The Impact of Second Step Child Protection Unit Teacher Training: Latent Moderated Structural Equation Modeling (SEM) Approach

Abstract

Given the high prevalence and harmful consequences of child sexual abuse (CSA), we need to empower teachers to play a critical role in prevention/intervention efforts. We therefore explored the potential of CSA teacher training to improve preventive outcomes based on teachers’ CSA knowledge. Analyzing the data from the implementation of a CSA prevention program using structural equation modeling (SEM) revealed a significant effect of CSA teacher training on improving teachers’ CSA knowledge, particularly in teachers with lower prior knowledge.

1. Introduction

Child sexual abuse (CSA) has traumatized millions of boys and girls worldwide [1]. In the United States alone, more than 57,000 child victims were reported during 2015 and it has been suggested that as many as 28% of youths aged from 14 to 17 are exposed to CSA [2, 3, 4]. The actual CSA occurrence rate may be even higher, given that only around one-third of child abuse cases are reported to authorities and CSA is considered to be greatly underreported [3, 5, 6].

These shocking statistics should be driving urgent action by administrators and policy-makers given the highly detrimental short- and long-term consequences of CSA, which include re-victimization, substance abuse, poor mental and physical health, lower academic achievement, higher school dropout levels, and suicidal ideation/attempts [3, 7, 8, 9, 10, 11, 12].

As part of our effort to prevent/intervene the occurrence of CSA and address both its high prevalence rate and serious adverse consequences, we explored the possible contribution of the teachers and other school staff members who play a critical role in educating children about sexual abuse and self-protection, and who are in an excellent position to notice sudden changes in children’s behavior that may indicate abuse [13, 14]. Despite their critical role and their mandatory requirement to report suspected abuse, research has shown that most teachers lack sufficient knowledge to identify potential cases of CSA and are unfamiliar with their schools’ procedures for reporting their suspicions [15, 16, 17]. Furthermore, one-third of teachers underreport child abuse [18, 19].

To address any issues that teachers might have in spite of their important roles, we investigated the effect of relevant teacher training based on the findings reported in several prior studies. Previous studies of the impact of providing CSA prevention programs for teachers have shown that they result in significant increases in knowledge, opinions, and anticipated behaviors when dealing with children who have been or are being sexually abused [15, 20]. Teachers were found to have high satisfaction levels related to this specialized training, increasing both their awareness of the problem and their readiness/inclination to develop prevention/intervention plans in the future [21].

In view of the potential role of teacher training in making teachers better prepared to execute CSA prevention/intervention plans, in this study we sought to investigate whether/how teacher training can increase important teacher outcomes, represented here by teachers’ CSA knowledge [22, 23, 24]. At the same time, we looked at the moderating role of teachers’ preexisting knowledge, by examining the effect of teacher training on those teachers who had lower initial levels of knowledge.

For this empirical investigation, we analyzed the data from teachers who participated in the Second Step Children Protection Unit (CPU). CPU is a comprehensive CSA prevention program developed by the Committee for Children (CfC) (http://www.cfchildren.org/child-protection) to offer specialized teacher training on this difficult subject. We applied a series of structural equation modeling (SEM) analyses to investigate the effect of this widely accepted teacher training intervention as well as any interaction effect between the baseline (teachers’ preexisting knowledge) and the intervention [25, 26]. We also investigated the association between teachers’ acceptance of the CSA interventions and their CSA knowledge after the CPU teacher training.

The following research questions guided our study.

1. Does the CPU teacher training result in any improvement in teacher knowledge of CSA?

2. Does the CPU teacher training interact with teachers’ prior knowledge to improve their CSA knowledge?

2. Methods

2.1 Sample and intervention

The participants consisted of 161 teachers from a randomized control study designed to evaluate CPU. These participating teachers taught students in grades PreK-5 in elementary schools in New York State and their teaching experience ranged from 1 year to more than 30 years. Participating teachers were assigned to either the intervention group or the control group. The teachers in the intervention group received the treatment by taking the CPU online teacher training, which consisted of two modules: Policies and Procedures (75–90 min) and Recognize, Respond, and Report Abuse (45–75 min). The intervention teachers also implemented the CPU 6-week lessons with children in their classrooms. Teachers in the intervention group completed pretest/posttest measure assessing their knowledge of CSA prior to receiving the training and afterwards. The same pretest/posttest measure was administered to the teachers in the control group who did not receive the training.

2.2 Measures

Participating teachers were asked to complete the measures described below. Teachers in both the intervention and control groups completed the measure on teachers’ preventive outcome in terms of knowledge.

Educators and Child Abuse Questionnaire (ECAQ;23): The ECAQ was used to assess knowledge of both CSA and policy components (e.g., reporting procedures). Out of the total of four subscales of ECAQ, this study examined one subscale of awareness of signs and symptoms of child abuse featuring good psychometric quality in our study settings. It has been used in several studies [16, 23, 27] to assess educators’ knowledge and competence with regard to CSA and policies. The measure includes 12 items on a 5-point Likert scale ranging from strongly agree (1) to strongly disagree (5) (Some items are reverse-coded so that higher scores represent more knowledge) (Table 1).

	ECAQ (Pretest)	ECAQ (Posttest)
Reliability (α)	0.77	0.86

Table 1.

Reliability of measures.

2.3 Analysis

We analyzed the data by building a series of SEM models utilizing Mplus 8.1 [28]. Particularly, we took a latent variable moderated SEM approach to test how teachers’ baseline knowledge measured at the pretest moderated the effect of the teacher training intervention on teacher knowledge at the posttest, in addition to the effect of the teacher training intervention on teacher knowledge at the posttest [29, 30, 31, 32].

Figure 1 shows a conceptual diagram of our SEM model (Intervention Model) with our teacher training intervention (train: coded 0 for the control group; 1 for the intervention group), as well as the indicator variables (awar1_1, awar1_2, and awar1_3) for the latent teacher knowledge at the pretest (awar1) and those (awar2_1, awar2_2, and awar2_3) for the latent teacher knowledge at the posttest (awar2). The latent moderation (aka, interaction) model included the interaction term (awar1tr) by adopting the latent moderator variable (awar1) and the teacher training intervention (train).

Figure 1.

Conceptual diagram of SEM model.

In addition, we calculated simple slopes in order to investigate further the latent moderation effect with focused attention on the intervention effect on teachers with lower initial knowledge. Specifically, the simple slopes were assessed at three values of the moderator (teacher knowledge construct at pretest: awar1) below, at, and above the baseline average (MODLO, MOD0, and MODHI, respectively). Moreover, we explored the latent interaction effect across the value range of the latent moderator (awar1) beyond its three specific values by producing a series of LOOP plots.

3. Results

As the results presented in Table 2 indicate, our Intervention Model had acceptable fit statistics: Χ²(10) = 15.48, p > 0.05; RMSEA = 0.06 (<0.08); CFI = 0.99 (>0.95); and TLI = 0.97 (>0.95) [31, 33, 34, 35, 36].

Fit indices
Chi-square (df)	15.48 (10), p = 0.12
RMSEA	0.06
CFI	0.99
TLI	0.97

Table 2.

Fit statistics.

RMSEA, root mean square error of approximation; CFI, comparative fit index; TLI, Tucker-Lewis index.

Figure 2 presents the conceptual model for the Intervention Model with parameter estimates from the Mplus outputs. The CPU teacher training (train) was effective in improving teacher knowledge (awar2) (β = 0.51, p < 0.01). There was a significant positive effect of teacher prior knowledge (awar1) on teacher knowledge at the posttest (awar2) (β = 0.54, p < 0.05).

Figure 2.

Parameter estimates for SEM model.

On the other hand, we detected insignificant interaction/moderation effect (awar1tr) with β = −0.20, p = 0.40. To understand more detailed implications of the moderation/interaction effect, we reviewed simple slope results, which were estimated in the sections of the New/Additional Parameters in Table 3. We identified significant effects of the teacher training on improving teachers’ CSA knowledge at three specific values: below, at, and above the mean (β = 0.59, p = 0.00, β = 0.51, p = 0.00, β = 0.44, p = 0.00, respectively). Among these three estimates for simple slopes, we found a greater magnitude at the lower moderator value (MODLO) compared with two other values (MOD0 and MODHI).

	Estimate	S.E.	p
AWAR1 BY
AWAR1_1	1.00	0.00	999.00
AWAR1_2	1.17	0.18	0.00
AWAR1_3	1.41	0.33	0.00
AWAR 2 BY
AWAR2_1	1.00	0.00	999.00
AWAR2_2	1.54	0.27	0.00
AWAR2_3	1.00	0.11	0.00
AWAR 2 ON
AWAR1	0.54	0.27	0.04
AWAR1TR	−0.20	0.24	0.40
TRAIN	0.51	0.09	0.00
New/additional parameters
MODLO	0.59	0.14	0.00
MOD0	0.51	0.09	0.00
MODHI	0.44	0.11	0.00

Table 3.

Parameter estimates for SEM model.

To visually represent how the teacher training effect was moderated by teacher knowledge at the pretest, we produced LOOP Plots. Figure 3 shows the range of the moderator values (awar1) along the x-axis for which the effect of teacher training (train) is significant. The straight line shown in red represents the estimated moderator function and the blue curves represent the confidence intervals. As the graph shows, the intervention effect has a greater positive value with decreasing values of baseline teacher knowledge. This suggests that the effect of the teacher training tended to be more pronounced among teachers with lower baseline scores, as would be expected.

Figure 3.

Estimated moderator function and confidence interval.

In Figure 4, the bold blue line (TX1) displays the estimate for preventive knowledge average at the posttest among the intervention group teachers, with the bold red line (TX0) representing that among the control group teachers. The effect of the teacher training denotes the gap between TX1 and TX0. The gap widens at the lower baseline scores, while narrowing at the higher baseline scores. Corresponding to those estimates for the above mentioned simple slopes, LOOP Plot results revealed that the CPU teacher training was more effective among teachers with lower CSA awareness at pretest.

Figure 4.

The posttest (teacher knowledge) means as a function of baseline for the intervention and the control groups.

4. Significance

Given the prevalence and serious adverse outcomes of CSA persistently observed despite international efforts such as the declaration of the Rights of the Child by the United Nations [37], we sought to evaluate the effectiveness of efforts to prevent and intervene to ameliorate the effects of CSA by focusing on teachers as they not only have great accessibility to children but also expertise in child education/development [13, 20]. To help teachers, who unfortunately have felt unprepared and unsure of how to intervene/protect their students from those who might be abusing them [17, 38], we explored the potential role of teacher training for improving teachers’ preventive outcomes of knowledge.

By analyzing empirical data from an implementation of the CPU, we found a significant effect of CPU teacher training on improving teachers’ preventive knowledge. These results are in accordance with those reported in prior studies of the positive association between teacher training and teacher outcomes such as CSA knowledge, attitudes, and confidence [15, 21]. Crucially, we found the effect of CPU teacher training was more pronounced for teachers with the lower baseline scores in terms of teacher knowledge. Our findings based on empirical data for elementary school teachers from the years 2017–2018 contribute to the field of CSA prevention programs by addressing concerns that existing evaluation studies of CSA prevention programs in this fast-moving field have become outdated [39, 40, 41].

Given these findings, we urge administrators and policy-makers to devote more funding to efforts to develop and/or offer more teacher training opportunities, including investing in CPU teacher training. Our results show that these teacher training programs would possibly help teachers confront this uncomfortable issue and deal with the consequences, particularly those who started with relatively low levels of CSA knowledge. This is important as it may empower teachers to respond and intervene more effectively.

To respond to the concerns teachers expressed regarding their lack of knowledge and competency, despite their critical role for CSA prevention/intervention efforts [15, 17], we sought to strengthen teachers’ individual preventive outcomes via teacher training. Also, our other studies showed the potential effects of teacher training on other preventive outcomes for teachers as well as students’ preventive outcomes [42, 43]. However, one limitation of our approach is that we only included educators working in schools. Given CSA occurrence in various organizations, such as sporting institutes and religious organizations, importantly, professionals, who are responsible for child-care and build rapport with the children across varying institutions, need to be equipped with an awareness of CSA signals and symptoms and knowledge to report suspected CSA cases [44, 45, 46]. Future studies should extend the findings of this study to develop and/or improve training that will empower child-care professionals across all such organizations with preventative knowledge to protect children from the threat of CSA. Providing networking opportunities via the training sessions for these child-care professionals could also contribute by building a community of child-care professionals where these professionals work together to develop more comprehensive prevention solutions.

Acknowledgments

This work was supported by the Committee for Children, Seattle, WA.

Conflict of interest

The authors declare no conflict of interest. Dr. Tia Kim is employed by the Committee for Children.