Self-rated child maltreatment, behavioural problems, and contacts with welfare and police authorities – longitudinal community data

ABSTRACT

This study examines how children report abuse, neglect and behavioural problems and what authorities they claim to have had contact with that are legally mandated to report to the Child Welfare Services, e.g. health services and police. It draws data from a longitudinal research programme, LoRDIA, in which four data collections followed adolescents’ development from 12 to 15 years (n = 1884). A total of 61 indicators of self-reported child abuse, neglect and behavioural problems were constructed to identify children with severe exposure so that each indicator would have prompted referral. The main finding is that 445 (25.3%) of the study population reported severe exposure of this magnitude. Among these self-reported severely exposed children (SSE), boys reported higher rates of child neglect and of overall behavioural problems, specifically criminal and other socially destructive behaviour. Poverty and living in single-parent households significantly increase the risk of abuse, neglect, and behavioural problems. Increased risk of neglect and criminal behaviour were found for children studying Swedish as second language. The SSE children more than other children report contact with all authorities with mandatory reporting. Odds ratios of contact were higher in the case of behavioural problems compared to the odds ratios for abuse and neglect.

ABSTRAKT

Denna studie undersöker hur barn rapporterar övergrepp, försummelse och beteendeproblem samt vilka verksamheter de uppger att de har haft kontakt med, vilka enligt lag är skyldiga att anmäla sin oro till den sociala barnavården, t.ex. hälso- och sjukvård samt polis. Studien är baserad på ett longitudinellt forskningsprogram, LoRDIA, som genom fyra datainsamlingar följde ungdomars utveckling från 12 till 15 års ålder (n = 1884). Totalt 61 indikatorer för barns självrapporterade övergrepp, försummelse samt utsatthet genom beteendeproblem konstruerades för att identifiera barn med allvarlig utsatthet i den grad att varje indikator skulle ha lett till en orosanmälan. Det huvudsakliga fyndet är att 445 (25,3%) av studiepopulationen rapporterade allvarlig utsatthet i denna grad. Bland dessa barn med självrapporterad allvarlig utsatthet (SSE) rapporterade fler pojkar än flickor försummelse och beteendeproblem, särskilt kriminellt och annat socialt nedbrytande beteende. Att leva i fattigdom och att bo med en ensamstående förälder ökar avsevärt risken för övergrepp, försummelse och beteendeproblem. Ökad risk för försummelse och kriminellt beteende konstaterades också för barn som studerar svenska som andra språk. Dessa allvarligt utsatta barn uppger, mer än andra barn, kontakt med anmälningsskyldiga myndigheter. Mest kontakt har de barn som uppger allvarlig utsatthet genom beteendeproblem jämfört med dem som uppger allvarlig utsatthet genom övergrepp och försummelse.

KEYWORDS:

• Child maltreatment
• behavioural problems
• self-rated
• severely exposed
• detection

NYCKELORD:

• Bristande omsorg
• beteendeproblem
• självskattning
• allvarligt utsatta
• upptäckt

Introduction

Children’s development can be at risk due to maltreatment, i.e. abuse and neglect, and due to behavioural problems, i.e. substance misuse, criminal, or other socially destructive behaviour. In Sweden, both maltreatment and behavioural problems are handled by the Child Welfare Services (CWS) according the Social Services Act. Authorities and providers of care and support with knowledge about children at risk (CAR) are mandated to refer their concerns to CWS, which is responsible for assessment and providing support and protection if needed. The threshold for referrals should, according to the law, be low, and concerns do not need to be substantiated (NBHW, 2019; Wiklund, 2006). Still, not even severely exposed children are always known to the CWS (Backlund et al., 2012; Stokes & Schmidt, 2011; Svensson & Janson, 2008; Woodruff & Lee, 2011).

This study concerns children who, according to self-ratings, are severely exposed to maltreatment and behavioural problems. These ‘Self-reported Severely Exposed’ (SSE) children are thus a severe subgroup within CAR. Severity concern both frequency and harmfulness of childhood adversities. For example, a child severely exposed to physical abuse can be slapped several times a week or be hit with a tool once, causing injury. In Swedish legislation, these kinds of severe exposure concerning both maltreatment and behavioural problems should unquestionably be addressed and referred to the CWS. Guidelines for referral state that a child may be at risk due to abuse, neglect, and exploitation, or due to own behaviour e.g. substance misuse, criminal behaviour, and other socially destructive behaviour (e.g. aggressive behaviour and truancy) (NBHW, 2014). It is crucial that the children most in need are identified and referred for support and protection (Andershed & Andershed, 2005; Gilbert et al., 2009).

In Sweden, referrals to the CWS on CAR are mandatory for most authorities. The most common referral sources are the police, schools, and healthcare authorities. The annual prevalence of CWS referrals in Sweden is 2–8% (NBHW, 2012, 2019; Sundell et al., 2007). However, most CAR are never referred. Self-ratings and parental reports give a CAR prevalence up to ten times higher than other sources (Gilbert et al., 2009; Östberg, 2010).

Swedish studies estimated prevalence of child maltreatment to 6–36%, physical abuse is most common, then emotional abuse, domestic violence, neglect, and lastly sexual abuse (Annerbäck, 2011; Jernbro et al., 2015; Jernbro & Janson, 2017). To our knowledge, no Swedish adolescence-restricted survey has been conducted on maltreatment. An international review found that maltreatment seem to be equally prevalent in adolescence (Trickett et al., 2011). According to Frenzel and Westerberg (2018), adolescent criminal behaviour in Sweden is estimated to 18–49%, and other socially destructive behaviour to 4–24%. Thor (2019) reports that 21% of adolescents have experienced alcohol-related negative consequences.

Thus, the estimated prevalence of CAR and the number of CWS referrals differ. One often applied explanation is that many professionals are unwilling to refer their concerns due to a perceived risk of harming relationships with the parents, lack of trust in the CWS, or a belief that their services can compensate for maltreatment (Gilbert et al., 2009; Kuruppu et al., 2020; Kvist et al., 2014; Svärd, 2017; Svensson & Janson, 2008).

Some groups are more frequently referred, e.g. children living in (1) single-parent households, or (2) with poor parents, or (3) or receiving monthly social assistance, or (4) with a foreign background (Bywaters et al., 2020; Kvist et al., 2017; Putnam-Hornstein et al., 2013; Sidebotham & Heron, 2006; Stranz & Wiklund, 2015; Svensson et al., 2015).

A growing body of research identified the importance of poly-victimisation, where combinations of different types of maltreatment are more severe than a single type (Aho et al., 2016; Cater et al., 2014; Finkelhor et al., 2007; H. A. Turner et al., 2010). Studies identified a small group of adolescents with multiple and severe behavioural problems (Soydan & Vinnerljung, 2002; R. Turner et al., 2020; Vaughn et al., 2014). Adolescents placed in secure accommodation for behavioural problems have to a large extent been severely exposed to child maltreatment (SiS, 2019; Vinnerljung et al., 2001).

Thus, the total number of CAR in Sweden is not known. It is important that at least the severely exposed children are identified. Professionals often fail to refer CAR, and it is not known if those not referred also include severely exposed children despite contact with authorities obliged to report. Knowledge is needed concerning overlap between different types of self-rated maltreatment and behavioural problems.

Our aim is to estimate the prevalence of SSE children, the types of maltreatment and behavioural problems they were exposed to, and the extent of contact with referral sources mandated to report. The following questions are examined: Which children rate themselves on indicators of severe exposure to abuse, neglect, substance misuse, criminal and socially destructive behaviour? Do such ratings differ based on gender, foreign background, poverty, or living in single-parent households? Are SSE children more likely to be in contact with the most common referral sources of CAR?

Method

The study is part of the interdisciplinary research programme LoRDIA (Longitudinal Research on Development In Adolescence) which, through repeated data collections, tracks adolescents’ development, including substance use, family and peer relations, social adaptation, and mental health. It started in 2013 in four small and medium-sized municipalities (10,000–36,000 inhabitants) in southern Sweden. Two municipalities are characterised as industrial, and two as commuter municipalities. One commuter municipality is located near a large city. The unemployment rate, annual income, educational level, and proportion of first-generation immigrants across the four municipalities were close to the national means (Statistics Sweden, 2019).

All children in two school-year cohorts were invited to participate: those who in autumn 2013 were in grades 6 and 7. This study is based on annual data collections conducted in three or four waves, respectively, up to grade 9. The research programme and each data collection were approved by the Regional Ethics Review Board in Gothenburg (No. 362-13; 2013-09-25; 2014-05-20; 2015-07-31).

Procedure

Letters were sent separately to parents with detailed information about the study as well as their right to decline participation on behalf of their child. For immigrant parents, the information was translated into their 32 respective home languages. The children received a letter in Swedish. If parents did not decide against participation, the children had the right thereafter to decide whether to participate. Out of 2150 invited, 1884 (88%) agreed to participate. Those who declined show no differences in demography (gender and studying Swedish as a second language) or school performance (merit rating and attendance) (Kapetanovic et al., 2017). Questionnaire data were collected annually in class by qualified research staff who were available to explain in case of problems in understanding. This was however less of a problem since the questionnaires were written in easy language. All forms were piloted to ensure that they were comprehensible, and that included scales had acceptable psychometric properties.

Study population

Out of the 1884 children in the study population, 1760 participated at least once (93.4%), with 70–85% participating in each wave. Those who never participated more often studied Swedish as a second language (26.6% vs. 10.9%, p < .001), and more often had a teenage parent (4.0% vs. 1.6%, p < .05). No participants living in poverty or with single parents were excluded. There was no difference between genders.

Measures

Fifteen indicators of maltreatment and 14 indicators of behavioural problems were constructed. Since certain indicators were measured in several waves, a total of 61 dichotomous indicators over time were generated. Indicators of maltreatment and behavioural problems were compiled into five main subtypes (see below). The indicators were designed so that each should have prompted a CWS referral, by using strict cut-offs to distil those severely exposed. Therefore, only one indicator is required for being categorised as SSE. For example, incidental drunkenness and drug use were not considered as indicators, only when, in waves 1–2, these had occurred many times in the past year, and in waves 3–4, when these had occurred every month.

For all scales and variables, cut-off values were chosen as the 99th most severe percentile. This is stricter than in most studies, e.g. in the Childhood Trauma Questionnaire (see below) the cut-off for severe problems mostly used is the 95th percentile (Bernstein & Fink, 2011). The indicators were organised in five subtypes under separate sub-headings. Categorical cut-offs are given in parentheses. More information about the indicators are given on Figshare (https://doi.org/10.6084/m9.figshare.13385303).

Indicators of child abuse

Frequent parental aggression (wave 1). The scale (Tilton-Weaver et al., 2010) consists of four statements for each parent concerning what happens ‘if you do something your mother [father] really dislikes?’, e.g. ‘Gets really angry and has an outburst’, replied using a Likert-scale. Three indicators were created. (1) High mean values between both parents’ angry outbursts. (2) Mother’s extremely angry outbursts. (3) Father’s extremely angry outbursts. Indicators 2 and 3 are independent of the other parent’s reactions. Thus, parental aggression is indicated if the home is characterised by aggression, or if one parent very often has extreme anger outbursts.

Severe emotional abuse (waves 2–4) is a sub-scale of the Childhood Trauma Questionnaire (CTQ; Gerdner & Allgulander, 2009), with items e.g. ‘When I was growing up, people in my family said hurtful or insulting things to me’. All CTQ scales have five items each and answers use five-grade frequency scales.

Severe physical abuse (waves 3–4). Sub-scale in CTQ, e.g. ‘When I was growing up, I was punished with a belt, a board, a cord, or some hard object’.

Severe sexual abuse (waves 3–4). Sub-scale in CTQ, e.g. ‘When I was growing up, someone threatened to hurt me or tell lies about me unless I did something sexual with them’.

Frequently witnessed domestic violence (waves 3–4). One statement, ‘When I was growing up, I witnessed violence between adults in my home’, answered using a frequency scale (Very often true).

Indicators of neglect

Low perceived parental support (wave 1). The scale (Tilton-Weaver, 2014) have five statements, e.g. ‘I know that mum [dad] are there when I need her [him]’. Parental scores were compiled to get an average.

Parental coldness and rejection (wave 1). The scale (Tilton-Weaver et al., 2010) have six statements concerning what happens ‘if you do something your mother [father] really dislikes?’, e.g. ‘Makes you feel guilty for a long time’.

Low parental understanding (wave 1). The scale (Tilton-Weaver et al., 2010) have four statements concerning what happens ‘if you do something your mother [father] really dislikes?’, e.g. ‘Honestly wants to understand why you did what you did’.

Low parental knowledge (waves 1, 3–4). The scale (Kerr et al., 2010) have six statements, e.g. ‘Do your parents know what you do during your free time?’

Severe emotional neglect (waves 2–4). Sub-scale in CTQ, e.g. ‘When I was growing up, there was someone in my family who helped me feel important or special’ (Non-agreement gets high scoring).

Severe physical neglect (waves 3–4). Sub-scale in CTQ, e.g. ‘When I was growing up, my parents were too drunk or high to take care of me’.

Perceived parental acceptance of drunkenness (waves 3–4). One statement, ‘For my parents it’s okay if I get drunk’ (Positive answer).

Perceived parental acceptance of cannabis use (waves 3–4). One statement, ‘For my parents, it’s okay if I smoke hashish/marijuana’ (Positive answer).

Indicators of criminal behaviour

Delinquency (waves 1–4). The scale (Ring, 2013) consists of 9–14 statements, with more severe crimes added in W3–4, e.g. ‘Have you ever carried a knife (as a weapon) or other weapon with you when you went out during the past year?’.

Contact with police due to crime at early age (waves 3–4). One question, ‘Have you ever been arrested by the police for any crime before the age of 12?’ (Positive answer).

CWS contact due to misconduct at early age (waves 3–4). One question, ‘Have you ever had contact with the CWS before the age of 12 for something bad that you did? (e.g. drug misuse, vandalism, fights or illegal acts)’ (Positive answer).

Indicators of substance misuse

Frequent drunkenness (waves 1–4) (Gripe, 2013). One question, ‘Have you ever been drunk at any time in the past year?’. Indicators: several times (wave 1–2); many times each month (waves 3–4).

Frequent sniffing/inhaling (wave 1) (Gripe, 2013). One question, ‘Have you been sniffing/inhaling volatile substances past year?’ (Several times).

Frequent drug use (waves 1–2) (Gripe, 2013). One question: ‘Have you used drugs past year? (e.g. hashish, marijuana, amphetamine, heroin, cocaine, ecstasy, GHB)’ (Several times).

Frequent cannabis use (waves 3–4). One question: ‘Have you used cannabis, marijuana, spice or other cannabis drugs past year?’ (Several times).

Frequent use of other drugs (waves 3–4). One question: ‘Have you used other drugs past year? (e.g. ecstasy, GHB, amphetamine, heroin, cocaine)’ (Several times).

Indicators of socially destructive behaviour

Severe conduct problems (waves 1, 3–4). Sub-scale in SDQ (Malmberg et al., 2003) with five statements, e.g. ‘I get very angry and often lose my temper’.

Frequent truancy (waves 1–4). (Magnusson et al., 1975). One question, ‘Have you been truant from school this term (been away from school for a whole day – unexcused absence)?’ (>10 times).

Frequent bullying (waves 1, 3–4). Scale (Özdemir & Stattin, 2011) consisting of five statements, e.g. ‘Have you (this semester) said nasty things, mocked or teased anyone in an unpleasant way at school?’.

Ethnic harassment as perpetrator (wave 2). From Bayram Özdemir et al. (2016). One question: ‘Have you said nasty things to anyone in school this semester, simply because that person was an immigrant?’ (Several times).

Selling sex (waves 3–4). One question, ‘Have you ever performed sexual services in any form for compensation (money or otherwise)?’ (Several times).

Sexual offending (waves 3–4). One question, ‘Have you forced or coerced anyone into sex in any form?’ (Several times).

Contact with welfare services

Contact with care and protection services were captured by the question ‘Have you, due to any kind of problem, had contact with any of these?’ specifying the following services (waves 3 or 4): Student Health Services, i.e. ‘School counsellor, school nurse or school psychologist’; Youth Guidance Centre; Child and Adolescent Psychiatry; the police as victim, i.e. ‘because of something someone did to you’, or as offender, i.e. ‘because of something you did’. Replies recorded no contact, previous contact, and ongoing contact (Previous and ongoing contact).

Family and child background

Poverty. Household income below 60% of the median income based on tax registry data.

Swedish as a second language (SSL), based on school registry (waves 1–2) was a proxy indicator for integration needs due to poor Swedish skills.

Single-parent household. One question (waves 3–4): ‘Which adults do you live with?’ (‘only dad’ or ‘only mum’).

Clarification

Although the study draws on data from a longitudinal research programme, this is not a longitudinal study. To get as comprehensive an image as possible, everyone who participated at least one wave was therefore included, thus they are not a panel. For the same reason, all available data relevant to the CAR type concepts were used, although measures sometimes differed between waves, and categorisations were deliberately changed to be age-adapted. Data are presented in relation to early and late waves for transparency, for having the best conditions for background group comparisons, and since a lot happens in these years. It should, however, be noted that eventual comparisons between waves – not an aim of the study – must take these differences in operationalisation into consideration.

Test-retest reliability of self-reports

The use of repeated questions in many waves made it possible to conduct test-retests, applied when change cannot explain disagreements between reports. Some variables concerned experiences before the age of 12, i.e. referring to the same time span. A parallel study (Hagborg Melander et al., 2020) reports from one-year test-retest on all CTQ items that systematic correlation, Gamma (γ), indicated that consistency was substantial (Emotional abuse, Emotional neglect, and Physical neglect; mean γ = 0.71–0.75) or almost perfect (Physical abuse and Sexual abuse; mean γ = 0.83 and 0.91). For this study we also tested three singular items which all had almost perfect consistency: Witnessed domestic violence (γ = 0.89); CWS contact due to misconduct (γ = 0.92); and police contact due to crime (γ = 0.94). The questions on substance misuse in this study did not refer to the same time span. However, another study from the same programme on early onset of substance misuse (Ander et al., 2020), reported almost perfect agreement between waves 1 and 2 concerning early onset in alcohol drunkenness (γ = 0.95) and cannabis use (γ = 0.97). Thus, all tested variables on abuse, neglect and behavioural problems showed substantial or almost perfect consistencies.

Replies to the questions on contact with health services and the police were highly consistent. Inconsistent replies (i.e. reporting contact in wave 3 and denying having had the same contact in wave 4) were <2% for Child and Adolescent Psychiatry and the police; 5.1% for Youth Guidance Centre, but higher (10.1%) for Student Health.

Results

Note that all types of exposure in the following concern the most severe cases. Table 1 is a compilation of a comprehensive table of all 61 SSE indicators. Table 1 presents the relative frequencies of children meeting some indicator for each of the five SSE subtypes separately for early and late data collection waves and combined for all waves. Differences based on gender, socioeconomic status, studying SSL, and single-parent households concerning these indicators, were tested with Chi-2.

Table 1. Distribution of SSE.

	Waves	% in waves W1–2 (n = 1700) W3–4 (n = 1440) W1–4 (n = 1760)	Boys (%) (n = 889)	In poverty (%) (n = 79)	Studying SSL (%) (n = 191)	Single-parent household (%) (n = 153)
Abuse	W1–2	7.4	7.3	5.5	5.5	11.7*
	W3–4	3.4	3.7	7.6*	5.1	7.8***
Neglect	W1–2	6.2	6.5	6.8	6.6	7.6
	W3–4	7.6	9.0*	16.5**	12.1*	17.0***
Substance misuse	W1–2	2.0	2.0	2.7	0.0*	2.8
	W3–4	3.5	4.0	7.6*	3.8	3.3
Criminal behaviour	W1–2	1.9	2.5	2.7	0.6	2.1
	W3–4	6.2	9.4***	17.7***	10.2*	13.7***
Socially destructive behaviour	W1–2	4.5	5.4	5.5	5.0	3.4
	W3–4	5.8	7.3*	13.9**	7.0	15.0***
Maltreatment (both types)	W1–4	17.5	18.7	30.4**	18.3	31.4***
Behavioural problems (all three types)	W1–4	13.8	17.4***	26.6***	15.7	26.1***
Total SSE group	W1–4	25.3	28.2**	44.3***	28.3	41.8***

Note: Percentage of adolescents with self-rated indications of exposure in early and late waves, and in relation to gender and background of family and child. Significance (Chi-2): *p < .05, **p < .01, ***p < .001.

Table 1 shows that 17.5% of children reported severe maltreatment and 13.8% severe behavioural problems, while 25.3% of all were SSE in at least one of these aspects. Boys to a greater extent than girls reported that they were exposed to neglect, and rated overall behavioural problems higher, especially criminal behaviour and socially destructive behaviour. Poverty and living in a single-parent household significantly increased frequencies of abuse, neglect, criminal behaviour, and socially destructive behaviour as well as overall SSE. Children studying SSL showed increased risk of neglect and criminal behavioural problems.

Correlations of exposure over time and between types were tested with Phi (Table 2).

Table 2. Associations between different CAR types in early and late waves.

	2.	3.	4.	5.	6.	7.	8.	9.	10.
1. Abuse W1–2	.18**	.19**	.09**	.04	.05*	.13**	.07*	.09**	.06*
2. W3–4		.07*	.22**	.07*	.23**	.11**	.19**	.15**	.22**
3. Neglect W1–2			.10**	.14**	.04	.16**	.13**	.13**	.08*
4. W3–4				.18**	.29**	.19**	.28**	.11**	.22**
5. Substance misuse W1–2					.19**	.44**	.23**	.31**	.19**
6. W3–4						.27**	.28**	.13**	.23**
7. Criminal behaviour W1–2							.32**	.32**	.17**
8. W3–4								.22**	.29**
9. Other socially destructive behaviour W1–2									.13**
10. W3–4									–

Note: Grey boxes show same-type, over time associations W1–2/W3–4. Between-type associations in W1–2 (n = 1700) are shown in italics, and in W3–4 (n = 1440) with bold figures. Significance of Phi: *p < .05, **p < .01, ***p < .001.

The findings separated by type and time show:

• Considerable significant associations over time in same type (grey boxes) with Phi-correlations <0.20, except for – as expected – criminal behaviour, which has an even stronger over time association (0.32).

• Considerable overlap between all different types in waves 3–4 (bold figures) – with all between-type associations >0.20, except for one just below that (abuse/criminal behaviour: 0.19).

• The greatest overlap, however, was recorded in the early waves (W1–2) for the behavioural problems, i.e. criminal behaviour/misuse (0.44), criminal behaviour/other behavioural (0.32), and misuse/other behavioural (0.31). There were more moderate early overlaps between abuse and neglect (0.19) and between neglect and criminal behaviour (0.16). All other early overlaps (W1–2) were lower.

Concerning the association between severe maltreatment (abuse and neglect combined, W1–4) and severe behavioural problems (all types combined, W1–4), Phi was 0.28 (p < .001). Of those 243 self-rating severe behavioural problems, 43.6% also rated themselves as subjects of severe maltreatment, while 34.4% of those 308 self-rating severe maltreatment also rated themselves as having severe behavioural problems. Of all 1760, 6.0% were rated as having severe exposure to both maltreatment and behaviour problems.

Society has far-reaching objectives to detect CAR. A prerequisite for this is that the affected children have contact with authorities and care services that are mandated to refer and therefore not hindered by confidentiality from contacting CWS about concerns. These authorities have staff specially trained to interact with CAR.

Table 3 examines the crude odds ratios (OR) of whether SSE children also have contact with authorities identified as frequent informers to CWS and mandated to refer CAR. Contacts with these authorities are thus handled as dependent variables.

Table 3. Contact with professionals subjected to mandatory reporting in W3–W4 × types of self-rated exposure from reports in W1–W2 (n = 1700) and in W3–W4 (n = 1440).

	Waves	Student Health Services	Youth Guidance Centre	Child and Adolescent Psychiatry	Police as victim	Police as offender
Type of maltreatment
Abuse	W1–2	1.42 (0.95–2.12)	0.86 (0.49–1.52)	2.19 (1.25–3.81)**	1.00 (0.47–2.12)	1.32 (0.56–3.15)
	W3–4	0.89 (0.49–1.60)	1.33 (0.65–2.70)	2.30 (1.09–4.85)*	2.44 (1.11–5.34)*	3.51 (1.52–8.13)**
Neglect	W1–2	1.09 (0.70–1.71)	0.68 (0.34–1.33)	2.18 (1.19–4.00)*	1.47 (0.71–3.02)	2.01 (0.89–4.57)
	W3–4	1.22 (0.83–1.81)	1.88 (1.19–2.96)**	2.60 (1.56–4.34)***	3.29 (1.95–5.55)***	9.16 (5.38–15.62)***
Type of behavioural problem
Substance misuse	W1–2	1.17 (0.51–2.69)	2.28 (0.93–5.60)	0.97 (0.23–4.20)	5.50 (2.21–13.68)***	8.05 (3.06–21.19)***
	W3–4	0.83 (0.46–1.49)	2.22 (1.19–4.12)*	2.86 (1.43–5.71)**	5.61 (2.97–10.61)***	11.30 (5.89–21.67)***
Criminal behaviour	W1–2	0.60 (0.23–1.57)	1.61 (0.58–4.44)	3.29 (1.18–9.13)*	7.83 (3.17–19.35)***	18.13 (7.32–44.85)***
	W3–4	1.33 (0.87–2.05)	2.26 (1.40–3.65)***	2.51 (1.43–4.41)***	7.01 (4.26–11.56)***	12.26 (7.10–21.18)***
Socially destructive behaviour	W1–2	1.43 (0.83–2.47)	2.69 (1.50–4.83)***	3.13 (1.60–6.11)***	2.51 (1.19–5.28)*	4.71 (2.19–10.13)***
	W3–4	0.96 (0.61–1.51)	1.69 (1.00–2.85)*	2.11 (1.15–3.87)*	7.81 (4.71–12.97)***	6.34 (3.48–11.54)***
Total
All SSE	W1–4	1.52*** (1.20–1.92)	1.73 (1.29–2.32)***	2.72*** (1.89–3.91)	4.06*** (2.73–6.03)	6.61*** (3.93–11.11)

Note: Logistic regression: Crude OR (95% CI). Significance *p < .05, **p < .01, ***p < .001.

From Table 3, no significantly elevated OR is found for SSE children to have contact with Student Health compared to other children regardless of the SSE category. This is expected since the Student Health is available to all students regardless of problem. Taken together over type and time, all SSE, however, reported an increase of about 50% in contact with Student Health.

Children exposed to neglect in waves 3–4 reported about 90% more contact with Youth Guidance Centres compared to other children. Moreover, compared to others, almost doubled ORs were found for substance misuse and criminal behaviour in waves 3–4. There was also a doubling for other forms of socially destructive behaviour in waves 1–2 and a 70% increase compared to others in waves 3–4.

The children who faced abuse and neglect – both in waves 1–2 and waves 3–4 – met Child and Adolescent Psychiatry twice as often as others, and two to three times more often than others if they showed any of the three behavioural problems (except for substance misuse in waves 1–2). Overall, all SSE met Child and Adolescent Psychiatry nearly three times more often than other children.

In waves 1–2, they had contact with the police twice as often as victims in case of socially destructive behaviour, five times more often in case of substance misuse and seven times more often in case of criminal behaviour. In waves 3–4, they had contacts with the police two to four times more often as victims if they faced abuse and neglect, and five to eight times more often in case of substance misuse, criminal, and socially destructive behaviour. Overall, all SSE children had contact with the police four times more often as victims.

In waves 1–2, the children had contact with the police four times more as offenders if they had socially destructive behaviour, eight times more often if they misused substances, and 18 times more often if they had criminal behaviour. In waves 3–4, they had contacts with the police three times more often as offenders if they had been abused, nine times more often if they had faced severe neglect, six times more often in case of socially destructive behaviour, and 11–12 times more often in case of substance misuse and criminal behaviour. Overall, all SSE had contacts with the police more than six times more often as offenders.

Discussion

This is – to our knowledge – the first longitudinal Swedish study based on self-reported data from children aged 12 to 16 years addressing the extent of severe child maltreatment and severe behavioural problems in the same study, and also addressing the contacts of affected children with health and protective authorities mandated to refer such problems to the CWS. The indicators used for identifying the most severe cases of problems are stricter than in any other studies we have seen, with cut-off’s corresponding to the 99th most severe percentiles. This operationalisation was chosen, so that there would be no question that the problems should have prompted referrals to CWS for assessment and interventions. Our main finding is that more than one in four children reported exposure of such severity.

Due to differences in design, it is difficult to compare this finding to previous studies. Still, the frequency of severe cases is astonishingly high, even if we consider that maltreatment and behavioural cases are combined. With such strict severity cut-offs, one would have expected lower prevalence of the various types of problems. While some other researchers estimate prevalence of CAR with less strict cut-offs, we estimate it with strict cut-offs. One explanation of the high SSE prevalence is that self-rated survey data tend to capture more CAR than official statistics and assessments by professionals (Gilbert et al., 2009). One explanation might be the number of indicators, and another the format of the data collection which based on concrete questions put in a direct and non-judgemental form made it easier for the children to respond. Yet another explanation could be the timing of reporting. Asking the children repeatedly from an early age in four waves should capture more problems, due to less availability bias and less minimisation. Few previous studies have included children`s experiences before they are 15 years old.

Another finding is that the maltreatment and behaviour problem overlap, resulting in correlations of about 0.20–0.30 between the severest cases of each type of problem, shows that no combinations of the most severe problems are rare. Surprisingly few studies have studied these problems together, despite the often-expected overlap. A fragmented approach to children’s exposure to risk may hamper the identification of multiple exposure, i.e. those at the greatest risk (Aho et al., 2016; Cater et al., 2014; Finkelhor et al., 2007; Vaughn et al., 2014).

The study confirms previous Scandinavian studies (Franzén et al., 2007; Vinnerljung et al., 2008), that male gender, poverty and living in a single-parent household increase the risk of being severely exposed to maltreatment and behavioural problems. In line with a previous study on immigrant children (Hällsten et al., 2013), studying SSL increased the risk of being severely exposed to neglect and criminal behaviour in W3–4, but not in W1–2, and not for other problem types. Even though there are elevated risks for being severely exposed, it is crucial to note that most children in poverty, living in single-parent household or studying SSL are not severely exposed. Note that these factors should not be interpreted as causal since great overlaps are expected between them.

The study also shows that severely exposed children, compared to other children, have more contacts with police, Youth Guidance Centres, and Child and Adolescent Psychiatry, all being mandated to refer to CWS. These findings indicate an increased chance of possible referrals to the CWS. The findings raise important new questions. Which children are referred to the CWS and why? Furthermore, if a great number of severely exposed children are not referred to the CWS, what does that mean for their future development? Further studies are important.

Methodological discussion

Having all children in two-year cohorts from four municipalities with similar socioeconomic factors as the country, with an 88% acceptance of inclusion in the study and the 12% opt-out group showing no differences in demography and school performance, and with a high participation rate (>93%) of the others, provided us with a fairly representative analytical sample of 1760 students as the basis for inference. We note some more attrition among children studying SSL and among children of young parents (<20 years at the time of birth). These differences are not, however, expected to have any significant impact on our findings since 85% of both these groups participated.

The indicators of SSE children are partly based on single-item questions and partly on validated scales concerning different aspects of maltreatment and behavioural problems. To both the single-item questions and the scales, strict cut-offs were applied corresponding to the 99th most severe percentile. Having such strict cut-offs should make it clear that the SSE-group is indeed a severe subgroup within CAR that should have prompted CWS referrals, if known to authorities mandated to report.

A possible limitation is that the study relies on self-reports only. Since no single data collection method can claim to give the full picture, multiple source analysis might have yielded different results in rating of abuse, neglect, and behavioural problems. On one hand, self-reports were found to produce higher prevalence of exposure than reports from CWS (Gilbert et al., 2009). On the other hand, previous studies on childhood maltreatment based on self-reports, suggest that underestimations should be expected (Hardt & Rutter, 2004). Their study, however, concerned self-reports of adults, looking back on their childhood decades later. We argue that self-reports collected in early adolescence should increase the likelihood of correct recollection. Still, the validity of retrospective self-reports from young adolescents on their experiences from early childhood should be discussed, due to the sensitive nature of the questions and the risk of bias. Could the non-anonymous character of reports affect the credibility in such sensitive areas? In this study, however, through using longitudinal data collected in four waves and starting from an earlier age than in previous research, we had the opportunity to address these problems by investigating consistencies over time concerning physical, emotional and sexual abuse, physical and emotional neglect, witnessing domestic violence, CWS contact due to misconduct, and police contact due to crime. The one-year test-retests on all these produced substantial or almost perfect systematic correlations, showing that the adolescents’ reports were consistent over time, which gave support to their credibility. In another study on the same study population (Ander et al., 2020), reported almost perfect systematic correlations concerning onset of drunkenness and cannabis use. Although early onset is not the same as the indicators of frequency in drunkenness and drug use used here, these findings, too, add support to the credibility of reports on substance misuse. We may therefor treat these replies as reliable.

Implications for practice and research

The findings of a strong overlap between maltreatment and behavioural problems, and that many children therefore are subjected to both, show the importance of studying these problems together as CAR indicators. We argue that knowledge of the co-occurrence of exposure is relevant when approaching children, regardless of the causal direction. Our study shows the importance of informing referral sources about indicators of maltreatment and behavioural problems, not only their co-occurrence with sociodemographic factors.

To protect children’s health and development, early detection of CAR is essential. The current study provides information about the extent to which SSE children also report having contact with relevant referral sources to the CWS. On the other hand, it is unknown whether the children receive relevant support and protection, and subsequently whether the identification and referral to the CWS will provide benefits to the children. These are questions for future studies.

Conclusion

This is the first Swedish study with a focus on the severest exposure of children to both maltreatment and behavioural problems. One in four children report experiences of severe exposure to child maltreatment and/or behavioural problems, and 6% report severe exposure to both. There is a large overlap between types of severe exposure. It is possible to identify such children since, to a large extent, they are in contact with common referral sources to the CWS. By addressing indicators of maltreatment and behavioural problems together, it is possible to identify CAR and refer them to CWS to a large extent. We recommend future studies to apply such a double focus, rather than treating maltreatment and behavioural problems as relevant to different categories of children.

Acknowledgements

We are grateful to all students who continued answering extensive and detailed questionnaires for many years, and to collaborating schools.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

A major financial contribution was granted in a combined decision (No. 259-2012-25) from four Swedish research foundations: Swedish Research Council (VR); Swedish Research Council for Health, Working Life and Welfare (FORTE); Sweden’s Innovation Agency (VINNOVA); and The Swedish Research Council Formas.

Notes on contributors

Thorbjörn Ahlgren

Thorbjörn Ahlgren is assistant professor in social work at the Department of Social Work, School of Health and Welfare, Jönköping University. His research focuses on children at risk, their way through the child welfare services, and their entry into adulthood. His research interest also includes non-institutional care for children and supervision of professionals in different fields. He is head of LoRDYA, Longitudinal Research on Development as Young Adults.

Torbjörn Kalin

Torbjörn Kalin is Master in social work and an adjunct teacher at the Department of Social Work, School of Health and Welfare, Jönköping University. He is also a PhD student in Welfare and Social Sciences at the Research School of Health and Welfare, Jönköping University. He has previously worked as a social worker in the Child Welfare Services. He is currently working on a dissertation about children at risk.

Arne Gerdner

Arne Gerdner is senior professor in social work at Jönköping University. His research has – among other topics – dealt with treatment of addictions, the impact of childhood maltreatment in adults, psychiatric comorbidity, social exclusion, ethnicity and integration. He has also studied adolescents and is the head of LoRDIA, Longitudinal Research on Development In Adolescence. He served as scientific advisor to several governmental task forces, as member of the project leadership for the national guidelines on treatment and support to alcohol and drug addicts, and of the scientific advisory council for Swedish Agency for Health Technology Assessment and Assessment of Social Services.