Self-Reported and Agency-Notified Child Abuse as Contributors to Suicidal Behaviour in a Population-Based Birth Cohort Study at 30-Year-Follow-Up

Abstract

This was a record-linkage analysis of a birth cohort to examine the association between self-reported self-harm in adulthood and childhood maltreatment (CM) as prospectively notified to authorities and self-reported on the Child Trauma Questionnaire. There were 2507 participants at 30-year follow-up with data on both CM and self-reported self-harm including an intent to die. Of the participants, 304 (12.1%) had self-harmed at some time in their lives while 150 (4.2%) had wanted to die. The prevalence of self- and agency-reported maltreatment was 513 (20.5%) and 143 (5.7%) respectively. On adjusted analyses, CM irrespective of reporting source showed significant associations with both suicidal outcomes. Physical and emotional abuse showed the strongest associations while findings for neglect were mixed. The only association for sexual abuse was for self-reported maltreatment and intent to die but numbers may have been under-powered.

Keywords

Child abuse child maltreatment adverse child experiences suicidal behaviour deliberate self-harm birth cohort

Introduction

Suicidal behaviour can range in severity from suicidal ideation through to actions with a serious intent to die (Molnar et al., 2001). Suicidal behaviour is far more prevalent than death by suicide, although there is a wide variation between countries and specific populations. For instance, in a population-based cross-sectional comparison of adults in five European countries (Norway, Spain, Britain, Finland and Ireland), rates ranged from 2.3% in Spain to 14.6% in Ireland (Casey et al., 2008). Lifetime suicidal ideation in adults as measured by community surveys was just under 10% in the United States (Baca-Garcia et al., 2010) and 13% in Australians aged between 16 and 85 years old (Johnston et al., 2009). More serious suicidal ideation or attempts are much rarer at around 1–3% depending on the country and timeframe (Baca-Garcia et al., 2010; Casey et al., 2008; Ivey-Stephenson et al., 2022; Johnston et al., 2009).

Of concern is the increase in incidence in recent years. For instance, there has been a 50% increase in self-harm in middle-aged men, and a 70% increase in young people in the United Kingdom (Kapur, 2020). In a report published in 2019, the risk of suicide after self-harm in England was 50 times higher in the first year after hospital attendance than in the general population, and nearly 200 times higher in the first month (Geulayov et al., 2019).

One potential risk factor for suicidal behaviour is child maltreatment (CM). Most information concerns sexual abuse where several studies have reported strong associations with subsequent suicide attempts (Angelakis et al., 2019; Ng et al., 2018; Zatti et al., 2017). However, other CM types such as physical and emotional abuse are also associated with both suicide attempts and ideation (Angelakis et al., 2019; Zatti et al., 2017). Evidence for neglect is less clear (Zatti et al., 2017). For instance, in one meta-analysis that included cross-sectional and longitudinal designs (k=68 studies), emotional neglect was associated with subsequent suicidal ideation or attempts (Angelakis et al., 2019), while in another smaller systematic review that was restricted to longitudinal studies, it was not (Zatti et al., 2017). However, this negative finding was based on just two studies. It is therefore unclear whether these disparities by CM type reflect true differences or are due to the variations in the number of relevant studies with some types, such as neglect, receiving less attention.

Following their findings that all types of CM were associated with suicidal ideation and attempts, Angelakis and colleagues (2019) proposed that there was a single mechanism for all forms of maltreatment. One possibility is through the association between CM and underlying psychiatric illness such as depression, anxiety, or post-traumatic stress disorder (Angelakis et al., 2019; Ng et al., 2018; Zatti et al., 2017). Externalising personality traits may also be related to suicide attempts (Brezo et al., 2008). However, this does not appear to be the only explanation as the relationship between CM and suicidal ideation or attempts persists after adjustment for comorbid psychiatric symptoms in most work (Brodsky & Stanley, 2008; Marques-Feixa et al., 2021; Molnar et al., 2001) with the exception of one study (Klonsky & Moyer, 2008).

Two sources of information on CM are retrospective reports of adults and prospectively collected notifications to child protection agencies. These sources appear to identify different populations and both have limitations (Baldwin et al., 2019; Najman et al., 2020). For instance, findings from retrospective studies are limited by the potential for recall bias, the use of clinical rather than population samples, and the possibility that childhood maltreatment might be both a cause and effect (Fergusson et al., 2000; Widom et al., 2004).

Studies using prospectively collected reports to child protection agencies are less common. For instance, in two recent comprehensive systematic reviews (Angelakis et al., 2019; Zatti et al., 2017), only one of the included studies looked at agency reported CM and this did not consider subtypes (Brown et al., 1999). However, although studies of agency reported or substantiated maltreatment may not be subject to recall bias, they may underestimate the true prevalence of CM. For instance, the self-reported prevalence of child sexual abuse in one Australian study was approximately 50% in boys and 30% girls (Dunne et al., 2003), whereas agency reported abuse in a prospective study reported a prevalence of only 2% (Mills et al., 2014). As noted previously, there is also poor agreement between self- and agency-reported CM. (Baldwin et al., 2019; Najman et al., 2020). For instance, a recent meta-analysis found that 52% of individuals with prospectively recorded CM did not retrospectively report it, while 56% of those who retrospectively reported maltreatment did not have prospectively recorded data (Baldwin et al., 2019). Similarly, we found a kappa score of only 0.18 for the agreement between the two reporting sources in a previous paper using data from this cohort (Najman et al., 2020).

After investigation, a smaller number of reports to statutory agencies are substantiated. For instance, Australian-wide data indicate that this occurs in 42% of notifications (Australian Institute of Health and Welfare, 2022). However, these may be the most severe cases and therefore an even greater underestimate than notifications of true CM prevalence.

To our knowledge, there has been no study that looked at suicidal behaviour using different reporting sources in the same population. We therefore used a longitudinal, population-based birth cohort to compare self- and agency-reported and/or substantiated CM of different types in terms of their association with self-harming behaviour by 30 years of age. Self-harm with an intent to die was considered separately. We adjusted for sociodemographic characteristics, as well as internalising or externalising behaviour in the young adult years and concurrent depression or PTSD at 30-year follow up. This is because any association between maltreatment and suicidal behaviour might be influenced by pre-existing mental illness, as well as concurrent depression or PTSD, for which childhood abuse is also a risk factor. We hypothesised all forms of CM would be associated with self-harming behaviour regardless of reporting source.

Method

Data Sources

This longitudinal birth cohort study of self-harming behaviour in 30-year-olds used data from the “Mater-University of Queensland Study of Pregnancy” (MUSP) in Queensland, Australia. (Najman et al., 2005). The University of Queensland Behavioural and Social Sciences Ethical Review Committee approved the study. As described elsewhere, baseline data were collected at the first antenatal visit from 7223 consecutive women who gave birth to live singleton babies at the main obstetrics hospital in Brisbane, Australia (Najman et al., 2005). Both offspring and their mothers were assessed at birth, as well as 6 months, five, fourteen, twenty-one and thirty years afterwards (Najman et al., 2015). We report on a selected sub-sample of 2507 adults with data on suicidal behaviour at 30-year follow-up. There was greater loss to follow-up from baseline in subjects who were male or non-White, as well as those whose parents were not living together at the time of their birth, and/or on a low income (Table 1). Attrition was largely due to the inability to locate some respondents (changes of address and phone number).

Table 1.

Factors significantly associated with loss to follow-up.

	30-year Follow-Up
	No (n = 4707)		Yes (n = 2507)		χ² Test (a)	p-Value
	Count	Column N, %	Count	Column N, %	χ² Test (a)	p-Value
Male gender	2675	56.8	1070	42.7	131.19	<0.0001
Non-white	585	12.4	166	6.6	58.76	<0.0001
Parents not living together at birth	747	15.8	244	9.7	51.57	<0.0001
Low family income prior to birth	1616	37.2	688	28.7	49.43	<0.0001

(a) df = 1.

Agency-Reported Child Maltreatment

Information on CM came from notifications to Queensland’s child protection agency, the Department of Families, Youth and Community Care (DFYCC) up to September 2000, at which time participants were approximately 16 years old. We used anonymised identification number to link these data to the MUSP dataset. (Strathearn et al., 2009). These reports were divided into the following subtypes: physical, sexual or emotional abuse, and/or neglect. If there was a reasonable belief of maltreatment after investigation, they were considered as substantiated. All variables were dichotomous in terms of the presence or absence of maltreatment.

Self-Reported CM Measures

We administered the Child Trauma Questionnaire (CTQ) at the 30-year follow-up as it has reliability, sensitivity and discriminant validity (Bernstein et al., 1997). This has 5 sub-scales with 5 items in each and covered events up 16 years old (Bernstein et al., 2003). We combined the physical and emotional neglect scales to enable comparisons with the agency-notified category of “neglect” in the present study, as well as previous MUSP studies on substance use and psychiatric outcomes (Kisely et al., 2020a, 2020b, 2021). This involved selecting the highest level of maltreatment scored on either the physical or emotional neglect sub-scales Both the Kolmogorov–Smirnov and Shapiro–Wilk tests indicated that none of the scores showed a normal distribution (p < 0.0001). In common with previous work on other outcomes in the same cohort, the continuous scores from the subscales were therefore divided into “none/low” and “moderate/severe” but sensitivity analyses were also performed of the effect of splitting the scores into “severe” versus all other categories (Kisely, Abajobir, et al., 2020; Kisely et al., 2021; Kisely et al., 2021b; Kisely et al., 2022; Najman et al., 2020). In particular, this approach was used in our comparison between self-reported and agency reported maltreatment in the same birth cohort, on which the CTQ designer was a co-author (Najman et al., 2020). However, we also undertook sensitivity analyses of using total continuous CTQ scores to facilitate comparisons with other work (Baker & Maiorino, 2010).

Self-Harm

Participants were asked two questions about self-harm coded as yes or no. The primary outcome was whether they had ever hurt themselves on purpose in any way such as by taking an overdose of pills. If so, they were asked if this had occurred in the previous 12 months. The second question was, if on any of the occasions they had self-harmed, they had ever seriously wanted to kill themselves. These questions are similar to those that have been used in other population surveys such as the National Survey on Drug Use and Health in the United States and the Australian National Survey of Mental Health and Wellbeing (Ivey-Stephenson et al., 2022; Pirkis et al., 2000).

Covariates

We adjusted for several variables given their possible association with either child maltreatment or self-harm. These included gender and parental racial origin, as was well as parental relationship status and family income at the time of study entry. This ensured these variables predated any subsequent maltreatment. To these we added internalising and externalising scores at 21 years old derived from the Young Adult Self Report (YASR) questionnaire. The YASR is widely used in the research of young people with most significant psychological symptoms included in one or the other scale. The internalising scale is the total of the subscales of withdrawal, anxious/depressed and somatic symptoms, while the externalising scale is the aggregation of aggression and delinquency symptoms. Both the youth and young adult forms show good reliability, internal consistency, and predictive validity (Achenbach, 1991; Achenbach et al., 1995). As in other work, we dichotomised the scores into the top 10% and bottom 90% (Murray et al., 2010; Spence et al., 2002). This was because of the non-normal distribution of the scores on the Kolmogorov–Smirnov and Shapiro–Wilk tests (p < 0.0001), as well as to simplify the presentation of results and ensure that differences were clinically significant (Farrington & Loeber, 2000; Spence et al., 2002). We undertook sensitivity analyses of the effect of dividing the scores into the top 20% and bottom 80%.

Finally, we adjusted for the following variables at the 30-year follow-up: the offspring’s employment status, their educational achievement, marital and employment status, as well as depression and post-traumatic stress disorder (PTSD) over the 12 months preceding the interview. The latter two were measured using the computerised version of the WHO Composite International Diagnostic Interview (CIDI), a diagnostic interview schedule based on DSM-IV criteria (Peters et al., 1998).

Statistical Analysis

We used logistic regression analysis to control for the covariates with all variables entered into each model at the same time with listwise deletion of missing data. Maltreatment in general was considered, as well as neglect and the emotional, physical, or sexual abuse categories. Each dichotomised CM category was run as a separate model for each outcome. This was because all the CM types showed highly significant associations on chi-square tests (p < 0.0001) and could not therefore be seen as truly independent variables, a requirement for inclusion in logistic regression (Ranganathan et al., 2017). We also undertook a propensity analysis of adding a variable representing the following baseline variables across the whole cohort at risk of exposure to identify possible effects of differential attrition on the association with outcomes at 30 years old: gender, parental race, family income and parental relationship status. Multiple imputation was not used, as this relies on data being missing at random (Najman et al., 2014), which could not be assumed in this case given a statistically significant result for Little's Missing Completely at Random (MCAR) Test (p < 0.0001) and previous findings from the same cohort (Mills et al., 2016).

Finally, we performed three sensitivity analyses. One was of the effect of dividing the internalising and externalising scores into the top 20% and bottom 80%. The second was of the effect of using continuous total CTQ scores. We used non-parametric statistics for bivariate analyses (the Mann-Whitney U test) and adjusted odds ratios for incremental changes in score as neither require a normal distribution. The final was of restricting analyses of our primary outcome of self-harm to events that had occurred in the previous 12 months. The Statistical Package for Social Sciences (SPSS version 27.0) was used for all analyses.

Results

Of the 2507 participants, 57% were females (n = 1437). A fifth of the cohort reported maltreatment at 30-year follow-up (n = 513), and this was severe in 274. This was most commonly neglect (n = 382), followed by emotional (n = 225), sexual (n = 198) and physical (n = 197) abuse. There was a high overlap of subtypes with 210 participants reporting more than one maltreatment category. One hundred and forty-three young adults (5.7%) had been the subject of a notification to Queensland’s child protection agency. Of the subtypes, physical abuse was the most common (n = 75), followed by neglect (n = 65), emotional (n = 61) and sexual abuse (n = 50). Of these, 84 individuals had been the subject of more than type of maltreatment notification. Eighty-three of the 143 notifications (58%) were substantiated (overall prevalence = 3.3%). In order of frequency this was for emotional (n = 39) or physical abuse (n = 34), followed by neglect (n = 31) and sexual abuse (n = 25). There was more than one substantiation type in 36 participants.

Self-Harm

Three hundred and four participants (12.1%) said they had self-harmed at some time in their lives, sixty (2.4%) in the previous 12 months. One hundred and five participants (4.2%) said they had wanted to die. Table 2 compares the results for our primary outcome by a range of socio-demographic and CM variables adjusted for all the other variables in the table. The covariate results are from the model for agency reported CM. Being of non-White origin was the only baseline variables associated with self-harm. In terms of subsequent variables, internalising and externalising scores at both 14 years and 21 years old showed a significant association with subsequent self-harm. At 30-year follow-up, self-harm was significantly associated with being single, unemployed or on a pension, as well as having depression or PTSD. Covariate results were similar for the agency-substantiated and self-reported CM models.

Table 2.

Factors associated with any self-harm at 30 years old and adjusted for all covariates.

		Any Self-Harm
		No (n = 2203)		Yes (n = 304)		Significance
		N	Column %	N	Column %	OR	95% CI	Adj OR	95% CI	P
Maltreatment Any agency-reported notifications	Yes	113	5.1%	30	9.9%	2.03	1.33–3.09	1.69	1.07–2.69	0.025
Any agency-substantiation	Yes	60	2.7%	23	7.6%	2.92	1.78–4.80	2.45	1.42–4.22	0.001
Any self-reported maltreatment	Yes	399	18.1%	114	37.5%	2.71	2.10–3.50	1.75	1.31–2.33	<0.0001
Covariates Gender	Female	1249	56.7%	188	61.8%	1.24	0.97–1.58	1.13	0.86–1.48	0.370
Racial origin	Non-white	138	6.3%	28	9.2%	1.52	0.99–2.32	1.63	1.04–2.55	0.034
Parental relationship at birth	Not living together	210	9.5%	34	11.2%	1.19	0.81–1.76	1.02	0.67–1.57	0.911
Family income prior to birth	Low income	601	27.3%	87	28.6%	1.07	0.82–1.39	0.89	0.66–1.19	0.418
Internalising at age 21	Yes	138	6.3%	68	22.4%	4.32	3.13–5.94	2.68	1.85–3.89	<0.0001
Externalising at age 21	Yes	135	6.1%	51	16.8%	3.09	2.18–4.37	1.69	1.12–2.54	0.013
Education at age 30	Tertiary level	539	24.5%	65	21.4%	0.94	0.63–1.12	0.85	0.63–1.16	0.317
Relationships at age 30	Single	494	22.4%	111	36.5%	1.99	1.54–2.57	1.44	1.08–1.91	0.013
Employment at age 30	Unemployed/pension	92	4.2%	49	16.1%	4.41	3.04–6.38	2.52	1.64–3.87	<0.0001
Depression at 30	Yes	167	7.6%	83	27.3%	4.58	3.40–6.17	2.96	2.13–4.13	<0.0001
PTSD at 30	Yes	92	4.2%	44	14.5%	3.88	2.65–5.69	2.03	1.31–3.14	0.001

Variables used in adjustment

At baseline: gender; parental race; family income; maternal relationship status

At 21-year follow-up: internalising & externalising scores

At 30-year follow-up: young adult’s employment status; young adult’s education level; young adult’s marital status; CIDI-defined concurrent depression or posttraumatic stress disorder over 12 months

Self- and agency-reported CM, particularly where reports had been substantiated, showed a significant association with self-harm. After adjusting for all the above socio-demographic and clinical factors, we found that young people who had experienced either self- or agency-reported abuse were approximately 70% more likely to report having self-harmed by 30 years of age, both being statistically significant results (Table 2). In the case of substantiated CM this rose to a 200% increase in likelihood (Table 2). There were similar significant results for all forms of self- and agency-reported maltreatment with the sole exceptions of agency-substantiated neglect, and sexual abuse irrespective of reporting source (Table 3). In general, associations were strongest for self-reported CM. Sensitivity analyses of the effect of splitting the self-reported scores into “severe” versus all other categories did not alter the results (Table 3).

Table 3.

Factors associated with any self-harm at 30 years old in separate models for each maltreatment category and adjusted for all covariates.

	Any Self-Harm			Any Intent to Kill Self
	Unadjusted OR	Adjusted OR	p- value	Unadjusted OR	Adjusted OR	p- value
Any notification	2.03 (1.33–3.09)	1.69 (1.07–2.69)	0.025	3.49 (2.02–6.04)	3.07 (1.67–5.92)	<0.0001
Physical abuse	2.99 (1.80–4.97)	2.27 (1.29–3.98)	0.004	5.08 (2.70–9.55)	3.86 (1.89–7.88)	<0.0001
Emotional abuse	3.48 (2.05–5.92)	2.79 (1.54–5.05)	0.001	4.90 (2.49–9.66)	3.81 (1.75–8.26)	0.001
Neglect	2.25 (1.27–3.99)	1.96 (1.05–3.64)	0.033	3.66 (1.76–7.59)	3.48 (1.58–7.67)	0.002
Sexual abuse	1.18 (0.53–2.66)	0.86 (0.34–2.14)	0.734	2.03 (0.71–5.74)	1.38 (0.44–4.635	0.582
Any substantiation	2.92 (1.78, 4.80)	2.45 (1.42–4.22)	0.001	5.20 (2.62–9.60)	4.61 (2.33–9.11)	<0.0001
Physical abuse	4.07 (1.99–8.31)	3.08 (1.41–6.72)	0.005	7.53 (3.32–17.1)	5.75 (2.29–14.2)	<0.0001
Emotional abuse	4.20 2.16–8.18)	3.65 (1.76–7.57)	<0.0001	6.31 (2.82–14.1)	5.49 (2.24–13.4)	<0.0001
Neglect	3.02 (1.38–6.63)	2.35 (0.99–5.57)	0.052	5.76 (2.31–14.4)	4.84 (1.76–13.3)	0.002
Sexual abuse	1.38 (0.47–4.06)	1.06 (0.33–3.46)	0.989	2.01 (0.47–8.63)	1.39 (0.29–6.61)	0.680
Any self-reported	2.05 (1.64–2.56)	1.75 (1.31–2.33)	<0.0001	3.51 (1.21–3.07)	1.94 (1.24–3.04)	0.004
Physical abuse	3.34 (2.30–4.83)	2.18 (1.43–3.32)	0.001	4.53 (2.71–7.57)	2.51 (1.39–4.51)	0.002
Emotional abuse	4.59 (3.29–6.43)	2.71 (1.84–3.98)	<0.0001	6.40 (4.01–10.2)	3.26 (1.87–5.68)	<0.0001
Neglect	3.51 (2.61–4.72)	2.52 (1.81–3.50)	<0.0001	3.57 (2.17–5.63)	2.06 (1.23–3.46)	0.006
Sexual abuse	2.04 (1.37–3.04)	1.31 (0.83–2.07)	0.239	3.22 (1.87–5.56)	2.02 (1.08–3.77)	0.027
Severe maltreatment of any type	3.32 (2.39–4.35)	2.08 (1.34–2.92)	<0.0001	3.54 (2.27–5.51)	1.82 (1.09–3.03)	0.023

Variables used in adjustment.

1. At baseline: gender; parental race; family income; maternal relationship status.

2. At 21-year follow-up: internalising & externalising scores.

3. At 30-year follow-up: young adult’s employment status; young adult’s education level; young adult’s marital status; CIDI-defined concurrent depression or posttraumatic stress disorder over 12 months.

There were similar results for self-harm with the intent to die where all forms of self- and agency-reported maltreatment, including neglect, showed a significant association with the sole exception of agency-reported or substantiated sexual abuse (Table 3). As before, the associations were strongest for self-reported CM, moderate or severe (Table 3).

Propensity and Sensitivity Analyses

Propensity score analyses taking into account baseline covariates across the entire cohort did not alter the significance of the results for the overall categories of self- or agency-reported maltreatment for either of the self-harm outcomes (Table 4). Similarly, sensitivity analyses of the effect of dividing the internalising and externalising scores into the top 20% and bottom 80% made little difference to findings, except that the results were no longer significant for the overall category of agency-notified CM (OR_adj = 1.52; 95% CIs = 0.95–2.44; p = 0.078) as well as the subcategory of neglect (OR_adj = 1.75; 95% CIs = 0.93–3.29; p = 0.081).

Table 4.

Propensity score analyses adjusted for all covariates.

	Any Self-Harm		Any Intent to Kill Self
	Adjusted OR	p- value	Adjusted OR	p- value
Any notification	1.69 (1.06–2.68)	0.026	2.58 (1.41–4.71)	0.002
Any substantiation	2.44 (1.42–4.19)	0.001	3.62 (1.86–7.05)	<0.0001
Any self-reported	1.76 (1.32–2.34)	<0.0001	1.91 (1.23–2.98)	0.004

Variables used in adjustment.

1. At baseline: gender; parental race; family income; maternal relationship status.

2. At 21-year follow-up: internalising & externalising scores.

In terms of continuous CTQ score, the median overall score was 31.00 (range = 25–106). Scores were significantly higher in the presence of self-harm, including with the intent to die (medians of 37 vs. 30; Mann-Whitney U = 144,751.00 p < 0.0001 and 41 vs. 31; Mann-Whitney U = 47,281.00 p < 0.0001; respectively). This was reflected in the respective adjusted results (OR_adj = 1.04; 95% CIs = 1.03–1.05; p < 0.0001 and OR_adj = 1.04; 95% CIs = 1.02–1.06; p < 0.0001). Finally, there were strong associations between self-harm behaviour in the previous 12 months and most forms of CM in spite of small numbers (Table 5). However, only agency-reported emotional abuse remained significant on adjusted analyses (OR_adj = 2.95; 95% CIs = 1.05–8.32; p = 0.041).

Table 5.

Factors associated with any self-harm in previous 12 months at 30 years old.

	Any Self-Harm in Previous 12 months
	No (n = 2447)		Yes (n = 60)
	N	Column %	N	Column, %	Unadjusted OR
Any notification	136	5.6%	7	11.7	2.24 (1.05–5.03)
Physical abuse	73	3.0%	5	8.3	2.96 (1.15–7.60)
Emotional abuse	61	2.5%	6	10.0	4.35 (1.18–10.5)
Neglect	67	2.7%	2	3.3	1.23 (0.29–5.12)
Sexual abuse	46	1.9%	4	6.7	3.73 (1.30–10.7)
Any substantiation	78	3.2%	5	8.3	2.76 (1.08–7.10)
Physical abuse	32	1.3%	2	3.3	2.60 (0.61–11.1)
Emotional abuse	36	1.5%	3	5.0	3.52 (1.05–11.8)
Neglect	30	1.2%	1	1.7	1.36 (0.18–10.2)
Sexual abuse	23	0.9%	2	3.3	3.63 (0.84–15.8)
Any self-reported	490	20.0%	23	38.3	2.48 (1.46–4.22)
Physical abuse	153	7.4%	10	20.4	3.21 (1.57–6.55)
Emotional abuse	181	8.7%	12	24.5	3.39 (1.74–6.62)
Neglect	303	14.6%	14	28.6	2.34 (1.24–4.40)
Sexual abuse	162	7.8%	10	20.4	3.01 (1.48–6.15)

Discussion

Most of the information on the association on childhood maltreatment and suicidal or self-harming behaviour comes from cross-sectional or longitudinal studies of adults who retrospectively report childhood abuse (Angelakis et al., 2019; Ng et al., 2018; Zatti et al., 2017). For instance, two recent systematic reviews of the subject found only one study that looked at agency reported CM (Angelakis et al., 2019; Zatti et al., 2017), and this did not consider the subtypes (Brown et al., 1999). To our knowledge, this is therefore the first paper to consider and compare the long-term effects of both self- and agency-reported CM of all types on self-harming behaviour in adulthood in the same population-based cohort.

In terms of prevalence, 12% of the sample reported that they had self-harmed up to the age of 30 years old, of whom 105 (4.2%) said they had wanted to die. This is consistent with lifetime rates of suicidal ideation/behaviour in United States but at the higher end of cross-sectional prevalence as reported in Europe (Baca-Garcia et al., 2010; Casey et al., 2008). The presence of maltreatment irrespective of reporting source showed a significant association with both measures of self-harming behaviour. In terms of the subtypes, physical and emotional abuse, whether self- or agency reported/substantiated, were also significantly associated with self-harm of both forms. Neglect, irrespective of reporting source, was associated with an intent to die. By contrast, associations for agency reported sexual abuse were non-significant, while self-reported sexual abuse was only associated with intent to die.

Importantly, we considered both self- and agency-reported CM. This is because prospective and retrospective measures of childhood identify different groups of individuals with comparatively little overlap between the two groups (Baldwin et al., 2019; Najman et al., 2020). These differences cannot be solely attributed to recall bias as there is evidence that people are generally accurate in their reports of childhood maltreatment experiences (Pinto et al., 2014). However, discrepancies can occur in one of two directions (Baldwin et al., 2019). Over-reporting may occur through suggestibility or negative bias as a result of depression at the time of assessment. Under-reporting may be due to deficits in memory coding, consolidation or retrieval because of impaired brain function or the absence of accompanying emotional distress. Despite these disadvantages, it is also possible that self-reported data may be more comprehensive and accurate than agency-reported, which may focus on the most extreme or severe event with physical evidence to support the claim. For example, agency reports may only be submitted when there is physical risk to the child, thereby missing experiences of abuse and neglect that are less easily recognisable by others (Baldwin et al., 2019).

The above concerns especially apply to the smaller number of substantiated cases, which accounted for 58% of notifications, a similar percentage to that reported in South Australia during the same reporting period (Winefield & Bradley, 1992). There may be several reasons for the relatively high level of unsubstantiated notifications. It might be that the criteria for substantiation are too stringent or there are insufficient resources to fully investigate reports (Winefield & Bradley, 1992). Conversely, some of the notifications may be inaccurate (Winefield & Bradley, 1992). Some work suggests that the caseworker's estimate of severity is an independent predictor of substantiation (Winefield & Bradley, 1992).The latter finding would be consistent with our finding that substantiated cases had worse outcomes than those that were only notified. By contrast, other authors have reported that the distinction between notification and substantiation does not greatly affect outcome, suggesting that notified cases can be used as an overall indicator of CM (Drake, 1996; Hussey et al., 2005; Kohl et al., 2009). We have chosen to present findings for using substantiation, both as a marker of severity, and to allow comparison with work using either substantiated or notified CM.

Our finding that self-harming behaviour was consistently associated with emotional and physical abuse, irrespective of source, is consistent with that from studies restricted to self-reported cases (Angelakis et al., 2019; Zatti et al., 2017). The more equivocal findings for neglect also mirror findings in self-reported maltreatment (Zatti et al., 2017). However, unlike other papers and contrary to expectations, sexual maltreatment showed the weakest association with the only significant result being for self-reported abuse and intent to die. One explanation for the non-significant findings for agency reported sexual abuse was that our study was underpowered to detect statistical significance. This is further discussed under limitations. However, this would not explain similar findings for self-reported sexual abuse where numbers were higher.

One strength of this paper is adjustment for concurrent PTSD and depression given these are associated with both childhood maltreatment and self-harming behaviour. We also adjusted for internalising and externalising scores at 21 years old. Adding these variables into our models did not abolish the association between CM and self-harming behaviour suggesting that these are unlikely to be the only explanation for our findings. One suggestion is that depression, anxiety, PTSD and suicidality show similar and independent associations with perceptions of defeat and entrapment (Siddaway et al., 2015).

Limitations

There are several limitations to the study. This is a 30-year follow-up and so attrition, which was associated with markers of social deprivation, may limit the study’s generalisability (Kisely, Strathearn, et al., 2020). Although we were able to test for MCAR using Little's test, we cannot exclude the possibility that data were missing at random (MAR). However, recent findings from this cohort suggest that the rate of loss to follow-up, for both dependent and independent variables, has little effect on estimates of association and so is less important as a source of bias (Saiepour et al., 2019). For instance, multiple imputation of data from this cohort changes neither the estimates nor their precision (Najman et al., 2014; Saiepour et al., 2019), and this is consistent with findings of the effects of differential attrition in other longitudinal studies (Wolke et al., 2009). In addition, our findings were generally the same on propensity score analyses that considered baseline covariates across the entire at-risk cohort.

Another limitation is that the relatively few agency-notified or substantiated cases, particularly those of sexual abuse, are probably an under-estimate of the true prevalence and therefore of insufficient power to detect statistical significance for some of the outcomes. All the CM categories showed highly significant associations with each other and could not therefore be seen as truly independent variables, a requirement for inclusion in a single logistic regression model. As a result, we had to run each dichotomised CM category in separate models and so were unable to account for any overlap between the CM types. In addition, we were unable to explore the effect of the age when the CM occurred. Notified or substantiated cases may also reflect the reporting practices of 15–20 years ago.

The self-harm outcomes were self-reported and not confirmed through other sources. We therefore have no indication of the severity. Neither did we consider death by suicide. Nevertheless, suicidal or self-harming behaviour is often a precursor to non-fatal repetition or death by suicide. For instance, about 1.6% of those seen in hospital after self-harm die by suicide in the following year (House & Owens, 2020). Conversely a quarter of all suicides in the UK have attended an acute hospital after self-harm in the previous 12 months (House & Owens, 2020). Use of lifetime self-harm as the primary outcome meant that it was impossible to determine whether this preceded or followed self- and agency-reported CM. There were also insufficient numbers to comprehensively look at people who said they had self-harmed in the previous 12 months.

Implications

At the individual level, clinicians should be aware of the increased possibility of a past history of CM in people presenting with self harming behaviour and be comfortable with screening for this possibility. Similarly, those who have been subjected to CM should be asked about previous or current ideas or acts of self-harm. At a population level, interventions to prevent maltreatment may help to reduce suicide rates. Examples include home visiting by nurses particularly for those families at greatest risk (Olds et al., 1997). This can reduce both CM and its mental health consequences such as depression in adolescence and adulthood.

Further research is indicated to compare the effects of agency- and self-reported maltreatment to better reflect the more recent policies for reporting and investigating CM given the present dataset reflects the practices of 15–20 years ago. Other areas of possible investigation include the effect of when the maltreatment occurred and the characteristics of the perpetrator, as well as the impact on death by suicide.

In conclusion, these findings suggest that suicidal behaviour may be a further adverse consequence, amongst many, of childhood maltreatment, so highlighting the public health importance of the issue. In particular, these results emphasise the importance of physical and emotional maltreatment in terms of suicidal ideation and behaviour. This suggests that reducing CM may have a role in the primary prevention of deliberate self-harm and suicide. In addition, people presenting with self-harming behaviour should be screened for the possibility of CM.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Steve Kisely DMedRes

References

Achenbach

T. M.

(1991). Manual for the youth self-report and 1991 profile. Department of Psychiatry, University of Vermont.

Achenbach

T. M.

Howell

C. T.

McConaughy

S. H.

Stanger

(1995). Six-year predictors of problems in a national sample: III. Transitions to young adult syndromes. J Am Acad Child Adolesc Psychiatry, 34(5), 658–669. https://doi.org/10.1097/00004583-199505000-00018

Angelakis

Gillespie

E. L.

Panagioti

(2019). Childhood maltreatment and adult suicidality: A comprehensive systematic review with meta-analysis. Psychological Medicine, 49(7), 1057–1078. https://doi.org/10.1017/s0033291718003823

Australian Institute of Health and Welfare (2022) Child protection Australia 2020–21, AIHW, Australian Government, accessed 15 June 2022.

Baca-Garcia

Perez-Rodriguez

M. M.

Keyes

K. M.

Oquendo

M. A.

Hasin

D. S.

Grant

B. F.

Blanco

(2010). Suicidal ideation and suicide attempts in the United States: 1991–1992 and 2001–2002. Molecular Psychiatry, 15(3), 250–259. https://doi.org/10.1038/mp.2008.98

Baker

A. J.

Maiorino

(2010). Assessments of emotional abuse and neglect with the CTQ: Issues and estimates. Children and Youth Services Review, 32(5), 740–748. https://doi.org/10.1016/j.childyouth.2010.01.011

Baldwin

J. R.

Reuben

Newbury

J. B.

Danese

(2019). Agreement between prospective and retrospective measures of childhood maltreatment: A systematic review and meta-analysis. JAMA Psychiatry, 76(6), 584–593. https://doi.org/10.1001/jamapsychiatry.2019.0097

Bernstein

D. P.

Ahluvalia

Pogge

Handelsman

(1997). Validity of the childhood trauma Questionnaire in an adolescent psychiatric population. Journal of the American Academy of Child and Adolescent Psychiatry, 36(3), 340–348. https://doi.org/10.1097/00004583-199703000-00012

Bernstein

D. P.

Stein

J. A.

Newcomb

M. D.

Walker

Pogge

Ahluvalia

Stokes

Handelsman

Medrano

Desmond

Zule

(2003). Development and validation of a brief screening version of the Childhood Trauma Questionnaire. Child Abuse and Neglect, 27(2), 169–190. https://doi.org/10.1016/s0145-2134(02)00541-0

10.

Brezo

Paris

Vitaro

Hebert

Tremblay

R. E.

Turecki

(2008). Predicting suicide attempts in young adults with histories of childhood abuse. The British Journal of Psychiatry, 193(2), 134–139. https://doi.org/10.1192/bjp.bp.107.037994

11.

Brodsky

B. S.

Stanley

(2008). Adverse childhood experiences and suicidal behavior. Psychiatric Clinics of North America, 31(2), 223–235. https://doi.org/10.1016/j.psc.2008.02.002

12.

Brown

Cohen

Johnson

J. G.

Smailes

E. M.

(1999). Childhood abuse and neglect: Specificity of effects on adolescent and young adult depression and suicidality. Journal of the American Academy of Child and Adolescent Psychiatry, 38(12), 1490–1496. https://doi.org/10.1097/00004583-199912000-00009

13.

Casey

Dunn

Kelly

B. D.

Lehtinen

Dalgard

O. S.

Dowrick

Ayuso-Mateos

J. L.

(2008). The prevalence of suicidal ideation in the general population: Results from the outcome of depression international network (ODIN) study. Social Psychiatry and Psychiatric Epidemiology, 43(4), 299–304. https://doi.org/10.1007/s00127-008-0313-5

14.

Drake

(1996). Unraveling “unsubstantiated”. Child Maltreatment, 1(3), 261–271. https://doi.org/10.1177/1077559596001003008

15.

Dunne

M. P.

Purdie

D. M.

Cook

M. D.

Boyle

F. M.

Najman

J. M.

(2003). Is child sexual abuse declining? Evidence from a population-based survey of men and women in Australia. Child Abuse and Neglect, 27(2), 141–152. https://doi.org/10.1016/s0145-2134(02)00539-2

16.

Farrington

D. P.

Loeber

(2000). Some benefits of dichotomization in psychiatric and criminological research. Criminal Behaviour and Mental Health, 10(2), 100–122. https://doi.org/10.1002/cbm.349

17.

Fergusson

D. M.

Horwood

L. J.

Woodward

L. J.

(2000). The stability of child abuse reports: A longitudinal study of the reporting behaviour of young adults. Psychological Medicine, 30(03), 529–544. https://doi.org/10.1017/s0033291799002111

18.

Geulayov

Casey

Bale

Brand

Clements

Farooq

Kapur

Ness

Waters

Tsiachristas

Hawton

(2019). Suicide following presentation to hospital for non-fatal self-harm in the multicentre study of self-harm: A long-term follow-up study. Lancet Psychiatry, 6(12), 1021–1030. https://doi.org/10.1016/s2215-0366(19)30402-x

19.

House

Owens

(2020). General hospital services in the UK for adults presenting after self-harm: Little evidence of progress in the past 25 years. British Journal of Psychiatry, 217(6), 661–662. https://doi.org/10.1192/bjp.2020.85

20.

Hussey

J. M.

Marshall

J. M.

English

D. J.

Knight

E. D.

Lau

A. S.

Dubowitz

Kotch

J. B.

(2005). Defining maltreatment according to substantiation: Distinction without a difference? Child Abuse and Neglect, 29(5), 479–492. https://doi.org/10.1016/j.chiabu.2003.12.005

21.

Ivey-Stephenson

A. Z.

Crosby

A. E.

Hoenig

J. M.

Gyawali

Park-Lee

Hedden

S. L.

(2022). Suicidal thoughts and behaviors among adults Aged≥ 18 Years—United States, 2015–2019. MMWR Surveillance Summaries, 71(1), 1–19. https://doi.org/10.15585/mmwr.ss7101a1

22.

Johnston

A. K.

Pirkis

J. E.

Burgess

P. M.

(2009). Suicidal thoughts and behaviours among Australian adults: Findings from the 2007 national survey of mental health and wellbeing. Australian and New Zealand Journal of Psychiatry, 43(7), 635–643. https://doi.org/10.1080/00048670902970874

23.

Kapur

(2020). Services for self-harm: Progress and promise? The British Journal of Psychiatry, 217(6), 663–664. https://doi.org/10.1192/bjp.2020.104

24.

Kisely

Abajobir

A. A.

Mills

Strathearn

Clavarino

Gartner

Najman

J. M.

(2020). Child maltreatment and persistent smoking from adolescence into adulthood: A birth cohort study. Nicotine & Tobacco Research: Official Journal of the Society for Research on Nicotine and Tobacco, 22(1), 66–73. https://doi.org/10.1093/ntr/ntz039

25.

Kisely

Strathearn

Mills

Najman

J. M.

(2021). A comparison of the psychological outcomes of self-reported and agency-notified child abuse in a population-based birth cohort at 30-year-follow-up. Journal of Affective Disorders, 280(Pt), 167–172. https://doi.org/10.1016/j.jad.2020.11.017

26.

Kisely

Strathearn

Najman

(2021a). The influence of child maltreatment on substance or alcohol use in 30-year-old adults: A birth cohort study. Drug and Alcohol Review, 40(4), 673–680. https://doi.org/10.1111/dar.13192

27.

Kisely

Strathearn

Najman

J. M.

(2020). Child maltreatment and mental health problems in 30-year-old adults: A birth cohort study. Journal of Psychiatric Research, 129, 111–117. https://doi.org/10.1016/j.jpsychires.2020.06.009

28.

Kisely

Strathearn

Najman

J. M.

(2021b). A comparison of oral health outcomes of self-reported and agency-notified child maltreatment in a population-based birth cohort at 30-year-old follow-up. Psychosomatic Medicine, 84(2), 179–187. https://doi.org/10.1097/psy.0000000000001029

29.

Kisely

Strathearn

Najman

J. M.

(2022). A comparison of psychosis-like symptoms following self-reported and agency-notified child abuse in a population-based birth cohort at 30-year-follow-up. Schizophrenia Research, 239, 116–122. https://doi.org/10.1016/j.schres.2021.11.029

30.

Klonsky

E. D.

Moyer

(2008). Childhood sexual abuse and non-suicidal self-injury: meta-analysis. British Journal of Psychiatry, 192(3), 166–170. https://doi.org/10.1192/bjp.bp.106.030650

31.

Kohl

P. L.

Jonson-Reid

Drake

(2009). Time to leave substantiation behind. Child Maltreatment, 14(1), 17–26. https://doi.org/10.1177/1077559508326030

32.

Marques-Feixa

Moya-Higueras

Romero

Santamarina-Pérez

Rapado-Castro

Zorrilla

Martín

Anglada

Lobato

M. J.

Ramírez

Moreno

Mayoral

Marín-Vila

Arias

Fañanás

Epi-Young Stress

(2021). Risk of suicidal behavior in children and adolescents exposed to maltreatment: The mediating role of borderline personality traits and recent stressful life events. J Clin Med, 10(22), 5293. https://doi.org/10.3390/jcm10225293

33.

Mills

Alati

Strathearn

Najman

J. M.

(2014). Alcohol and tobacco use among maltreated and non-maltreated adolescents in a birth cohort. Addiction (Abingdon, England), 109(4), 672–680. https://doi.org/10.1111/add.12447

34.

Mills

Kisely

Alati

Strathearn

Najman

(2016). Self-reported and agency-notified child sexual abuse in a population-based birth cohort. Journal of Psychiatric Research, 74, 87–93. https://doi.org/10.1016/j.jpsychires.2015.12.021

35.

Molnar

B. E.

Berkman

L. F.

Buka

S. L.

(2001). Psychopathology, childhood sexual abuse and other childhood adversities: Relative links to subsequent suicidal behaviour in the US. Psychological Medicine, 31(6), 965–977. https://doi.org/10.1017/S0033291701004329

36.

Murray

Irving

Farrington

D. P.

Colman

Bloxsom

C. A.

(2010). Very early predictors of conduct problems and crime: Results from a national cohort study. Journal of Child Psychology and Psychiatry, 51(11), 1198–1207. https://doi.org/10.1111/j.1469-7610.2010.02287.x

37.

Najman

J. M.

Alati

Bor

Clavarino

Mamun

McGrath

J. J.

McIntyre

O’callaghan

Scott

Shuttlewood

Williams

G. M.

Wray

(2014). Cohort profile update: The Mater-University of queensland study of pregnancy (MUSP). International Journal of Epidemiology, 44(1), 78–78f. https://doi.org/10.1093/ije/dyu234

38.

Najman

J. M.

Alati

Bor

Clavarino

Mamun

McGrath

J. J.

McIntyre

O’Callaghan

Scott

Shuttlewood

Williams

G. M.

Wray

(2015). Cohort profile update: The mater-university of queensland study of pregnancy (MUSP). International Journal of Epidemiology, 44(1), 78–78f. https://doi.org/10.1093/ije/dyu234

39.

Najman

J. M.

Bor

O'Callaghan

Williams

G. M.

Aird

Shuttlewood

(2005). Cohort profile: The mater-university of queensland study of pregnancy (MUSP). International Journal of Epidemiology, 34(5), 992–997. https://doi.org/10.1093/ije/dyi119

40.

Najman

J. M.

Kisely

Scott

J. G.

Strathearn

Clavarino

Williams

G. M.

Middeldorp

Bernstein

(2020). Agency notification and retrospective self-reports of childhood maltreatment in a 30-Year cohort: Estimating population prevalence from different data sources. Child Abuse and Neglect, 109, 104744. https://doi.org/10.1016/j.chiabu.2020.104744

41.

Q. X.

Yong

B. Z. J.

C. Y. X.

Lim

D. Y.

Yeo

W. S.

(2018). Early life sexual abuse is associated with increased suicide attempts: An update meta-analysis. Journal of Psychiatric Research, 99, 129–141. https://doi.org/10.1016/j.jpsychires.2018.02.001

42.

Olds

D. L.

Eckenrode

Henderson

C. R.

Jr. Kitzman

Powers

Cole

Sidora

Morris

Pettitt

L. M.

Luckey

(1997). Long-term effects of home visitation on maternal life course and child abuse and neglect. Fifteen-year follow-up of a randomized trial. JAMA, 278(8), 637–643. https://doi.org/10.1001/jama.1997.03550080047038

43.

Peters

Clark

Carroll

(1998). Are computerized interviews equivalent to human interviewers? CIDI-auto versus CIDI in anxiety and depressive disorders. Psychological Medicine, 28(4), 893–901. https://doi.org/10.1017/s0033291798006655

44.

Pinto

Correia

Maia

Â.

(2014). Assessing the reliability of retrospective reports of adverse childhood experiences among adolescents with documented childhood maltreatment. Journal of Family Violence, 29(4), 431–438. https://doi.org/10.1007/s10896-014-9602-9

45.

Pirkis

Burgess

Dunt

(2000). Suicidal ideation and suicide attempts among Australian adults. Crisis: The Journal of Crisis Intervention and Suicide Prevention, 21(1), 16–25. https://doi.org/10.1027//0227-5910.21.1.16

46.

Ranganathan

Pramesh

C. S.

Aggarwal

(2017). Common pitfalls in statistical analysis: Logistic regression. Perspectives in Clinical Research, 8(3), 148–151. https://doi.org/10.4103/picr.PICR_87_17

47.

Saiepour

Najman

Ware

Baker

Clavarino

Williams

(2019). Does attrition affect estimates of association: A longitudinal study. Journal of Psychiatric Research, 110, 127–142. https://doi.org/10.1016/j.jpsychires.2018.12.022

48.

Siddaway

A. P.

Taylor

P. J.

Wood

A. M.

Schulz

(2015). A meta-analysis of perceptions of defeat and entrapment in depression, anxiety problems, posttraumatic stress disorder, and suicidality. Journal of Affective Disorders, 184, 149–159. https://doi.org/10.1016/j.jad.2015.05.046

49.

Spence

S. H.

Najman

J. M.

Bor

O'Callaghan

M. J.

Williams

G. M.

(2002). Maternal anxiety and depression, poverty and marital relationship factors during early childhood as predictors of anxiety and depressive symptoms in adolescence. Journal of Child Psychology and Psychiatry, 43(4), 457–469. https://doi.org/10.1111/1469-7610.00037

50.

Strathearn

Mamun

A. A.

Najman

J. M.

O'Callaghan

M. J.

(2009). Does breastfeeding protect against substantiated child abuse and neglect? A 15-year cohort study. Pediatrics, 123(2), 483–493. https://doi.org/10.1542/peds.2007-3546

51.

Widom

C. S.

Raphael

K. G.

DuMont

K. A.

(2004). The case for prospective longitudinal studies in child maltreatment research: Commentary on dube, williamson, thompson, felitti, and anda (2004). Child Abuse and Neglect, 28(7), 715–722. https://doi.org/10.1016/j.chiabu.2004.03.009

52.

Winefield

H. R.

Bradley

P. W.

(1992). Susstantiation of reported child abuse or neglect: Predictors and implications. Child Abuse and Neglect, 16(5), 661–671. https://doi.org/10.1016/0145-2134(92)90104-y

53.

Wolke

Waylen

Samara

Steer

Goodman

Ford

Lamberts

(2009). Selective drop-out in longitudinal studies and non-biased prediction of behaviour disorders. The British Journal of Psychiatry, 195(3), 249–256. https://doi.org/10.1192/bjp.bp.108.053751

54.

Zatti

Rosa

Barros

Valdivia

Calegaro

V. C.

Freitas

L. H.

Ceresér

K. M. M.

Rocha

N. S. D.

Bastos

A. G.

Schuch

F. B.

(2017). Childhood trauma and suicide attempt: A meta-analysis of longitudinal studies from the last decade. Psychiatry Research, 256, 353–358. https://doi.org/10.1016/j.psychres.2017.06.082