Adverse Life Events and Psychopathology and Prosocial Behavior in Late Adolescence: Testing the Timing, Specificity, Accumulation, Gradient, and Moderation of Contextual Risk

Copyright @ 2007 American Academy of Child and Adolescent Psychiatry. Unauthorized reproduction of this article is prohibited.

Adverse Life Events and Psychopathologyand Prosocial Behavior in Late Adolescence: Testingthe Timing, Specificity, Accumulation, Gradient,

and Moderation of Contextual Risk

EIRINI FLOURI, PH.D., AND CONSTANTINOS KALLIS, PH.D.

ABSTRACT

Objective: To explore the timing, specificity, accumulation, gradient, and moderation of contextual risk in psychopathology

and prosocial behavior in late adolescence. Method: In 2006, three hundred eighty-one 16- to 18-year-olds in Britain

reported on the number and type of adverse life events experienced in the past month, when they were age 15, and when

they were age 10, and on their concurrent psychopathology and prosocial behavior. They also undertook a reasoning

ability test. Control variables were sex, age, and mothers’ and fathers’ educational attainment when participants were age

10 years. Results: Although the number of adverse life events irrespective of their timing was associated with emotional

and behavioral problems, the number of proximal adverse life events experienced was associated with psychopathology

over and above the association of contextual risk in late childhood and in middle adolescence with psychopathology. The

cumulative risk model was more parsimonious than the specific risk model. The relationship between proximal cumulative

adversity and psychopathology was monotonic, and reasoning ability buffered the association of proximal cumulative

adversity with psychopathology, mainly because it moderated the association of proximal cumulative adversity with

hyperactivity. Conclusions: This study highlighted the importance of exploring risk accumulation rather than specificity in

explaining psychopathology and showed that the number of adverse life events experienced has a nonmultiplicative

association with psychopathology. J. Am. Acad. Child Adolesc. Psychiatry, 2007;46(12):1651Y1659. Key Words:

contextual risk, multivariate response models, Strengths and Difficulties Questionnaire.

RISK ACCUMULATION

Contextual risk factors do not occur in isolation, andit is the combination of various risk factors thatportends numerous negative child outcomes. InRutter’s (1979) approach to risk in child psychiatry,organismic characteristics as well as proximal and distalqualities of the environment are modeled collectively.

Cumulative risk is then calculated by a simplesummation of the multiple risk categories. In thisapproach, therefore, no one risk factor is seen as moreimportant than another, which is in line with thefoundation of cumulative risk theory that the con-fluence of risk factors rather than any singular risk,regardless of its context, is what leads to dysfunctionbecause it overwhelms the adaptive capacities of theorganism. This depicts the theoretical notion of massaccumulation or the idea that the total effect ofindividual risk factors is greater than the sum of theirindividual effects. Simmons et al. (1987), for example,found evidence that cumulative life transitions affectedchild outcomes in a curvilinear, accelerated manner,indicating that a high level of life change is especiallydifficult to manage.This theoretical justification aside, the methodolog-

ical benefits of using the cumulative risk approach are

Accepted July 5, 2007.Dr. Flouri is with the School of Psychology and Human Development,

Institute of Education, University of London; and Dr. Kallis is with the Centrefor Longitudinal Studies, Bedford Group for Lifecourse and Statistical Studies,Institute of Education, University of London.

The authors gratefully acknowledge Rina Siyani, who collected the data.Correspondence to Dr. Eirini Flouri, School of Psychology and Human

Development, Institute of Education, University of London, 25 Woburn Square,London WC1H 0AA, UK; e-mail: [email protected].

0890-8567/07/4612-1651�2007 by the American Academy of Childand Adolescent Psychiatry.

DOI: 10.1097/chi.0b013e318156a81a

1651J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:12, DECEMBER 2007


that cumulative risk indexes can capture the naturalcovariation of risk factors (e.g., physical risk factors suchas poor housing quality, noise, and pollution are highlyinterrelated as are psychosocial risk parameters such asfamily turmoil and violence [Evans, 2003]), thataggregate variables of risk are more stable than anyindividual measure, and that there is increased power todetect effects because errors of measurement decrease asscores are summed and degrees of freedom arepreserved (Burchinal et al., 2000). Cumulative riskmeasures are consistently found to explain morevariance in children’s outcomes than single factors(Atzaba-Poria et al., 2004; Deater-Deckard et al., 1998;Sameroff et al., 1993). However, it is not clear whetherthe relationship between cumulative risk and childmental health problems is linear or nonlinear. Moststudies have identified a linear relationship wherebyincrements in risk factors have a steady, additive effecton child psychopathology (e.g., Deater-Deckard et al.,1998). Because few researchers (e.g., Gerard andBuehler, 2004) actually report whether their investiga-tions included appropriate tests for nonlinear patternsof cumulative risk, this ignores the possibility of anonlinear relationship that may manifest itself, forexample, as an acceleration of mental health problemsat a critical level of risk.

Specificity

A drawback of the cumulative risk approach is theassumption that each risk factor carries the same weightin children’s lives and that risk factors are interchange-able. Although several investigators have shown that theexplanatory power of the cumulative risk metric inexplaining psychopathology and adjustment is fargreater than individual risk effects (e.g., Ackermanet al., 1999), individual risk factors do vary in therespective impacts that they have. Therefore, one needsto also assess the influence of individual risk factors sothat they could be examined simultaneously withoutlosing their particular salience. In the aforementionedAckerman et al. (1999) study, for instance, althoughtheir 11 risk factor indicators together were significantlyassociated with child psychopathology, parental alco-hol/drug abuse was the only individual risk factor thataccounted for a significant amount of unique variance.Therefore, the importance of testing for stressorspecificity in child psychopathology and adjustmentshould, equally, not be underestimated (McMahon

et al., 2003). At the same time, risk indicatorsunderlying the development of problem behavior inone child adjustment domain may not underlie thedevelopment of problem behavior in another childadjustment domain (Tiet et al., 2001), which also raisesquestions about the legitimacy of the cumulative riskperspective, and calls for tests of outcome specificityas well.

Resilience

Although numerous recent studies in child andadolescent psychiatry have paid attention to thephenomenon of resilience, meaning a degree ofresistance to adversities, operationally defined as rela-tively good outcomes despite experiencing major risks,few studies (e.g., Ackerman et al., 1999) have actuallyexamined factors that protect against cumulative risk.Various protective factors, usually conceptualized at theindividual level (e.g., intelligence, self-regulation,positive temperament) because resilience suggests theindividual’s response to risk factors, have been identifiedin the literature as moderating the impact of contextualrisk on children’s psychopathology and adjustment(Masten, 2001), with cognitive ability featuring as anadaptive resource with importance for positive adultoutcomes as well (Masten et al., 2004). More recently,Grant et al. (2006), reviewing the studies that havetested for moderators of the relationship betweenstressors and child and adolescent psychopathology,showed that of the at least 11 studies that have examinedintellectual/academic competence as a moderatingvariable, 55% found such an effect. However, studieshave yet to explore (see also Tiet et al., 1998) ability asmoderating the impact of cumulative risk on psycho-pathology and adjustment.

Rationale for the Proposed Study

This study was undertaken to address the concernsraised above. It was designed to extend in several waysprevious work on the role of contextual risk inpsychopathology and adjustment. First, following thesuggestion of Grant et al. (2003) that future studiesshould aim to develop a taxonomy of stressors similar tothe taxonomies developed for child and adolescentpsychopathology, it used a well-validated measure offamily risk. This is important because the variability incumulative stressor measurement is often such thatmakes comparisons of studies almost meaningless, with

FLOURI AND KALLIS

1652 J . AM. ACAD. CHILD ADOLESC. PSYCHIATRY, 46:12, DECEMBER 2007


only a small proportion of studies that explorecumulative risk using well-validated measures. Accord-ing to Grant et al. (2003), this proportion is less than10%, with 45% of studies reporting that theydeveloped their own measure and the remaining onesusing one of the approximately 50 currently availablemeasures of cumulative stressors. What is more,psychometric data on most of these measures were notprovided, and only a few of the authors who developedtheir own scales provided any information about theirmethod of measurement development or even itemsincluded in their scales. Second, by measuring the sameaspects of family risk in three ages, it explored the role ofthe timing of risk in psychopathology and adjustment.Third, it compared the cumulative risk with the specificrisks models to test for risk specificity. Fourth, itexplored the gradient of risk, and fifth, it tested whetherability moderates the impact of adverse life events onpsychopathology and adjustment.

METHOD

Participants

For the purpose of this study, data provided from 381participants (195 of whom were male) ages between 16 and 18years (mean age 16.97 years, SD 0.71) were used. The participantswere year 12 and 13 students from two state secondary schools in aninner London borough chosen to have a similar current Office forStandards in Education report. Office for Standards in Education isa nonministerial U.K. government department, set up onSeptember 1, 1992, whose main aim is to help improve the qualityand standards of education and child care through independentinspection and regulation and provide advice to the Secretary ofState. Schools are normally inspected on a 6-year cycle. A team ofinspectors led by a Registered Inspector spends a few days in theschool observing lessons and speaking to teachers and pupils togather evidence on how well the school is performing. The length ofthe school inspection is related to the size of the school and the typeof inspection. Office for Standards in Education inspectorssummarized the teaching and learning in both schools to be goodoverall and reported both schools to provide an enriching andstimulating learning environment. In both schools, the number ofstudents with special educational needs was in line with the nationalaverage, whereas the number of students entitled to free schoolmeals was slightly higher than the national average. In addition, inboth schools, the number of students achieving five or more A to Cgrades in the national General Certificate of Secondary Education(GCSE) exams was in line with the national average and above theaverage of similar schools. The GCSE was introduced in 1986 withfirst examinations in 1988. It is the principal means of assessingpupil attainment at the end of compulsory secondary education. A(advanced) levels remain the gold standard academic qualificationfor the 16- to 19-year-old age group. These will normally be takenover 2 years (full time) at a sixth form college, school sixth form, atertiary college, or a college of further education and can give entry

to higher education. Normally two or three A levels are studied (bythose with the appropriate entry qualifications, usually five or moreGCSEs at grades A to C).

Materials and Procedure

The questionnaire, designed to collect information from theadolescents who agreed to take part in the study, was piloted twice.Ethical approval for the study was obtained from the DepartmentalEthics Committee, and consent for participation was gained fromthe Head of Sixth Form at each school. The questionnaires wereadministered during a Personal Social Health Education lesson inthe classroom. The adolescents were told that they were allowed toopt out of the study at any point, were reassured that thequestionnaires were anonymous and confidential, and wereinformed of the process of the questionnaire administration.First, a 3-minute timed reasoning ability test was administered.

Next, the adolescents were informed that they would complete aquestionnaire about themselves and their family. They werereminded once again that the questionnaires were anonymous andconfidential and were encouraged to be honest when answering thequestions. They were reminded to answer all of the questions andwere given the opportunity to raise any questions about the study.There was no time limit for this part of the questionnaire.

Measures

Reasoning Ability. Reasoning ability was measured with theBaddeley Reasoning Test, a 3-minute reasoning test based ongrammatical transformation (Baddeley, 1968). The BaddeleyReasoning Test is a quick and reliable (Carter and Kennedy,1981) measurement of fluid intelligence through logical reasoning.The test’s construct validity has also been supported (Kyllonen andChristal, 1990). Scores range from 0 to 64. Each of the 64 items inthe scale is presented in the form of a grammatical transformationthat is either true or false, for example, BA follows B Y BA^ (true) orBB precedes A Y AB^ (false). In this study, the participants’ scoresranged from 0 to 62. The average score was 22.16 (SD 13.73), andthe median was 20. Skewness was 0.82 (SE 0.13).

Contextual and Structural Factors. The contextual and structuralfactors included sex, age, and family socioeconomic status(measured with father’s and mother’s educational attainment).

Psychopathology and Adjustment

The Strengths and Difficulties Questionnaire (SDQ), a 25-item3-point (ranging from 0 to 2) scale measuring four difficulties(hyperactivity, emotional symptoms, conduct problems, and peerproblems), as well as prosocial behavior (Goodman, 1994, 1997)was used to measure adolescents’ problem behavior. The reliabilityand validity of SDQ have been well established (e.g., Goodman,2001). Each SDQ subscale had five items such as Bconstantlyfidgeting or squirming^ (hyperactivity), Bmany worries, often seemsworried^ (emotional symptoms), Bsteals from home, school orelsewhere^ (conduct problems), Brather solitary, tends to playalone^ (peer problems), and Bhelpful if someone is hurt, upset orfeeling ill^ (prosocial behavior). A total difficulties scale is calculatedby summing the scores for hyperactivity, emotional symptoms,conduct problems, and peer problems ( www.sdqinfo.com ). Cutoffscores for the borderline/abnormal range (the SDQ cutoff scoreidentifies 20% of the population) are 16+ for total difficulties, 6+for emotional symptoms, 4+ for conduct problems, 6+ for

ADVERSE LIFE EVENTS AND PSYCHOPATHOLOGY



hyperactivity, and 4+ for peer problems, whereas the borderline/abnormal range for prosocial behavior is 0 to 5. In this study, 18% ofthe respondents were high scorers in total difficulties, 20% inemotional symptoms, 19% in conduct problems, 19% in hyper-activity, 14% in peer problems, and 13% in prosocial behavior.

Adverse Life Events. The Adverse Life Events scale of Tiet et al.(2001), a modification of the Life Events Checklist (Brand andJohnson, 1982; Coddington, 1972a,b), which has acceptablevalidity and test-retest reliability (Brand and Johnson, 1982), wasused to measure contextual risk. The Life Events Checklist is ameasure of exposure to potentially traumatic events developed at theNational Center for Posttraumatic Stress Disorder (PTSD) tofacilitate the diagnosis of PTSD. The Adverse Life Events scale iscomposed of 25 possible events occurring within a specified timeframe. For most of these, youths had no control (e.g., Bsomeone inthe family was arrested,^ Bnegative change in parents’ financialsituation^). In this study, adolescents were asked to complete thescale for three time periods in their life: when they were 10 years old,when they were in year 11 (i.e., their GCSE year) at age 15, andduring the past month. Adverse life events ranged from 0 to 16 atages 10 and age 15 and from 0 to 14 in the past month. Althoughapproximately one third of the participants had not experienced anyadverse life events in the past month (30.2%), one fourth hadexperienced no adverse life events at age 10 (23.1%) and one fifthhad experienced no adverse life events at age 15 (21%). The medianwas two for adverse life events at age 10 and in the past month(mean 2.89, SD 3.08 and mean 2.19, SD 2.29, respectively) andthree for adverse life events at age 15 (mean 3.27, SD 3.04).

RESULTS

First we investigated the association between cumu-lative contextual risk (number of adverse life eventsexperienced) and adolescents’ total difficulties usingordinary linear regression models. It was found thatalthough cumulative contextual risk at age 10, at age 15,and in the past month were related to total difficulties,the association of total difficulties with proximal (i.e., inthe past month) adverse life events was the strongest.This baseline model (without confounders) indicatedthat there was a strong positive effect of proximal lifeadversities on total difficulties (b = .661, SE 0.116). Inthe next step, the full model was introduced. Thisincluded the following variables as possible controlvariables: adverse life events at age 15, adverse life eventsat age 10, age, sex, whether the father had a universitydegree when the youth was age 10, whether the motherhad a university degree when the youth was age 10, andreasoning ability. The effect of cumulative contextualrisk in the past month was slightly smaller (b = .625, SE0.151) but still statistically significant at 1% level. Noneof the control variables in the full model mentionedabove had a statistically significant effect on the totaldifficulties score.

Next, we compared the cumulative with the specificrisk models. Applying a Bonferroni correction (" of0.05/25 adverse events = .002), we concluded thatTable 1 suggests that losing a close friend and havingfamily with a drug or alcohol problem were the onlyspecific proximal risks significantly associated withadolescents’ current total difficulties. Adolescents whoreported that someone in their family had died in thepast month appeared to score lower on totaldifficulties than those without this experience, whichappears counterintuitive. Further analyses on the SDQsubscales (analyses available from the authors) showedthat this was because this item was significant inpredicting negatively hyperactivity in both the baseline(b = j1.131, SE 0.495) and the full (b = j1.205,SE 0.495) model, which is in line with previousresearch (Schmiege et al., 2006).To compare the goodness of fit of the cumulative risk

and the specific risks models, we used an appropriatestatistic that takes into account that the models arenot nested. For this purpose, we used the Bayesianinformation criterion, which is a function of thelikelihood, the number of observations, and thenumber of free parameters of each model (Schwarz,1978). As can be seen in Table 1 the Bayesianinformation criterion for the cumulative risk modelwas much lower than that for the specific risksmodel. Thus, we conclude that the cumulative riskmodel specification should be preferred. For thisreason, we used the number of adverse life eventsexperienced in the past month risk specification for theremainder of the statistical analysis.Therefore, to test for the gradient of the contextual

risk index, we entered total number of adverse lifeevents in the past month squared in the cumulative riskmodel presented in Table 1. Its effect was, however, notdifferent from zero (b = j.013, SE 0.035), whichsuggests that the relationship between number ofproximal adverse life events experienced and totaldifficulties is linear.Next, we explored the association between the

number of proximal adverse life events experienced inthe past month and the four difficulties (hyperactivity,emotional symptoms, conduct problems, and peerproblems) and prosocial behavior problem binary (dueto the subscale scores’ skewed distributions) indicators.Because the five behavior problem indicators are likelyto be correlated, we fitted a multivariate response

FLOURI AND KALLIS



logistic regression model allowing for the correlation ofthe error terms using the statistical software packageMLwiN (Goldstein, 2003). This was because it ispossible that the effect of the total number of adverselife events on the five SDQ behavior problem indicatorsmay be confounded by correlated unobserved factors.Therefore, by running a multivariate response modelwith correlated error terms, we control for the effect ofthese unobserved confounders.

The results of the analysis are shown in Tables 2 and3. The baseline model (without any observed con-founders) indicated that number of adverse life eventsin the past month was associated with the fourdifficulties but not with prosocial behavior problemsat the 5% significance level (analyses available from theauthors). Even when we added the observed con-founders included in the total difficulties full model,the effect of the number of adverse life events in the

TABLE 1Cumulative Risk and Specific Risks Model Specification: Total Difficulties

Adverse Life Events in Past Month Model Specification Cumulative Risk Specific Risks

Variables Coeff. SE p Coeff. SE p

No. of adverse life events in past month 0.624 0.150 .000 NA NA NAType of adverse life events in past monthFamily member died (n = 59) NA NA NA j2.536 0.779 .001Family member was seriously injured (n = 62) NA NA NA 0.363 0.804 .652Saw crime or an accident (n = 85) NA NA NA j0.121 0.712 .865Lost a close friend (broke/split up; n = 54) NA NA NA 3.028 0.836 .000Close friend was seriously sick or injured (n = 43) NA NA NA j0.240 0.893 .788Negative change in parent’s financial situation (n = 69) NA NA NA 0.146 0.776 0.851Family had drug/alcohol problem (n = 33) NA NA NA 3.531 1.060 .001Got seriously sick or injured (n = 46) NA NA NA j0.492 0.916 .592Parents argued more than previously (n = 48) NA NA NA 2.243 0.845 .008Mother/father figure lost job (n = 19) NA NA NA j0.780 1.277 .542One parent was away from home more often (n = 58) NA NA NA 0.819 0.800 .307Someone in the family was arrested (n = 16) NA NA NA 1.212 1.443 .401Close friend died (n = 18) NA NA NA 2.511 1.432 .080Family member had mental/emotional problem (n = 42) NA NA NA 1.063 0.944 .261Brother or sister left home (n = 26) NA NA NA 0.699 1.190 .557Being a victim of crime/violence/assault (n = 15) NA NA NA 2.008 1.402 .153Parents separated (n = 16) NA NA NA 0.788 1.487 .596Parent(s) got in trouble with the law (n = 14) NA NA NA j1.567 1.640 .340Attended a new school (n = 12) NA NA NA j3.000 1.622 .065Family moved (n = 17) NA NA NA 2.317 1.473 .117Parents got divorced (n = 13) NA NA NA 0.761 1.688 .652One of the parents went to jail (n = 7) NA NA NA j1.233 2.437 .613Got a new stepmother or stepfather (n = 11) NA NA NA 0.897 1.786 .616Parent got a new job (n = 29) NA NA NA 0.105 1.110 .924Got a new brother or sister (n = 18) NA NA NA 2.094 1.348 .121No. of adverse life events at age 15 y 0.069 0.119 .561 0.092 0.124 .460No. of adverse life events at age 10 y j0.007 0.102 .940 j0.094 0.103 .365Age 17 y j0.245 0.649 .705 j0.280 0.643 .663Age 18 y j1.141 0.762 .135 j1.522 0.764 .047Girl 0.263 0.541 .626 0.435 0.544 .425Father at youth’s age 10 y did not have university degree 0.165 0.674 .806 j0.438 0.669 .513Mother at youth’s age 10 y did not have university degree j0.354 0.695 .611 0.157 0.694 .820Reasoning ability j0.009 0.109 .644 j0.030 0.019 .123BIC 125.579 211.056

Note: BIC = Bayesian information criterion.




past month on the four difficulties was significant atthe 5% level (Table 2). From Table 3 we can see thatall but three error term correlations were statisticallysignificant at the 5% level. This suggests that it isadvantageous to model the five behavior outcomessimultaneously. As the full multivariate responsemodel shows, there is a significant positive correlationbetween time-constant unobserved factors affectingprosocial and conduct problems, prosocial and peerproblems, emotional symptoms and hyperactivityproblems, emotional symptoms and peer problems,emotional symptoms and conduct problems, andfinally hyperactivity and peer problems.To see whether the relationship of number of

proximal adverse life events experienced with thespecific problem behaviors is linear or nonlinear, weran these multivariate response models with correlatederror terms having entered total number of adverse lifeevents in the past month squared. We found that, aswith the relationship between number of proximaladverse life events and total difficulties, the relationshipbetween number of proximal adverse life events andeach of the five problem behaviors was linear. None of

the effects of total number of proximal adverse lifeevents squared on prosocial behavior, emotionalsymptoms, conduct problems, hyperactivity, and peerproblems were different from zero (b = .026, SE 0.020;b = j.023, SE 0.020; b = .029, SE 0.022; b = j.030,SE 0.019, and b = j.001, SE 0.016, respectively).Finally, to explore the possible moderator effect of

reasoning ability on the relationship between number ofadverse life events in the past month and problembehavior, we added to the full total difficulties model(presented in Table 1 and discussed above) aninteraction between number of adverse life events inthe past month and reasoning ability. This interactionwas statistically significant at the 5% level (b = .018,SE 0.008), thus indicating that reasoning ability was animportant moderator of the relationship of number ofproximal adverse life events with total difficulties.When this analysis was carried out separately for the fiveSDQ subscales on the full multivariate response model,a statistically significant moderating effect of reasoningability on the effect of total number of adverse lifeevents in the past month was also found but only forhyperactivity (b = .008, SE 0.004). The interaction

TABLE 3Correlation Matrix of the Behavior Problem Error Terms

Behavior ProblemProsocial Behavior

Coeff. (SE)Emotional Symptoms

Coeff. (SE)Conduct Problems

Coeff. (SE)HyperactivityCoeff. (SE)

Prosocial behavior VEmotional symptoms j0.083 (0.051) VConduct problems 0.116 (0.050) j0.022 (0.051) VHyperactivity 0.009 (0.051) 0.275 (0.046) 0.138 (0.050) VPeer problems 0.110 (0.050) 0.389 (0.041) 0.076 (0.051) 0.240 (0.047)

TABLE 2Adverse Life Events on Behavior Problems

ProsocialBehavior

EmotionalSymptoms

ConductProblems Hyperactivity

PeerProblems

Variables Coeff. SE Coeff. SE Coeff. SE Coeff. SE Coeff. SE

No. of adverse life events in past month 0.094 0.082 0.118 0.070 0.371 0.077 0.174 0.070 0.171 0.076No. of adverse life events at age 15 y 0.008 0.067 0.042 0.055 j0.101 0.060j0.109 0.058 0.007 0.062No. of adverse life events at age 10 y 0.055 0.054j0.069 0.052 j0.007 0.053 0.065 0.049j0.091 0.060Age 17 y 0.602 0.438 0.070 0.312 j0.180 0.320j0.047 0.310 0.213 0.376Age 18 y 0.764 0.483j0.188 0.381 j0.934 0.439j0.469 0.389 0.403 0.421Father at youth’s age 10 y did not have university degree j0.819 0.373 0.307 0.342 0.375 0.379j0.517 0.326 0.114 0.380Mother at youth’s age 10 y did not have university degree 0.268 0.396 0.017 0.347 j0.136 0.371 0.563 0.363 0.068 0.391Girl j1.000 0.344 0.502 0.265 j0.608 0.289 0.123 0.267j0.139 0.298Reasoning ability j0.013 0.012j0.001 0.010 j0.015 0.011 0.010 0.010 0.010 0.010

FLOURI AND KALLIS



between reasoning ability and number of proximaladverse life events was not different from zero inexplaining prosocial behavior (b = .000, SE 0.005),emotional symptoms (b = .002, SE 0.004), conductproblems (b = .006, SE 0.004), or peer problems(b = .004, SE 0.004).

DISCUSSION

This study was carried out to explore the role ofcontextual risk in adolescents’ psychopathology andadjustment. Assessing family risk with a well-validatedmeasure at three time points (age 10, age 15, andduring the past month) in a sample of 16- to 18-year-old adolescents, it showed, in line with previousresearch, that although the number of adverse lifeevents irrespective of their timing was associated withemotional and behavioral problems, the number ofproximal adverse life events experienced was associatedwith emotional and behavioral problems over andabove the association of emotional and behavioralproblems and contextual risk in late childhood andin middle adolescence. However, the number ofproximal adverse life events was not related to prosocialbehavior.

The study also contributed to the evidence base byinvestigating the association between risk specificity andbroad psychopathology in adolescence. By showing thatthe most parsimonious model was the cumulative riskrather than the specific risks model, it highlighted theimportance of exploring risk accumulation rather thanspecificity in explaining adolescent psychopathology.Previous studies (e.g., Sandler et al., 1992; Tiet et al.,2001) have tested stressor-specific, outcome-specificmodels.These included a heterogeneous sample ofstressors (i.e., single risk factors) and a range ofpsychological outcomes, therefore allowing for specifi-city of both stressor and outcome to be determined:each of the stressors was examined in relation to each ofthe outcomes. The present study did test stressor-specific but not stressor-specific, outcome-specificmodels because it found that the cumulative riskmodel should be preferred over the specific risk modelin explaining broad psychopathology. Furthermore, bytesting for the gradient of the cumulative contextual riskindex, this study showed that the relationship betweenproximal contextual adversity and psychopathology ismonotonic. This suggests that increments in the

number of proximal adverse life events experiencedincrease psychopathology scores, which highlights theimportance of protecting those at risk from further riskexposure. Finally, it showed that reasoning abilitymoderated the association of proximal cumulativeadversity with psychopathology, mainly because itbuffered the association of contextual risk withhyperactivity.The correlations between the error terms of the

various problem behaviors suggest that variables beyondthose indexed by a history of life adversities, sex, age,reasoning ability, and parental socioeconomic SESpotentially contribute to a shared vulnerability process.We use Bpotentially^ because this correlation couldreflect processes other than shared vulnerability, suchas shared method variance or reciprocal influences.Put simply, this study showed that any left-outregressors of the psychiatric outcomes examined arecorrelated.

Limitations

The strengths of this study should be seen in light ofits limitations. Selection bias is likely as a smaller thanaverage proportion of the study participants scoredabove cutoff for peer and prosocial behavior problems.This was to be expected, however, given that the studyparticipants remained in school after the minimum agethat children leave school (16 years), which suggeststhat they were likely to be well adjusted bothacademically and psychologically. In addition, this, asthe well-known Adverse Childhood Experiences Studyin the United States, is a retrospective longitudinalstudy of contextual risk. Although retrospective designsmay suggest possible risk factors for outcomes, the testof the validity of these hypothetical relationships lies inprospective designs (Widom et al., 2004) and experi-ments (Costello et al., 2003). Reporting of recentadverse events may be highly related to current state ofbehavioral or emotional difficulties due only to issues ofrecency and coloring of recall by current state.Furthermore, our models necessarily treated life eventsas if they were independent events when in fact theymay not be. In addition, our methods have a problemwith single source confounding: the SDQ and theadverse life events have the same reporter, the youth.This could falsely raise correlations. This study wouldhave been much stronger if parents had also reportedon life events. Finally, the threat to reliability and validity




of using retrospective reporting of life events becomesan increasing problem as the reporting intervallengthens.To develop the field further, future studies should

extend prior work on the role of contextual risk inproblem behavior in several ways. First, studies lookingat family contextual risk particularly should disentanglethe family-wide from the person-specific risk factors.With few exceptions (e.g., Ackerman et al., 1999),previous studies group in the same cumulative riskindex both behaviors, which can be child-specific (e.g.,parenting, abuse) and family-wide factors (e.g., house-hold dysfunction, poverty). For example, the AdverseChildhood Experiences Study operationalized cumula-tive contextual risk (e.g., Anda et al., 1999; Chapmanet al., 2004; Dube et al., 2001; Whitfield et al., 2005)on the basis of the presence/absence of eight adversechildhood experiences: emotional, physical, and sexualabuse; a battered mother; parental separation ordivorce; and growing up with a substance-abusing,mentally ill, or incarcerated household member.Second, they should test whether contextual risk atthe neighborhood or the family level is a betterpredictor of problem behavior and should explorearea effects. Third, and related to this, they shouldexplore within-family effects of cumulative contextualrisk on psychopathology (O’Connor et al., 2001). Thiscan be done easily by using sibling samples. Fourth,they should use experimental and prospective long-itudinal designs, and they should also explore whetherthere is evidence of specificity and accumulation ofprotection.

Clinical Implications

This study showed that the number of adverse lifeevents experienced has a nonmultiplicative associationwith adolescent psychopathology. Clinical assessmentof emotional and behavioral problems should, there-fore, include a comprehensive assessment of environ-mental components, and efforts should be made toprotect youths from further risk exposure. In addition,this study showed that the number was more importantthan the type of risk factors in explaining problembehavior in late adolescence. This suggests thatprevalence estimates or identification of high-riskyouths may be underestimated if based solely onexposure to a single extreme risk factor. This means thatby identifying solely extreme risk on the basis of single

risk factors, those who may be at higher risk due toexperience of multiple medium-level risks areneglected.

Disclosure: The authors have no financial relationships to disclose.

REFERENCES

Ackerman BP, Izard CE, Schoff K, Youngstrom EA, Kogos J (1999),Contextual risk, caregiver emotionality, and the problem behaviors ofsix-and seven-year-old children from economically disadvantagedfamilies. Child Dev 70:1415Y1427

Anda RF, Croft JB, Felitti VJ et al. (1999), Adverse childhood experiencesand smoking during adolescence and adulthood. JAMA 282:1652Y1658

Atzaba-Poria N, Pike A, Deater-Deckard K (2004), Do risk factors forproblem behaviour act in a cumulative manner? An examination ofethnic minority and majority children through an ecological perspective.J Child Psychol Psychiatry 45:707Y718

Baddeley AD (1968), A three-minute reasoning test based on grammaticaltransformation. Psychometric Sci 10:341Y342

Brand AH, Johnson JH (1982), Note on reliability of the Life EventsChecklist. Psychol Rep 50:1274

Burchinal M, Roberts J, Hooper S, Zeisel S. (2000), Cumulative risk andearly cognitive development: a comparison of statistical risk models. DevPsychol 36:793Y807

Carter RC, Kennedy RS, Bittner AC Jr (1981), Grammatical reasoning: astable performance yardstick. Hum Factors 23:587Y591

Chapman DP, Whitfield CL, Felitti VJ (2004), Adverse childhoodexperiences and the risk of depressive disorders in adulthood. J AffectDisord 82:217Y225

Coddington RD (1972a), The significance of life events as etiologic factorsin the diseases of children: I. A survey of professional workers.J Psychosom Res 16:7Y18

Coddington RD. (1972b), The significance of life events as etiologic factorsin the diseases of children: II. A study of a normal population.J Psychosom Res 16:205Y213

Costello EJ, Compton SN, Keeler G (2003), Relationships between povertyand psychopathology: a natural experiment. JAMA 290:2023Y2029

Deater-Deckard K, Dodge KA, Bates JE, Pettit GS (1998), Multiple riskfactors in the development of externalizing behavior problems: groupand individual differences. Dev Psychopathol 10:469Y493

Dube SR, Anda RF, Felitti VJ, Chapman DP, Williamson DF, Giles WH(2001), Childhood abuse, household dysfunction, and the risk ofattempted suicide throughout the life span: findings from the AdverseChildhood Experiences study. JAMA 286:3089Y3096

Evans GW (2003), A multimethodological analysis of cumulative risk andallostatic load among rural children. Dev Psychol 39:924Y933

Gerard JM, Buehler C (2004), Cumulative environmental risk and youthproblem behavior. J Marriage Fam 66:702Y720

Goldstein H (2003), Multilevel Statistical Models, 3rd ed. London: ArnoldGoodman R (1994), A modified version of the Rutter parent questionnaire

including extra items on children’s strengths: a research note. J ChildPsychol Psychiatry 35:1483Y1494

Goodman R (1997), The Strengths and Difficulties Questionnaire: aresearch note. J Child Psychol Psychiatry 38:581Y586

Goodman R (2001), Psychometric properties of the Strengths andDifficulties Questionnaire. J Am Acad Child Adolesc Psychiatry 40:1337Y1345

Grant KE, Compas BE, Stuhlmacher AF, Thurm AE, McMahon SD(2003), Stress and child and adolescent psychopathology: moving frommarkers to mechanisms of risk. Psychol Bull 129:447Y466

Grant KE, Compas BE, Thurn AE et al. (2006), Stressors and child andadolescent psychopathology: evidence of moderating and mediatingeffects. Clin Psychol Rev 26:257Y283

FLOURI AND KALLIS



Kyllonen PC, Christal RE (1990), Reasoning ability is (little more than)working-memory capacity? Intelligence 14:389Y433

Masten AS (2001), Ordinary magic: resilience processes in development. AmPsychol 56:227Y238

Masten AS, Burt KB, Roisman GI, Obradovic J, Long JD, Tellegen A(2004), Resources and resilience in the transition to adulthood:continuity and change. Dev Psychopathol 16:1071Y1094

McMahon SD, Grant KE, Compas BE (2003), Stress and psychopathologyin children and adolescents: is there evidence of specificity? J ChildPsychol Psychiatry 44:107Y133

O’Connor TG, Dunn J, Jenkins JM, Pickering K, Rasbash J (2001), Familysettings and children’s adjustment: differential adjustment within andacross families. Br J Psychiatry 179:110Y115

Rutter M (1979), Protective factors in children’s responses to stress anddisadvantage. In: Primary Prevention of Psychopathology: III. PromotingSocial Competence and Coping in Children, Kent MW, Rolf JE, eds.Hanover, NH: University Press of New England, pp 49Y74

Sameroff AJ, Seifer R, Baldwin A, Baldwin C (1993), Stability of intelligencefrom preschool to adolescence: the influence of social and family riskfactors. Child Dev 64:80Y97

Sandler IN, Reynolds KD, Kliewer W, Ramirez R (1992), Specificity of therelation between life events and psychological symptomatology. J ClinChild Psychol 21:240Y248

Schmiege SJ, Khoo ST, Sandler IN, Ayers TS, Wolchik SA (2006),Symptoms of internalizing and externalizing problems: modelingrecovery curves after the death of a parent. Am J Prev Med 31:S152YS160

Schwarz G (1978), Estimating the dimension of a model. Ann Stat 6:461Y464Simmons RG, Burgeson R, Carlton-Ford S, Blyth DA (1987), The impact

of cumulative change in early adolescence. Child Dev 58:1220Y1234Tiet QQ, Bird HR, Davies M et al. (1998), Adverse life events and resilience.

J Am Acad Child Adolesc Psychiatry 37:1191Y1200Tiet QQ, Bird HR, Hoven CW et al. (2001), Relationship between specific

adverse life events and psychiatric disorders. J Abnorm Child Psychol29:153Y164

Whitfield CL, Dube SR, Felitti VJ (2005), Adverse childhood experiencesand hallucinations. Child Abuse Negl 29:797Y810

Widom CS, Raphael KG, DuMont KA (2004), The case for prospectivelongitudinal studies in child maltreatment research: commentary onDube, Williamson, Thompson, Felitti, and Anda (2004). Child AbuseNegl 28:715Y722

Trajectories of Socioeconomic Status Across Children_s Lifetime Predict Health Edith Chen, PhD, Andrew D. Martin,PhD, Karen A. Matthews, PhD

Objective: Socioeconomic status is one of the most robust social factors associated with health, but the dynamics of howsocioeconomic status over time affects children_s health remains unclear. This study tested how various models of childhoodsocioeconomic status (accumulation, change, variability, and critical periods of family income) would predict health outcomes ata final time point in childhood. Methods: This was a prospective, longitudinal study of 6306 children who were aged 10 to 11years and whose families were interviewed every other year from birth onward. The sample came from the US NationalLongitudinal Survey of Youth-Children. In the same data set, a replication sample of 4305 14- to 15-year-old children was alsoexamined. Primary outcomes included parent report of asthma and conditions that limited activity and school and requiredphysician treatment. Results: Lower cumulative family income was associated with higher odds for having a condition that limitedchildhood activities, as well as a condition that required treatment by a physician at ages 10 to 11. Cumulative family income wasa stronger predictor than change in income or variability in income. Lower family income early in life (ages 0Y5 years) wasassociated with higher odds for having a condition that limited activities and a condition that required treatment by a physician atages 10 to 11, independent of current socioeconomic status. Findings were replicated in the 14- to 15-year-old sample.Conclusions: These findings suggest that the accumulation of socioeconomic status in terms of family income across childhood ismore important than social mobility or variability in socioeconomic status, although there may be certain periods of time (earlylife) that have stronger effects on health. These findings suggest the importance of childhood interventions for reducing healthdisparities. Pediatrics 2007;120:e297Ye303.



Documents

Adverse Life Events and Psychopathology and Prosocial Behavior in Late Adolescence: Testing the Timing, Specificity, Accumulation, Gradient, and Moderation of Contextual Risk