Use of self-figure drawing as a screening tool to identify child sexual abuse, eating disorders, and suicidality

ABSTRACT

Childhood sexual abuse is a significant risk factor for adverse psychological outcomes such as eating disorders, as well as suicidal attempts and ideation. Detecting these phenomena at an early stage is crucial for improving prognosis. The current study investigated indicators of childhood sexual abuse, eating disorders, and suicidal tendencies as reflected in the self-figure drawings of 130 Slovakian adolescents treated for eating disorders or hospitalized after attempted suicide and a group of healthy controls. Kruskal Wallis test findings indicated that the drawings made by the hospitalized adolescents contained a higher number of indicators associated with childhood sexual abuse, eating disorders, and suicidal tendencies than those from the healthy control group. Similarly, the hospitalized adolescents had higher somatic dissociation scores on the MSDQ than the healthy controls. Logistic regression and decision tree models indicated that child sexual abuse could be predicted by both the drawing indicators and the somatic dissociation scores.

KEYWORDS:

• Childhood sexual abuse
• eating disorders
• suicidality
• self-figure drawing
• screening

Introduction

Childhood sexual abuse (CSA) is a social phenomenon with long-term physical and mental consequences, including difficulties with emotional regulation, depression, anxiety, dissociation, post-traumatic stress disorder, addiction, eating disorders, and suicidal tendencies (Chang et al., 2017). Due to the high frequency and the severe consequences of sexual abuse worldwide, which affects 8 to 31% of all girls and 3 to 17% of all boys (Euser et al., 2016; Finkelhor et al., 2014), detecting sexual abuse as early as possible is crucial to putting an end to the abuse, referring the victims for treatment, and preventing further deterioration. However, professionals often fail to identify or report child sexual abuse when there is no apparent corroborative physical evidence (Pelisoli et al., 2015; Powell & Barnett, 2015). CSA is a significant risk factor for two widespread problems in adolescence: eating disorders and suicidality (Barakat et al., 2023; Collin-Vézina et al., 2021).

The suicide rate of adolescents is 5.18/100,000 individuals (Sheftall et al., 2016). According to Shain et al. (2016), adolescent boys (aged 15 to 19) have a three-fold higher risk of committing suicide than girls of the same age. However, the rate of attempted suicide is twice as high in girls than in boys. The prevalence of eating disorders in adolescents is 0.3% for anorexia nervosa and 0.9% for bulimia nervosa but lower than in the adult population (Swanson et al., 2010). Roughly 23 to 48% of all adolescents with eating disorders have comorbid depression, with a higher prevalence rate in patients with bulimia nervosa (50%) than in those with anorexia nervosa (11%) (Watson et al., 2014). Preti et al. (2009) reported that most eating disorders (ED) begin between the ages of 10 and 20, but many individuals who suffer from ED are only diagnosed and treated after drastic weight loss or after suffering severe distress (Preti et al., 2009). Anorexia nervosa, for example, is the third most common chronic illness among adolescents. Young women suffering from this illness have a more than 12-fold higher mortality rate than average, making it the mental health illness with the highest premature mortality rate. If untreated, eating disorders become more severe and less receptive to treatment (Becker et al., 2004; Fichter et al., 2006). Similar figures have been reported for attempted suicides in adolescents (Bedi et al., 2011). Despite the high number of incidents, adolescents are often referred to treatment after an attempt rather than before for depression or withdrawal behaviours.

There is a consensus among clinicians that the earlier the screening for child sexual abuse, an eating disorder, or suicidality (often the consequences of CSA), the better the prognosis (Singh et al., 2014). Thus, the early detection, identification, and screening of victims of CSA are essential for halting further negative consequences. Whereas structured instruments, self-report measures, forensic medical examination, and psychological assessment can be harnessed to diagnose and treat these problems, there are few screening tools combining CSA, ED, or suicidality. Self-report instruments are often used to verify the ED diagnosis in patients already under observation, but require the individuals’ cooperation at a time when they often attempt to conceal the problem.

By contrast, techniques adapted from the creative arts therapies may help overcome these shortcomings. In particular, the Draw a Person (DAP) test addresses the issue of conscious and unconscious conflicts about one’s perception and body image (Levy, 1950). Like the SCOFF questionnaire (Morgan et al., 2000), the DAP test can be easily administered to a large group of individuals.

The current study was designed to detect indicators in self-figure drawings that reflect CSA, ED, and suicidality. Specifically, it examined whether the Draw-Yourself technique (a version of the Draw-A-Person DAP test; Machover, 1949) would be an effective screening tool to detect adolescents at risk for CSA, ED, and suicidality. The participants were asked to make a self-figure drawing (which requires relating to one’s own body) and complete the Medical Somatic Dissociation Questionnaire (MSDQ) that was found to successfully differentiate between CSA survivors and non-CSA survivors (Balla et al., 2023).

Background

Child Sexual Abuse (CSA) – consequences and detection tools

Child sexual abuse is a worldwide phenomenon. However, there is a vast gap between the prevalence and the reporting of CSA (Winters et al., 2020). Facilitating disclosure is complicated by the victims’ experiences (Alaggia et al., 2019). Demographics such as age and gender affect children’s ability to readily disclose abusive events (Yurteri et al., 2022). Adolescent sexual abuse has multiple consequences, such as depression, post-traumatic stress disorder, and suicidal tendencies (Paolucci et al., 2001), difficulties in interpersonal relationships, and risky sexual behaviours (Homma et al., 2012). When child sexual abuse is suspected, medical or physical evaluation is one of the crucial steps in the clinical and legal process (Adams et al., 2016). Often, a forensic interviewer takes the history which includes physical and emotional/behavioural symptoms (Finkel & Alexander, 2011).

Three main tools or techniques are used to detect CSA: the victimized child’s testimony, medical forensic examination, and psychological tests. However, each has its own limitations. Most victimized children refrain from disclosure, and their testimony in court is often incoherent out of fear or their desire to protect the family or perpetrator (Efrati, 2018; Author). Forensic medical examination often does not reveal evidence about the abuse since the victim is often seen more than 72 hours after the abuse (Herrmann et al., 2014). Finally, psychological tests assess the level of symptoms severity rather than their source. Recently, the Medical Somatic Dissociation Questionnaire (Daphna-Tekoah et al., 2019), which consists of three factors of somatization, dissociation, and depression was found to differentiate between CSA victims and non-victims at high symptom severity (i.e. after prolonged CSA) (Balla et al., 2023).

Eating disorders – consequences and detection tools

Eating disorders (ED) include anorexia nervosa (AN), bulimia nervosa (BN), binge eating disorder (BED), and purging disorder (PD) (Stice et al., 2010). Though the prevalence of eating disorders and their mortality and morbidity rates are high, they often go undiagnosed by professionals (Campbell & Peebles, 2014). These disorders involve biological, familial, and psychosocial factors (Stice et al., 2010). Studies suggest that roughly 80–97% of all individuals with an eating disorder do not receive appropriate medical attention (Swanson et al., 2010). In addition, eating disorders have been associated with CSA (Carter et al., 2006; Caslini et al., 2016; Dominé et al., 2009; Fischer et al., 2010) and CSA was found to be a risk factor for developing an eating disorder in adolescence (Neumark-Sztainer et al., 2000; Pollert et al., 2013).

Self-report questionnaires such as the Eating Disorder Inventory (ED-12) and the Eating Attitudes Test (EAT-26) are designed to assess whether an individual is at risk for developing an eating disorder. However, since many patients are in denial or try to conceal their eating disorders (Vitousek et al., 1991), these self-report instruments have low reliability.

Suicidality – consequences and detection tools

Suicidal ideations are defined as the thought and desire to die, or actually taking steps to end one’s life (Posner et al., 2011). Suicide prevention is hampered by the lack of sufficient assessment and screening (Posner et al., 2011), although specific assessment procedures have been shown to identify high risk better than standard clinical interviews (Bongiovi-Garcia et al., 2009). Several studies have found that adolescents who have been sexually abused are at higher risk of suicide and suicidal tendencies (Dube et al., 2001; Sapp & Vandeven, 2005). Extreme suicidal ideations with actual attempts have been associated with a traumatic history (Dube et al., 2001; Ullman & Najdowski, 2009). Note that a suicide attempt resulting in death is 50 times higher among those with anorexia nervosa (Keel et al., 2003).

Self-figure drawing for assessment

A popular projective drawing technique for psychological assessment is the Draw-A-Person (DAP) by Machover (1949), a human figure drawing test. The figure is thought to represent the subject, and the paper the person’s environment. This technique is designed to gain insights into a person’s emotional and cognitive functioning and elicit impulses, anxieties, and conflicts (Furth, 1988; Gillespie, 1994; Kaplan, 1994). For example, numerous studies have identified indicators in self-figure drawings that point to forms of abuse (Girish et al., 2023; Jaroenkajornkij et al., 2022; Author), depression (shaded eyes, detailed clothing other personal belongings) (Deng et al., 2022), child sexual abuse (Face Line – shaded or emphasized cheek or chin), Eyes – hollowed, shaded or omitted, Hands and Arms -emphasized, disconnected or omitted, and shaded, blocked off, disconnected or omitted lower body) (Kissos et al., 2020; Lev-Wiesel, 1999), eating disorders (Guez et al., 2010) (Neck -omitted, doubled or disconnected, the mouth is emphasized or omitted, and the feet are missing or disconnected), Suicidality (slash lines on the body, empty face or omission of head or face, and shading or X’s over the body figure (Jacobs-Kayam et al., 2013; Zalsman et al., 2000).

Slovakian adolescence population, CSA, ED, suicidality

Adolescence is a period of multiple psychological changes. Several socioeconomic crises, such as the COVID pandemic, the climate crisis, and the war in Ukraine, in addition to feelings of helpless, lonely, insecurity, and insignificance (Everall et al., 2006) have impacted Slovakian adolescents. Specific adverse childhood experiences are also known to contribute to severe mental health issues, and increase the need for psychosocial support and care (Lackova Rebicova et al., 2022).

There is little research on child abuse and violence against minors in Slovakia. One study reported that roughly two-thirds of all Slovakians have affective disorders, and a considerable proportion have anxiety disorders (Brazinova et al., 2019). Another recent study found a significant association between age and anxiety and depressive symptoms, where the younger the individual, the greater the symptoms in Slovakian youth (Kralova et al., 2022).

There are very few studies that have examined on the prevalence of child sexual abuse (Karkošková & Ropovik, 2018), eating disorders, or suicidal ideation in Slovakian adolescents. Suicide is considered to be one of the leading causes of death among adolescents in Europe (Kokkevi et al., 2012). Nearly 17.9% of all Slovakian adolescents in a large sample of students reported attempting suicide (Kokkevi et al., 2012). It is also a risk factor for developing an eating disorder (Levická et al., 2014). Adverse life events such as sexual abuse, issues in relationships, experiencing loss, or even moving from one place to another have also been reported to trigger eating disorders (Levická et al., 2014). A study on children in Bratislava showed that 32% of the girls and 11% of the boys in the sample were at risk of developing an eating disorder (Osaďan & Drgoňová, 2013).

To fill these gaps in the literature, the current study was designed to validate the Self-Figure drawing tool for detecting sexual abuse in hospitalized adolescents that would also predict eating disorders and suicidal ideation. All three often have no external indications (Guez et al., 2010; Author) and can be masked, thus blurring diagnosis and screening. Most clinical assessments also involve self-reporting, which can be inaccurate because there is a tendency to conceal their presence out of shame, guilt, or disappointment (Nock & Banaji, 2007). By contrast, an art-based assessment tool such as the self-figure drawing is less intrusive and may be more straightforward in detecting sexual abuse, eating disorders, or suicidality.

Specifically, the first objective of the present study was to investigate variations in pictorial indicators and scores on the MSDQ in a group of adolescents with a history of attempted suicide, a second group with eating disorders, and a healthy control group. The potential of self-figure drawing indicators to detect CSA, as reported by the adolescents in the Traumatic Events Questionnaire, was compared to these same adolescents’ MSDQ scores. Hypothesis 1 posited that adolescents with a history of attempted suicide or a diagnosis of eating disorders would have more drawing indicators associated with sexual abuse, eating disorders, and suicidality than healthy controls. Hypothesis 2 posited that elevated scores for the self-figure indicators linked to sexual abuse, eating disorders, and suicidality would positively correlate with the MSDQ scores that predict adolescents’ CSA.

Method

Participants and procedure

One hundred thirty adolescents aged 11 to 17 years (MED = 14.5), of whom 97 were girls (74.6%), were divided into three groups: adolescents who had attempted suicide, adolescents with a diagnosed eating disorder, and healthy controls. Participants in the suicide attempts and eating disorders groups were approached between January and November 2022 to participate in this study while hospitalized in a paediatric psychiatry department or during a visit to the department’s outpatient facility. All adolescents in the treatment groups had been diagnosed according to ICD-10 criteria. The healthy controls were adolescents with no psychiatric history who, in the same timeframe, had consulted a local paediatric outpatient facility for common transitional health issues (flu, for example).

The eating disorders group consisted of 37 participants (36 girls, 97.3%) with a median age of 15 years (range 14 to 16). The attempted suicide group consisted of 43 participants (41 girls, 95.3%), with a median age of 15 years (range 13 to 15), and the healthy control group consisted of 50 participants (20 girls, 40%) with a median age of 14 years (range 13 to 16).

Upon approval from the Ethics Committee and as a prerequisite for participation, the participant’s parents or legal guardians were required to provide written informed consent to participate in the study. All the participants gave their verbal informed consent for involvement in the study. This study was exempted by the institutional review board.

Measures

This study employed a mixed-method design. The qualitative part was composed of the analysis of the self-figure drawing. The quantitative part included demographics and the administration of the MSDQ and the Traumatic Events Questionnaire. The translation to Slovak was conducted according to the principles of translated and back-translated measure lingual adaptation.

Self-figure drawing

A self-figure drawing version of the Draw-A-Person projective tool (Machover, 1949) was used in this study. Participants were asked to draw themselves on a blank sheet of A4-sized paper with a pencil. No further instructions were given. Questions were addressed by encouraging participants to proceed with the drawing based on what they understood from the instructions. Once the drawing was completed, they were rated by two art therapists based on a list of indicators reflecting CSA, eating disorders, and suicidality on a four-point Likert-type scale ranging from 1 (not at all obvious) to 4 (very obvious). Disagreements were resolved by discussion. To assess the probability of suicidal ideation or eating disorders as manifested in the drawings, a composite score for each condition was created.

The Sexual Abuse score (SA) was computed by adding the scores of indicators related to the eyes (hollowed, shadowed, dots or omitted) cheek/chin (emphasized or doubled), arms/hands (shadowed, disconnected, or omitted), and genital area (disconnected upper body from lower body, shadowed, omission of lower body, or emphasized) (for the indicators, see Kissos et al., 2020; Lev-Wiesel, 1999). The scores ranged from 1 to 16, with a higher score indicating a greater risk of having experienced sexual abuse. The Eating Disorder score (ED) was generated by summing the scores of indicators related to the mouth (enlarged, shadowed, or hollowed), neck (doubled or shadowed), and feet (shadowed, disconnected or omitted) (for the indicators, see Guez et al., 2010). These scores also ranged from 1 to 12, with a higher score indicating a higher risk of having an eating disorder. The Suicidal Attempts/Ideations score (SUID) was constructed by summing the scores of indicators related to the face/head (shadowed or omitted), shading the body contour, and adding X’s on the whole figure face or body organs (for the indicators, see Jacobs-Kayam et al., 2013; Zalsman et al., 2000). These scores ranged from 1 to 12, with a higher score suggesting a greater risk of suicidal attempts or thoughts. Previous cross-cultural studies using self-figure drawing have provided evidence for the reliability and validity of these scores (e.g. Girish et al., 2023; Jacobs-Kayam et al., 2013, Jaroenkajornkij et al., 2022; Kissos et al., 2020; Lev-Wiesel et al., 2022).

Medical Somatic Dissociation Questionnaire (MSDQ) (daphna-tekoah et al., 2019)

This questionnaire consists of 30 items covering all three somatic dissociation symptoms commonly seen in sexual abuse. The three subscales are dissociation, depression, and somatization. Participants rate each item on a 5-point Likert-type scale ranging from 0 (not at all) to 4 (extremely). The internal consistency (Cronbach’s alpha) reported for the measure was .93 (Daphna-Tekoah et al., 2019). Robust convergent validity has also been reported (Balla et al., 2023; Daphna-Tekoah et al., 2019).

The traumatic life events questionnaire

The TLEQ assesses experiences related to traumatic events (e.g. sexual abuse,

accidents, and crime-related incidents) that are considered potential triggers of Post-

Traumatic Stress Disorder (PTSD) and various symptoms (Vrana & Lauterbach, 1994).

Respondents provide information for each event experienced (physical abuse, emotional abuse, sexual abuse, car accidents, loss of a family member, terror attacks, war, hospitalization or illness, parent’s divorce, and bullying), as well as their age at the

time of the event; items are measured on a 9-point Likert scale, with responses ranging

from ‘not at all’ to ‘severely/extremely.’ The TLEQ has previously been used in Israel (e.g. Balla et al., 2023).

Results

Table 1 presents the descriptive statistics for the sample, encompassing the traumatic events reported by participants, the medians of the drawing indicators, and MSDQ scores. Because the continuous variables did not have a normal distribution, the scores are reported as the medians and interquartile ranges (IQR). Categorical variables are presented using frequencies and proportions.

Table 1. Descriptive statistics.

Variable	n (%)/MED, IQR
Group
Healthy controls	50 (38.5%)
Eating disorder	37 (28.5%)
Suicide Attempts/Ideation	43 (33.1%)
Sex
Boys	33 (25.4%)
Girls	97 (74.6%)
Age	14.5 (13.0–16.0)
Traumatic events
Car Accident	13 (10.0%)
Physical Assault	65 (50.0%)
Terror attack or war	4 (3.08%)
Hospitalization due to illness	95 (73.1%)
Sexual Abuse	21 (16.2%)
Loss of a family member	85 (65.4%)
Verbal emotional abuse	42 (32.3%)
Bullying	67 (51.5%)
Parental divorce	37 (28.5%)
None of the above	4 (3.08%)
Other traumatic events	122 (93.8%)
Drawing indicators
SA score	11.00 (10.00, 14.00)
Eyes	4.00 (3.00, 4.00)
Cheek Chin	4.00 (2.00, 4.00)
Arms Hands	4.00 (2.00, 4.00)
Genital Area	2.00 (1.00, 3.75)
ED score	8.00 (6.00, 10.0)
Mouth Emphasized or Omitted	3.00 (2.00, 4.00)
Neck	3.00 (2.00, 4.00)
Feet omission or disconnected	2.00 (1.00, 4.00)
SUID score	6.00 (4.00, 9.00)
Empty face, no head or face	1.00 (1.00, 3.00)
Shading	3.00 (1.25, 4.00)
Adding X	1.00 (1.00, 3.00)
MSDQ score	2.41 (1.79, 2.96)

Preliminary analysis

Table 2 provides a comparison of demographic variables across the three groups. As shown in the table, the chi-square analysis revealed differences in gender distribution across groups (X²(2) = 2.01). Notably, the eating disorders and suicide attempts groups were predominantly composed of females (>95%), while the healthy control group comprised of 40% females. In addition, 81.4% of the attempted suicide group, 51.4% of the eating disorders group, and 22.0% of the healthy control group reported experiencing physical assault. Furthermore, 94.6% of the eating disorders group, 81.4% of the attempted suicide group, and 80.0% of the healthy control group reported illness-related hospitalization. Lastly, 34.9% of the attempted suicide group, 16.2% of the eating disorders group, but none of the healthy control group reported experiencing sexual abuse.

Table 2. Comparisons between the study groups.

	Healthy controls (HC) n = 50	Eating disorders (ED) n  = 37	Suicide attempts/ ideations (SAI) n  = 43	X^2(2)	ES	p overall	p HC vs ED	P HC vs SAT	p ED vs SAT
Sex Female n/%	20 (40.0%)	36 (97.3%)	41 95.3%)	2.01	<.001	<.001	<.001	<.001	ns
Age MED (IQR)	14.0 (13.00,16.00)	15.0 14.00,16.00	15.0 (13.00,15.00)	60.85	0.46	.367	.355	.858	.355
MSDQ score MED (IQR)	1.68 (1.37,2.06)	2.47 2.00,3.00	2.93 (2.75,3.19)	15.97	0.11	<.001	<.001	<.001	.002
Traumatic event n/%
Car Accident	4 (8.00%)	2 (5.41%)	7 (16.3%)	2.97	.11	.282	ns	.544	.498
Physical Assault	11 (22.0%)	19 (51.4%)	35 (81.4%)	32.66	.35	<.001	.009	<.001	.009
Terror attack or war	1 (2.00%)	1 (2.70%)	2 (4.65%)	.57	.05	.828	ns	ns	ns
Hospitalization	25 (50.0%)	35 (94.6%)	35 (81.4%)	23.75	.30	<.001	<.001	.005	.097
Sexual Abuse	0 (0.00%)	6 (16.2%)	15 (34.9%)	20.77	.28	<.001	.007	<.001	.102
Loss of a family member	31 (62.0%)	24 (64.9%)	30 (69.8%)	.62	.05	.733	.961	.961	.961
Verbal emotional abuse	7 (14.0%)	14 (37.8%)	21 (48.8%)	13.55	.23	.001	.031	.002	.446
Bullying	17 (34.0%)	15 (40.5%)	35 (81.4%)	23.3	.30	<.001	.689	<.001	.001
Parental divorce	4 (8.00%)	15 (40.5%)	18 (41.9%)	16.72	.25	<.001	.001	.001	ns
None of the above	4 (8.00%)	0 (0.00%)	0 (0.00%)	6.6	.16	.037	.133	.133	ns
Other traumatic events	44 (88.0%)	35 (94.6%)	43 (100%)	5.81	.15	.046	.458	.087	.316
Drawing Indicators MED (IQR)
SA score	10.00 (8.00,11.0)	12.00 (10.0,14.0)	15.00 (13.0,16.0)	15.97	.11	<.001	.001	<.001	<.001
Eyes	3.50 (1.00,4.00)	4.00 (3.004.00)	4.00 (4.00,4.00)	10.50	.07	<.001	.049	<.001	.058
Cheek Chin	3.00 (1.00,4.00)	4.00 (2.00, 4.00)	4.00 (3.50,4.00)	15.52	.11	.005	.271	.004	.067
Arms Hands	3.00 (2.00,4.00)	4.00 (2.00;4.00)	4.00 (4.00,4.00)	40.64	.30	<.001	.13	<.001	.034
Genital Area	1.00 (1.00,2.00)	2.00 (1.00,3.00)	4.00 (3.00,4.00)	31.00	.23	<.001	.005	<.001	<.001
ED score	6.00 (5.00,7.75)	9.00 (7.00,10.00)	10.00 (9.00,12.00)	20.35	.14	<.001	<.001	<.001	.002
Mouth	2.00 (1.00,3.00)	3.00 (2.00,4.00)	4.00 (4.00,4.00)	23.33	.17	<.001	.005	<.001	.005
Neck	2.00 (1.25,3.00)	3.00 (2.00,4.00)	4.00 (3.00,4.00)	31.35	.23	<.001	.004	<.001	.128
Feet	1.00 (1.00,2.00)	3.00 (2.00,3.00)	4.00 [2.00,4.0)	56.99	.43	<.001	.001	<.001	.143
SUID score	3.50 (3.00,4.00)	6.00 (5.00;7.00)	9.00 (8.00,12.00)	61.32	.47	<.001	<.001	<.001	<.001
Face/Head	1.00 (1.00,1.00)	1.00 [1.00,4.00]	2.00 (1.00,4.00)	49.68	.38	<.001	<.001	<.001	.118
Shading	1.50 (1.00,2.00)	3.00 [2.00,4.00]	4.00 (4.00;4.00)	47.93	.36	<.001	<.001	<.001	<.001
Adding X	1.00 (1.00,1.00)	1.00 [1.00,2.00]	4.00 (2.00;4.00)	72.52	.56	<.001	.001	<.001	<.001

MSDQ = Medical Somatic Dissociation Questionnaire; SA = sexual abuse; ED = eating disorders; SUID = suicide attempts, HC=healthy controls, SAT = suicidal attempts, ES= effect size, in categorical variables, the static test is Pearson Chi-square and the effect size is Cramer V which interpretation is :0–0.21 small, .21–.35 medium, and ≥ .35 large. In continuous variables, the statistic test in Kruskall Wallis and the effect size is .01- < .06 (small effect), .06 - < 0.14 (moderate effect), and ≥ .14 (large effect), Cohen (1988).

Differences between variables across groups

To investigate the differences in the pictorial indicator scores and the MSDQ scores in the three groups, a series of Kruskal Wallis tests were conducted, followed by post-hoc pairwise comparisons. These comparisons were adjusted for Tukey’s multiple testing to assess differences between pairs of groups. As anticipated and shown in Table 2, the sexual abuse, eating disorder, and suicidal attempts/ideations pictorial indicators scores, as well as the MSDQ scores were the highest in the attempted suicide group and the lowest in the healthy control group. Note that only three out of four pictorial indicators of SA (previously reported in studies) were found significant in the current study.

Predicting child sexual abuse

To predict child sexual abuse according to the drawing scores (i.e. sexual abuse, eating disorder, and suicidality pictorial indicator scores) and the MSDQ scores, logistic regression models and decision tree models were employed. A series of models were constructed for the logistic regression models, each with a different combination of the four predictors: sexual abuse, eating disorder, and suicidality pictorial indicators and MSDQ scores. The predictive performance of each model was assessed using the area under the curve (AUC) value. The AUC values were then compared using the non-parametric approach described by DeLong et al. (1988). AUC values ranging from .60 to .75 indicate moderate-level prediction accuracy, .75 to .90 represent good accuracy, .90 to .97 indicate excellent accuracy, and .97 to 1.00 represent optimum accuracy (Swets, 1988).

The decision tree models then explored how the MSDQ and pictorial indicator scores interacted in predicting the history of sexual abuse. The decision tree model provides a transparent and easily interpretable representation of the decision-making process. The splits and branches of the tree serve to identify which scores have the most significant impact on the outcome. Two decision trees were produced for each outcome: one with MSDQ and the self-drawing indicator scores as potential predictors and the other using only the pictorial indicator scores. This was done to test whether the results could be predicted solely based on the pictorial indicators. The Recursive Partitioning (RPART) analysis generated the decision trees. All potential predictors were entered into the regression tree analysis using the R package ‘part’ (Breiman et al., 1984). The final decision tree was obtained by determining the tree’s optimal maximum depth parameter based on cross-validation estimated prediction error.

Table 3 presents the logistic regression prediction performance of the different combinations of variables predicting sexual abuse experience. The model that achieved the highest performance included the MSDQ score and three indicator scores; however, a comparison to the MSDQ predictor model alone did not yield a statistically significant difference in AUC. Models without MSDQ scores had significantly lower AUC values than models with MSDQ.

Table 3. Performance of logistic regression models predicting sexual abuse.

MODEL	AUC
MSDQ + SA score+ ED score+ SUID score	.82
MSDQ + ED score	.81
MSDQ + SUID score	.80
MSDQ	.80
MSDQ + SA score	.80
ED score	.71
SA +ED+SUID scores	.71
SUID score	.65
SA score	.62

MSDQ = Medical Somatic Dissociation Questionnaire; SA = sexual abuse; ED = eating disorders; SUID = suicide attempts.

The tree decision analysis (Figure 1) for predicting sexual abuse from the MSDQ and indicators’ scores revealed that adolescents with a MSDQ ≥ 2.9 and an eating disorder score ≥ 11 were likelier to have experienced sexual abuse. The AUC for this predictive analysis was 0.78.

Figure 1. Decision tree for predicting sexual abuse (1), candidate predictors are MSDQ and drawing indicator scores.

The decision tree analysis (Figures 2–5) for predicting sexual abuse solely from the indicator scores revealed that patients with eating disorder score ≥ 11 and a suicide score = 12 were likelier to have experienced sexual abuse. The AUC for this predictive analysis was .70

Figure 2. Decision tree for predicting sexual abuse (1), candidate predictors are drawing indicator scores.

Figure 3. Female aged 14, diagnosed with eating disorders, three indicators of ED (shaded mouth, doubly emphasized neck, no feet), 4 indicators of CSA (shaded eyes, doubled cheek, shaded arms, and hands, emphasized shaded lower body- genitals).

Figure 4. Female aged 15, hospitalized after attempted suicide three indicators of suicidality (shaded face/head, shaded feet, X on mouth), and 4 indicators of CSA (no eyes, no hands, doubled chin, belt/emphasized boundary between upper and lower body).

Figure 5. Healthy female aged 14 (controls), no indicators of CSA, ED, or suicidality.

Discussion

The main aim of this study was to investigate the extent to which the self-figure drawing reflects CSA, eating disorders, and suicidal ideation in adolescents. The analysis showed that the drawing indicators and MSDQ scores achieved good prediction accuracy for sexual abuse. The MSDQ has previously been shown to be a highly reliable and valid tool in assessing child sexual abuse (Balla et al., 2023; Daphna-Tekoah et al., 2019). In the present study, the indicators for sexual abuse (Author), such as eyes, arms/hands, and the genital area, remained significant and enabled comparison between groups, except for emphasized cheek/chin for all three groups. This particular indicator, which is considered to reflect the individual’s shame/guilt at having been abused as seen in adult survivors of child sexual abuse (Author), may not apply to this particular population of adolescents, who were found to have disclosed to peers (Schönbucher et al., 2012). Cheek or chin emphasis is considered to signify the feeling of disgust and choking on unbearable material. The indicators for ED (mouth, neck, feet) and suicidality (empty face/head, shading, marking an X) were significant in the current study, as seen in other studies (Guez et al., 2010; Jacobs-Kayam et al., 2013; Zalsman et al., 2000).

The decision tree prediction model revealed a consistent finding: somatic dissociation requires high scores for at least 50% of the items and at least two very clear indicators for eating disorders. The prediction accuracy, again, was only considered ‘good,’ thus warranting further studies with a larger sample. When the drawing indicators were used alone, the ED and suicidality indicators needed to be highly obvious to predict sexual abuse. Other studies have reported that ED and suicidality may be long-term outcomes of sexual abuse (Hailes et al., 2019).

Despite its limitations, relatively small sample size, unequal gender distribution, and the fact that the ED and suicidal groups were hospitalized and hence were high-risk patients, the results of this study contribute to the assessment and screening of CSA, ED, and suicidal tendencies. The self-figure drawing, a brief, non-intrusive screening tool, can overcome the shortcomings of self-report tools. The self-figure drawing can also be used to prompt a discussion with the adolescent about suicidal ideations in therapeutic settings. Thus, practitioners and professionals working with adolescents may find the self-figure drawing useful in assessing ED and suicidality or a tendency to develop either, thereby averting tragic consequences. Few studies have explored screening of CSA as a possible source for the two serious problems of ED and suicidality using a single tool, making this study novel with significant practical contributions.

Authorship

Rachel Lev-Wiesel, writing – original draft, writing – review & editing, conceptualization; Jana Trabaticka, conceptualization: Matzová Zuzana, Vatrál Martin, Vášáryová Dorota, Žemberová Hana data collection; Nava Weiner, Meghna Girish, Yael Zaksh Irit Oryan, Drawing data curation; Dana Hadar, Limor Goldner, Statistical analysis.

All authors approved the final version of the article.

Disclosure statement

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