School Psychology Quarterly 2014, Vol. 29, No. 4, 395-408
© 2014 American Psychological Association 1045-3830/14/$ 12.00 http://dx.doi.org/10.1037/spq0000069
The Role of ADHD in Academic Adversity: Disentangling ADHD Effects From Other Personal and Contextual Factors Andrew J. Martin University of New South Wales Students with attention-deficit/hyperactivity disorder (ADHD) experience significant academic difficulties that can lead to numerous negative academic consequences. With a focus on adverse academic outcomes, this study seeks to disentangle variance attributable to ADHD from variance attributable to salient personal and contextual covariates. Responses from 136 students with ADHD and 3,779 non-ADHD peers from 9 high schools were analyzed using logistic regression. Dependent measures included academic failure, grade repetition, school refusal, changing classes and school, school exclusion, and schoolwork noncompletion. Covariates comprised personal (e.g., so ciodemographics, personality, prior achievement, specific learning disabilities, moti vation) and contextual (e.g., school size, school socioeconomic status, school average achievement) factors. Findings indicated that, after accounting for personal and con textual covariates, ADHD explained significant variance in numerous adversities (schoolwork noncompletion, school suspension, school expulsion, changing schools, grade repetition). Thus, beyond the effects of numerous personal and contextual covariates, ADHD has a distinct presence in students’ academic adversity. Also interesting, after accounting for other personal and contextual factors, was academic adversity with which ADHD was not associated. Findings provide direction for educational intervention targeting ADHD and associated factors found to be significant in the study. Keywords: attention-deficit/hyperactivity disorder, ADHD, academic adversity, academic risk, academic outcomes
Students with attention-deficit/hyperactivity disorder (ADHD) experience significant aca demic difficulties, leading to problematic aca demic outcomes (Barkley, 2006; DuPaul & Stoner, 2003). In better understanding ADHD and its effects, it is important to examine the disorder in the context of diverse academic ad versities (e.g., grade repetition, academic fail ure, school refusal, school exclusion, school-
This article was published Online First May 12, 2014. Thanks are extended to Brad Papworth, Harry Nejad, Farideh Nejad, and Marianne Mansour for data collection and data entry, Gregoiy Liem for data management, the Australian Research Council for funding, and participating schools and students. Some of the subscales used in the study are from a published instrument (i.e., the Motivation and Engagement Scale) attracting a small fee. No fee was involved in its use in this research. Correspondence concerning this article should be ad dressed to Andrew J. Martin, School of Education, Univer sity of New South Wales, NSW 2052, Australia. E-mail: [email protected]
395
work noncompletion) that may result (Loe & Feldman, 2007). There is also a need to distin guish ADHD effects from personal (e.g., so ciodemographics, personality, prior achieve ment, specific learning disability, motivation) and contextual (e.g., school) factors that are associated with adverse academic outcomes (Bauermeister et al., 2007). This study explored the extent of unique influence on multidimen sional academic adversity attributable to ADHD, disentangling its effects from variance attributable to other personal and contextual factors. The proposed contribution of this inves tigation lay in the integration of diverse per sonal and contextual factors predicting varying academic adversity outcomes. In so doing, the study sought to identify the unique effects of ADHD (controlling for other personal contex tual factors), alongside the unique of effects of personal and contextual factors (controlling for ADHD) on a wider range of academic adversity outcomes than studied previously.
MARTIN
396
ADHD and Multidimensional Academic Risk
Consistent with recent conceptual frame works focusing on at-risk youth (Coleman & Hagell, 2007), students with ADHD are consid ered to be at particular academic risk on a potentially frequent and ongoing basis (e.g., Barkley, 2006). Coleman and Hagell (2007) articulate major dimensions of risk, including risk factors and risk behavior, both of which have relevance to students with ADHD. Ac cording to Coleman and Hagell, risk factors refer to factors that contribute to poor outcomes, including illness, dysfunction, and disorders. ADHD has been defined as “a persistent pattern of inattention and/or hyperactivity-impulsivity that interferes with functioning or develop ment” (American Psychiatric Association, DSM-5, 2013, p. 59). Dominant conceptual models of ADHD highlight impairments to ex ecutive function and self-regulation (Loe & Feldman, 2007; Nigg, 2001). Other psycholog ical models of ADHD also emphasize cognitive and neuropsychological risk (Brown, 2005; Gray & McNaughton, 2003; Sergeant, 2005), as well as neurological and biochemical risk (see Barkley, 2006; Chandler, 2010 for summaries). In addition to these is research demonstrating comorbidities (e.g., specific learning disabili ties, anxiety) with ADHD that further com pound difficulties with academic outcomes (Bauermeister et al., 2007). In all perspectives, there are risk factors relevant to ADHD that increase the possibility of problematic pro cesses that are likely to compromise and disrupt academic outcomes (Martin, 2014; Vostanis, 2007; Wilmshurst, Peele, & Wilmshurst, 2011). Coleman and Hagell’s (2007) second di mension of risk relates to risk behaviors. This refers to potentially harmful and challenging behavior in which young people engage. Adapting this definition to ADHD and the academic domain, the ADHD condition has troubling behavioral effects across the spec trum of educational development (Barkley, Murphy, & Kwasnik, 1996; DuPaul, Ruther ford, & Hosterman, 2008; Martin, 2012a). For students with ADHD, there are high rates of off-task behavior, attention seeking, poor task completion, and poor self-regulation (Vile Junod, DuPaul, Jitendra, Volpe, & Cleary, 2006). As a result of these factors, there are
significantly poorer outcomes on numerous academic performance indicators, such as poor achievement, school exclusion, schoolwork noncompletion, school refusal, grade repetition, and changing classrooms and/or schools (DuPaul et al., 2008; DuPaul & Stoner, 2003; Pliszka, 2009; Purdie, Hattie, & Carroll, 2002). Thus, there are behavioral challenges experienced by students with ADHD (see Purdie et al., 2002 for a meta analysis; Wilmshurst et al., 2011) that qualify as risk behaviors under a risk-conceptual framework. Given this multiple risk profile associated with ADHD, this study adopts a multidimen sional approach to academic risk and adver sity. Because ADHD has been associated with diverse negative academic outcomes, it is im portant to examine the disorder in the context of such multiple academic outcomes. As de scribed in the methodology, these outcomes traverse a diverse range of academic adversi ties and provide a more comprehensive basis upon which to understand the unique variance in academic adversity that can be attributable to ADHD. Disentangling ADHD From Other Salient Factors Relevant to Academic Adversity
Across a vast body of psychoeducational re search it is also clear that many other factors predict academically adverse outcomes such as low achievement, school exclusion, and school refusal or avoidance. For example, sociodemo graphic factors have been associated with prob lematic academic pathways in low socioeco nomic status (SES) boys, (e.g., those with low parental education and occupations) and ethnic minorities at greater risk for academic poor performance and school exclusion (e.g., Martin, 2004; OECD, 2006; Sirin, 2005). In addition, students low in motivation and prior achieve ment are more prone to failure, grade repetition, and noncompletion of schoolwork (e.g., Finn & Rock, 1997; Jimerson, 2001; Martin, 2011). Specific learning disabilities (e.g., dyslexia, dysgraphia, dyscalculia) are associated with failure, grade repetition, school exclusion, and school refusal (Johnson, 2012; Scanlon & Mellard, 2002). Aspects of personality have also been linked to low levels of academic achieve ment (Chamorro-Premuzic & Furnham, 2003),
ADHD AND ACADEMIC ADVERSITY
lower academic buoyancy (Martin, Nejad, Col mar, & Liem, 2013) and academic disengage ment (Martin et al., 2013). Indeed, research has found ADHD students lower on the personality factors Conscientiousness, Agreeableness and Emotional Stability, and higher on Neuroticism (De Pauw & Mervielde, 2011; Martel, Nigg, & Lucas, 2008). In addition, there may be schoollevel factors associated with negative academic outcomes, with some schools having higher rates of academic failure, problem behavior, and school exclusions (e.g., low SES schools, schools low in average ability, schools with higher proportions of minority students, schools with inadequate staff-to-student ratios; Martin, Anderson, Bobis, Way, & Veller, 2012; Perry & McConney, 2010). Aims of the Study This study sought to disentangle the effects of ADHD on academic adversity from the effects of personal (e.g., sociodemographics, personal ity, prior achievement, specific learning disabil ities, motivation) and contextual (e.g., school factors) covariates. For example, including spe cific learning disability alongside ADHD as a predictor of academic adversity enables us to understand unique variance attributable to ADHD, controlling for learning disability, as well as unique variance attributable to specific learning disability, controlling for ADHD. Such research provides important insights for inter vention. If the effects of ADHD are largely ameliorated once these covariates are entered, then clearly covariates are pivotal factors and intervention is to be directed to the covariates of relevance. To the extent that ADHD effects remain significant after entering covariates, then intervention is also to be directed at ADHD itself (e.g., toward executive functioning, med ication, or self-regulation, which have been shown to be central to ADHD; Chandler, 2010; DuPaul & Stoner, 2003; Pliszka, 2009). In ad dition, because multiple dimensions of aca demic adversity (e.g., failure, grade repetition, school refusal, school exclusion, schoolwork noncompletion) are included as outcome fac tors, the study aims to shed expanded light on the particular mix of ADHD, personal, and con textual factors that best predicts particular ad verse outcomes.
397
Method Sample and Procedure Students with and without ADHD were drawn from “regular” (or “m ainstream ”) schools of mixed ability. Four schools were coeducational, three were single-sex boys’ schools, and two were single-sex girls’ schools. Schools were from the independent school and the systemic Catholic sectors in major urban areas on the east coast of Australia. They com prised students from a range of SES levels that tended to be slightly higher SES than the na tional average (see Materials for School SES section). School achievement scores ranged from below the national average to above the national average as indicated by data from Aus tralian Curriculum and Assessment Authority (ACARA, 2012), though the average school achievement score was slightly higher than the national average. Ethics approval was provided by the re searcher’s university and parental consent was required. With few exceptions, targeted stu dents in attendance on the day of the testing participated in the survey. Students were al lowed to decline participation or withdraw from the process at any time, but none did so (to the author’s knowledge). Standardized instructions were provided to each teacher who administered the survey. The teachers explained the rating scale and students were presented with a sample item. Students were instructed not to identify themselves to ensure anonymity and allow full and frank responses. Students were also in structed to complete the survey on their own but were allowed to ask teachers for help with any survey items they had difficulty understanding or reading. Schools in Australia reside in one of three sectors, government, independent, and systemic Catholic. Most schools operate under curricu lum and assessment determined by their state or territory. In the main, Australian schools are academically comprehensive (i.e., little or no academic selection for entry) and include stu dents with ADHD (and many students with specific learning disabilities such as dyslexia). Although schools are academically nonselective, there is often streaming of classes within the school (usually in mathematics and Eng lish), with many students at academic risk (in-
398
MARTIN
eluding those with ADHD) often in the lower and middle classes. There is some additional funding available to schools to assist students with disorders such as ADHD and there are published standards relevant to instruction and assessment accommodations, classroom adjust ments, and targeted services with which schools are expected to comply (e.g., New South Wales Education and Communities, 2012), though, in reality the onus tends to be on individual schools and teachers to secure such funding and to ensure that standards are implemented appro priately and effectively. Students With ADHD The sample of students with ADHD (N = 136) were in junior high school (11-14 years, 59%) and senior high school (15-19 years, 41%). This sample of students with ADHD was built on the sample from two earlier studies that comprised 87 students with ADHD (Martin, 2012b, 2014). These students were the sub sample of students from the larger survey sam ple who reported (in the survey) that they had received a formal medical diagnosis of ADHD (thus, they were not specifically targeted stu dents). Although the validity of self-reports has received some support in relation to adults with ADHD (Kessler, Adler, Gruber, Sarawate, Spencer, & Van Brunt, 2007), self-report valid ity among adolescents is unclear. It is noted that N = 136 corresponds to a 3.5% incidence rel ative to the total sample, a prevalence broadly aligning with estimates of adolescents with ADHD (Barkley, 2006), however the present findings must be interpreted with the limitations of self-report in mind. The average age was 14.32 years (SD = 1.38). The average time since diagnosis was 5.07 years (SD = 4.61). Based on data available, the study was unable to identify whether students were mainly inatten tive type, hyperactive type, or both. Thus, the broad ADHD designation was adopted. Just un der half the sample (46%) was on medication to help manage the condition, a figure generally lower than population estimates (Visser, Bitsko, Danielson, Perou, & Blumberg, 2010). Consis tent with population statistics is the prevalence of boys (68%) compared with girls (32%) with ADHD, X2(l) = 12.97, p < .001. A total of 9% of students with ADHD were from a nonEnglish-speaking background (NESB); how
ever, students with ADHD were no more (or less) likely to come from an NESB, x2(l) = 2.45, p > .05. Students Without ADHD The non-ADHD group comprised a large sample of 3,779 students from the same schools and year levels as the ADHD group. Although the ADHD sample was augmented from the previous Martin (2012b, 2014) studies, this non-ADHD group was largely the same as that reported in the earlier studies (this sample was already large enough, and thus required no aug mentation). The non-ADHD group included students in junior high school (11-14 years, 58%) and senior high school (15-19 years, 42%). Just under half (47%) of the participants were female and 53% were male. The average age was 14.20 years (SD = 1.53). Of the sam ple, 14% were from an NESB. The non-ADHD sample was much larger than the ADHD sample because the study aimed to comprise a sample reflecting a more typical ratio of ADHD-to-nonADHD students and because a larger sample size would generate less biased standard-error estimates, the study retained this comprehensive representation of non-ADHD students. Materials The dependent variables (academic adver sity outcomes). The dependent measures were the adversity items from the Academic Risk and Resilience Scale (Martin, 2013). There are 10 adversity items, eight of which were used in this study. These were repeated a grade (4% of students); failed a school subject in the endof-year report card (18%); did not hand in most assignments (7%); moved to another class be cause of difficulties with work, teacher, or stu dents (8%); suspended from school (5%); changed schools (26%; numerous students tran sitioned from elementary to secondary school); expelled from school (1%); and avoided or re fused to go to school (11%). To each of these items, students answered No (coded 0) or Yes (coded 1). Two items were excluded. These were major illness (as this was not deemed a direct consequence of ADHD) and learning disability (as this was collinear with one of the predictors). This item set has been effec tively implemented in previous research into academic adversity (Martin, 2013). Correia-
ADHD AND ACADEMIC ADVERSITY
tions among outcome variables are presented in Table 1.
The Independent Variables (Personal and Contextual Covariates) Sociodemographics. S ociodem ographic variables com prised age, gender, language background, and SES indicated by parent edu cation and parent occupation. G ender was scored 1 for boys and 0 for girls. Age was analyzed as a continuous variable. For language background, students were asked if they spoke English (0) or another language (1; NESB) at home. Students were asked to report their moth ers’ (or female caregivers’) and fathers’ (or male caregivers’) education and occupation on an ordinal scale developed by the Australian Bureau of Statistics (ABS). Parent occupation and parent education factors were developed from the ordinal scales through deriving the average of m others’ and fathers’ education scores (for parent education; M = 4.59, SD = 1.33, range = 1 -6 ) and the average of parents’ occupation scores (for parent occupation; M = 5.28, SD = 1.75, range = 1-8). These two factors are recognized indicators of SES (Dun can, Daly, McDonough, & Williams, 2002). Correlations among all predictor variables (in cluding personality, academic, and school fac tors) are presented in Table 2. Personality. Five personality factors— Openness to Experience (Cronbach’s a = .73; M = 5.04; SD = .93), Extraversion (a = .81; M = 4.92; SD = 1.09), Neuroticism (a = .75; M = 3.74; SD = 1.04), Conscientiousness (a = .84; M = 4.75; SD = 1.15), and Agreeableness (a = .79; M = 5.57; SD = .88; 8 items per
Table 1
Correlations Among Outcomes Outcome variable 1. Repeated grade 2. Failed subject 3. Uncompleted work 4. Changed classes 5. Suspended 6. Changed schools 7. Expelled 8. School refusal
1
2
3
4
5
6
7
8
—
10 08 13 05 19 10 06
—
26 20 20 11 11 28
—
18 16 09 14 26
—
21 — 10 10 — 15 30 09 — 22 18 11 11 —
Note. Decimals omitted. Correlations > 1.041 significant at p < .05.
399
factor)— were measured with the 40-item Inter national English Big-Five Mini-Markers instru ment (IEBM; Thompson, 2008). Students rated (1 = Very inaccurate to 7 = Very accurate) the extent to which 40 one-word trait adjectives reflected accurate descriptions of themselves. Sample words for each factor are as follows: “creative” (Openness), “talkative” (Extraver sion), “efficient” (Conscientiousness), “moody” (Neuroticism), and “warm” (Agreeableness). Based on the acceptable reliability (Cronbach’s a ; Kline, 2000), five variables were developed from the means of each of the five-factor sets. Academic factors. Prior achievement was drawn from students’ results in the National Assessment Program in Literacy and Numeracy (NAPLAN) administered by the ACARA (ACARA, 2012). NAPLAN is a nationally standardized test on which school students receive scores each for literacy and numeracy. Students re ported their numeracy and literacy NAPLAN scores. Based on an acceptable reliability (a = .80; Kline, 2000), one achievement score was developed that was the mean of the two literacy and numeracy scores {M = 7.28, SD = 1.44, range = 1-10). Motivation was assessed via the Motivation and Engagement Scale (MES; Mar tin, 2010). Motivation comprises self-efficacy (e.g., “If I try hard, I believe I can do my schoolwork well”), mastery orientation (e.g., “I feel very pleased with myself when I do well at school by working hard”), valuing school (e.g., “Learning at school is important”), persistence (e.g., “If I don’t give up, I believe I can do difficult schoolwork”), planning (e.g., “I try to plan things out before I start working on my homework or assignments”), and task manage ment (e.g., “When I study, I usually try to find a place where I can study well”). For each item, students rated themselves on a scale of 1 {Strongly disagree) to 7 {Strongly agree). Based on the acceptable reliability (a = .86; Kline, 2000), one motivation-scale score was devel oped that was the mean of the six motivation indicators {M = 5.30, SD = .86). Specific learn ing disabilities are defined as formally determined/diagnosed problems with specific as pects, processes, or skills that affect learning (e.g., dyslexia; Australian Psychological Soci ety, 2014). This study assessed specific learning disabilities by asking all students in the survey sample if they experienced any formally deter mined or diagnosed reading difficulty/dyslexia,
MARTIN
400
o CN CN 1 1 1 l
o I' O CN ve VO 1
1 1
44
VO O Ov CN o CN 1 1 •'t1 1 cn VO m CN t1 1 cn VO r-- cn in in cn 1 T 1
in cn
CN CN CN cn Ov 00 O m cn Ov
■g
oW ° 3 t/3
CN Q *■§ ro H U
CN cn'st'invO t-'O O O v©
,-2 oo jS Q N 3 o -b W O 00 bo 00 CQ O Z < cn cn ■'fr >n vo i>
22. Indigenous %
Oo c o
Decimals omitted. Correlations > 1.041 significant at p < .05.
— e n o c N - cn I I I I
—13
—
-
—12
—o o I I I
0 0
I
-H M N N oo n 0 0 0 0 0 —0 II II
VO CO CO
CN CN 's f
o o c n o
41
o —o I
Note.
^ D ' O M n O N ' O I ' O H O o i n i n N i n o o
1
—21
00 oo cn CN vO r- t-» r-- in O o O o o o o o —o O s
ADHD AND ACADEMIC ADVERSITY
mathematics difficulty/dyscalculia, or writing difficulty/dysgraphia. For each student, a spe cific learning disability index was scored 1 for the presence of one or more of these specific disabilities and 0 for no such specific disability (n = 274 with specific learning disabilities; 7% of sample). School factors. School factors assessed were school-average achievement in national numeracy tests, school SES, school size, staffto-student ratio, ethnic and indigenous compo sition, and gender composition. School size (number of students relative to the number of grades in the school, M = 101 students; SD = 31.98) and full-time (equivalent) staff-tostudent ratio (higher scores reflecting more full time staff per student, M = 10.89; SD = 2.47), and gender composition (n = 3 boys’ schools with n = 1,268 students; n = 2 girls’ schools with n = 1182 students, n = 4 coeducational schools with n = 1465 students; coeducational was the reference category for dummy coding) are from ACARA, 2012 records. The School SES Index is a nationally assigned score (na tional M = 1,000) by the ABS comprising in formation on students’ household income, parent/caregiver occupation, and parent/caregiver education (higher scores reflect higher schoollevel SES; M = 1045, SD = 83). Schoolaverage achievement is drawn from ACARA (2012) and reflects the schools’ performance in annual national literacy and mathematics as sessment (M = 598.21, SD = 26.52). School NESB and school indigenous composition are the percentage of students in the school from an NESB (M = 25.17%, SD = 22.28) or indige nous background (M = 1.44%, SD = 1.59), based on ACARA records. Data Analysis Data were analyzed to identify the role of ADHD status in predicting academic adversity, controlling for sociodemographic, personality, academic, and school factors. Such modeling, which comprises numerous predictors, brings into consideration multiple regression tech niques. As each dependent variable was scored on a dichotomous scale, logistic regression was the analytical tool of choice. The method of entry was to enter ADHD in the first step to determine its initial variance for each academic adversity factor. This was followed by a for
401
ward stepwise procedure for the covariate set with entry testing based on the significance of the score statistic, and removal testing based on the probability of the Wald statistic (Tabachnick & Fidell, 2012). The probability of the score statistic for variable entry was .05. The probability of the Wald statistic to remove a variable w a s. 10. In this stepwise procedure, the role of covariates was determined. Of interest in these regressions was the initial role of ADHD in predicting academic adversity and then its subsequent role once covariates were entered; this latter step indicated unique variance in ac ademic adversity explained by ADHD and which factors do and do not mitigate its effects. Logistic regression was selected over dis criminant analysis as the former approach re quires fewer assumptions than the latter (Hosmer, Lemeshow, & Sturdivant, 2013). Core assumptions relevant to logistic regression were met (e.g., independent variables were interval, ratio, or dichotomous; relevant predictors were included; form of the relationships were linear; no autocorrelation; and no correlation between error and independent variables). Maximum likelihood (with the logit link function) was the method of estimation. A test of Cook’s distance indicated no particularly influential observa tions or outliers, using the criterion (Cook’s D > 1.00) suggested by Cook and Weisberg (1982). Missing data can pose problems if ex ceeding 5% (e.g., Graham & Hoffer, 2000). There are concerns about listwise, pairwise, and mean substitution approaches to missing data (Graham & Hoffer, 2000), leading to recom mendation of the expectation-maximization (EM) algorithm, implemented in this study us ing LISREL 8.80 (Joreskog & Sorbom, 2006). Imputation identified that less than 5% of the data were missing, so the EM algorithm was employed to manage this. Results Results of the logistic regressions are pre sented in Table 3 (comprising B values, statis tical significance, the step number at which a particular covariate significantly entered the model, and model fit). Initially, ADHD was predictive of grade repetition (B = 1.04, p < .01, OR = 2.83, 95% Cl [1.39, 5.76]); however, once Specific Learning Disabilities, Prior Achievement, and Indigenous School Composi-
MARTIN
402
II I I
o o 43
03
gd U
- D 43
•S 3 03 >.£? S 3
"3 *3
3 oM >p.2. Z CJ T3 ."S «
g fc — —.
o „ OQ W
© 2014 American Psychological Association 1045-3830/14/$ 12.00 http://dx.doi.org/10.1037/spq0000069
The Role of ADHD in Academic Adversity: Disentangling ADHD Effects From Other Personal and Contextual Factors Andrew J. Martin University of New South Wales Students with attention-deficit/hyperactivity disorder (ADHD) experience significant academic difficulties that can lead to numerous negative academic consequences. With a focus on adverse academic outcomes, this study seeks to disentangle variance attributable to ADHD from variance attributable to salient personal and contextual covariates. Responses from 136 students with ADHD and 3,779 non-ADHD peers from 9 high schools were analyzed using logistic regression. Dependent measures included academic failure, grade repetition, school refusal, changing classes and school, school exclusion, and schoolwork noncompletion. Covariates comprised personal (e.g., so ciodemographics, personality, prior achievement, specific learning disabilities, moti vation) and contextual (e.g., school size, school socioeconomic status, school average achievement) factors. Findings indicated that, after accounting for personal and con textual covariates, ADHD explained significant variance in numerous adversities (schoolwork noncompletion, school suspension, school expulsion, changing schools, grade repetition). Thus, beyond the effects of numerous personal and contextual covariates, ADHD has a distinct presence in students’ academic adversity. Also interesting, after accounting for other personal and contextual factors, was academic adversity with which ADHD was not associated. Findings provide direction for educational intervention targeting ADHD and associated factors found to be significant in the study. Keywords: attention-deficit/hyperactivity disorder, ADHD, academic adversity, academic risk, academic outcomes
Students with attention-deficit/hyperactivity disorder (ADHD) experience significant aca demic difficulties, leading to problematic aca demic outcomes (Barkley, 2006; DuPaul & Stoner, 2003). In better understanding ADHD and its effects, it is important to examine the disorder in the context of diverse academic ad versities (e.g., grade repetition, academic fail ure, school refusal, school exclusion, school-
This article was published Online First May 12, 2014. Thanks are extended to Brad Papworth, Harry Nejad, Farideh Nejad, and Marianne Mansour for data collection and data entry, Gregoiy Liem for data management, the Australian Research Council for funding, and participating schools and students. Some of the subscales used in the study are from a published instrument (i.e., the Motivation and Engagement Scale) attracting a small fee. No fee was involved in its use in this research. Correspondence concerning this article should be ad dressed to Andrew J. Martin, School of Education, Univer sity of New South Wales, NSW 2052, Australia. E-mail: [email protected]
395
work noncompletion) that may result (Loe & Feldman, 2007). There is also a need to distin guish ADHD effects from personal (e.g., so ciodemographics, personality, prior achieve ment, specific learning disability, motivation) and contextual (e.g., school) factors that are associated with adverse academic outcomes (Bauermeister et al., 2007). This study explored the extent of unique influence on multidimen sional academic adversity attributable to ADHD, disentangling its effects from variance attributable to other personal and contextual factors. The proposed contribution of this inves tigation lay in the integration of diverse per sonal and contextual factors predicting varying academic adversity outcomes. In so doing, the study sought to identify the unique effects of ADHD (controlling for other personal contex tual factors), alongside the unique of effects of personal and contextual factors (controlling for ADHD) on a wider range of academic adversity outcomes than studied previously.
MARTIN
396
ADHD and Multidimensional Academic Risk
Consistent with recent conceptual frame works focusing on at-risk youth (Coleman & Hagell, 2007), students with ADHD are consid ered to be at particular academic risk on a potentially frequent and ongoing basis (e.g., Barkley, 2006). Coleman and Hagell (2007) articulate major dimensions of risk, including risk factors and risk behavior, both of which have relevance to students with ADHD. Ac cording to Coleman and Hagell, risk factors refer to factors that contribute to poor outcomes, including illness, dysfunction, and disorders. ADHD has been defined as “a persistent pattern of inattention and/or hyperactivity-impulsivity that interferes with functioning or develop ment” (American Psychiatric Association, DSM-5, 2013, p. 59). Dominant conceptual models of ADHD highlight impairments to ex ecutive function and self-regulation (Loe & Feldman, 2007; Nigg, 2001). Other psycholog ical models of ADHD also emphasize cognitive and neuropsychological risk (Brown, 2005; Gray & McNaughton, 2003; Sergeant, 2005), as well as neurological and biochemical risk (see Barkley, 2006; Chandler, 2010 for summaries). In addition to these is research demonstrating comorbidities (e.g., specific learning disabili ties, anxiety) with ADHD that further com pound difficulties with academic outcomes (Bauermeister et al., 2007). In all perspectives, there are risk factors relevant to ADHD that increase the possibility of problematic pro cesses that are likely to compromise and disrupt academic outcomes (Martin, 2014; Vostanis, 2007; Wilmshurst, Peele, & Wilmshurst, 2011). Coleman and Hagell’s (2007) second di mension of risk relates to risk behaviors. This refers to potentially harmful and challenging behavior in which young people engage. Adapting this definition to ADHD and the academic domain, the ADHD condition has troubling behavioral effects across the spec trum of educational development (Barkley, Murphy, & Kwasnik, 1996; DuPaul, Ruther ford, & Hosterman, 2008; Martin, 2012a). For students with ADHD, there are high rates of off-task behavior, attention seeking, poor task completion, and poor self-regulation (Vile Junod, DuPaul, Jitendra, Volpe, & Cleary, 2006). As a result of these factors, there are
significantly poorer outcomes on numerous academic performance indicators, such as poor achievement, school exclusion, schoolwork noncompletion, school refusal, grade repetition, and changing classrooms and/or schools (DuPaul et al., 2008; DuPaul & Stoner, 2003; Pliszka, 2009; Purdie, Hattie, & Carroll, 2002). Thus, there are behavioral challenges experienced by students with ADHD (see Purdie et al., 2002 for a meta analysis; Wilmshurst et al., 2011) that qualify as risk behaviors under a risk-conceptual framework. Given this multiple risk profile associated with ADHD, this study adopts a multidimen sional approach to academic risk and adver sity. Because ADHD has been associated with diverse negative academic outcomes, it is im portant to examine the disorder in the context of such multiple academic outcomes. As de scribed in the methodology, these outcomes traverse a diverse range of academic adversi ties and provide a more comprehensive basis upon which to understand the unique variance in academic adversity that can be attributable to ADHD. Disentangling ADHD From Other Salient Factors Relevant to Academic Adversity
Across a vast body of psychoeducational re search it is also clear that many other factors predict academically adverse outcomes such as low achievement, school exclusion, and school refusal or avoidance. For example, sociodemo graphic factors have been associated with prob lematic academic pathways in low socioeco nomic status (SES) boys, (e.g., those with low parental education and occupations) and ethnic minorities at greater risk for academic poor performance and school exclusion (e.g., Martin, 2004; OECD, 2006; Sirin, 2005). In addition, students low in motivation and prior achieve ment are more prone to failure, grade repetition, and noncompletion of schoolwork (e.g., Finn & Rock, 1997; Jimerson, 2001; Martin, 2011). Specific learning disabilities (e.g., dyslexia, dysgraphia, dyscalculia) are associated with failure, grade repetition, school exclusion, and school refusal (Johnson, 2012; Scanlon & Mellard, 2002). Aspects of personality have also been linked to low levels of academic achieve ment (Chamorro-Premuzic & Furnham, 2003),
ADHD AND ACADEMIC ADVERSITY
lower academic buoyancy (Martin, Nejad, Col mar, & Liem, 2013) and academic disengage ment (Martin et al., 2013). Indeed, research has found ADHD students lower on the personality factors Conscientiousness, Agreeableness and Emotional Stability, and higher on Neuroticism (De Pauw & Mervielde, 2011; Martel, Nigg, & Lucas, 2008). In addition, there may be schoollevel factors associated with negative academic outcomes, with some schools having higher rates of academic failure, problem behavior, and school exclusions (e.g., low SES schools, schools low in average ability, schools with higher proportions of minority students, schools with inadequate staff-to-student ratios; Martin, Anderson, Bobis, Way, & Veller, 2012; Perry & McConney, 2010). Aims of the Study This study sought to disentangle the effects of ADHD on academic adversity from the effects of personal (e.g., sociodemographics, personal ity, prior achievement, specific learning disabil ities, motivation) and contextual (e.g., school factors) covariates. For example, including spe cific learning disability alongside ADHD as a predictor of academic adversity enables us to understand unique variance attributable to ADHD, controlling for learning disability, as well as unique variance attributable to specific learning disability, controlling for ADHD. Such research provides important insights for inter vention. If the effects of ADHD are largely ameliorated once these covariates are entered, then clearly covariates are pivotal factors and intervention is to be directed to the covariates of relevance. To the extent that ADHD effects remain significant after entering covariates, then intervention is also to be directed at ADHD itself (e.g., toward executive functioning, med ication, or self-regulation, which have been shown to be central to ADHD; Chandler, 2010; DuPaul & Stoner, 2003; Pliszka, 2009). In ad dition, because multiple dimensions of aca demic adversity (e.g., failure, grade repetition, school refusal, school exclusion, schoolwork noncompletion) are included as outcome fac tors, the study aims to shed expanded light on the particular mix of ADHD, personal, and con textual factors that best predicts particular ad verse outcomes.
397
Method Sample and Procedure Students with and without ADHD were drawn from “regular” (or “m ainstream ”) schools of mixed ability. Four schools were coeducational, three were single-sex boys’ schools, and two were single-sex girls’ schools. Schools were from the independent school and the systemic Catholic sectors in major urban areas on the east coast of Australia. They com prised students from a range of SES levels that tended to be slightly higher SES than the na tional average (see Materials for School SES section). School achievement scores ranged from below the national average to above the national average as indicated by data from Aus tralian Curriculum and Assessment Authority (ACARA, 2012), though the average school achievement score was slightly higher than the national average. Ethics approval was provided by the re searcher’s university and parental consent was required. With few exceptions, targeted stu dents in attendance on the day of the testing participated in the survey. Students were al lowed to decline participation or withdraw from the process at any time, but none did so (to the author’s knowledge). Standardized instructions were provided to each teacher who administered the survey. The teachers explained the rating scale and students were presented with a sample item. Students were instructed not to identify themselves to ensure anonymity and allow full and frank responses. Students were also in structed to complete the survey on their own but were allowed to ask teachers for help with any survey items they had difficulty understanding or reading. Schools in Australia reside in one of three sectors, government, independent, and systemic Catholic. Most schools operate under curricu lum and assessment determined by their state or territory. In the main, Australian schools are academically comprehensive (i.e., little or no academic selection for entry) and include stu dents with ADHD (and many students with specific learning disabilities such as dyslexia). Although schools are academically nonselective, there is often streaming of classes within the school (usually in mathematics and Eng lish), with many students at academic risk (in-
398
MARTIN
eluding those with ADHD) often in the lower and middle classes. There is some additional funding available to schools to assist students with disorders such as ADHD and there are published standards relevant to instruction and assessment accommodations, classroom adjust ments, and targeted services with which schools are expected to comply (e.g., New South Wales Education and Communities, 2012), though, in reality the onus tends to be on individual schools and teachers to secure such funding and to ensure that standards are implemented appro priately and effectively. Students With ADHD The sample of students with ADHD (N = 136) were in junior high school (11-14 years, 59%) and senior high school (15-19 years, 41%). This sample of students with ADHD was built on the sample from two earlier studies that comprised 87 students with ADHD (Martin, 2012b, 2014). These students were the sub sample of students from the larger survey sam ple who reported (in the survey) that they had received a formal medical diagnosis of ADHD (thus, they were not specifically targeted stu dents). Although the validity of self-reports has received some support in relation to adults with ADHD (Kessler, Adler, Gruber, Sarawate, Spencer, & Van Brunt, 2007), self-report valid ity among adolescents is unclear. It is noted that N = 136 corresponds to a 3.5% incidence rel ative to the total sample, a prevalence broadly aligning with estimates of adolescents with ADHD (Barkley, 2006), however the present findings must be interpreted with the limitations of self-report in mind. The average age was 14.32 years (SD = 1.38). The average time since diagnosis was 5.07 years (SD = 4.61). Based on data available, the study was unable to identify whether students were mainly inatten tive type, hyperactive type, or both. Thus, the broad ADHD designation was adopted. Just un der half the sample (46%) was on medication to help manage the condition, a figure generally lower than population estimates (Visser, Bitsko, Danielson, Perou, & Blumberg, 2010). Consis tent with population statistics is the prevalence of boys (68%) compared with girls (32%) with ADHD, X2(l) = 12.97, p < .001. A total of 9% of students with ADHD were from a nonEnglish-speaking background (NESB); how
ever, students with ADHD were no more (or less) likely to come from an NESB, x2(l) = 2.45, p > .05. Students Without ADHD The non-ADHD group comprised a large sample of 3,779 students from the same schools and year levels as the ADHD group. Although the ADHD sample was augmented from the previous Martin (2012b, 2014) studies, this non-ADHD group was largely the same as that reported in the earlier studies (this sample was already large enough, and thus required no aug mentation). The non-ADHD group included students in junior high school (11-14 years, 58%) and senior high school (15-19 years, 42%). Just under half (47%) of the participants were female and 53% were male. The average age was 14.20 years (SD = 1.53). Of the sam ple, 14% were from an NESB. The non-ADHD sample was much larger than the ADHD sample because the study aimed to comprise a sample reflecting a more typical ratio of ADHD-to-nonADHD students and because a larger sample size would generate less biased standard-error estimates, the study retained this comprehensive representation of non-ADHD students. Materials The dependent variables (academic adver sity outcomes). The dependent measures were the adversity items from the Academic Risk and Resilience Scale (Martin, 2013). There are 10 adversity items, eight of which were used in this study. These were repeated a grade (4% of students); failed a school subject in the endof-year report card (18%); did not hand in most assignments (7%); moved to another class be cause of difficulties with work, teacher, or stu dents (8%); suspended from school (5%); changed schools (26%; numerous students tran sitioned from elementary to secondary school); expelled from school (1%); and avoided or re fused to go to school (11%). To each of these items, students answered No (coded 0) or Yes (coded 1). Two items were excluded. These were major illness (as this was not deemed a direct consequence of ADHD) and learning disability (as this was collinear with one of the predictors). This item set has been effec tively implemented in previous research into academic adversity (Martin, 2013). Correia-
ADHD AND ACADEMIC ADVERSITY
tions among outcome variables are presented in Table 1.
The Independent Variables (Personal and Contextual Covariates) Sociodemographics. S ociodem ographic variables com prised age, gender, language background, and SES indicated by parent edu cation and parent occupation. G ender was scored 1 for boys and 0 for girls. Age was analyzed as a continuous variable. For language background, students were asked if they spoke English (0) or another language (1; NESB) at home. Students were asked to report their moth ers’ (or female caregivers’) and fathers’ (or male caregivers’) education and occupation on an ordinal scale developed by the Australian Bureau of Statistics (ABS). Parent occupation and parent education factors were developed from the ordinal scales through deriving the average of m others’ and fathers’ education scores (for parent education; M = 4.59, SD = 1.33, range = 1 -6 ) and the average of parents’ occupation scores (for parent occupation; M = 5.28, SD = 1.75, range = 1-8). These two factors are recognized indicators of SES (Dun can, Daly, McDonough, & Williams, 2002). Correlations among all predictor variables (in cluding personality, academic, and school fac tors) are presented in Table 2. Personality. Five personality factors— Openness to Experience (Cronbach’s a = .73; M = 5.04; SD = .93), Extraversion (a = .81; M = 4.92; SD = 1.09), Neuroticism (a = .75; M = 3.74; SD = 1.04), Conscientiousness (a = .84; M = 4.75; SD = 1.15), and Agreeableness (a = .79; M = 5.57; SD = .88; 8 items per
Table 1
Correlations Among Outcomes Outcome variable 1. Repeated grade 2. Failed subject 3. Uncompleted work 4. Changed classes 5. Suspended 6. Changed schools 7. Expelled 8. School refusal
1
2
3
4
5
6
7
8
—
10 08 13 05 19 10 06
—
26 20 20 11 11 28
—
18 16 09 14 26
—
21 — 10 10 — 15 30 09 — 22 18 11 11 —
Note. Decimals omitted. Correlations > 1.041 significant at p < .05.
399
factor)— were measured with the 40-item Inter national English Big-Five Mini-Markers instru ment (IEBM; Thompson, 2008). Students rated (1 = Very inaccurate to 7 = Very accurate) the extent to which 40 one-word trait adjectives reflected accurate descriptions of themselves. Sample words for each factor are as follows: “creative” (Openness), “talkative” (Extraver sion), “efficient” (Conscientiousness), “moody” (Neuroticism), and “warm” (Agreeableness). Based on the acceptable reliability (Cronbach’s a ; Kline, 2000), five variables were developed from the means of each of the five-factor sets. Academic factors. Prior achievement was drawn from students’ results in the National Assessment Program in Literacy and Numeracy (NAPLAN) administered by the ACARA (ACARA, 2012). NAPLAN is a nationally standardized test on which school students receive scores each for literacy and numeracy. Students re ported their numeracy and literacy NAPLAN scores. Based on an acceptable reliability (a = .80; Kline, 2000), one achievement score was developed that was the mean of the two literacy and numeracy scores {M = 7.28, SD = 1.44, range = 1-10). Motivation was assessed via the Motivation and Engagement Scale (MES; Mar tin, 2010). Motivation comprises self-efficacy (e.g., “If I try hard, I believe I can do my schoolwork well”), mastery orientation (e.g., “I feel very pleased with myself when I do well at school by working hard”), valuing school (e.g., “Learning at school is important”), persistence (e.g., “If I don’t give up, I believe I can do difficult schoolwork”), planning (e.g., “I try to plan things out before I start working on my homework or assignments”), and task manage ment (e.g., “When I study, I usually try to find a place where I can study well”). For each item, students rated themselves on a scale of 1 {Strongly disagree) to 7 {Strongly agree). Based on the acceptable reliability (a = .86; Kline, 2000), one motivation-scale score was devel oped that was the mean of the six motivation indicators {M = 5.30, SD = .86). Specific learn ing disabilities are defined as formally determined/diagnosed problems with specific as pects, processes, or skills that affect learning (e.g., dyslexia; Australian Psychological Soci ety, 2014). This study assessed specific learning disabilities by asking all students in the survey sample if they experienced any formally deter mined or diagnosed reading difficulty/dyslexia,
MARTIN
400
o CN CN 1 1 1 l
o I' O CN ve VO 1
1 1
44
VO O Ov CN o CN 1 1 •'t1 1 cn VO m CN t1 1 cn VO r-- cn in in cn 1 T 1
in cn
CN CN CN cn Ov 00 O m cn Ov
■g
oW ° 3 t/3
CN Q *■§ ro H U
CN cn'st'invO t-'O O O v©
,-2 oo jS Q N 3 o -b W O 00 bo 00 CQ O Z < cn cn ■'fr >n vo i>
22. Indigenous %
Oo c o
Decimals omitted. Correlations > 1.041 significant at p < .05.
— e n o c N - cn I I I I
—13
—
-
—12
—o o I I I
0 0
I
-H M N N oo n 0 0 0 0 0 —0 II II
VO CO CO
CN CN 's f
o o c n o
41
o —o I
Note.
^ D ' O M n O N ' O I ' O H O o i n i n N i n o o
1
—21
00 oo cn CN vO r- t-» r-- in O o O o o o o o —o O s
ADHD AND ACADEMIC ADVERSITY
mathematics difficulty/dyscalculia, or writing difficulty/dysgraphia. For each student, a spe cific learning disability index was scored 1 for the presence of one or more of these specific disabilities and 0 for no such specific disability (n = 274 with specific learning disabilities; 7% of sample). School factors. School factors assessed were school-average achievement in national numeracy tests, school SES, school size, staffto-student ratio, ethnic and indigenous compo sition, and gender composition. School size (number of students relative to the number of grades in the school, M = 101 students; SD = 31.98) and full-time (equivalent) staff-tostudent ratio (higher scores reflecting more full time staff per student, M = 10.89; SD = 2.47), and gender composition (n = 3 boys’ schools with n = 1,268 students; n = 2 girls’ schools with n = 1182 students, n = 4 coeducational schools with n = 1465 students; coeducational was the reference category for dummy coding) are from ACARA, 2012 records. The School SES Index is a nationally assigned score (na tional M = 1,000) by the ABS comprising in formation on students’ household income, parent/caregiver occupation, and parent/caregiver education (higher scores reflect higher schoollevel SES; M = 1045, SD = 83). Schoolaverage achievement is drawn from ACARA (2012) and reflects the schools’ performance in annual national literacy and mathematics as sessment (M = 598.21, SD = 26.52). School NESB and school indigenous composition are the percentage of students in the school from an NESB (M = 25.17%, SD = 22.28) or indige nous background (M = 1.44%, SD = 1.59), based on ACARA records. Data Analysis Data were analyzed to identify the role of ADHD status in predicting academic adversity, controlling for sociodemographic, personality, academic, and school factors. Such modeling, which comprises numerous predictors, brings into consideration multiple regression tech niques. As each dependent variable was scored on a dichotomous scale, logistic regression was the analytical tool of choice. The method of entry was to enter ADHD in the first step to determine its initial variance for each academic adversity factor. This was followed by a for
401
ward stepwise procedure for the covariate set with entry testing based on the significance of the score statistic, and removal testing based on the probability of the Wald statistic (Tabachnick & Fidell, 2012). The probability of the score statistic for variable entry was .05. The probability of the Wald statistic to remove a variable w a s. 10. In this stepwise procedure, the role of covariates was determined. Of interest in these regressions was the initial role of ADHD in predicting academic adversity and then its subsequent role once covariates were entered; this latter step indicated unique variance in ac ademic adversity explained by ADHD and which factors do and do not mitigate its effects. Logistic regression was selected over dis criminant analysis as the former approach re quires fewer assumptions than the latter (Hosmer, Lemeshow, & Sturdivant, 2013). Core assumptions relevant to logistic regression were met (e.g., independent variables were interval, ratio, or dichotomous; relevant predictors were included; form of the relationships were linear; no autocorrelation; and no correlation between error and independent variables). Maximum likelihood (with the logit link function) was the method of estimation. A test of Cook’s distance indicated no particularly influential observa tions or outliers, using the criterion (Cook’s D > 1.00) suggested by Cook and Weisberg (1982). Missing data can pose problems if ex ceeding 5% (e.g., Graham & Hoffer, 2000). There are concerns about listwise, pairwise, and mean substitution approaches to missing data (Graham & Hoffer, 2000), leading to recom mendation of the expectation-maximization (EM) algorithm, implemented in this study us ing LISREL 8.80 (Joreskog & Sorbom, 2006). Imputation identified that less than 5% of the data were missing, so the EM algorithm was employed to manage this. Results Results of the logistic regressions are pre sented in Table 3 (comprising B values, statis tical significance, the step number at which a particular covariate significantly entered the model, and model fit). Initially, ADHD was predictive of grade repetition (B = 1.04, p < .01, OR = 2.83, 95% Cl [1.39, 5.76]); however, once Specific Learning Disabilities, Prior Achievement, and Indigenous School Composi-
MARTIN
402
II I I
o o 43
03
gd U
- D 43
•S 3 03 >.£? S 3
"3 *3
3 oM >p.2. Z CJ T3 ."S «
g fc — —.
o „ OQ W
