiver of consent), risk behavior, adolescen*. Results: Fifteen studies were identified with a total number of 104,074 children. Results showed (1) response rates were significantly lower for studies using active consent procedure than those using passive consent procedure (Z ¼ 3.05, p ¼ .002); (2) more females, younger participants, and less African-Americans were included in studies using active consent procedures than studies using passive procedures (Z ¼ 2.73, p ¼ .006; Z ¼ 12.06, p < .00001; Z ¼ 2.19, p ¼ .03, respectively); (3) studies with passive consent procedures showed higher rates of self-reported substance use than studies using active consent procedures (Z ¼ 3.07, p ¼ .002). Conclusions: Requiring active parental consent can lead to a systematic bias in the sample where the population under study is misrepresented. Institutional review board committees should collaborate with researchers to find solutions that protect minors without silencing the voice of high-risk youth in the literature. Ó 2017 Society for Adolescent Health and Medicine. All rights reserved.
Parental informed consent is an established federal regulation for protecting minors from potential harms or risks introduced by research. These procedures ensure that parents or guardians
Conflicts of Interest: The authors have no conflicts of interest to disclose. * Address correspondence to: Chao Liu, M.A., Department of Human Development and Family Science, Oklahoma State University, Stillwater, OK 74075. E-mail address: [email protected] (C. Liu). 1054-139X/Ó 2017 Society for Adolescent Health and Medicine. All rights reserved. http://dx.doi.org/10.1016/j.jadohealth.2017.01.015
IMPLICATIONS AND CONTRIBUTION
Selection biases can result from requiring parental consent when studying adolescents’ risk behaviors related to substance use, limiting the generalizability of research findings. Alternative consent procedures such as passive parental consent may be used when research involves substance use behaviors among adolescents.
are given adequate information about the purpose, benefits, and risks associated with a research study to make an informed decision about their child’s participation. The U.S. Department of Health and Human Services (HHS) regulations at Code of Federal Regulations 46.116 state that “no investigator may involve a human being as a subject in research covered by this policy unless the investigator has obtained the legally effective informed consent of the subject or the subject’s legally
2
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
authorized representative.” However, these regulations also acknowledge that under some circumstances an institutional review board (IRB) may approve a consent procedure that waives or alters parts of this requirement [1]. Although not formally designated in HHS regulations, researchers in practice often refer to two forms of parental consent procedures: active and passive. In “active consent” parents must be informed of the research and provide their permission before a minor child is allowed to participate in any aspect of the research project. “Passive consent” procedures generally describe situations in which the requirement for written permission is waived unless a parent restricts their child’s participation via the opt out method specified by the researcher. Under passive consent procedures, parental nonresponse is considered as a permission to participate in the research study. Potential biases of active consent procedure The U.S. Department of Education, IRBs, and local school boards have increasingly required more stringent active consent procedures as the result of environmental and statutory changes. A gradual trend toward requiring active parental consent has been observed in 1995 when the House passed the Family Privacy Act that sought to require active parental consent for any government-funded program or activity. Although this Act failed to pass in the Senate, similar provisions were passed in five states (California, Indiana, Kansas, Michigan, and Texas) with several other states putting it under consideration [2]. The increasing preference of active consent over passive consent among school administrators echoes these regulatory changes, probably reflecting a “free of troubles” line of thought. For example, although DHS regulations allow waiving parental consents under certain conditions such as the case that the waiver will not adversely affect the rights and welfare of the child participating in the research or the situation when research could not be practicably carried out without a waiver of parent permission, studies meeting these conditions that have been granted a waiver of parental consent are very limited [3]. When requirements for consent become more stringent, however, science runs the risk of losing the very subjects that are the target of their investigation or intervention [4] and of stifling the voice of those most in need of answers. For example, requiring parental consent may jeopardize the opportunity to conduct research on risk behaviors that begin in adolescence such as suicide, sexually transmitted diseases, pregnancy, and substance use which are the major causes of morbidity and mortality during this period of time [5,6]. Several potential problems may arise from following an active consent procedure, not the least of which is a low response rate. Previous research has reported response rates using active consent ranging from 29% to 60% [7e11], much lower than the response rates of 79%e100% under passive consent procedures [7e9,12e14]. Although various retrieval methods and delivery strategies can boost the response rate under active consent procedure, these practices are often labor intensive and costly. For example, Ellickson and Hawes [7] made intensive follow-up efforts to improve parental consent rates by 52%, but the high cost of the procedures significantly reduced the size of the study the authors were able to conduct. A low response rate arising from active consent procedures could compromise the validity of research findings due to the introduction of sampling biases. Sampling bias results from nonrandom selection into a study. The low-response rates
reported in active consent studies reflect a systematic bias of sampling in a way that over-represents some portions of the population while under-representing others. For example, active consent procedures yielded study samples that over-represented female and Caucasian students [15,16] and students with high academic achievement [10], and under-represented minority groups such as African-American and Asian American students, Hispanic youth [9,17], children who were low achievers, and children whose parents were less well-educated [15,16]. Significant differences between respondents and nonrespondents may also emerge in the prevalence of risk behaviors when active consent procedures are used [12,17]. For example, Severson and Ary [18] reported a significantly higher number of risk behaviors (e.g., smoke tobacco, marijuana, and drink alcohol) reported by students whose parents did not provide consent compared to students with consenting parents. Estimates of prevalence rates may also be affected by an interaction between consent procedures and subject characteristics. For example, age of the student has also been found to influence self-report of some risk behaviors between samples requiring parental consent versus those that do not. Requiring written parental consent significantly reduced self-report of smoking among ninth grade students when compared with a sample in the same grade not requiring written parental consent, although this difference disappeared among 12th grade students [15]. Counterargument to biases of active consent procedure Notwithstanding the problems associated with active consent procedures, others debate the extent and significance of the bias. For example, Dent, Sussman, and Stacy [19] examined the differences between a full school-based sample and subsamples restricted by the requirement of active parental consent on several outcomes of a substance use survey. Little or no bias was found on measures related to mental health, drug use, or violence, despite some small differences in demographic variables. And although males and African-American students were under-represented in the sample, behavioral outcomes and mental health indicators were generally unaffected by the consent format (i.e., within statistical sampling error). Similarly, in another study [15] where a sensitive health survey was used to examine the impact of the active consent procedure, the authors found no significant differences between active and passive consent groups for self-report of alcohol or illicit drug use although several demographic differences (e.g., ethnicity and family structure) still existed. Given these inconsistencies, this study aims to systematically examine the role of parental consent in research involving risk behaviors among adolescents by conducting a meta-analysis on studies that compared an active parental consent procedure with a passive parental consent procedure. Specifically, we examine whether an active consent procedure leads to differences in response rate, demographic characteristics, and estimates of substance use and nonesubstance use risk behaviors among adolescents. Methods Selection of studies The following steps guided our approach to locate studies and to maximize the chance of including all relevant studies. First, we
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
3
577 articles identified through database searching ERIC: 44 ProQuest Dissertation: 175 PsycINFO: 326 PubMed: 10 Social Services Abstract: 22
482 articles retained after duplicates were removed 432 articles excluded based on screening of targeted keywords in titles and/or abstracts 50 potentially relevant articles identified for further assessment
1 article identified from references lists
51 full-text articles assessed for eligibility
37 full-text articles excluded with followed-up inclusion criteria 14 articles included in quantitative synthesis (meta-analysis) Figure 1. Procedure of study selection.
conducted computerized literature searches using five online databases: ERIC, PsycINFO, PubMed, Social Services Abstract, and ProQuest (for unpublished dissertations). The following keywords were used to guide the search of articles published from 1975 (The year 1975 was chosen because General Requirements for Informed Consent [C.F.R. Sec. 46.116] was established in 1974) to 2016: (active consent OR informed consent OR parental consent) AND (passive consent OR waiver of consent [Although waiver of consent could be an alternative option to active and passive consent procedure, we found no studies under this category probably because the focus of our study was on general survey research of adolescents’ risk behaviors which is typically not qualified for the waiver of parental consent]) AND (risk behavior) AND (adolescen*). Figure 1 shows the study selection procedure. Original search criteria across the five databases yielded 577 results. After removing the duplicates, 482 articles were identified. Initial screening of titles and abstracts excluded 432 articles because those articles did not explicitly state their consent procedures or did not involve any adolescent’s risk behaviors, reducing the total to 50 articles. We then examined the references cited in these articles to include any article we may have previously missed. One additional study was found [20]. This yielded a pool of 51 studies. Three inclusion criteria were further used to select studies from this pool for the meta-analysis: (1) studies that were quantitative; (2) studies that included independent samples of
active and passive consent (Only studies with independent samples for active and passive consent conditions were selected because they would not introduce the potential confounding effects from repeated measure designs. For example, parents may decline their child’s participation in a study under active consent procedure but later give no response when the same study switches to using a passive consent procedure because they think they have already clearly expressed their intention); and (3) studies that were conducted in the United States to reflect consistency in policy and implementation. First, because our focus was on quantitative studies, 13 qualitative studies were excluded, leaving 38 studies. Second, 19 studies were excluded because of nonindependent samples, leaving 19 studies. Third, five non-U.S. studies were excluded. These filters produced a final list of 14 studies with a total number of 104,074 children (see Table 1 for sample study characteristics).
Coding of studies All studies were coded on information from both active and passive consent groups, which include sample size, response rate, age, gender, ethnicity, and relevant outcomes of risk behaviors. We also coded the effect sizes on each of those variables. If effect sizes were not provided, we calculated them by the following procedures:
4
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
Table 1 Descriptive summary of sample study characteristics Study
Passive consent (N)
Active consent (N)
Age or grade
Ethnicity
Gender (% girl)
Setting
Study design
Variable
White and others
51.20
School
Obs
White, black, Hispanics, Asian, and Native American White, black, and Hispanics
75.50
School
Obs
Substance use; health statusa Psychological riska
48.23
School
Expb
Anderman et al. (1995) [15]
8,645
8,645
Chartier et al. (2008) [21]
1,011
1,021
9th and 12th grade 6th grade
Courser et al. (2009) [11]
3,750
2,190
14.90 years
769
2,727
6th grade
White, black, Hispanics, and others
48.50
School
Obs
Depue (2009) [20]
13,930
4,093
6th, 9th, and 12th grade
51.00
School
Obs
Eaton et al. (2004) [13]
10,118
5,722
9th through 12th grade 7th grade
White, black, Hispanics, Asian, American Indian/ Alaskan Native, and Native Hawaiian or Other Pacific Islander White, black, Hispanics, Asian, and others d
51.30
School
Obs
d
School
Obs
White, black, Hispanics, and others White, black, Hispanics, Asian/Pacific Islander, Native American, and others or mixed White, black, Hispanics, Asian, American Indian, and Native Hawaiian, and others Black, Hispanics, and others d
50.00
School
Expb
52.00
School
Obs
Attitude measurea; antisocial behaviora Substance use
55.70
School
Obs
Substance use
48.80 47.00
School School
Obs Obs
Health statusa Substance use
70.80
Clinic
Obs
44.50
School
Obs
Substance use; delinquent behaviora Substance use
Dent et al. (1993) [19]
Ellickson and Hawes (1989) [7] Esbensen et al. (1999) [17]
117
86
1,083
1,411
12.12 years
Frissell et al. (2004) [22]
5,903
3,176
9th through 12th grade
Horn et al. (2009) [23]
4,922
968
Mellor et al. (2008) [24] Pokorny et al. (2001) [12]
2,104 7,127
2,330 7,138
Rojas et al. (2008) [25]
670
411
Unger et al. (2004) [26]
649
3,358
16.05 years
11.41 years 6th through 8th grade 16.55 years 6th grade
White, black, Hispanics, Asian, and others White, black, Hispanics, Asian/Pacific Islander, and others
Substance use; antisocial behaviora Substance use; socialitya; personalitya Response rate
Substance use; psychological riska Response rate
Obs ¼ observation; Exp ¼ experiment. a Variables that were categorized as nonesubstance use risk behaviors’ for the purpose of subsequent analysis. b More weights were placed on these studies in calculating the overall effect size given the experimental design they used.
1). If the mean and standard deviation were reported in the study (i.e., continuous outcome), then standardized mean difference (SMD) was calculated
SMD ¼
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðn1 1Þs21 þ ðn2 1Þs22 ; s ¼ n1 þ n2 2 where n1 is the sample size of the passive group and n2 is the sample size of the active group and s21 is the variance of the passive group and s22 is the variance of the active group.
x1 x2 ; s
where x1 is the mean of the passive group and x2 is the mean of the active group and
Odds Ratio ðORÞ ¼
2). For those studies that reported percentages (i.e., dichotomous outcome), we calculated the odds ratio using the formula:
% of children who display the behavior in passive group % of children who do not display the behavior in passive group % of children who display the behavior in active group % of children who do not display the behavior in active group
:
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
To combine studies with dichotomous and continuous outcomes, we converted SMD to OR using the formula [27]:
SMD ¼
pffiffiffi 3
p
InOR
Results Before examining the effects of requiring parental consent on the outcome variables of interest, we inspected the distribution of effect sizes by checking the Q and I2 statistics (Table 2). The Q statistic is distributed as a chi-square with k-1 degrees of freedom, where k is the number of effect sizes. Heterogeneity indicated by a p value of less than .05 warrants the use of random effects model. Random effects model assumes that the variability of effect sizes beyond subject-level sampling error is random and therefore cannot be identified [28]. A random effects model is more conservative under this circumstance because the unidentified sources of variability would lead to wider confidence intervals (CIs) for tests of significance which further increase the uncertainty of estimating the population mean. Because significant heterogeneity was found across studies on all the outcome variables except age, we applied random effects models to all outcomes with significant heterogeneity and a fixed effect model to age. We used Cochrane Review Manager (RevMan) 5.3 to derive all the inferential statistics (i.e., Z-score, p value, OR, and CI) as well as the forest plot. Response rate and demographic characteristics in active and passive consent groups There were 11 studies that reported the association between response rate and consent procedure. As shown in Figure 2, the response rate of passive consent groups was significantly higher than that of active consent groups (Z ¼ 3.05, p ¼ .002; OR ¼ 2.39; 95% CI ¼ 1.37e4.20). Significant selection biases were also found in the gender makeup (Z ¼ 2.73, p ¼ .006; OR ¼ .80; 95% CI ¼ .68e.94) and age composition of participants (Z ¼ 12.06, p < .00001; OR ¼ .53; 95% CI ¼ .48e.59): the active consent group contained more females and younger participants than the passive consent group (See Supplementary data for the forest plots of gender and age difference.). There were also more Black participants in the passive consent group (Z ¼ 2.19, p ¼ .03; OR ¼ 1.53; 95% CI ¼ 1.05, 2.23; see Supplementary data for the forest plot of difference of AfricanAmerican participants.). More white participants were found in the active consent group, whereas more Hispanic participants were found in the passive consent group but these differences were not significant (Z ¼ .37, p ¼ .71 and Z ¼ .15, p ¼ .88, respectively). Table 2 Heterogeneity test statistics Q (c2) Response rate Gender Age Ethnic composition-black Ethnic composition-white Ethnic composition-Hispanic a
p < .001.
I2 a
2294.52 211.99a .02 105.75a 364.56a 138.27a
100% 95% 0% 93% 98% 95%
5
Substance use and nonesubstance use risk behaviors in active and passive consent groups We also assessed how the two consent procedures are differentially correlated with the report of substance use and nonesubstance use risk behaviors. Participants in the passive consent groups reported a higher level of risk for substance use than those in the active consent groups (Z ¼ 3.07, p ¼ .002; OR ¼ 1.21; 95% CI ¼ 1.07, 1.37; see Supplementary data for the forest plot of substance use risk behaviors.). The reported rate of nonesubstance use risk behaviors was also higher in passive consent groups than that in active consent groups, but the difference was not significant (Z ¼ 1.65, p ¼ .10). Publication bias assessment We used Stata 13 to assess publication bias by examining the relationship between effect size and its associated variance. The presence of publication bias is indicated by the high correlation between large variances (caused by studies with small sample sizes) and large effect sizes [29]. We conducted Begg and Mazumdar’s rank correlation test and Egger’s regression for this correlation assessment. While these two tests are similar in terms of quantifying the relationship between effect size and precision, Egger’s regression uses the actual values instead of ranks and thus has higher power [29]. Nevertheless, both tests can detect the presence of publication bias based on the significant correlation or regression. None of the outcomes was significant on these two tests (see Table 3), showing no sign of publication bias. Discussion To our knowledge this is the first meta-analysis that systematically examines the impacts of requiring active parental consent, despite decades of debate regarding the effects of this practice on adolescent research. There were three primary results from this study: (1) response rates were significantly lower for studies using active consent procedure than those using passive consent procedure; (2) demographic makeup of the samples varied substantially between studies using active and passive consent procedures; and (3) rates of self-reported substance use were higher in studies using passive consent than those in active consent procedures. At least two reasons may help explain the cause of low response rates under an active consent procedure particularly when it involves self-report of risk behaviors. First, an active consent procedure requires more effort from parents. Parent involvement varies widely and has been linked to family characteristics such as structure (e.g., lower among single-parent households), race (e.g., higher involvement for European American parents than Hispanic and African-American parents) [30], and with an array of negative child outcomes such as poor academic performance, substance use, teen pregnancy, poor emotion regulation, and antisocial behaviors [31,32]. Requiring active parental consent from parents whose involvement is already low may introduce sampling bias that systematically limits access to those children whose parents are not engaged enough to read, sign, and return a consent form in a timely manner. Consequently, some youth may be excluded from research by virtue of family characteristics, which creates the
6
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
Study Unger g et al. ((2004)) Mellor et al. (2008) Anderman et al. (1995) ( ) Pokorny y et al. ((2001) Eaton et al. ((2004)) Ellickson et al. (1989) ( ) Chartier et al. (2008) Depue p ((2009)) j et al. ((2008)) Rojas Courser et al. ((2009)) Frissell et al. (2004)
Odds Ratio [95% CI] 0.17 [[0.13,, 0.21]] 1.05 [0.92, 1.20] 1.52 [[1.43,, 1.62]] 1.61 [[1.46,, 1.78]] 1.91 [[1.76,, 2.08]] 2.21 [[0.86,, 5.67]] 2.98 [[2.40,, 3.70]] 4.46 [[3.91,, 5.10]] 5.90 [[4.48,, 7.75]] 8.95 [[7.93,, 10.10]] 11.50 [10.30, 12.85] 2.39 [1.37, 4.20]
Overall effect size
0.05
0.2 1 5 20 Odds Ratio, 95% CI
Figure 2. Effect sizes and confident intervals of response rate.
potential for requiring parental consent to be in opposition to the principles of justice and inclusiveness. A second reason for lower response rates when requiring parental consent is that adolescents may have concerns that their desire to participate in a study could be interpreted as admission of engaging in the behavior [5,6]. In this case, the low response rate may be due to adolescents’ failure to notify their parents when some sensitive information is included in the study. This may also help explain the significant difference of substance use rather than nonesubstance use risk behaviors in the active and passive consent conditions. Because others have found that parental consent varies by the topic matter and scope of the research [33], it may be that youth are less likely to pursue parental consent for sensitive study topics such as substance use for which there is risk that participation could create a problem with their parents. The fact that more female adolescent participants were included in studies using active parental consent procedure may further contribute to selection bias in research involving risk behaviors. The underestimated prevalence of substance use in the active consent condition may result from an underrepresentative population of male adolescent participants, a population of which a higher prevalence of substance use has been well documented compared with girls [34]. This type of systematic bias on sample selection has been speculated to distort the relation between predictors and the observed outcomes [12]. For example, a study under active consent procedure
containing more female participants may fail to find any significant association between substance use and aggressive behaviors that is in fact present. The failure to detect an existed effect can be ascribed to high homogeneity of the sample (i.e., the prevalence of aggressive behaviors and substance use are both quite low for girls). This creates high intraclass correlations which tend to increase type II error. Increasing sample size to adjust for the increased intraclass correlation is one solution, but it leads to higher study costs, thereby creating barriers for research [35]. The higher proportion of younger participants in active consent procedure underlines the assumption that younger adolescents are not capable to consent to research involving risk behaviors, and therefore, a more capable representative is needed. This assumption, however, is ungrounded given previous evidence showing that children as young as 7 demonstrated an adequate understanding of their research rights such as freedom to ask questions and withdraw, time involved, and potential harms and benefits of their participation [36]. Other studies have shown that adolescents around the age of 13 years had the similar ability to make reasonable health-related decisions as adults [37,38]. These competencies of young adolescents seem to be underestimated in reality, and the pervasive implementation of active parental consent procedure will further reinforce this underestimation, making the voice of minors less likely to be heard. It is not our intention to push for the complete replacement of active consent procedure with
Table 3 Publication bias assessment
Response rate Gender Age Ethnic composition-black Ethnic composition-white Ethnic composition-Hispanic Substance use risk behaviors Nonesubstance use risk behaviors
Begg and Mazumdar’s rank correlation (s)
p
Egger’s regression(b, [95% CI])
p
1.25 .08 1.00 .74 1.16 .00 .15 .52
.21 .94 .32 .46 .25 1.00 .88 .60
4.06 [5.82 to 13.92] 1.61 [4.07 to 7.29] NAa 2.46 [3.21 to 8.13] .03 [11.65 to 11.58] 4.03 [4.19 to 12.26] 1.34 [4.92 to 7.60] 4.30 [65.52 to 74.12]
.38 .54 NA .33 1.00 .28 .61 .58
CI ¼ confidence interval. a Egger’s regression test was not available because the minimum number of studies required for this test is three.
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
passive consent procedure, but there should be decreasing emphasis on parental consent and a corresponding increasing influence from the autonomous decisions made by adolescents. Social justice The potential to exclude specific categories of adolescents from research, particularly that which is focused on the use of substances, raises an issue of social justice. The practice of requiring active consent appears to limit generalizability of research findings due to sampling bias. The findings of this study highlight the likelihood that the population most likely not appropriately represented in active consent studies will be female and/or African-American. This becomes an issue of social injustice as any policy developed from substance use research on adolescents will be based on a population that has been systematically (although not intentionally) over-represented, resulting in a lack of attention to other certain populations. The potential for excluding some adolescents from research is exacerbated because many investigators are reluctant to include minors in research due to additional steps involved in the IRB process, complications in obtaining parental consent, and legality concerns associated with research when a waiver of parental consent is granted [39]. What’s more, if research has no prospect of direct benefits to the child, but knowledge vital to understanding a disorder may be gained, the risk category may be adjusted to “greater than minimal risk.” Under this risk category, HHS guidelines for research require permission be obtained from both parents, unless researchers can establish that one parent is deceased, unknown, incompetent, or not reasonably available, or when only one parent has legal responsibility for the care and custody of the child [26]. Social injustice may also reflect in intervention efforts. For example, intervention effects found in samples using active consent procedure may not be generalizable to a larger population because active consent procedures are more likely to screen out socially disadvantaged groups who have higher needs for the intervention, resulting in an underestimated effect for populations with greater diversity [15]. Accordingly, resources such as the distribution of federal funds may be misallocated away from populations most in need of help. Therefore, rather than dogmatically applying active parental consent procedures for all studies involving minors, members of IRBs and researchers should carefully evaluate the costs and benefits of both active and passive consent procedures on a case-by-case basis. Limitations Several limitations of this study should be noted. First, the number of studies included in our meta-analysis is small. Despite of a large pool of individual participants that affords plenty of within-study variances, between-study variances are limited which make it difficult to generalize based on the findings. For example, although different ethnic composition was found for studies using active versus passive consent procedure, it is likely that some other sociodemographic factors such as geographic location, income, education, and immigration status may further contribute to these ethnic differences. The lower number of black participants in particular could also be due to the distrust arising from prior history of some unethical research conduct performed on black minority subjects [40]. These sources of information, however, were missing from those studies included in our
7
meta-analysis and so we cannot compare studies on these levels. Future studies need to incorporate more information to gain more insights into this area. A second limitation of this metaanalysis is that we only included studies conducted in the United States, and therefore, it is impossible for us to make an inference about the situation in other countries. Nevertheless, the procedure required for obtaining parental permission could vary from country to country and is thus the reason that we excluded studies from other countries. More studies are certainly needed to explore the impact of different consent procedures for research conducted in other countries. Supplementary Data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jadohealth.2017.01.015. References [1] Services USDoHaH. Can consent or parental permission ever be “passive” or “implied?”. 2011. Available at: http://www.hhs.gov/ohrp/policy/faq/ informed-consent/can-consent-be-passive-or-implied.html. Accessed February 22, 2016. [2] Ross JG, Sundberg EC, Flint KH. Informed consent in school health research: Why, how, and making it easy. J Sch Health 1999;69:171e6. [3] Iltis AS. Parents, adolescents, and consent for research participation. J Med Philos 2013;38:332e46. [4] Baker JR, Yardley JK, McCaul K. Characteristics of responding-, nonresponding- and refusing-parents in an adolescent lifestyle choice study. Eval Rev 2001;25:605e18. [5] Flicker S, Guta A. Ethical approaches to adolescent participation in sexual health research. J Adolesc Health 2008;42:3e10. [6] Rew L, Taylor-Seehafer M, Thomas N. Without parental consent: Conducting research with homeless adolescents. J Soc Pediatr Nurs 2000;5: 131e8. [7] Ellickson PL, Hawes JA. An assessment of active versus passive methods for obtaining parental consent. Eval Rev 1989;13:45e55. [8] Esbensen F-A, Deschenes EP, Vogel RE, et al. Active parental consent in school-based research. An examination of ethical and methodological issues. Eval Rev 1996;20:737e53. [9] Kearney KA, Hopkins RH, Mauss AL, Weisheit RA. Sample bias resulting from a requirement for written parental consent. Public Opin Q 1983;47: 96e102. [10] Lueptow L, Mueller SA, Hammes RR, Master LS. The impact of informed consent regulations on response rate and response bias. Sociol Methods Res 1977;6:183e204. [11] Courser MW, Shamblen SR, Lavrakas PJ, Collins D, Ditterline P. The Impact of active consent procedures on nonresponse and nonresponse error in youth survey data: Evidence from a new experiment. Eval Rev 2009;33: 370e95. [12] Pokorny SB, Jason LA, Schoeny ME, et al. Do participation rates change when active consent procedures replace passive consent. Eval Rev 2001; 25:567e80. [13] Eaton DK, Lowry R, Brener ND, et al. Passive versus active parental permission in school-based survey research: Does the type of permission affect prevalence estimates of risk behaviors? Eval Rev 2004;28:564e77. [14] Langhinrichsen-Rohling J, Arata C, O’Brien N, et al. Sensitive research with adolescents: Just how upsetting are self-report surveys anyway? Violence Vict 2006;21:425e44. [15] Anderman C, Cheadle A, Curry S, et al. Selection bias related to parental consent in school-based survey research. Eval Rev 1995;19:663e74. [16] Dent CW, Galaif J, Sussman S, et al. Demographic, psychosocial and behavioral differences in samples of actively and passively consented adolescents. Addict Behav 1993;18:51e6. [17] Esbensen F-A, Miller MH, Taylor T, et al. Differential attrition rates and active parental consent. Eval Rev 1999;23:316e35. [18] Severson HH, Ary DV. Sampling bias due to consent procedures with adolescents. Addict Behav 1983;8:433e7. [19] Dent CW, Sussman SY, Stacy AW. The impact of a written parental consent policy on estimates from a school-based drug use survey. Eval Rev 1997; 21:698e712. [20] Depue SM. Psychometrics of the Missouri student Survey: Examining validity, reliability and consent. St. Louis, MO: University of Missouri-Saint Louis; 2009.
8
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
[21] Chartier M, Stoep AV, McCauley E, et al. Passive versus active parental permission: Implications for the ability of school-based depression screening to reach youth at risk. J Sch Health 2008;78:157e64. [22] Frissell KC, McCarthy DM, D Amico EJ, et al. Impact of consent procedures on reported levels of adolescent alcohol use. Psychol Addict Behav 2004; 18:307. [23] Horn KA, Branstetter SA, Dino GA, et al. Potential effects of active parental consent: Enrolling teen smokers into a school-based cessation program. Nicotine Tob Res 2009;11:1359e67. [24] Mellor JM, Rapoport RB, Maliniak D. The impact of child obesity on active parental consent in school-based survey research on healthy eating and physical activity. Eval Rev 2008;32:298e312. [25] Rojas NL, Sherrit L, Harris S, Knight JR. The role of parental consent in adolescent substance use research. J Adolesc Health 2008;42:192e7. [26] Services USDoHaH. Code of Federal Regulations: Requirements for permission by parents or guardians and for assent by children. 2010. Available at: http://www.hhs.gov/ohrp/humansubjects/guidance/45cfr46 .html#46.408. Accessed February 22, 2016. [27] Chinn S. A simple method for converting an odds ratio to effect size for use in meta-analysis. Stat Med 2000;19:3127e31. [28] Lipsey MW, Wilson DB. Practical meta-analysis. Thousand Oaks, CA: Sage Publications, Inc; 2001. [29] Rothstein HR, Sutton AJ, Borenstein M. Publication bias in meta-analysis: Prevention, assessment and adjustments. Chichester, UK: John Wiley & Sons; 2006. [30] Lee J-S, Bowen NK. Parent involvement, cultural capital, and the achievement gap among elementary school children. Am Educ Res J 2006;43: 193e218.
[31] Griffin KW, Botvin GJ, Scheier LM, et al. Parenting practices as predictors of substance use, delinquency, and aggression among urban minority youth: Moderating effects of family structure and gender. Psychol Addict Behav 2000;14:174. [32] Morris AS, Silk JS, Steinberg L, et al. The role of the family context in the development of emotion regulation. Social Dev 2007;16:361e88. [33] Pasternak RH, Geller G, Parrish C, Cheng TL. Adolescent and parent perceptions on youth participation in risk behavior research. Arch Pediatr Adolesc Med 2006;160:1159e66. [34] Johnston LD, O’malley PM, Bachman JG, Schulenberg JE. Monitoring the Future national survey results on drug use, 1975-2010. Vol II, College Students & Adults Ages 19e50. Ann Arbor, MI: Institute for Social Research; 2011. [35] White VM, Hill DJ, Effendi Y. How does active parental consent influence the findings of drug-use surveys in schools? Eval Rev 2004;28:246e60. [36] Susman EJ, Dorn LD, Fletcher JC. Participation in biomedical research: The consent process as viewed by children, adolescents, young adults, and physicians. J Pediatr 1992;121:547e52. [37] Tait AR, Voepel-Lewis T, Malviya S. Do they understand? (part II): Assent of children participating in clinical anesthesia and surgery research. Anesthesiology 2003;98:609e14. [38] Weithorn LA, Campbell SB. The competency of children and adolescents to make informed treatment decisions. Child Development 1982;53:1589e98. [39] Hester CJ. Adolescent consent: Choosing the right path. Issues Compr Pediatr Nurs 2004;27:27e37. [40] Shavers VL, Lynch CF, Burmeister LF. Racial differences in factors that influence the willingness to participate in medical research studies. Ann Epidemiol 2002;12:248e56.
www.jahonline.org Original article
The Effects of Requiring Parental Consent for Research on Adolescents’ Risk Behaviors: A Meta-analysis Chao Liu, M.A. a, *, Ronald B. Cox Jr., Ph.D. a, Isaac J. Washburn, Ph.D. a, Julie M. Croff, Ph.D. b, and Hugh C. Crethar, Ph.D. b a b
Department of Human Development and Family Science, Oklahoma State University, Stillwater, Oklahoma School of Applied Health and Educational Psychology, Oklahoma State University, Stillwater, Oklahoma
Article history: Received August 17, 2016; Accepted January 31, 2017 Keywords: Parental consent; Active consent; Passive consent; Risk behaviors; Adolescents; Meta-analysis
A B S T R A C T
Purpose: Requiring parental consent may result in sampling biases that confound scientific conclusions and stifle the representation of children most at risk for adverse outcomes. This study aims to investigate whether active parental consent, compared with passive parental consent, creates a bias in response rate, demographic makeup, and adverse outcomes in adolescent samples. Methods: A meta-analysis was performed on peer-reviewed articles and unpublished dissertations from 1975 to 2016 in five computerized databases ERIC, PsycINFO, MEDLINE, PubMed and ProQuest. Quantitative studies were retained if they included the following keywords: active consent (or informed consent or parental consent), passive consent (or waiver of consent), risk behavior, adolescen*. Results: Fifteen studies were identified with a total number of 104,074 children. Results showed (1) response rates were significantly lower for studies using active consent procedure than those using passive consent procedure (Z ¼ 3.05, p ¼ .002); (2) more females, younger participants, and less African-Americans were included in studies using active consent procedures than studies using passive procedures (Z ¼ 2.73, p ¼ .006; Z ¼ 12.06, p < .00001; Z ¼ 2.19, p ¼ .03, respectively); (3) studies with passive consent procedures showed higher rates of self-reported substance use than studies using active consent procedures (Z ¼ 3.07, p ¼ .002). Conclusions: Requiring active parental consent can lead to a systematic bias in the sample where the population under study is misrepresented. Institutional review board committees should collaborate with researchers to find solutions that protect minors without silencing the voice of high-risk youth in the literature. Ó 2017 Society for Adolescent Health and Medicine. All rights reserved.
Parental informed consent is an established federal regulation for protecting minors from potential harms or risks introduced by research. These procedures ensure that parents or guardians
Conflicts of Interest: The authors have no conflicts of interest to disclose. * Address correspondence to: Chao Liu, M.A., Department of Human Development and Family Science, Oklahoma State University, Stillwater, OK 74075. E-mail address: [email protected] (C. Liu). 1054-139X/Ó 2017 Society for Adolescent Health and Medicine. All rights reserved. http://dx.doi.org/10.1016/j.jadohealth.2017.01.015
IMPLICATIONS AND CONTRIBUTION
Selection biases can result from requiring parental consent when studying adolescents’ risk behaviors related to substance use, limiting the generalizability of research findings. Alternative consent procedures such as passive parental consent may be used when research involves substance use behaviors among adolescents.
are given adequate information about the purpose, benefits, and risks associated with a research study to make an informed decision about their child’s participation. The U.S. Department of Health and Human Services (HHS) regulations at Code of Federal Regulations 46.116 state that “no investigator may involve a human being as a subject in research covered by this policy unless the investigator has obtained the legally effective informed consent of the subject or the subject’s legally
2
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
authorized representative.” However, these regulations also acknowledge that under some circumstances an institutional review board (IRB) may approve a consent procedure that waives or alters parts of this requirement [1]. Although not formally designated in HHS regulations, researchers in practice often refer to two forms of parental consent procedures: active and passive. In “active consent” parents must be informed of the research and provide their permission before a minor child is allowed to participate in any aspect of the research project. “Passive consent” procedures generally describe situations in which the requirement for written permission is waived unless a parent restricts their child’s participation via the opt out method specified by the researcher. Under passive consent procedures, parental nonresponse is considered as a permission to participate in the research study. Potential biases of active consent procedure The U.S. Department of Education, IRBs, and local school boards have increasingly required more stringent active consent procedures as the result of environmental and statutory changes. A gradual trend toward requiring active parental consent has been observed in 1995 when the House passed the Family Privacy Act that sought to require active parental consent for any government-funded program or activity. Although this Act failed to pass in the Senate, similar provisions were passed in five states (California, Indiana, Kansas, Michigan, and Texas) with several other states putting it under consideration [2]. The increasing preference of active consent over passive consent among school administrators echoes these regulatory changes, probably reflecting a “free of troubles” line of thought. For example, although DHS regulations allow waiving parental consents under certain conditions such as the case that the waiver will not adversely affect the rights and welfare of the child participating in the research or the situation when research could not be practicably carried out without a waiver of parent permission, studies meeting these conditions that have been granted a waiver of parental consent are very limited [3]. When requirements for consent become more stringent, however, science runs the risk of losing the very subjects that are the target of their investigation or intervention [4] and of stifling the voice of those most in need of answers. For example, requiring parental consent may jeopardize the opportunity to conduct research on risk behaviors that begin in adolescence such as suicide, sexually transmitted diseases, pregnancy, and substance use which are the major causes of morbidity and mortality during this period of time [5,6]. Several potential problems may arise from following an active consent procedure, not the least of which is a low response rate. Previous research has reported response rates using active consent ranging from 29% to 60% [7e11], much lower than the response rates of 79%e100% under passive consent procedures [7e9,12e14]. Although various retrieval methods and delivery strategies can boost the response rate under active consent procedure, these practices are often labor intensive and costly. For example, Ellickson and Hawes [7] made intensive follow-up efforts to improve parental consent rates by 52%, but the high cost of the procedures significantly reduced the size of the study the authors were able to conduct. A low response rate arising from active consent procedures could compromise the validity of research findings due to the introduction of sampling biases. Sampling bias results from nonrandom selection into a study. The low-response rates
reported in active consent studies reflect a systematic bias of sampling in a way that over-represents some portions of the population while under-representing others. For example, active consent procedures yielded study samples that over-represented female and Caucasian students [15,16] and students with high academic achievement [10], and under-represented minority groups such as African-American and Asian American students, Hispanic youth [9,17], children who were low achievers, and children whose parents were less well-educated [15,16]. Significant differences between respondents and nonrespondents may also emerge in the prevalence of risk behaviors when active consent procedures are used [12,17]. For example, Severson and Ary [18] reported a significantly higher number of risk behaviors (e.g., smoke tobacco, marijuana, and drink alcohol) reported by students whose parents did not provide consent compared to students with consenting parents. Estimates of prevalence rates may also be affected by an interaction between consent procedures and subject characteristics. For example, age of the student has also been found to influence self-report of some risk behaviors between samples requiring parental consent versus those that do not. Requiring written parental consent significantly reduced self-report of smoking among ninth grade students when compared with a sample in the same grade not requiring written parental consent, although this difference disappeared among 12th grade students [15]. Counterargument to biases of active consent procedure Notwithstanding the problems associated with active consent procedures, others debate the extent and significance of the bias. For example, Dent, Sussman, and Stacy [19] examined the differences between a full school-based sample and subsamples restricted by the requirement of active parental consent on several outcomes of a substance use survey. Little or no bias was found on measures related to mental health, drug use, or violence, despite some small differences in demographic variables. And although males and African-American students were under-represented in the sample, behavioral outcomes and mental health indicators were generally unaffected by the consent format (i.e., within statistical sampling error). Similarly, in another study [15] where a sensitive health survey was used to examine the impact of the active consent procedure, the authors found no significant differences between active and passive consent groups for self-report of alcohol or illicit drug use although several demographic differences (e.g., ethnicity and family structure) still existed. Given these inconsistencies, this study aims to systematically examine the role of parental consent in research involving risk behaviors among adolescents by conducting a meta-analysis on studies that compared an active parental consent procedure with a passive parental consent procedure. Specifically, we examine whether an active consent procedure leads to differences in response rate, demographic characteristics, and estimates of substance use and nonesubstance use risk behaviors among adolescents. Methods Selection of studies The following steps guided our approach to locate studies and to maximize the chance of including all relevant studies. First, we
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
3
577 articles identified through database searching ERIC: 44 ProQuest Dissertation: 175 PsycINFO: 326 PubMed: 10 Social Services Abstract: 22
482 articles retained after duplicates were removed 432 articles excluded based on screening of targeted keywords in titles and/or abstracts 50 potentially relevant articles identified for further assessment
1 article identified from references lists
51 full-text articles assessed for eligibility
37 full-text articles excluded with followed-up inclusion criteria 14 articles included in quantitative synthesis (meta-analysis) Figure 1. Procedure of study selection.
conducted computerized literature searches using five online databases: ERIC, PsycINFO, PubMed, Social Services Abstract, and ProQuest (for unpublished dissertations). The following keywords were used to guide the search of articles published from 1975 (The year 1975 was chosen because General Requirements for Informed Consent [C.F.R. Sec. 46.116] was established in 1974) to 2016: (active consent OR informed consent OR parental consent) AND (passive consent OR waiver of consent [Although waiver of consent could be an alternative option to active and passive consent procedure, we found no studies under this category probably because the focus of our study was on general survey research of adolescents’ risk behaviors which is typically not qualified for the waiver of parental consent]) AND (risk behavior) AND (adolescen*). Figure 1 shows the study selection procedure. Original search criteria across the five databases yielded 577 results. After removing the duplicates, 482 articles were identified. Initial screening of titles and abstracts excluded 432 articles because those articles did not explicitly state their consent procedures or did not involve any adolescent’s risk behaviors, reducing the total to 50 articles. We then examined the references cited in these articles to include any article we may have previously missed. One additional study was found [20]. This yielded a pool of 51 studies. Three inclusion criteria were further used to select studies from this pool for the meta-analysis: (1) studies that were quantitative; (2) studies that included independent samples of
active and passive consent (Only studies with independent samples for active and passive consent conditions were selected because they would not introduce the potential confounding effects from repeated measure designs. For example, parents may decline their child’s participation in a study under active consent procedure but later give no response when the same study switches to using a passive consent procedure because they think they have already clearly expressed their intention); and (3) studies that were conducted in the United States to reflect consistency in policy and implementation. First, because our focus was on quantitative studies, 13 qualitative studies were excluded, leaving 38 studies. Second, 19 studies were excluded because of nonindependent samples, leaving 19 studies. Third, five non-U.S. studies were excluded. These filters produced a final list of 14 studies with a total number of 104,074 children (see Table 1 for sample study characteristics).
Coding of studies All studies were coded on information from both active and passive consent groups, which include sample size, response rate, age, gender, ethnicity, and relevant outcomes of risk behaviors. We also coded the effect sizes on each of those variables. If effect sizes were not provided, we calculated them by the following procedures:
4
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
Table 1 Descriptive summary of sample study characteristics Study
Passive consent (N)
Active consent (N)
Age or grade
Ethnicity
Gender (% girl)
Setting
Study design
Variable
White and others
51.20
School
Obs
White, black, Hispanics, Asian, and Native American White, black, and Hispanics
75.50
School
Obs
Substance use; health statusa Psychological riska
48.23
School
Expb
Anderman et al. (1995) [15]
8,645
8,645
Chartier et al. (2008) [21]
1,011
1,021
9th and 12th grade 6th grade
Courser et al. (2009) [11]
3,750
2,190
14.90 years
769
2,727
6th grade
White, black, Hispanics, and others
48.50
School
Obs
Depue (2009) [20]
13,930
4,093
6th, 9th, and 12th grade
51.00
School
Obs
Eaton et al. (2004) [13]
10,118
5,722
9th through 12th grade 7th grade
White, black, Hispanics, Asian, American Indian/ Alaskan Native, and Native Hawaiian or Other Pacific Islander White, black, Hispanics, Asian, and others d
51.30
School
Obs
d
School
Obs
White, black, Hispanics, and others White, black, Hispanics, Asian/Pacific Islander, Native American, and others or mixed White, black, Hispanics, Asian, American Indian, and Native Hawaiian, and others Black, Hispanics, and others d
50.00
School
Expb
52.00
School
Obs
Attitude measurea; antisocial behaviora Substance use
55.70
School
Obs
Substance use
48.80 47.00
School School
Obs Obs
Health statusa Substance use
70.80
Clinic
Obs
44.50
School
Obs
Substance use; delinquent behaviora Substance use
Dent et al. (1993) [19]
Ellickson and Hawes (1989) [7] Esbensen et al. (1999) [17]
117
86
1,083
1,411
12.12 years
Frissell et al. (2004) [22]
5,903
3,176
9th through 12th grade
Horn et al. (2009) [23]
4,922
968
Mellor et al. (2008) [24] Pokorny et al. (2001) [12]
2,104 7,127
2,330 7,138
Rojas et al. (2008) [25]
670
411
Unger et al. (2004) [26]
649
3,358
16.05 years
11.41 years 6th through 8th grade 16.55 years 6th grade
White, black, Hispanics, Asian, and others White, black, Hispanics, Asian/Pacific Islander, and others
Substance use; antisocial behaviora Substance use; socialitya; personalitya Response rate
Substance use; psychological riska Response rate
Obs ¼ observation; Exp ¼ experiment. a Variables that were categorized as nonesubstance use risk behaviors’ for the purpose of subsequent analysis. b More weights were placed on these studies in calculating the overall effect size given the experimental design they used.
1). If the mean and standard deviation were reported in the study (i.e., continuous outcome), then standardized mean difference (SMD) was calculated
SMD ¼
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðn1 1Þs21 þ ðn2 1Þs22 ; s ¼ n1 þ n2 2 where n1 is the sample size of the passive group and n2 is the sample size of the active group and s21 is the variance of the passive group and s22 is the variance of the active group.
x1 x2 ; s
where x1 is the mean of the passive group and x2 is the mean of the active group and
Odds Ratio ðORÞ ¼
2). For those studies that reported percentages (i.e., dichotomous outcome), we calculated the odds ratio using the formula:
% of children who display the behavior in passive group % of children who do not display the behavior in passive group % of children who display the behavior in active group % of children who do not display the behavior in active group
:
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
To combine studies with dichotomous and continuous outcomes, we converted SMD to OR using the formula [27]:
SMD ¼
pffiffiffi 3
p
InOR
Results Before examining the effects of requiring parental consent on the outcome variables of interest, we inspected the distribution of effect sizes by checking the Q and I2 statistics (Table 2). The Q statistic is distributed as a chi-square with k-1 degrees of freedom, where k is the number of effect sizes. Heterogeneity indicated by a p value of less than .05 warrants the use of random effects model. Random effects model assumes that the variability of effect sizes beyond subject-level sampling error is random and therefore cannot be identified [28]. A random effects model is more conservative under this circumstance because the unidentified sources of variability would lead to wider confidence intervals (CIs) for tests of significance which further increase the uncertainty of estimating the population mean. Because significant heterogeneity was found across studies on all the outcome variables except age, we applied random effects models to all outcomes with significant heterogeneity and a fixed effect model to age. We used Cochrane Review Manager (RevMan) 5.3 to derive all the inferential statistics (i.e., Z-score, p value, OR, and CI) as well as the forest plot. Response rate and demographic characteristics in active and passive consent groups There were 11 studies that reported the association between response rate and consent procedure. As shown in Figure 2, the response rate of passive consent groups was significantly higher than that of active consent groups (Z ¼ 3.05, p ¼ .002; OR ¼ 2.39; 95% CI ¼ 1.37e4.20). Significant selection biases were also found in the gender makeup (Z ¼ 2.73, p ¼ .006; OR ¼ .80; 95% CI ¼ .68e.94) and age composition of participants (Z ¼ 12.06, p < .00001; OR ¼ .53; 95% CI ¼ .48e.59): the active consent group contained more females and younger participants than the passive consent group (See Supplementary data for the forest plots of gender and age difference.). There were also more Black participants in the passive consent group (Z ¼ 2.19, p ¼ .03; OR ¼ 1.53; 95% CI ¼ 1.05, 2.23; see Supplementary data for the forest plot of difference of AfricanAmerican participants.). More white participants were found in the active consent group, whereas more Hispanic participants were found in the passive consent group but these differences were not significant (Z ¼ .37, p ¼ .71 and Z ¼ .15, p ¼ .88, respectively). Table 2 Heterogeneity test statistics Q (c2) Response rate Gender Age Ethnic composition-black Ethnic composition-white Ethnic composition-Hispanic a
p < .001.
I2 a
2294.52 211.99a .02 105.75a 364.56a 138.27a
100% 95% 0% 93% 98% 95%
5
Substance use and nonesubstance use risk behaviors in active and passive consent groups We also assessed how the two consent procedures are differentially correlated with the report of substance use and nonesubstance use risk behaviors. Participants in the passive consent groups reported a higher level of risk for substance use than those in the active consent groups (Z ¼ 3.07, p ¼ .002; OR ¼ 1.21; 95% CI ¼ 1.07, 1.37; see Supplementary data for the forest plot of substance use risk behaviors.). The reported rate of nonesubstance use risk behaviors was also higher in passive consent groups than that in active consent groups, but the difference was not significant (Z ¼ 1.65, p ¼ .10). Publication bias assessment We used Stata 13 to assess publication bias by examining the relationship between effect size and its associated variance. The presence of publication bias is indicated by the high correlation between large variances (caused by studies with small sample sizes) and large effect sizes [29]. We conducted Begg and Mazumdar’s rank correlation test and Egger’s regression for this correlation assessment. While these two tests are similar in terms of quantifying the relationship between effect size and precision, Egger’s regression uses the actual values instead of ranks and thus has higher power [29]. Nevertheless, both tests can detect the presence of publication bias based on the significant correlation or regression. None of the outcomes was significant on these two tests (see Table 3), showing no sign of publication bias. Discussion To our knowledge this is the first meta-analysis that systematically examines the impacts of requiring active parental consent, despite decades of debate regarding the effects of this practice on adolescent research. There were three primary results from this study: (1) response rates were significantly lower for studies using active consent procedure than those using passive consent procedure; (2) demographic makeup of the samples varied substantially between studies using active and passive consent procedures; and (3) rates of self-reported substance use were higher in studies using passive consent than those in active consent procedures. At least two reasons may help explain the cause of low response rates under an active consent procedure particularly when it involves self-report of risk behaviors. First, an active consent procedure requires more effort from parents. Parent involvement varies widely and has been linked to family characteristics such as structure (e.g., lower among single-parent households), race (e.g., higher involvement for European American parents than Hispanic and African-American parents) [30], and with an array of negative child outcomes such as poor academic performance, substance use, teen pregnancy, poor emotion regulation, and antisocial behaviors [31,32]. Requiring active parental consent from parents whose involvement is already low may introduce sampling bias that systematically limits access to those children whose parents are not engaged enough to read, sign, and return a consent form in a timely manner. Consequently, some youth may be excluded from research by virtue of family characteristics, which creates the
6
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
Study Unger g et al. ((2004)) Mellor et al. (2008) Anderman et al. (1995) ( ) Pokorny y et al. ((2001) Eaton et al. ((2004)) Ellickson et al. (1989) ( ) Chartier et al. (2008) Depue p ((2009)) j et al. ((2008)) Rojas Courser et al. ((2009)) Frissell et al. (2004)
Odds Ratio [95% CI] 0.17 [[0.13,, 0.21]] 1.05 [0.92, 1.20] 1.52 [[1.43,, 1.62]] 1.61 [[1.46,, 1.78]] 1.91 [[1.76,, 2.08]] 2.21 [[0.86,, 5.67]] 2.98 [[2.40,, 3.70]] 4.46 [[3.91,, 5.10]] 5.90 [[4.48,, 7.75]] 8.95 [[7.93,, 10.10]] 11.50 [10.30, 12.85] 2.39 [1.37, 4.20]
Overall effect size
0.05
0.2 1 5 20 Odds Ratio, 95% CI
Figure 2. Effect sizes and confident intervals of response rate.
potential for requiring parental consent to be in opposition to the principles of justice and inclusiveness. A second reason for lower response rates when requiring parental consent is that adolescents may have concerns that their desire to participate in a study could be interpreted as admission of engaging in the behavior [5,6]. In this case, the low response rate may be due to adolescents’ failure to notify their parents when some sensitive information is included in the study. This may also help explain the significant difference of substance use rather than nonesubstance use risk behaviors in the active and passive consent conditions. Because others have found that parental consent varies by the topic matter and scope of the research [33], it may be that youth are less likely to pursue parental consent for sensitive study topics such as substance use for which there is risk that participation could create a problem with their parents. The fact that more female adolescent participants were included in studies using active parental consent procedure may further contribute to selection bias in research involving risk behaviors. The underestimated prevalence of substance use in the active consent condition may result from an underrepresentative population of male adolescent participants, a population of which a higher prevalence of substance use has been well documented compared with girls [34]. This type of systematic bias on sample selection has been speculated to distort the relation between predictors and the observed outcomes [12]. For example, a study under active consent procedure
containing more female participants may fail to find any significant association between substance use and aggressive behaviors that is in fact present. The failure to detect an existed effect can be ascribed to high homogeneity of the sample (i.e., the prevalence of aggressive behaviors and substance use are both quite low for girls). This creates high intraclass correlations which tend to increase type II error. Increasing sample size to adjust for the increased intraclass correlation is one solution, but it leads to higher study costs, thereby creating barriers for research [35]. The higher proportion of younger participants in active consent procedure underlines the assumption that younger adolescents are not capable to consent to research involving risk behaviors, and therefore, a more capable representative is needed. This assumption, however, is ungrounded given previous evidence showing that children as young as 7 demonstrated an adequate understanding of their research rights such as freedom to ask questions and withdraw, time involved, and potential harms and benefits of their participation [36]. Other studies have shown that adolescents around the age of 13 years had the similar ability to make reasonable health-related decisions as adults [37,38]. These competencies of young adolescents seem to be underestimated in reality, and the pervasive implementation of active parental consent procedure will further reinforce this underestimation, making the voice of minors less likely to be heard. It is not our intention to push for the complete replacement of active consent procedure with
Table 3 Publication bias assessment
Response rate Gender Age Ethnic composition-black Ethnic composition-white Ethnic composition-Hispanic Substance use risk behaviors Nonesubstance use risk behaviors
Begg and Mazumdar’s rank correlation (s)
p
Egger’s regression(b, [95% CI])
p
1.25 .08 1.00 .74 1.16 .00 .15 .52
.21 .94 .32 .46 .25 1.00 .88 .60
4.06 [5.82 to 13.92] 1.61 [4.07 to 7.29] NAa 2.46 [3.21 to 8.13] .03 [11.65 to 11.58] 4.03 [4.19 to 12.26] 1.34 [4.92 to 7.60] 4.30 [65.52 to 74.12]
.38 .54 NA .33 1.00 .28 .61 .58
CI ¼ confidence interval. a Egger’s regression test was not available because the minimum number of studies required for this test is three.
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
passive consent procedure, but there should be decreasing emphasis on parental consent and a corresponding increasing influence from the autonomous decisions made by adolescents. Social justice The potential to exclude specific categories of adolescents from research, particularly that which is focused on the use of substances, raises an issue of social justice. The practice of requiring active consent appears to limit generalizability of research findings due to sampling bias. The findings of this study highlight the likelihood that the population most likely not appropriately represented in active consent studies will be female and/or African-American. This becomes an issue of social injustice as any policy developed from substance use research on adolescents will be based on a population that has been systematically (although not intentionally) over-represented, resulting in a lack of attention to other certain populations. The potential for excluding some adolescents from research is exacerbated because many investigators are reluctant to include minors in research due to additional steps involved in the IRB process, complications in obtaining parental consent, and legality concerns associated with research when a waiver of parental consent is granted [39]. What’s more, if research has no prospect of direct benefits to the child, but knowledge vital to understanding a disorder may be gained, the risk category may be adjusted to “greater than minimal risk.” Under this risk category, HHS guidelines for research require permission be obtained from both parents, unless researchers can establish that one parent is deceased, unknown, incompetent, or not reasonably available, or when only one parent has legal responsibility for the care and custody of the child [26]. Social injustice may also reflect in intervention efforts. For example, intervention effects found in samples using active consent procedure may not be generalizable to a larger population because active consent procedures are more likely to screen out socially disadvantaged groups who have higher needs for the intervention, resulting in an underestimated effect for populations with greater diversity [15]. Accordingly, resources such as the distribution of federal funds may be misallocated away from populations most in need of help. Therefore, rather than dogmatically applying active parental consent procedures for all studies involving minors, members of IRBs and researchers should carefully evaluate the costs and benefits of both active and passive consent procedures on a case-by-case basis. Limitations Several limitations of this study should be noted. First, the number of studies included in our meta-analysis is small. Despite of a large pool of individual participants that affords plenty of within-study variances, between-study variances are limited which make it difficult to generalize based on the findings. For example, although different ethnic composition was found for studies using active versus passive consent procedure, it is likely that some other sociodemographic factors such as geographic location, income, education, and immigration status may further contribute to these ethnic differences. The lower number of black participants in particular could also be due to the distrust arising from prior history of some unethical research conduct performed on black minority subjects [40]. These sources of information, however, were missing from those studies included in our
7
meta-analysis and so we cannot compare studies on these levels. Future studies need to incorporate more information to gain more insights into this area. A second limitation of this metaanalysis is that we only included studies conducted in the United States, and therefore, it is impossible for us to make an inference about the situation in other countries. Nevertheless, the procedure required for obtaining parental permission could vary from country to country and is thus the reason that we excluded studies from other countries. More studies are certainly needed to explore the impact of different consent procedures for research conducted in other countries. Supplementary Data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jadohealth.2017.01.015. References [1] Services USDoHaH. Can consent or parental permission ever be “passive” or “implied?”. 2011. Available at: http://www.hhs.gov/ohrp/policy/faq/ informed-consent/can-consent-be-passive-or-implied.html. Accessed February 22, 2016. [2] Ross JG, Sundberg EC, Flint KH. Informed consent in school health research: Why, how, and making it easy. J Sch Health 1999;69:171e6. [3] Iltis AS. Parents, adolescents, and consent for research participation. J Med Philos 2013;38:332e46. [4] Baker JR, Yardley JK, McCaul K. Characteristics of responding-, nonresponding- and refusing-parents in an adolescent lifestyle choice study. Eval Rev 2001;25:605e18. [5] Flicker S, Guta A. Ethical approaches to adolescent participation in sexual health research. J Adolesc Health 2008;42:3e10. [6] Rew L, Taylor-Seehafer M, Thomas N. Without parental consent: Conducting research with homeless adolescents. J Soc Pediatr Nurs 2000;5: 131e8. [7] Ellickson PL, Hawes JA. An assessment of active versus passive methods for obtaining parental consent. Eval Rev 1989;13:45e55. [8] Esbensen F-A, Deschenes EP, Vogel RE, et al. Active parental consent in school-based research. An examination of ethical and methodological issues. Eval Rev 1996;20:737e53. [9] Kearney KA, Hopkins RH, Mauss AL, Weisheit RA. Sample bias resulting from a requirement for written parental consent. Public Opin Q 1983;47: 96e102. [10] Lueptow L, Mueller SA, Hammes RR, Master LS. The impact of informed consent regulations on response rate and response bias. Sociol Methods Res 1977;6:183e204. [11] Courser MW, Shamblen SR, Lavrakas PJ, Collins D, Ditterline P. The Impact of active consent procedures on nonresponse and nonresponse error in youth survey data: Evidence from a new experiment. Eval Rev 2009;33: 370e95. [12] Pokorny SB, Jason LA, Schoeny ME, et al. Do participation rates change when active consent procedures replace passive consent. Eval Rev 2001; 25:567e80. [13] Eaton DK, Lowry R, Brener ND, et al. Passive versus active parental permission in school-based survey research: Does the type of permission affect prevalence estimates of risk behaviors? Eval Rev 2004;28:564e77. [14] Langhinrichsen-Rohling J, Arata C, O’Brien N, et al. Sensitive research with adolescents: Just how upsetting are self-report surveys anyway? Violence Vict 2006;21:425e44. [15] Anderman C, Cheadle A, Curry S, et al. Selection bias related to parental consent in school-based survey research. Eval Rev 1995;19:663e74. [16] Dent CW, Galaif J, Sussman S, et al. Demographic, psychosocial and behavioral differences in samples of actively and passively consented adolescents. Addict Behav 1993;18:51e6. [17] Esbensen F-A, Miller MH, Taylor T, et al. Differential attrition rates and active parental consent. Eval Rev 1999;23:316e35. [18] Severson HH, Ary DV. Sampling bias due to consent procedures with adolescents. Addict Behav 1983;8:433e7. [19] Dent CW, Sussman SY, Stacy AW. The impact of a written parental consent policy on estimates from a school-based drug use survey. Eval Rev 1997; 21:698e712. [20] Depue SM. Psychometrics of the Missouri student Survey: Examining validity, reliability and consent. St. Louis, MO: University of Missouri-Saint Louis; 2009.
8
C. Liu et al. / Journal of Adolescent Health xxx (2017) 1e8
[21] Chartier M, Stoep AV, McCauley E, et al. Passive versus active parental permission: Implications for the ability of school-based depression screening to reach youth at risk. J Sch Health 2008;78:157e64. [22] Frissell KC, McCarthy DM, D Amico EJ, et al. Impact of consent procedures on reported levels of adolescent alcohol use. Psychol Addict Behav 2004; 18:307. [23] Horn KA, Branstetter SA, Dino GA, et al. Potential effects of active parental consent: Enrolling teen smokers into a school-based cessation program. Nicotine Tob Res 2009;11:1359e67. [24] Mellor JM, Rapoport RB, Maliniak D. The impact of child obesity on active parental consent in school-based survey research on healthy eating and physical activity. Eval Rev 2008;32:298e312. [25] Rojas NL, Sherrit L, Harris S, Knight JR. The role of parental consent in adolescent substance use research. J Adolesc Health 2008;42:192e7. [26] Services USDoHaH. Code of Federal Regulations: Requirements for permission by parents or guardians and for assent by children. 2010. Available at: http://www.hhs.gov/ohrp/humansubjects/guidance/45cfr46 .html#46.408. Accessed February 22, 2016. [27] Chinn S. A simple method for converting an odds ratio to effect size for use in meta-analysis. Stat Med 2000;19:3127e31. [28] Lipsey MW, Wilson DB. Practical meta-analysis. Thousand Oaks, CA: Sage Publications, Inc; 2001. [29] Rothstein HR, Sutton AJ, Borenstein M. Publication bias in meta-analysis: Prevention, assessment and adjustments. Chichester, UK: John Wiley & Sons; 2006. [30] Lee J-S, Bowen NK. Parent involvement, cultural capital, and the achievement gap among elementary school children. Am Educ Res J 2006;43: 193e218.
[31] Griffin KW, Botvin GJ, Scheier LM, et al. Parenting practices as predictors of substance use, delinquency, and aggression among urban minority youth: Moderating effects of family structure and gender. Psychol Addict Behav 2000;14:174. [32] Morris AS, Silk JS, Steinberg L, et al. The role of the family context in the development of emotion regulation. Social Dev 2007;16:361e88. [33] Pasternak RH, Geller G, Parrish C, Cheng TL. Adolescent and parent perceptions on youth participation in risk behavior research. Arch Pediatr Adolesc Med 2006;160:1159e66. [34] Johnston LD, O’malley PM, Bachman JG, Schulenberg JE. Monitoring the Future national survey results on drug use, 1975-2010. Vol II, College Students & Adults Ages 19e50. Ann Arbor, MI: Institute for Social Research; 2011. [35] White VM, Hill DJ, Effendi Y. How does active parental consent influence the findings of drug-use surveys in schools? Eval Rev 2004;28:246e60. [36] Susman EJ, Dorn LD, Fletcher JC. Participation in biomedical research: The consent process as viewed by children, adolescents, young adults, and physicians. J Pediatr 1992;121:547e52. [37] Tait AR, Voepel-Lewis T, Malviya S. Do they understand? (part II): Assent of children participating in clinical anesthesia and surgery research. Anesthesiology 2003;98:609e14. [38] Weithorn LA, Campbell SB. The competency of children and adolescents to make informed treatment decisions. Child Development 1982;53:1589e98. [39] Hester CJ. Adolescent consent: Choosing the right path. Issues Compr Pediatr Nurs 2004;27:27e37. [40] Shavers VL, Lynch CF, Burmeister LF. Racial differences in factors that influence the willingness to participate in medical research studies. Ann Epidemiol 2002;12:248e56.