View in EDS HTML Full Text PDF Full Text

Very Early Predictors of Conduct Problems and Crime: Results from a National Cohort Study

Saved in:
Bibliographic Details
Title: Very Early Predictors of Conduct Problems and Crime: Results from a National Cohort Study
Language: English
Authors: Murray, Joseph, Irving, Barrie, Farrington, David P., Colman, Ian, Bloxsom, Claire A. J.
Source: Journal of Child Psychology and Psychiatry. Nov 2010 51(11):1198-1207.
Availability: Wiley-Blackwell. 350 Main Street, Malden, MA 02148. Tel: 800-835-6770; Tel: 781-388-8598; Fax: 781-388-8232; e-mail: cs-journals@wiley.com; Web site: http://www.wiley.com/WileyCDA/
Peer Reviewed: Y
Page Count: 10
Publication Date: 2010
Document Type: Journal Articles
Reports - Research
Descriptors: Behavior Problems, Crime, Criminals, Pregnancy, Risk, Longitudinal Studies, Foreign Countries, Predictor Variables, Surveys, Socioeconomic Background, Prenatal Influences, Mothers, Smoking, Interviews, Birth, Parent Attitudes, Medical Evaluation, Records (Forms), Scores, Antisocial Behavior, Children, Identification
Geographic Terms: United Kingdom
DOI: 10.1111/j.1469-7610.2010.02287.x
ISSN: 0021-9630
Abstract: Background: Longitudinal research has produced a wealth of knowledge about individual, family, and social predictors of crime. However, nearly all studies have started after children are age 5, and little is known about earlier risk factors. Methods: The 1970 British Cohort Study is a prospective population survey of more than 16,000 children born in 1970. Pregnancy, birth, child, parent, and socioeconomic characteristics were measured from medical records, parent interviews, and child assessments at birth and age 5. Conduct problems were reported by parents at age 10, and criminal convictions were self-reported by study members at ages 30-34. Results: Early (up to age 5) psychosocial risk factors were strong predictors of conduct problems and criminal conviction. Among pregnancy and birth measures, only prenatal maternal smoking was strongly predictive. Risk factors were similar for girls and boys. Additive risk scores predicted antisocial behaviour quite strongly. Conclusions: Risk factors from pregnancy to age 5 are quite strong predictors of conduct problems and crime. New risk assessment tools could be developed to identify young children at high risk for later antisocial behaviour.
Abstractor: As Provided
Entry Date: 2010
Accession Number: EJ900714
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwF6-lieZMi6lZCmh-EhjOvwAAAA4TCB3gYJKoZIhvcNAQcGoIHQMIHNAgEAMIHHBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDJSKHQwJvm6HVxCfTAIBEICBmRdxX1kDNgCob0t6O3JzeZyFVaxui_r1w8pSosmxYoWveYi05r-7zX6q6cpuQSvF6DZVAsG7L6ssOkMLnwU73cf-HNQo7o8sOrv4_UfRZphS3hVP5tHJESCepa26XEosVEALUrw75Mh-p07tqySBtaAg5zoR8iSL_KRmnwn27yPX6CSwgWcqn9KoQmViR3W_W2u4sr6zyM3WIA==
Text:
  Availability: 1
  Value: <anid>AN0054300664;jyy01nov.10;2024Jun04.07:35;v2.2.500</anid> <title id="AN0054300664-1">Very early predictors of conduct problems and crime: results from a national cohort study. </title> <p>Background:  Longitudinal research has produced a wealth of knowledge about individual, family, and social predictors of crime. However, nearly all studies have started after children are age 5, and little is known about earlier risk factors. Methods:  The 1970 British Cohort Study is a prospective population survey of more than 16,000 children born in 1970. Pregnancy, birth, child, parent, and socioeconomic characteristics were measured from medical records, parent interviews, and child assessments at birth and age 5. Conduct problems were reported by parents at age 10, and criminal convictions were self‐reported by study members at ages 30–34. Results:  Early (up to age 5) psychosocial risk factors were strong predictors of conduct problems and criminal conviction. Among pregnancy and birth measures, only prenatal maternal smoking was strongly predictive. Risk factors were similar for girls and boys. Additive risk scores predicted antisocial behaviour quite strongly. Conclusions:  Risk factors from pregnancy to age 5 are quite strong predictors of conduct problems and crime. New risk assessment tools could be developed to identify young children at high risk for later antisocial behaviour.</p> <p>Keywords: conduct problems; crime; prediction; birth cohort; longitudinal study; Antisocial behaviour</p> <p></p> <ulist> <item> Abbreviations:</item> <p></p> <item> BCS70 1970 British Cohort Study</item> <p></p> <item> OR odds ratio</item> <p></p> <item> CI 95% confidence interval</item> <p></p> <item> AUC Area under the ROC (Receiver Operating Characteristic) curve</item> </ulist> <p>International longitudinal research shows that the roots of serious antisocial behaviour can be found in childhood ([<reflink idref="bib14" id="ref1">14</reflink>]; [<reflink idref="bib33" id="ref2">33</reflink>]). From middle childhood (ages 6–10 years) onwards, the most important predictors include: child hyperactivity, low IQ, and poor school performance; parental criminality, poor supervision, parent–child separation, and hostile parenting; low family income, large family size, and low parental education; antisocial peers, and high delinquency rate schools and neighbourhoods. However, little is known about earlier (pregnancy to age 5) predictors of conduct problems and crime. While the next stage of research on middle childhood must focus on risk mechanisms, more basic research is needed on predictors in early childhood. Predictors are variables statistically associated with future antisocial behaviour; they do not necessarily cause antisocial behaviour.</p> <p>For two main reasons, researchers and policy makers are increasingly interested in early prediction. First, early onset antisocial behaviour is associated with marked continuity through time, and preventing chronic problems may require very early intervention. Second, recent theories, such as [<reflink idref="bib25" id="ref3">25</reflink>] account of life‐course persistent offending, emphasise the importance of early influences on more serious antisocial development. According to Moffitt, perinatal insults to the brain can cause neuropsychological deficits that increase the risk for early and persistent antisocial behaviour, especially when combined with adverse social environments. In particular, maternal smoking in pregnancy, birth complications, low birthweight, and minor physical anomalies might initiate these processes (see [<reflink idref="bib29" id="ref4">29</reflink>], for a review). Thus, in addition to substantial genetic effects ([<reflink idref="bib31" id="ref5">31</reflink>]), very early environmental influences might predict antisocial behaviour.</p> <p>Out of over 30 major prospective longitudinal studies of offending (for a review, see [<reflink idref="bib14" id="ref6">14</reflink>]) only 10 started before children were age 5 (Table 1). Among these, fewer focused on early influences and examined a wide range of predictors in multivariate models. In the Mater University Study of Pregnancy ([<reflink idref="bib2" id="ref7">2</reflink>]), adolescent antisocial behaviour was predicted quite strongly by parenting and family characteristics and early problem behaviour measured up to age 5, but perinatal factors were not predictive in multivariate models. In the Australian Temperament Project ([<reflink idref="bib35" id="ref8">35</reflink>]), adolescents were classified into groups of low/no‐antisocial behaviour, experimental, and persistent antisocial behaviour. Out of 23 variables measured between birth and age 4, none was significantly associated with this classification. In the Dunedin Longitudinal Study ([<reflink idref="bib26" id="ref9">26</reflink>]; [<reflink idref="bib38" id="ref10">38</reflink>]), adolescent delinquency was predicted with modest accuracy by poor motor coordination, low non‐verbal cognitive ability, being difficult to manage, and externalising and internalising problems, measured up to age 5 (other early predictors were also found in bivariate analyses).</p> <p>1  Major prospective studies of early risk factors for antisocial behaviour</p> <p> <ephtml> <table><thead valign="bottom"><tr><th>Study</th><th>Sample</th><th>Outcome</th><th>Significant risk factors measured between pregnancy and age 5</th></tr></thead><tbody valign="top"><tr><td><xref ref-type="bibr" rid="bibr2">Bor, McGee, & Fagan (2004)</xref> Brisbane, Australia</td><td>5,278</td><td>Antisocial behaviour age 14</td><td>Male sex, aggression, attention problems/restlessness, accidents, low IQ, low birthweight, teenage mother, low family income, single parent, marital changes, marital conflict, parental arrest/imprisonment, heavy maternal smoking and drinking, low maternal affect, doesn't teach baby, allows much freedom, physical punishment</td></tr><tr><td><xref ref-type="bibr" rid="bibr3">Brennan, Grekin, & Mednick (1999)</xref> Copenhagen, Denmark</td><td>3,728 males</td><td>Arrested up to age 34</td><td>Maternal smoking in pregnancy, young mother, low social class. Maternal (licit) drug use in pregnancy and delivery complications (negative associations)</td></tr><tr><td><xref ref-type="bibr" rid="bibr8">Denno (1990)</xref> Philadelphia, United States</td><td>987</td><td>Police contacts up to age 22</td><td>Females: none significant. Males: low maternal education, left‐handedness</td></tr><tr><td valign="bottom"><xref ref-type="bibr" rid="bibr18">Fergusson, Woodward, & Horwood (1998)</xref>; <xref ref-type="bibr" rid="bibr16">Fergusson, Lynskey, & Horwood (1995)</xref> Christchurch, New Zealand</td><td valign="bottom">1,022</td><td>Conduct disorder ages 16–18</td><td>Maternal smoking in pregnancy</td></tr><tr><td>Delinquent ages 14–16</td><td>Adoption, single parent family</td></tr><tr><td><xref ref-type="bibr" rid="bibr20">Kolvin, Miller, Scott, Gatzanis, & Fleeting (1990)</xref> Newcastle, England</td><td>847</td><td>Convicted up to age 33</td><td>Marital disruption, parental illness, poor physical care, social dependency, overcrowding, poor mothering, semi‐skilled father, large family, teenage mother at time of marriage</td></tr><tr><td valign="bottom"><xref ref-type="bibr" rid="bibr26">Moffitt, Caspi, Rutter, & Silva (2001)</xref>; <xref ref-type="bibr" rid="bibr38">White et al. (1990)</xref> Dunedin, New Zealand</td><td valign="bottom">956</td><td>Antisocial behaviour ages 13–18</td><td>Females and males: difficult baby (mother report), undercontrolled temperament, low IQ, young mother, low reading level of mother, mother's neuroticism, deviant mother–child interactions. Males only: low mother IQ</td></tr><tr><td>Delinquency age 15</td><td>Externalising and internalising problems, poor motor coordination, low non‐verbal cognitive ability, difficult to manage</td></tr><tr><td><xref ref-type="bibr" rid="bibr30">Raine, Reynolds, Venables, Mednick, & Farrington (1998)</xref> Mauritius</td><td>1,130</td><td>Aggression age 11</td><td>Stimulation seeking, fearlessness, heavy, tall, high body bulk</td></tr><tr><td><xref ref-type="bibr" rid="bibr35">Vassallo et al. (2002)</xref> Victoria, Australia</td><td>1,063</td><td>Antisocial behaviour ages 13–18</td><td>No significant risk factors identified out of 23 pregnancy/birth, temperament, behaviour, mother–child relationship, and family variables</td></tr><tr><td><xref ref-type="bibr" rid="bibr36">Wadsworth (1979)</xref> Britain</td><td>2,196 males</td><td>Convicted up to age 21</td><td>Mother married under age 20, child late in birth order</td></tr><tr><td><xref ref-type="bibr" rid="bibr37">Werner & Smith (1982)</xref> Kauai</td><td>698</td><td>Delinquent up to age 18</td><td>Females and males: Mother pregnant or new sibling by age 2. Females only: Serious marital discord, mother mental health problems, low IQ, congenital deficit. Males only: low family stability</td></tr></tbody></table> </ephtml> </p> <p>We build on this work on early prediction in the 1970 British Cohort Study (BCS70). Previously in the BCS70, [<reflink idref="bib23" id="ref11">23</reflink>] found that maternal smoking in pregnancy was associated with child conduct problems at ages 5 and 10, especially when mothers smoked over 15 cigarettes per day. Using data on Study males, [<reflink idref="bib15" id="ref12">15</reflink>] found that cognitive functioning at age 5 was predictive of having two or more criminal convictions by age 30. There are many other medical, child, and family data available in the BCS70, which we use in the present study to investigate four questions about early risk factors for antisocial behaviour:</p> <p></p> <p>• 1</p> <p></p> <ulist> <item> Which variables measured up to age 5 predict conduct problems and crime?</item> <p></p> </ulist> <p>• 2</p> <p></p> <ulist> <item> Are early risk factors similar for girls and boys?</item> <p></p> </ulist> <p>• 3</p> <p></p> <ulist> <item> Which early risk factors predict conduct problems and crime independently of other risk factors?</item> <p></p> </ulist> <p>• 4</p> <p></p> <ulist> <item> What is the predictive accuracy of cumulative risk scores?</item> </ulist> <hd id="AN0054300664-2">Methods</hd> <p></p> <hd id="AN0054300664-3">Sample</hd> <p>The BCS70 is an ongoing prospective population survey of 17,287 people born in Britain in a single week in 1970 ([<reflink idref="bib9" id="ref13">9</reflink>]). The study has used medical records and interviews with parents from birth, and interviews with cohort members from age 5. Informed consent was obtained from study participants. For the present analyses, we selected the 16,401 singleton children who were alive at age 5 (521 singletons had died by age 5; 365 multiple births were excluded from analyses to reduce dependency in the data). The ethnic composition of this sample is 95.8% European UK, 1.0% European other, 1.2% West Indian, 1.4% Indian‐Pakistani, and.5% other.</p> <hd id="AN0054300664-4">Measures</hd> <p>We dichotomised several continuous measures of risk factors and outcomes to simplify the presentation of results, facilitate tests of interaction, calculate cumulative risk scores, and because scores were often not normally distributed (see [<reflink idref="bib13" id="ref14">13</reflink>], on the benefits of dichotomisation). Cut‐points were based on prior research and non‐linear relationships observed in histograms, and otherwise dichotomised as close as possible to the 90th percentile.</p> <p> <bold>Outcomes. </bold> Child conduct problems were assessed using the Rutter A2 behaviour rating scale ([<reflink idref="bib34" id="ref15">34</reflink>]), completed by parents when children were age 5 (<emph>N</emph> = 12,446) and 10 (<emph>N</emph> = 12,612) years. Total scores on 31 items (coded no problem, somewhat, and definite) have shown good retest reliability (<emph>r</emph> =.74) and discriminant validity between clinic and non‐clinic children ([<reflink idref="bib34" id="ref16">34</reflink>]: Appendix). In factor analyses of 19 child behaviour items, we identified three main factors: conduct problems, hyperactivity, and internalising problems. At age 5 and at age 10, six items loaded on conduct problems: destroys belongings, fights with other children, takes others' belongings, disobedient, tells lies, and bullies (α =.72 both ages). Conduct problem scores were dichotomised into the top 13% (<emph>n</emph> = 1,606) at age 5, and the top 11% (<emph>n</emph> = 1,334) at age 10, versus the remainders.</p> <p>Criminal convictions were self‐reported at age 30, when participants were asked if they had been found guilty in a criminal court since age 16. If the participant was not interviewed at 30, at age 34 they were asked the same question. Otherwise, participants were asked at age 34 if they had been convicted since the previous interview. To reduce missing data, we used responses at both interviews to make a single combined variable (<emph>N</emph> = 11,169): convicted by age 30 or age 34 (<emph>n</emph> = 1,556). We compared rates of conviction reported in the BCS70 with those found in another British cohort using official records. In the BCS70, 4% of females and 24% of males reported having been convicted between ages 16 and 30–34. In a large cohort (<emph>N</emph> = 63,000) born in England and Wales in 1968, official records show that 5% of females and 30% of males were convicted for a standard list offence up to age 30,1 giving reassurance that the BCS70 data are reasonably accurate. Although data on type of offence are not available in the BCS70, data on the 1968 cohort show that conviction most commonly occurred for the following four types of crime: theft (other than vehicle/shop theft: 47% males, 11% females), violence against the person (41% males, 10% females) criminal damage (37% males, 9% females) and burglary (35% males, 8% females) ([<reflink idref="bib27" id="ref17">27</reflink>]).</p> <p> <bold>Pregnancy‐birth risk factors. </bold> Maternal smoking in pregnancy was assessed by mid‐wives in interviews with mothers soon after birth (<emph>N</emph> = 15,608). High risk was defined as having a mother who smoked 15 or more cigarettes per day during pregnancy (<emph>n</emph> = 2,061).</p> <p>Birth complications (<emph>N</emph> = 15,685), birthweight (<emph>N</emph> = 15,669), and congenital abnormalities (<emph>N</emph> = 15,640) were measured using medical records at birth. Eighteen birth complications were coded 1–5 for severity (see [<reflink idref="bib1" id="ref18">1</reflink>]). Having a severe birth complication (scored 4–5 for events such as eclampsia, prolapsed cord, ruptured uterus, and breech delivery) was defined as high risk (<emph>n</emph> = 512). Low birthweight was defined as weighing under 2,500g ([<reflink idref="bib21" id="ref19">21</reflink>]) (<emph>n</emph> = 865). Having a congenital abnormality referred to having at least one abnormality out of 63 types specified in the records (<emph>n</emph> = 716).</p> <p> <bold>Child risk factors. </bold> Hyperactivity (<emph>N</emph> = 12,419) and internalising problems (<emph>N</emph> = 12,421) were measured on the Rutter A2 scale when children were age 5 (see conduct problems above). In factor analyses, items loading on hyperactivity were: restless, squirmy‐fidgety, and cannot settle (α =.67). Items loading on internalising problems were: worried, solitary, miserable‐tearful, fearful‐afraid of new situations, and fussy‐particular (α =.54). High risk was defined as being in the top 8% (<emph>n</emph> = 1,034) on hyperactivity scores and in the top 10% (<emph>n</emph> = 1,300) on internalising scores.</p> <p>Visual‐motor skills were measured at age 5 (<emph>N</emph> = 12,621). Children made copies of eight designs, which were coded good or poor according to nine criteria. The total score equals the number of good copies. Interrater reliability was high (<emph>r</emph> =.7). Having poor visual‐motor skills was defined as being in the lowest 31% (<emph>n</emph> = 3,914) on test scores.</p> <p>IQ was measured using the English Picture Vocabulary test ([<reflink idref="bib4" id="ref20">4</reflink>]), completed by children at age 5 (<emph>N</emph> = 11,855), with scores ranging between 0 and 56. Interrater reliability in scoring of test booklets was excellent (<emph>r</emph> =.96). Low IQ was defined as being in the lowest 10% (<emph>n</emph> = 1,241) on test scores.</p> <p> <bold>Parent risk factors. </bold> Parental loss (<emph>N</emph> = 12,487) was defined as having a biological parent living in the household at birth but not at age 5 (<emph>n</emph> = 818).</p> <p>Low cognitive stimulation was measured by asking parents who reads to the child most often at age 5 (<emph>N</emph> = 11,624). Book‐reading to children is associated with higher parental education, stronger parental verbal skills, and predicts improved child verbal skills ([<reflink idref="bib28" id="ref21">28</reflink>]). Seven hundred and ninety‐two (<reflink idref="bib792" id="ref22">792</reflink>) parents responded that no‐one reads to the child, which was considered an indicator of low cognitive stimulation.</p> <p>Authoritarian parenting was measured using 13 five‐point Likert items about attitudes to child‐rearing (e.g., 'children should always accept what their parents say as true' and 'a child should not be allowed to talk back') at age 5 (<emph>N</emph> = 11,356). The BCS70 study team derived a weighted authoritarian child‐rearing attitude score based on factor loadings in principal component analyses. High authoritarian parenting was defined as being in the top 10% (<emph>n</emph> = 1,234) on these scores.</p> <p>Maternal depression was measured when children were age 5 (<emph>N</emph> = 12,522) using the Malaise Inventory, a 24‐item self‐report questionnaire ([<reflink idref="bib34" id="ref23">34</reflink>]). The Inventory shows good psychometric properties, and high scores indicate risk for case‐level depression ([<reflink idref="bib32" id="ref24">32</reflink>]). Maternal depression was defined as having a score over seven on the Inventory (<emph>n</emph> = 2,260).</p> <p> <bold>Socioeconomic risk factors. </bold> Teenage mother and single mother at birth were assessed in the birth sweep (<emph>N</emph> = 15,586 and 15,668, respectively). Teenage mother was defined as the mother being under age 20 at delivery (<emph>n</emph> = 1,550). Single mother was defined as being single, widowed, divorced or separated at delivery (<emph>n</emph> = 1,139).</p> <p>Large family size was assessed in interviews with parents at age 5 (<emph>N</emph> = 12,674). A large family was defined as having four or more children living in the household (<emph>n</emph> = 923).</p> <p>Family deprivation was assessed using a composite score of low social class, low parental education, crowded household and poor household (<emph>N</emph> = 12,674). Low social class was defined by the husband's occupation before pregnancy (or the mother's occupation, if the husband's data were missing) being manual class V (on the Registrar General's 1966 Classification of Occupations; [<reflink idref="bib19" id="ref25">19</reflink>]). Low parental education (neither parent having educational qualifications), crowded household (≥ 1.2 persons‐per‐room ratio), and poor household (having fewer than four appliances out of the following six: telephone, fridge, colour TV, washing machine, dryer, and car) were measured at age 5. Overall, family deprivation was defined as having three or more of these indicators of deprivation (<emph>n</emph> = 923).</p> <p>Poor neighbourhood was rated by interviewers at age 5 (<emph>N</emph> = 12,316). Children's neighbourhoods were classified as: poor, average, well‐to‐do, and rural. Three hundred and twenty‐two (<reflink idref="bib322" id="ref26">322</reflink>) neighbourhoods were independently rated by two interviewers. Interrater reliability was good (Kendall's Tau B =.70; <emph>p</emph> <.001). One thousand and six (<reflink idref="bib1" id="ref27">1</reflink>,006) neighbourhoods were classified as poor.</p> <hd id="AN0054300664-5">Statistical methods</hd> <p>To investigate which variables were risk factors for child conduct problems and conviction, odds ratios (OR) and 95% confidence intervals (CI) were computed. We used logistic regression to investigate which variables were independently predictive in multivariate models. The forward stepwise entry method was used to avoid potential problems of multicollinearity. To assess the predictive accuracy of multiple risk scores, we calculated areas under ROC (Receiver Operating Characteristic) curves (AUC). All significance tests were two‐tailed.</p> <hd id="AN0054300664-6">Missing data</hd> <p>Of the 16,401 children targeted for inclusion in the study from age 5, 68% had data on convictions. Conviction data were missing because of death/emigration (<emph>n</emph> = 501), failure to make contact (<emph>n</emph> = 3,793), refusal to participate (<emph>n</emph> = 909) and non‐response on that question (<emph>n</emph> = 8). Missing data were associated with numerous early childhood variables. Recent evidence suggests that predictive models are quite robust to missing data, even when they are common and systematically associated with predictors ([<reflink idref="bib39" id="ref28">39</reflink>]). However, we used multiple imputation to examine possible influences of missing data on study results. Five imputed datasets were created in PASW statistics 18.0.0, using birth to age 5 variables as predictors. Each imputed dataset (<emph>N</emph> = 16,349) had complete data for all variables in the study. Pooled results from imputed datasets were then compared with results from the original data (which we report here).</p> <hd id="AN0054300664-7">Results</hd> <p>There was significant continuity between child conduct problems and criminal conviction in the BCS70. Conduct problems at age 5 strongly predicted conduct problems at age 10 (OR<subs>girls</subs> = 6.6, CI = 5.2‐8.5; OR<subs>boys</subs> = 5.5, CI = 4.6–6.5). Conduct problems at age 5 (OR<subs>girls</subs> = 2.4, CI = 1.6–3.7; OR<subs>boys</subs> = 1.9, CI = 1.6–2.2) and at age 10 (OR<subs>girls</subs> = 2.8, CI = 1.8–4.2; OR<subs>boys</subs> = 2.5, CI = 2.1‐3.0) both strongly predicted having a criminal conviction.</p> <p>Almost all psychosocial risk factors significantly predicted conduct problems at age 10 with large effect sizes (Table 2). According to [<reflink idref="bib5" id="ref29">5</reflink>], an OR of 2.0 or greater indicates a strong relationship. Psychosocial risk factors were much stronger predictors of conduct problems than pregnancy and birth risk factors. Only mother smoked in pregnancy was strongly predictive.2</p> <p>2  Risk factors for conduct problems and conviction</p> <p> <ephtml> <table><thead valign="bottom"><tr><th valign="bottom"><bold> </bold></th><th>Conduct problems age 10</th><th>Conviction ages 16–34</th></tr><tr><th>Girls</th><th>Boys</th><th>Girls</th><th>Boys</th></tr><tr><th>OR (95% CI)</th><th>OR (95% CI)</th><th>OR (95% CI)</th><th>OR (95% CI)</th></tr></thead><tbody valign="top"><tr><td>Pregnancy‐birth</td></tr><tr><td> Mother smoked in pregnancy</td><td>2.4* (1.9–3.0)</td><td>2.2* (1.8–2.7)</td><td>2.0* (1.4–2.9)</td><td>1.6* (1.3–1.9)</td></tr><tr><td> Birth complications</td><td>1.0 (.6–1.7)</td><td>1.1 (.8–1.7)</td><td>1.0 (.5–2.1)</td><td>1.1 (.8–1.6)</td></tr><tr><td> Low birthweight</td><td>1.7* (1.2–2.4)</td><td>1.6* (1.2–2.1)</td><td>1.4 (.9–2.4)</td><td>.8 (.6–1.2)</td></tr><tr><td> Congenital abnormality</td><td>1.0 (.6–1.5)</td><td>1.0 (.7–1.5)</td><td>.5 (.2–1.2)</td><td>.9 (.6–1.2)</td></tr><tr><td>Child</td></tr><tr><td> Hyperactive</td><td>3.7* (2.8–4.9)</td><td>2.8* (2.2–3.4)</td><td>1.3 (.7–2.2)</td><td>1.4* (1.2–1.8)</td></tr><tr><td> Internalising</td><td>1.5* (1.1–2.0)</td><td>1.4* (1.1–1.8)</td><td>.9 (.5–1.5)</td><td>.7* (.6–.9)</td></tr><tr><td> Poor visual‐motor skills</td><td>2.3* (1.8–2.8)</td><td>2.3* (1.9–2.7)</td><td>1.1 (.8–1.5)</td><td>1.7* (1.5–1.9)</td></tr><tr><td> Low IQ</td><td>2.3* (1.7–3.0)</td><td>2.2* (1.7–2.8)</td><td>1.2 (.8–1.9)</td><td>1.5* (1.2–1.9)</td></tr><tr><td>Parent</td></tr><tr><td> Parental loss</td><td>1.9* (1.4–2.7)</td><td>2.0* (1.5–2.6)</td><td>1.6 (.9–2.7)</td><td>1.7* (1.3–2.3)</td></tr><tr><td> Low cognitive stimulation</td><td>2.6* (1.8–3.6)</td><td>3.1* (2.4–4.0)</td><td>1.0 (.5–2.0)</td><td>1.8* (1.4–2.3)</td></tr><tr><td> Authoritarian parenting</td><td>2.0* (1.5–2.7)</td><td>1.8* (1.5–2.3)</td><td>.8 (.4–1.4)</td><td>1.3* (1.0–1.6)</td></tr><tr><td> Maternal depression</td><td>2.8* (2.3–3.5)</td><td>2.5* (2.1–3.0)</td><td>1.3 (.9–1.8)</td><td>1.5* (1.2–1.7)</td></tr><tr><td>Socioeconomic</td></tr><tr><td> Teenage mother</td><td>2.4* (1.8–3.1)</td><td>1.9* (1.5–2.3)</td><td>2.1* (1.4–3.0)</td><td>2.0* (1.6–2.4)</td></tr><tr><td> Single mother at birth</td><td>2.4* (1.8–3.2)</td><td>2.1* (1.6–2.8)</td><td>3.1* (2.0–4.6)</td><td>1.9* (1.4–2.5)</td></tr><tr><td> Large family</td><td>2.6* (2.0–3.2)</td><td>2.0* (1.7–2.5)</td><td>1.2 (.8–1.8)</td><td>1.7* (1.4–2.0)</td></tr><tr><td> Family deprivation</td><td>2.9* (2.1–3.9)</td><td>2.4* (1.9–3.1)</td><td>1.9* (1.2–3.1)</td><td>2.2* (1.7–2.8)</td></tr><tr><td> Poor neighbourhood</td><td>2.3* (1.7–3.2)</td><td>2.0* (1.6–2.6)</td><td>1.9* (1.2–3.0)</td><td>1.4* (1.1–1.8)</td></tr></tbody></table> </ephtml> </p> <p>1 <emph>Note</emph>: OR = odds ratio; CI = confidence interval; *<emph>p</emph> <.05.</p> <p>Table 2 also shows predictors of conviction. All socioeconomic factors were significantly predictive (except large family for girls). The only pregnancy–birth risk factor that was significantly predictive was mother smoked in pregnancy. Early risk factors were generally less predictive of conviction than conduct problems. Average odds ratios (weighted mean ORs across 17 risk factors) for conviction (girls = 1.5, CI = 1.1–2.0; boys = 1.5, CI = 1.3–1.7)3 were significantly (<emph>p</emph> <.01) smaller than average odds ratios for conduct problems (girls = 2.3, CI = 1.9–2.8; boys = 2.1, CI = 1.8–2.5), for both girls and boys.</p> <p>We investigated whether risk factors were different for girls and boys by testing for interactions between each risk factor and child sex in predicting antisocial outcomes. There were no significant interactions for conduct problems (all interaction terms <emph>p</emph> >.05). The only two significant interactions in the prediction of conviction were for poor visual motor coordination (which predicted more strongly for boys) and single mother (which predicted more strongly for girls).</p> <p>To quantify how similarly risk factors affected girls and boys, we calculated correlations between the 17 (natural logarithm) ORs for girls and the 17 (natural logarithm) ORs for boys, for conduct problems (<emph>r</emph> =.94, <emph>p</emph> <.001), and for convictions (<emph>r</emph> =.61, <emph>p</emph> <.01). These large correlations show that early risk factors affect girls and boys very similarly, especially in relation to conduct problems.</p> <p>Several studies have found biosocial interactions in the prediction of antisocial behaviour and violent crime (e.g., [<reflink idref="bib1" id="ref30">1</reflink>]; see also [<reflink idref="bib29" id="ref31">29</reflink>]). We tested whether conduct problems or crime were predicted by interactions between neurological deficits (indicated by birth complications, low birthweight, and low IQ) and environmental adversity (indicated by teenage mother and family deprivation). No interaction was significant. We also explored whether children's own behaviour (hyperactive or internalising) might interact with parenting styles (low cognitive stimulation or authoritarian parenting), in predicting conduct problems and crime. None of these interactions was significant.</p> <hd id="AN0054300664-8">Multivariate results</hd> <p>We investigated which variables significantly predicted antisocial outcomes using multivariate logistic regression. In the first models, all variables, except age 5 conduct problems, were included. In the second models, early conduct problems were also included to maximise predictive efficacy, and to investigate whether other risk factors were predictive independently of early problem behaviour. Variables significant at <emph>p</emph> <.05 were retained in the models.</p> <p>Predictors of conduct problems at age 10 are shown in Table 3. Age 5 conduct problems were the strongest predictor. Psychosocial risk factors dominated over pregnancy and birth risk factors. Table 4 shows the results of the same logistic regression analyses for conviction. Socioeconomic factors and early conduct problems were particularly predictive. For both age 10 conduct problems and later conviction, early risk factors had very similar odds ratios before and after controlling for early conduct problems (compare Models 1 and 2, Tables 3 and 4).4</p> <p>3  Risk factors for conduct problems at age 10 in multivariate logistic regression models</p> <p> <ephtml> <table><thead valign="bottom"><tr><th valign="bottom"><bold> </bold></th><th>Girls</th><th>Boys</th></tr><tr><th>Model 1</th><th>Model 2</th><th>Model 1</th><th>Model 2</th></tr><tr><th>POR (95% CI)</th><th>POR (95% CI)</th><th>POR (95% CI)</th><th>POR (95% CI)</th></tr></thead><tbody valign="top"><tr><td>Early conduct problems</td></tr><tr><td> Conduct problems at age 5</td><td>–</td><td>4.4 (3.3–5.9)</td><td>–</td><td>3.9 (3.2–4.8)</td></tr><tr><td>Pregnancy–birth</td></tr><tr><td> Mother smoked in pregnancy</td><td>2.1 (1.6–2.8)</td><td>1.8 (1.3–2.5)</td><td>1.8 (1.4–2.3)</td><td>1.7 (1.4–2.2)</td></tr><tr><td>Child</td></tr><tr><td> Hyperactive</td><td>2.8 (2.1–3.9)</td><td>2.0 (1.4–2.8)</td><td>2.2 (1.7–2.8)</td><td>1.6 (1.2–2.0)</td></tr><tr><td> Poor visual–motor skills</td><td>1.7 (1.3–2.2)</td><td>1.7 (1.3–2.1)</td><td>1.8 (1.5–2.2)</td><td>1.7 (1.4–2.1)</td></tr><tr><td>Parent</td></tr><tr><td> Parental loss</td><td /><td /><td>1.6 (1.2–2.3)</td><td>1.6 (1.1–2.3)</td></tr><tr><td> Low cognitive stimulation</td><td /><td /><td>2.4 (1.8–3.1)</td><td>2.1 (1.6–2.8)</td></tr><tr><td> Maternal depression</td><td>2.1 (1.6–2.7)</td><td>1.8 (1.3–2.3)</td><td>2.0 (1.6–2.4)</td><td>1.7 (1.3–2.1)</td></tr><tr><td>Socioeconomic</td></tr><tr><td> Teenage mother</td><td>2.3 (1.6–3.2)</td><td>2.2 (1.5–3.1)</td><td>1.4 (1.1–1.9)</td><td> </td></tr><tr><td> Single mother at birth</td><td>1.8 (1.2–2.7)</td><td>1.8 (1.2–2.7)</td><td>1.7 (1.2–2.5)</td><td>2.0 (1.4–2.8)</td></tr><tr><td> Large family</td><td>2.1 (1.6–2.8)</td><td>2.0 (1.5–2.7)</td><td>1.5 (1.2–1.9)</td><td>1.4 (1.1–1.7)</td></tr><tr><td> Poor neighbourhood</td><td>1.7 (1.2–2.3)</td><td>1.6 (1.1–2.3)</td><td> </td><td> </td></tr><tr><td>N in final model</td><td>4,711</td><td>4,676</td><td>4,697</td><td>4,687</td></tr></tbody></table> </ephtml> </p> <ulist> <item>2 <emph>Notes</emph>: POR = partial odds ratio; CI = confidence interval; All variables <emph>p</emph> <.05.</item> <item>4  Risk factors for conviction in multivariate logistic regression models</item> </ulist> <p> <ephtml> <table><thead valign="bottom"><tr><th valign="bottom"> </th><th>Girls</th><th>Boys</th></tr><tr><th>Model 1</th><th>Model 2</th><th>Model 1</th><th>Model 2</th></tr><tr><th>POR (95% CI)</th><th>POR (95% CI)</th><th>POR (95% CI)</th><th>POR (95% CI)</th></tr></thead><tbody valign="top"><tr><td>Early conduct problems</td></tr><tr><td> Conduct problems at age 5</td><td>–</td><td>2.0 (1.3–3.1)</td><td>–</td><td>1.6 (1.3–1.9)</td></tr><tr><td>Pregnancy–birth</td></tr><tr><td> Mother smoked in pregnancy</td><td>2.0 (1.4–2.9)</td><td>1.8 (1.2–2.7)</td><td>1.4 (1.2–1.8)</td><td>1.4 (1.1–1.7)</td></tr><tr><td>Child</td></tr><tr><td> Internalising</td><td /><td /><td>.6 (.5–.9)</td><td>.6 (.5–.8)</td></tr><tr><td> Poor visual–motor skills</td><td /><td /><td>1.6 (1.3–1.8)</td><td>1.5 (1.3–1.8)</td></tr><tr><td>Parent</td></tr><tr><td> Low cognitive stimulation</td><td /><td /><td>1.5 (1.1–2.0)</td><td>1.5 (1.1–1.9)</td></tr><tr><td>Socioeconomic</td></tr><tr><td> Teenage mother</td><td /><td /><td>1.8 (1.4–2.2)</td><td>1.7 (1.3–2.2)</td></tr><tr><td> Single mother at birth</td><td>2.8 (1.8–4.5)</td><td>2.8 (1.8–4.5)</td><td>1.8 (1.3–2.5)</td><td>1.8 (1.3–2.5)</td></tr><tr><td> Large family</td><td /><td /><td>1.5 (1.2–1.9)</td><td>1.5 (1.2–1.9)</td></tr><tr><td> Family deprivation</td><td /><td /><td>1.5 (1.1–2.0)</td><td>1.5 (1.1–2.1)</td></tr><tr><td> Poor neighbourhood</td><td>1.7 (1.1–2.7)</td><td>1.6 (1.0–2.6)</td><td /><td /></tr><tr><td><italic>N</italic> in final model</td><td>4,503</td><td>4,422</td><td>4,002</td><td>3,964</td></tr></tbody></table> </ephtml> </p> <p>3 <emph>Notes:</emph> POR = partial odds ratio; CI = confidence interval; All variables <emph>p</emph> <.05.</p> <hd id="AN0054300664-9">Cumulative risk scores</hd> <p>Finally, we investigated the predictive accuracy of cumulative risk scores based on the most important risk factors. Risk factors were selected if they predicted (<emph>p</emph> <.05) an increase in conduct problems or crime in the multivariate models above (Models 2, Tables 3 and 4). Risk scores were calculated as the number of risk factors (including early conduct problems) a child had for each outcome (prorated if a child was missing data on up to one quarter of the relevant risk factors). Thus, for girls, risk scores could range from 0 to 9 for conduct problems and 0 to 4 for conviction. For boys, risk scores could range from 0 to 9 for conduct problems and 0 to 8 for conviction. Figure 1 shows that, with increasing numbers of early risk factors, more children developed conduct problems: increasing from 2% to 38% for girls and from 5% to 54% for boys. Similarly, Figure 2 shows the increasing chance of conviction with higher risk scores: increasing from 3% to 11% for girls and from 17% to 44% for boys.</p> <p>Graph: 1 Risk scores for conduct problems at age 10</p> <p>Graph: 2 Risk scores for conviction</p> <p>The overall predictive accuracy of the early risk scores was assessed by computing areas under ROC curves (AUC), which can range from.5 (no prediction) to 1.0 (perfect prediction). For conduct problems, predictive accuracy was substantial (AUC<subs>girls</subs> =.76, CI =.73–.78; AUC<subs>boys</subs> =.72, CI =.70–.75). For conviction, predictive accuracy was more modest (AUC<subs>girls</subs> =.62, CI =.58–.67; AUC<subs>boys</subs> =.62, CI =.60–.64). We were interested in how this compared to predictive accuracy for older children. A similar risk score was calculated by [<reflink idref="bib10" id="ref32">10</reflink>] for South London boys aged 8–10 years, in the Cambridge Study in Delinquent Development. For each boy, seven risk factors were summed (troublesomeness, conduct disorder, acting out, low intelligence, poor parenting, social deprivation, and convicted parents). This risk score predicted being convicted by age 32 with an AUC of.71 (CI =.66–.76), indicating better prediction than for younger boys in the BCS70. Nevertheless, prediction in the BCS70 was significantly greater than chance, even for risk factors measured in the first five years of life.</p> <hd id="AN0054300664-10">Results using multiple imputation</hd> <p>We compared the results reported here (in Tables 2, 3 and 4) with results based on imputed data with no missing values. Correlations between 17 pairs of natural logarithm ORs (as in Table 2) from original and imputed data were.98 for conduct problems (for both girls and boys), and.84 and.99 for girls and boys respectively, for conviction. To compare multivariate results (Models 2, Tables 3 and 4) from original and imputed data, all 17 risk factors were forced into logistic regression models, first using original data and then imputed data. Correlations between original and imputed results were.91 and.88 for girls and boys respectively, for conduct problems, and.64 and.90, respectively, for conviction. Thus, results based on original and imputed data were generally extremely similar.</p> <hd id="AN0054300664-11">Discussion</hd> <p>In this prospective population study, many risk factors from pregnancy to age 5 predicted later conduct problems and criminal conviction. The study had several important strengths for investigating early predictors. The BCS70 is a national cohort with a wide range of potential early predictors studied prospectively from birth. Measures were taken from multiple sources and spanned over three decades of life. The sample is large, which allowed for the investigation of rare risk factors such as low birthweight in multivariate models, and antisocial behaviour among girls as well as boys.</p> <p>Before discussing the implications of the findings, it is important to consider the study limitations. Regarding the outcome measure of crime, we were limited to studying retrospective reports of criminal convictions. These are almost certainly less accurate than prospective or official measures, and did not allow us to investigate how results compared for criminal behaviour versus criminal conviction (which involves getting caught, going to trial and being found guilty). Many participants had missing outcome data, which were associated with exposure to early risk factors. Thus the strength of association between predictors and conviction may have been underestimated because of missing data, although some evidence suggests that this may not have impacted strongly on the results. The BCS70 does not have a genetically sensitive design and lacks measures of parent antisocial behaviour. Many associations between risk factors and outcomes may have been driven by these or other, unmeasured confounds. Some of the measures were rather crude (e.g., 'reading to children' and 'parental authoritarian attitude' were used as indicators of parenting behaviours), and this also might mean that risk factor–outcome associations were underestimated. It would be ideal include measures of other possible early risk factors, such as infant attachment patterns, or experiences of abuse or neglect, which were not examined here.</p> <p>Given these limitations, it is impressive that quite strong prediction was observed for conduct problems and crime in this study. In general, psychosocial risk factors were much stronger predictors than perinatal factors. Among perinatal factors, only maternal smoking in pregnancy was a strong predictor. Although this association might reflect effects of toxic nicotine exposure on foetal brain development ([<reflink idref="bib29" id="ref33">29</reflink>]), recent evidence suggests that genetic and other environmental risks might fully account for the association ([<reflink idref="bib7" id="ref34">7</reflink>]; [<reflink idref="bib22" id="ref35">22</reflink>]). In a similar way, all predictors identified in the current study might represent risk markers, rather than causal influences on antisocial behaviour.</p> <p>The findings that hyperactivity and poor visual‐motor skills were strong predictors concur with a wide criminological literature suggesting that neuropsychological deficits might affect antisocial behaviour through difficulties in behaviour control and educational failure ([<reflink idref="bib24" id="ref36">24</reflink>]). Internalising problems were associated with higher rates of childhood conduct problems but slightly lower rates of later criminal conviction. Although this might be a chance finding, it could reflect clustering of child psychopathology syndromes, and inhibition of delinquent behaviour among depressed or anxious people at later ages of criminal responsibility, as some previous work has suggested ([<reflink idref="bib11" id="ref37">11</reflink>]).</p> <p>Among the parent variables, low cognitive stimulation and maternal depression were particularly strong risk factors. These might hamper children's cognitive development and scholastic success, or indicate poor parent–child attachment, or impaired quality of child care, which are implicated in the development of antisocial behaviour ([<reflink idref="bib33" id="ref38">33</reflink>]). Caution is required in interpreting the strong association between maternal depression and child conduct problems, because depressed mothers might over‐report child behaviour problems ([<reflink idref="bib6" id="ref39">6</reflink>]), and strong effects were not found for criminal convictions.</p> <p>There were large associations between indicators of early socioeconomic risk and later antisocial behaviour. A long tradition of theory and research suggests that economic stress might contribute to antisocial behaviour indirectly through parental depression, marital stress, disrupted attachment relations, poor child‐rearing practices, poor school performance, and association with deviant peers ([<reflink idref="bib17" id="ref40">17</reflink>]; [<reflink idref="bib33" id="ref41">33</reflink>]).</p> <p>Despite the much higher prevalence of antisocial behaviour among boys, associations between early risk factors and antisocial outcomes were very similar for girls and boys. This is consistent with the results of the most comprehensive study of sex differences in antisocial behaviour ([<reflink idref="bib26" id="ref42">26</reflink>]), which showed that the higher prevalence of male antisocial behaviour is primarily explained by higher risk exposure among boys, rather than stronger associations between risk factors and behaviour for boys.</p> <p>The predictive accuracy of cumulative risk scores, based on risk factors up to age 5, was substantial for conduct problems and quite impressive for conviction, measured three decades later. Although these findings point towards the possibility of identifying high‐risk children for early prevention programmes, new risk assessment tools would need to be tested in contemporary samples for this purpose. Because of the low prevalence of criminal convictions, prediction almost necessarily involves a high rate of false positives, and any prevention efforts must be extremely careful not to stigmatise young children. Nevertheless, early interventions that focus on early positive outcomes, such as improved cognitive development and parent–child relations, have shown promising long‐lasting effects in reducing later antisocial behaviour ([<reflink idref="bib14" id="ref43">14</reflink>]). Given the continuum of early risk observed in this study, universal prevention programmes that aim to reduce early risk across the population might complement more focused efforts among high‐risk groups.</p> <p>There is limited evidence with which to assess the generalisability of these findings. In contrast to one study concluding no predictive power of early risk factors ([<reflink idref="bib35" id="ref44">35</reflink>]), our findings are commensurate with evidence from Brisbane ([<reflink idref="bib2" id="ref45">2</reflink>]) and Dunedin ([<reflink idref="bib38" id="ref46">38</reflink>]) suggesting moderate to strong prediction from a very early age, and the predominance of early psychosocial variables over perinatal factors. It is possible that similar results would not obtain in other countries such as the United States, where rates of conviction (e.g., for robbery) are much higher. However, in samples studied from mid‐childhood, there is general similarity in risk factors between England and the US ([<reflink idref="bib12" id="ref47">12</reflink>]), and this might also apply to earlier predictors. Some risk factors measured in the current study might not predict so strongly among contemporary British cohorts. For example, having an unmarried mother at birth might carry considerably less risk today, when many more unmarried parents are cohabiting than in 1970. Other measures, such as those of economic disadvantage, would also need updating to develop risk assessment instruments for contemporary children.</p> <p>The study has several implications for future research on early risk factors. First, it is important to distinguish which of the identified risk factors are causal, and which merely mark other risk mechanisms. This requires focused analyses on specific risk factors in genetically sensitive studies (e.g., [<reflink idref="bib7" id="ref48">7</reflink>]; [<reflink idref="bib22" id="ref49">22</reflink>]). Second, when resources permit, more refined measures of early social environment and later criminal outcomes should be used to see whether predictive accuracy improves with better measurement procedures. Third, advances could be made by examining risk factors in early childhood at the same time as in other developmental periods. Studies should simultaneously examine very early risk factors, later childhood experiences, and turning points in adolescence and adult life, to identify the most important moderators and mediators of early risk in determining antisocial outcomes.</p> <hd id="AN0054300664-12">Key points</hd> <p></p> <p>• •</p> <p></p> <ulist> <item> Many risk factors for antisocial behaviour have been identified in mid‐childhood, particularly for boys, but little is known about risk factors from pregnancy to age 5.</item> <p></p> </ulist> <p>• •</p> <p></p> <ulist> <item> Findings from a national population survey show that, from pregnancy to age 5, child, parent, family and socioeconomic factors predict later conduct problems and crime.</item> <p></p> </ulist> <p>• •</p> <p></p> <ulist> <item> Early predictors of antisocial behaviour are similar for girls and boys.</item> <p></p> </ulist> <p>• •</p> <p></p> <ulist> <item> The predictive strength of early risk scores is quite strong.</item> </ulist> <hd id="AN0054300664-13">Acknowledgements</hd> <p>This paper develops a preliminary report written by Barrie Irving and Claire A.J. Bloxsom for the Police Foundation. The authors are grateful to the BCS70 study team, Brian Dodgeon, Hélène O'Connor, Joy Beck, three anonymous reviewers, and the late Neville Butler, founder of the BCS70, for their helpful advice, and Gerry Rose for statistical assistance. We are grateful to John MacLeod (Economic and Resource Analysis Unit, Home Office) for data and advice on convictions in the 1968 cohort. Sources of support: British Academy (JM), Economic and Social Research Council (JM and DPF: grant RES‐000‐22‐2311), Alberta Heritage Foundation for Medical Research as a Population Health Investigator (IC), and the International Centre for Child Studies.</p> <hd id="AN0054300664-14">Correspondence to</hd> <p>Joseph Murray, Institute of Criminology, University of Cambridge, Sidgwick Avenue, Cambridge, CB3 9DA, England; Email: jm335@cam.ac.uk</p> <ref id="AN0054300664-15"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref18" type="bt">1</bibl> <bibtext> These are our calculations based on data from the 1968 cohort study, provided by John MacLeod at the Economic and Resource Analysis Unit, Home Office.</bibtext> </blist> <blist> <bibl id="bib2" idref="ref7" type="bt">2</bibl> <bibtext> It is possible that associations might have been stronger if we had analysed continuous variables. To investigate this, we calculated correlations (<emph>r</emph>) between nine continuous risk measures (mother smoked in pregnancy, birthweight, hyperactivity, internalising, visual‐motor skills, IQ, authoritarian parenting, maternal depression, and maternal age) and a continuous measure of conduct problems at age 10. We then converted these <emph>r</emph> correlations to ORs to see whether the continuous variables produced stronger effect sizes than dichotomous variables. None of the ORs based on continuous variables was larger than corresponding ORs for dichotomous variables. Average (weighted mean) ORs for dichotomous variables (girls = 2.3; boys = 2.1) were larger than average ORs for continuous variables (girls = 1.6; boys = 1.7). Thus, the results presented here, based on dichotomous variables, do not appear to underestimate associations, consistent with the results of Farrington and Loeber (2000).</bibtext> </blist> <blist> <bibl id="bib3" type="bt">3</bibl> <bibtext> To calculate a confidence interval for the weighted mean odds ratio, the smallest standard error of the individual odds ratios was used.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref20" type="bt">4</bibl> <bibtext> To consider whether significant risk factors were missed because variables were dichotomised, we calculated the final models again (Models 2, Tables 3 and 4) using continuous measures of the following variables: conduct problems at age 5, mother smoked in pregnancy, birthweight, hyperactive, internalising, visual‐motor skills, IQ, authoritarian parenting, maternal depression, and mother's age. The only additional significant (<emph>p</emph> <.05) risk factors identified using continuous measures were mother's age (predicting boys' conduct problems at age 10), and hyperactivity (predicting boys' conviction). We conclude that dichotomous and continuous methods produced only slightly different results.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref29" type="bt">5</bibl> <bibtext> Conflict of interest statement: No conflicts declared.</bibtext> </blist> </ref> <ref id="AN0054300664-16"> <title> References </title> <blist> <bibtext> Beck, J.E., & Shaw, D.S. (2005). The influence of perinatal complications and environmental adversity on boys' antisocial behavior. Journal of Child Psychology and Psychiatry, 46, 35 – 46.</bibtext> </blist> <blist> <bibtext> Bor, W., McGee, T.R., & Fagan, A.A. (2004). Early risk factors for adolescent antisocial behaviour: An Australian longitudinal study. Australian and New Zealand Journal of Psychiatry, 38, 365 – 372.</bibtext> </blist> <blist> <bibtext> Brennan, P.A., Grekin, E.R., & Mednick, S.A. (1999). Maternal smoking during pregnancy and adult male criminal outcomes. Archives of General Psychiatry, 56, 215 – 219.</bibtext> </blist> <blist> <bibtext> Brimer, M.A., & Dunn, L.M. (1962). English Picture Vocabulary Test. Bristol : Educational Evaluation Enterprises.</bibtext> </blist> <blist> <bibtext> Cohen, P. (1996). Childhood risks for young adult symptoms of personality disorder: Method and substance. Multivariate Behavioral Research, 31, 121 – 148.</bibtext> </blist> <blist> <bibl id="bib6" idref="ref39" type="bt">6</bibl> <bibtext> Collishaw, S., Goodman, R., Ford, T., Rabe‐Hesketh, S., & Pickles, A. (2009). How far are associations between child, family and community factors and child psychopathology informant‐specific and informant‐general? Journal of Child Psychology and Psychiatry, 50, 571 – 580.</bibtext> </blist> <blist> <bibl id="bib7" idref="ref34" type="bt">7</bibl> <bibtext> D'Onofrio, B.M., Van Hulle, C.A., Waldman, I.D., Rodgers, J.L., Harden, K.P., Rathouz, P.J., et al. (2008). Smoking during pregnancy and offspring externalizing problems: An exploration of genetic and environmental confounds. Development and Psychopathology, 20, 139 – 164.</bibtext> </blist> <blist> <bibl id="bib8" type="bt">8</bibl> <bibtext> Denno, D.W. (1990). Biology and violence: From birth to adulthood. Cambridge : Cambridge University Press.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref13" type="bt">9</bibl> <bibtext> Elliott, J., & Shepherd, P. (2006). Cohort profile: 1970 British Birth Cohort (BCS70). International Journal of Epidemiology, 35, 836 – 843.</bibtext> </blist> <blist> <bibtext> Farrington, D.P. (1989). Long‐term prediction of offending and other life outcomes. In H. Wegener, F. Lösel, & J. Haisch, (Eds), Criminal behavior and the justice system: Psychological perspectives. New York : Springer‐Verlag.</bibtext> </blist> <blist> <bibtext> Farrington, D.P., Gallagher, B., Morley, L., St. Ledger, R.J., & West, D.J. (1988). Are there any successful men from criminogenic backgrounds? Psychiatry, 51, 116 – 130.</bibtext> </blist> <blist> <bibtext> Farrington, D.P., & Loeber, R. (1999). Transatlantic replicability of risk factors in the development of delinquency. In P. Cohen, C. Slomkowski, & L.N. Robins, (Eds), Historical and geographical influences on psychopathology (pp. 299 – 329). Mahwah, NJ : Lawrence Erlbaum.</bibtext> </blist> <blist> <bibtext> Farrington, D.P., & Loeber, R. (2000). Some benefits of dichotomization in psychiatric and criminological research. Criminal Behaviour and Mental Health, 10, 100 – 122.</bibtext> </blist> <blist> <bibtext> Farrington, D.P., & Welsh, B.C. (2007). Saving children from a life of crime: Early risk factors and effective interventions. Oxford : Oxford University Press.</bibtext> </blist> <blist> <bibtext> Feinstein, L., & Bynner, J. (2004). The importance of cognitive development in middle childhood for adulthood socioeconomic status, mental health, and problem behavior. Child Development, 75, 1329 – 1339.</bibtext> </blist> <blist> <bibtext> Fergusson, D.M., Lynskey, M., & Horwood, L.J. (1995). The adolescent outcomes of adoption: A 16‐year longitudinal study. Journal of Child Psychology and Psychiatry, 36, 597 – 615.</bibtext> </blist> <blist> <bibtext> Fergusson, D.M., Swain‐Campbell, N., & Horwood, J. (2004). How does childhood economic disadvantage lead to crime? Journal of Child Psychology and Psychiatry, 45, 956 – 966.</bibtext> </blist> <blist> <bibtext> Fergusson, D.M., Woodward, L.J., & Horwood, L.J. (1998). Maternal smoking during pregnancy and psychiatric adjustment in late adolescence. Archives of General Psychiatry, 55, 721 – 727.</bibtext> </blist> <blist> <bibtext> General Register Office. (1966). Classification of occupations. London : H. M. Stationery Office.</bibtext> </blist> <blist> <bibtext> Kolvin, I., Miller, F.J.W., Scott, D.M., Gatzanis, S.R.M., & Fleeting, M. (1990). Continuities of deprivation? The Newcastle Thousand Family Study. Aldershot, England : Avebury.</bibtext> </blist> <blist> <bibtext> Kramer, M.S. (1987). Determinants of low birth weight: Methodological assessment and meta‐analysis. Bulletin of the World Health Organization (WHO), 65, 663 – 737.</bibtext> </blist> <blist> <bibtext> Maughan, B., Taylor, A., Caspi, A., & Moffitt, T.E. (2004). Prenatal smoking and early childhood conduct problems: Testing genetic and environmental explanations of the association. American Journal of Psychiatry, 61, 836 – 843.</bibtext> </blist> <blist> <bibtext> Maughan, B., Taylor, C., Taylor, A., Butler, N., & Bynner, J. (2001). Pregnancy smoking and childhood conduct problems: A causal association? Journal of Child Psychology and Psychiatry, 42, 1021 – 1028.</bibtext> </blist> <blist> <bibtext> Moffitt, T.E. (1990). The neuropsychology of juvenile delinquency: A critical review. In M. Tonry, & N. Morris (Eds.), Crime and justice: An annual review of research (vol. 12, pp. 99 – 169). Chicago, IL : University of Chicago Press.</bibtext> </blist> <blist> <bibtext> Moffitt, T.E. (1993). Adolescence‐limited and life‐course‐persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674 – 701.</bibtext> </blist> <blist> <bibtext> Moffitt, T.E., Caspi, A., Rutter, M., & Silva, P.A. (2001). Sex differences in antisocial behaviour: Conduct disorder, delinquency, and violence in the Dunedin Longitudinal Study. Cambridge : Cambridge University Press.</bibtext> </blist> <blist> <bibtext> Prime, J., White, S., Liriano, S., & Patel, K. (2001). Criminal careers of those born between 1953 and 1978 (Home Office Statistical Bulletin 4/01). London : Home Office.</bibtext> </blist> <blist> <bibtext> Raikes, H., Alexander Pan, B., Luze, G., Tamis‐LeMonda, C.S., Brooks‐Gunn, J., Constantine, J., et al. (2006). Mother–child bookreading in low‐income families: Correlates and outcomes during the first three years of life. Child Development, 77, 924 – 953.</bibtext> </blist> <blist> <bibtext> Raine, A. (2002). Annotation: The role of prefrontal deficits, low autonomic arousal, and early health factors in the development of antisocial and aggressive behavior in children. Journal of Child Psychology and Psychiatry, 43, 417 – 434.</bibtext> </blist> <blist> <bibtext> Raine, A., Reynolds, C., Venables, P.H., Mednick, S.A., & Farrington, D.P. (1998). Fearlessness, stimulation‐seeking, and large body size at age 3 years as early predispositions to childhood aggression at age 11 years. Archives of General Psychiatry, 55, 745 – 751.</bibtext> </blist> <blist> <bibtext> Rhee, S.H., & Waldman, I.D. (2002). Genetic and environmental influences on antisocial behavior: A meta‐analysis of twin and adoption studies. Psychological Bulletin, 128, 49 – 529.</bibtext> </blist> <blist> <bibtext> Rodgers, B., Pickles, A., Power, C., Collishaw, S., & Maughan, B. (1999). Validity of the Malaise Inventory in general population samples. Social Psychiatry and Psychiatric Epidemiology, 34, 333 – 341.</bibtext> </blist> <blist> <bibtext> Rutter, M., Giller, H., & Hagell, A. (1998). Antisocial behavior by young people. Cambridge : Cambridge University Press.</bibtext> </blist> <blist> <bibtext> Rutter, M., Tizard, J., & Whitmore, K. (Eds.). (1970). Education, health and behaviour. London : Longman.</bibtext> </blist> <blist> <bibtext> Vassallo, S., Smart, D., Sanson, A., Dussuyer, I., McKendry, B., & the Australian Temperament Project team. (2002). Patterns and precursors of adolescent antisocial behaviour. Melbourne, Australia : Crime Prevention Victoria.</bibtext> </blist> <blist> <bibtext> Wadsworth, M. (1979). Roots of delinquency: Infancy, adolescence, and crime. Oxford : Martin Robertson.</bibtext> </blist> <blist> <bibtext> Werner, E.E., & Smith, R.S. (1982). Vulnerable but invincible. New York : McGraw‐Hill.</bibtext> </blist> <blist> <bibtext> White, J.L., Moffitt, T.E., Earls, F., Robins, L., & Silva, P.A. (1990). How early can we tell? Predictors of childhood conduct disorder and adolescent delinquency Criminology, 28, 507 – 533.</bibtext> </blist> <blist> <bibtext> Wolke, D., Waylen, A., Samara, M., Steer, C., Goodman, R., Ford, T., et al. (2009). Selective drop‐out in longitudinal studies and non‐biased prediction of behaviour disorders. The British Journal of Psychiatry, 195, 249 – 256.</bibtext> </blist> </ref> <aug> <p>By Joseph Murray; Barrie Irving; David P. Farrington; Ian Colman and Claire A.J. Bloxsom</p> <p>Reported by Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib14" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib33" firstref="ref2"></nolink> <nolink nlid="nl3" bibid="bib25" firstref="ref3"></nolink> <nolink nlid="nl4" bibid="bib29" firstref="ref4"></nolink> <nolink nlid="nl5" bibid="bib31" firstref="ref5"></nolink> <nolink nlid="nl6" bibid="bib35" firstref="ref8"></nolink> <nolink nlid="nl7" bibid="bib26" firstref="ref9"></nolink> <nolink nlid="nl8" bibid="bib38" firstref="ref10"></nolink> <nolink nlid="nl9" bibid="bib23" firstref="ref11"></nolink> <nolink nlid="nl10" bibid="bib15" firstref="ref12"></nolink> <nolink nlid="nl11" bibid="bib13" firstref="ref14"></nolink> <nolink nlid="nl12" bibid="bib34" firstref="ref15"></nolink> <nolink nlid="nl13" bibid="bib27" firstref="ref17"></nolink> <nolink nlid="nl14" bibid="bib21" firstref="ref19"></nolink> <nolink nlid="nl15" bibid="bib28" firstref="ref21"></nolink> <nolink nlid="nl16" bibid="bib792" firstref="ref22"></nolink> <nolink nlid="nl17" bibid="bib32" firstref="ref24"></nolink> <nolink nlid="nl18" bibid="bib19" firstref="ref25"></nolink> <nolink nlid="nl19" bibid="bib322" firstref="ref26"></nolink> <nolink nlid="nl20" bibid="bib39" firstref="ref28"></nolink> <nolink nlid="nl21" bibid="bib10" firstref="ref32"></nolink> <nolink nlid="nl22" bibid="bib22" firstref="ref35"></nolink> <nolink nlid="nl23" bibid="bib24" firstref="ref36"></nolink> <nolink nlid="nl24" bibid="bib11" firstref="ref37"></nolink> <nolink nlid="nl25" bibid="bib17" firstref="ref40"></nolink> <nolink nlid="nl26" bibid="bib12" firstref="ref47"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ900714
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Very Early Predictors of Conduct Problems and Crime: Results from a National Cohort Study
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Murray%2C+Joseph%22">Murray, Joseph</searchLink><br /><searchLink fieldCode="AR" term="%22Irving%2C+Barrie%22">Irving, Barrie</searchLink><br /><searchLink fieldCode="AR" term="%22Farrington%2C+David+P%2E%22">Farrington, David P.</searchLink><br /><searchLink fieldCode="AR" term="%22Colman%2C+Ian%22">Colman, Ian</searchLink><br /><searchLink fieldCode="AR" term="%22Bloxsom%2C+Claire+A%2E+J%2E%22">Bloxsom, Claire A. J.</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Journal+of+Child+Psychology+and+Psychiatry%22"><i>Journal of Child Psychology and Psychiatry</i></searchLink>. Nov 2010 51(11):1198-1207.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley-Blackwell. 350 Main Street, Malden, MA 02148. Tel: 800-835-6770; Tel: 781-388-8598; Fax: 781-388-8232; e-mail: cs-journals@wiley.com; Web site: http://www.wiley.com/WileyCDA/
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 10
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2010
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Behavior+Problems%22">Behavior Problems</searchLink><br /><searchLink fieldCode="DE" term="%22Crime%22">Crime</searchLink><br /><searchLink fieldCode="DE" term="%22Criminals%22">Criminals</searchLink><br /><searchLink fieldCode="DE" term="%22Pregnancy%22">Pregnancy</searchLink><br /><searchLink fieldCode="DE" term="%22Risk%22">Risk</searchLink><br /><searchLink fieldCode="DE" term="%22Longitudinal+Studies%22">Longitudinal Studies</searchLink><br /><searchLink fieldCode="DE" term="%22Foreign+Countries%22">Foreign Countries</searchLink><br /><searchLink fieldCode="DE" term="%22Predictor+Variables%22">Predictor Variables</searchLink><br /><searchLink fieldCode="DE" term="%22Surveys%22">Surveys</searchLink><br /><searchLink fieldCode="DE" term="%22Socioeconomic+Background%22">Socioeconomic Background</searchLink><br /><searchLink fieldCode="DE" term="%22Prenatal+Influences%22">Prenatal Influences</searchLink><br /><searchLink fieldCode="DE" term="%22Mothers%22">Mothers</searchLink><br /><searchLink fieldCode="DE" term="%22Smoking%22">Smoking</searchLink><br /><searchLink fieldCode="DE" term="%22Interviews%22">Interviews</searchLink><br /><searchLink fieldCode="DE" term="%22Birth%22">Birth</searchLink><br /><searchLink fieldCode="DE" term="%22Parent+Attitudes%22">Parent Attitudes</searchLink><br /><searchLink fieldCode="DE" term="%22Medical+Evaluation%22">Medical Evaluation</searchLink><br /><searchLink fieldCode="DE" term="%22Records+%28Forms%29%22">Records (Forms)</searchLink><br /><searchLink fieldCode="DE" term="%22Scores%22">Scores</searchLink><br /><searchLink fieldCode="DE" term="%22Antisocial+Behavior%22">Antisocial Behavior</searchLink><br /><searchLink fieldCode="DE" term="%22Children%22">Children</searchLink><br /><searchLink fieldCode="DE" term="%22Identification%22">Identification</searchLink>
– Name: Subject
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22United+Kingdom%22">United Kingdom</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1111/j.1469-7610.2010.02287.x
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0021-9630
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Background: Longitudinal research has produced a wealth of knowledge about individual, family, and social predictors of crime. However, nearly all studies have started after children are age 5, and little is known about earlier risk factors. Methods: The 1970 British Cohort Study is a prospective population survey of more than 16,000 children born in 1970. Pregnancy, birth, child, parent, and socioeconomic characteristics were measured from medical records, parent interviews, and child assessments at birth and age 5. Conduct problems were reported by parents at age 10, and criminal convictions were self-reported by study members at ages 30-34. Results: Early (up to age 5) psychosocial risk factors were strong predictors of conduct problems and criminal conviction. Among pregnancy and birth measures, only prenatal maternal smoking was strongly predictive. Risk factors were similar for girls and boys. Additive risk scores predicted antisocial behaviour quite strongly. Conclusions: Risk factors from pregnancy to age 5 are quite strong predictors of conduct problems and crime. New risk assessment tools could be developed to identify young children at high risk for later antisocial behaviour.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2010
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ900714
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ900714
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1111/j.1469-7610.2010.02287.x
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 10
        StartPage: 1198
    Subjects:
      – SubjectFull: Behavior Problems
        Type: general
      – SubjectFull: Crime
        Type: general
      – SubjectFull: Criminals
        Type: general
      – SubjectFull: Pregnancy
        Type: general
      – SubjectFull: Risk
        Type: general
      – SubjectFull: Longitudinal Studies
        Type: general
      – SubjectFull: Foreign Countries
        Type: general
      – SubjectFull: Predictor Variables
        Type: general
      – SubjectFull: Surveys
        Type: general
      – SubjectFull: Socioeconomic Background
        Type: general
      – SubjectFull: Prenatal Influences
        Type: general
      – SubjectFull: Mothers
        Type: general
      – SubjectFull: Smoking
        Type: general
      – SubjectFull: Interviews
        Type: general
      – SubjectFull: Birth
        Type: general
      – SubjectFull: Parent Attitudes
        Type: general
      – SubjectFull: Medical Evaluation
        Type: general
      – SubjectFull: Records (Forms)
        Type: general
      – SubjectFull: Scores
        Type: general
      – SubjectFull: Antisocial Behavior
        Type: general
      – SubjectFull: Children
        Type: general
      – SubjectFull: Identification
        Type: general
      – SubjectFull: United Kingdom
        Type: general
    Titles:
      – TitleFull: Very Early Predictors of Conduct Problems and Crime: Results from a National Cohort Study
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Murray, Joseph
      – PersonEntity:
          Name:
            NameFull: Irving, Barrie
      – PersonEntity:
          Name:
            NameFull: Farrington, David P.
      – PersonEntity:
          Name:
            NameFull: Colman, Ian
      – PersonEntity:
          Name:
            NameFull: Bloxsom, Claire A. J.
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 11
              Type: published
              Y: 2010
          Identifiers:
            – Type: issn-print
              Value: 0021-9630
          Numbering:
            – Type: volume
              Value: 51
            – Type: issue
              Value: 11
          Titles:
            – TitleFull: Journal of Child Psychology and Psychiatry
              Type: main
ResultId 1