Why Do School Staff Sometimes Fail to Report Potential Victimization Cases? A Mixed-methods Study

Schools are in a position to connect children and adolescents suspected of being victims of violence with an external source of support by making referrals to external agencies. However, several studies have identified obstacles that hinder early reporting among school staff members. Very few studies have applied a mixed-method approach to try to understand this sensitive issue. The current study used this approach to analyze to what extent the students suspected of being victimized match the ones reported by active school staff members in Spain (n = 453, 83.5% females, age: M = 42.23, SD = 9.46). We classified the reasons given for not reporting the potential victimization cases encountered and made comparisons to determine whether there were differences in the level of knowledge, or in the sociodemographic characteristics, of respondents who gave different reasons for not reporting. Although 73.5% of school staff members had detected at least one potential case, 40.8% of them referred it to an external agency. The most common reasons for lack of reporting included deciding not to do so once concerns had been shared within the school and believing that one must be certain or that only serious violence should be reported. The findings of this study may help to further understand the decisions not to report certain suspicions of potential victimization cases to external agencies by school staff. There is an urgent need to raise awareness about the duty to report these concerns to external agencies, even in the absence of agreement from the school management team. Members of school staff need to be strongly encouraged to become familiar with the existing protocols.


Introduction
With the ultimate aim of protecting children, in many locations (such as the one where this study was performed) adults are obliged to report to the authorities not only victimizations that they are directly aware of but also any suspicions they may have that a child is at risk of suffering interpersonal violence. Several authors have proposed that reporting potential cases of victimization (as just defined) to specialized agencies such as social services is the best way to connect victims with a source of help (Mathews & Bross, 2008;Wekerle, 2013). Nevertheless, most people tend not to report concerns to these agencies because of misconceptions, like thinking that a child would be automatically removed from home if they were being maltreated (Walsh & Jones, 2016). This has been confirmed by professionals working with children on an everyday basis (Feng et al., 2010).
Within this context, schools have been defined as one of the governments' main tools to ensure that children's rights are upheld (McGarry & Buckley, 2013) by making the corresponding authorities aware of children at risk of suffering interpersonal violence. However, the education sector tends to present very low rates of reporting (Dinehart & Kenny, 2015) or fails to report suspicions properly (King & Scott, 2014). Several studies performed in school settings reported that teachers were able to identify potential victims of violence but not able to respond effectively (Gilbert et al., 2009). Victims of different types of violence expressed regret that their teachers had not tried to reach out to them more (Buckley et al., 2007). Some authors have proposed that unseen victims, who are actually detected but receive no official action, tend to develop more complex traumas (Smyth et al., 2012) as they need to cope not only with the victimization itself but with the silence and complicity of society, and maybe their loved ones (Münzer et al., 2016). This also makes victims more likely not to disclose their experience nor ask for help (McElvaney et al., 2014). We thus need further understanding of the response given to potential cases of child and youth victimization encountered at school (Gilbert et al., 2009).
So why are school staff members finding it so hard to report children that they suspect to be at risk of violence? As a recent review has highlighted (Alazri & Hanna, 2020), it seems that different variables in the reporter (e.g., lack of knowledge), the victims (e.g., type of victim's family), or the system (e.g., school setting) play a role. Other factors reported in the literature have been the link and communication between social services and schools (Nohilly, 2019), and the challenges regarding the definition of what constitutes a suspicion (Crowell & Levi, 2012).
When addressing such a sensitive topic, some authors have suggested that traditional quantitative research may be somehow unable to capture the meanings that people attach to complex phenomena (Collingridge & Gantt, 2019). Silber et al. (2013) found that open-ended questions led to higher response rates than closed-ended questions when participants were asked about sensitive issues. The authors proposed that respondents may not find a response option that is in line with what they believe can explain their behavior when the answers are restricted to multiple choices. The consequential loss of significant information and quality of the responses given may affect the conclusions of research targeting the detection and early reporting of child and youth victimization.
Some qualitative studies have addressed this issue with different types of school staff members, such as teachers (Falkiner et al., 2017), counselors (Jenkins & Palmer, 2012), and social school workers (Weegar & Romano, 2019). These studies have reported findings in line with quantitative results, like a tendency to believe that there is a need for certainty before making a referral (Falkiner et al., 2017) and a lack of awareness about the reporting protocols (Jenkins & Palmer, 2012).
Nevertheless, scientific literature mixing up these approaches in order to gain insights integrating both qualitative and quantitative types of data is scarce. Mixed-methods designs have been widely used to address similar reporting issues in the health sector (e.g., Feng et al., 2010) but few have addressed the reporting of potential victims of violence in the school sector (Bryant & Baldwin, 2010). Evidently, new studies using this methodology could confirm, question, or enrich the conclusions obtained from previously conducted research.
The present study aimed to apply a mixed-methods approach to explore to what extent detected cases match cases reported by school staff members and the reasons behind deciding not to report the potential victimization cases encountered. This issue is even more relevant in a context in which all adults (but especially those working with children on an everyday basis) are required to report situations of potential risk of maltreatment or neglect to the authorities, Journal of Interpersonal Violence as mandated by the Minors' Legal Protection Act of 1996 and the Modification of the Child Protection Services Act of 2015. This is why in our context the term "cases" is used to refer to situations that adults are obliged to report; that is, both when there is actual evidence of violence, and when there is a suspicion. More specifically, we wished to study the reasons that school staff give to explain their behavior when choosing not to communicate a potential case to an agency outside school. Since both quantitative (Alazri & Hanna, 2020) and qualitative studies (Falkiner et al., 2017) have proposed that the level of knowledge regarding child and youth victimization, its detection and the procedures to report it differ among respondents with different reasons for reporting, we also wanted to check whether the level of knowledge differs across respondents with reasons for not reporting their suspicions. Finally, we tested whether some of the respondents' characteristics, such as gender or years of work experience, had any influence on the reasons given for lack of reporting.

Sample
A total of 459 respondents returned or submitted questionnaires. Two responses were duplicated, and so were removed from the dataset, and 4 participants with over 50% of missing data were excluded, based on previous recommendations (Johansson & Karlsson, 2013). Thus, the final sample comprised 453 school staff members (83.53% females) between 22 and 65 years old (M = 42.23, SD = 9.46). Years of experience working with children ranged from 0 to 48 years (M = 17.25, SD = 9.94).
The respondents' distribution by demographic and professional variables is shown in Table 1.

Procedure
All procedures were conducted in accordance with the basic ethical principles of the Declaration of Helsinki (World Medical Assembly, 2013), respecting the ethical standards drawn up by the university's Committees on Bioethics (IRB00003099). All participants were informed of the aims and conditions of the study by means of a brief written invitation before consenting to respond to the survey. They were also told that they could withdraw from the study at any point. It was emphasized that participation was voluntary and that all data would remain confidential and accessible only to the research team. Other ethical aspects considered are detailed later in the descriptions of the data collection procedures in the two participating cities, Barcelona and Santander. In Santander, schools were contacted through the governmental agency in charge of training active teachers in February 2018. The invitation summarizing the purposes and conditions of the study was sent by e-mail directly to potential participants registered in the agency, including a link to fill in the online version of the questionnaire. Participants were also informed that they had a month to complete the survey and that they could contact the governmental agency if they had any questions or wished to make any comments, or if they wished to withdraw from the study. In Barcelona, schools were contacted by phone in February 2016. The study was presented to a referent person by the first author, discussing aims, conditions, and compensations personally or by phone. Following previous suggestions by other authors (Hardesty et al., 2019), flexibility along the procedure was allowed within the defined research structure, to adapt properly to the school preferences and value participants' time and collaboration. Each designated school contact chose whether they preferred a printed or an online survey (49.45% of participants used the online format) and in which language (Spanish or Catalan: 79.35% chose to answer in Catalan). Deadlines and locations for data collection were agreed upon with the designated school contact. Participants using the online version were proposed to answer the questionnaire in their spare Journal of Interpersonal Violence time and in the location of their choice, whereas schools that chose the printed version either left copies in the teachers' room or proposed a concrete space and time to fill in the surveys. All school staff members were warned about the deadlines for completing the questionnaires and provided with a contact phone-line and e-mail address, in case they had any questions or wished to make any comments, or if they wished to withdraw from the study. In return for their collaboration, schools were entitled to a report and a training session on child and youth victimization for all staff members.

Instrument
Based on previous studies with similar aims and drawing on current official conventions and protocols, a self-administered questionnaire was designed. The full protocol is available in its original language version at Greco et al. (2018). The questionnaire included a definition of victimization based on Finkelhor's framework (2007) but also adapted to the mandatory requirements in the context of this research. The textual definition was "potential or actual harm (psychological or physical) caused by the intentional behavior (whether by action or omission) of individuals or groups of individuals towards someone younger than 18 years old, which interferes or might interfere with their optimal development in the short or the long-term." The questionnaire also included 45 items measuring knowledge and experience regarding child and youth victimization, its detection, and its reporting. "Knowledge" was conceived as beliefs that are correct and justified with some degree of confidence, and which influence behavior, as they are also linked to the capacity of the individual to act in situations related to these beliefs (Hunt, 2003). In the framework of this research, "experience" was defined as the personal event that took place in the individual's life when confronted with the possibility of reporting a student at risk of victimization. The instrument went through a multi-stage pre-test process involving a focus group with target population (n = 8 teachers) to ensure comprehensibility, clarity, and relevance; cognitive interviews with experienced professionals (n = 5 psychologists or teachers specializing in vulnerable students) and reviews by two academic experts to assess content validity. Further information on the pretesting and previous studies performed with the instrument is available in Greco et al. (2017).

Detection of potential cases.
Participants were asked to answer the question "how many times during your career did you suspect that a minor might be being victimized?" by choosing among 5 response options that varied between "never" and "over 20 times," including the option "I do not know."

Reporting experience.
Participants answered the question "have you ever reported a child suspected to be a victim of violence to external agencies outside school (e.g., Social Services)?" by choosing among three response options (i.e., "yes," "no" and "I never had any suspicions").

Perception of the intervention in reported cases.
Only participants who had answered "yes" to the previous item were asked if they knew how the case or most cases had proceeded and if they believed the intervention had resulted in a good outcome for the minor.

Reasons for lack of reporting.
Only participants who had answered "no" to the item about reporting experience were asked to answer the open-ended question "if you answered no, please describe briefly why you did not report your suspicions?" This item was designed in an open-ended format because this format has been shown to elicit more socially undesirable answers to threatening questions about sensitive issues than closed questions (Singer & Couper, 2017).

Knowledge.
Knowledge was assessed using 10 statements about victimization (e.g., "Victimization affects less than 10% of children in Spain"), 10 statements about detection (e.g., "Most of the signs regarding child abuse are directly observable"), and 10 statements about reporting (e.g., "Reporting a suspicion is legally mandated in Spain"). References for each knowledge statement are provided as supplemental material. Participants responded to each statement by choosing between "yes," "no" or "I don't know." Every statement correctly classified was awarded 1 point and 0 points were given for wrong or "I don't know" responses. A total score of 10 points could be obtained for each section (i.e., victimization, detection, and reporting). Each statement was based on current official protocols (Ajuntament de Barcelona, 2007;Ministerio de Educación, Cultura y Deporte, 2006) and on previous projects with similar aims in Spain (Cerezo & Pons-Salvador, 2004). The design was based on previous research performed in the international context (Dinehart & Kenny, 2015;Mathews et al., 2017;Walsh & Jones, 2016).

Other variables.
Sociodemographic and professional information was also gathered (i.e., gender, age, whether if they were working in kindergarten, elementary or highschool level, and years of experience working with minors). Journal of Interpersonal Violence Data Analysis Missing data. Before running the analysis, we found that 2.11% of data were missing due to non-response, varying between .02 and .07% missing data per variable. We then explored the missing data pattern using Little's Missing Completely at Random test (MCAR), which was not significant (p > .05), suggesting that the pattern was MCAR. Thus, we chose the imputation method of fully conditional specification, since it has been shown to perform satisfactorily with our amount and type of missing data (Johansson & Karlsson, 2013). Data were imputed through the mice package (Buuren & Groothuis-Oudshoorn, 2011) in R (R Core Team, 2019) and the imputed dataset was compared to the original one through plots, tables, and chi-square comparisons, finding no statistically significant differences.

Quantitative analysis.
To address our first research question (i.e., to what extent detected cases match cases reported by school staff members), we obtained descriptive statistics for the experience of detection and reporting of potential victimization cases for all respondents (n = 453). Then, we focused on respondents that had had suspicions but had never referred a case outside school and responded to the item that asked the reasons why (n = 124). We analyzed the reasons why they chose not to report the case (refer to the section "Qualitative Analysis"). Once the responses were classified, we obtained all the descriptive statistics. To address the third and fourth research question (i.e., whether the level of knowledge or the sociodemographic characteristics differed among respondents with different reasons for lack of reporting) we compared the level of knowledge of respondents in each category through global and pairwise comparisons, using the Kruskal-Wallis test, based on recommendations for our type of study and sample (Lantz, 2013). In order to compare the sociodemographic characteristics of the respondents for each category, we used the Chisquare test for frequencies and Kruskal-Wallis test for age and years of experience. These comparisons would also provide evidence regarding the criterion validity of the category system created by means of the qualitative analysis. All statistical analyses were performed in R (R Core Team, 2019).

Qualitative analysis.
In order to address our second research question (i.e., what are the reasons behind deciding not to report the potential victimization cases encountered), we created a system to categorize all the answers about why a suspicion had not been reported outside school (n = 124), based on the principals of conventional content analysis. Considering the instrument used, we relied on the guidelines developed by Singer and Couper (2017). To create the system of categories, the first author (AG) read all the responses using open coding to propose a first draft of a system with 7 descriptive categories, with 12 subcategories to be more specific when the data allowed. Another author (EGP) checked the categories, reviewed the classification, and proposed an extra category, as well as five subcategories. Both authors discussed the categories system and reached an agreement with six main categories and nine subcategories, considering the integrity of the research and the fit between the research purpose and the qualitative technique used (Hardesty et al., 2019). They agreed that the categories would not be mutually exclusive, that is, an answer could be categorized in more than one category if the content provided enough evidence to do so.
Following this agreed classification system, the three authors (AG, EGP, and NP) independently categorized all responses. With the aim of assessing the reliability of this analysis, we tested the inter-rater degree of agreement for each category through Kappa coefficients of concordance. The inter-rater agreement between authors ranged from substantial (K = .71, CI = .62-.81) to almost perfect agreement (K = .84, CI = .76-.92), which was considered satisfactory. For the answers in which the three coders differed in opinion (n = 2), one was solved by considering the classification of other similar cases and the other was discussed by the whole team until consensus was reached. The system of categories is displayed in Table 2 and the classification of all answers is available upon request. Since the categories were not mutually exclusive, we calculated the percentages based on all fragments classified (n = 145).

Detection and Reporting of Potential Victimization Cases
With regard to our first research question (i.e., to what extent the cases detected matched the cases reported to external agencies), almost three-quarters of the sample (n = 333, 73.5%) reported having at least one suspicion of a potential victimization case during their career. Most of them (80.1%) reported having had between 1 and 10 suspicions, while 7.3% reported having encountered between 11 and 20 cases and 6.8% over 20 potential cases. Approximately 40% (n = 136) of these participants said they reported their suspicions to an external agency outside school. Out of the respondents who said that they knew what had happened to the student following referral (n = 101), the majority considered that the referral had been good for the minor's well-being (39.0%), only 8.8% considered that it had worsened the situation, 22.1% thought it had made no significant difference and 4.4% did not know or felt unsure about it. All these percentages are illustrated in Figure 1. Journal of Interpersonal Violence

Reasons for Lack of Reporting
To answer our second research question regarding the reasons why respondents had not made a referral outside school, we focused only on participants who had not done so (n = 124). As shown in Table 2, respondents most commonly said that they had decided not to make a referral outside the school once they had shared their concerns within the school (42.1% of answers), followed by a feeling that the suspicions needed to satisfy certain criteria, such as being certain or being serious (25.5%). Lack of knowledge, feeling that making a referral outside school was not their responsibility, and reporting inconsistencies (i.e., answering that they had never had a suspicion when they had answered that they had had at least one in the previous item) each represented between 12 and 14% of responses. Other reasons, such as fear of the consequences or intervening in the case in a different way, were identified in 6.2% of the responses. Each of the categories is described in more detail later.  c. It had already been referred "Child welfare services were already aware of the case" (Respondent 266) d. Other people would make the referral "Other people with higher responsibility did it in my place" (Respondent 209) e. We solved the problem within the school framework "It was not considered necessary since the safeguarding team solved it" (Respondent 212) f. We agreed not to take further actions "Because we talked it over within the school and with the family (warning)" (Respondent 119) g. My superiors did not support me "The investigation and application of the child maltreatment protocol was not carried out, because the management team had blocked it" (Respondent 351) Thought the suspicion must satisfy certain criteria a. Being certain "Lack of evidence" (Respondent 241) 37(25.5) b. The case needs to be serious or severe "The case was never as serious as it needed to be for referral" (Respondent 165) Other "High social controversy" (Respondent 355) 8(5.4)

Decided not to report the case outside the school after sharing concerns within school
In over 42% of cases (n = 61), teachers and other school staff members reported sharing their concerns about a suspicion of a potential victimization case with other institutional members before they took the decision to communicate them to an agency outside the school framework. When they did this, several situations may have led them to decide not to report their suspicions to any external services, as described in the following subcategories. a. Reporting to a superior or child safeguarding team. A large proportion of respondents (n = 19) communicated their concerns to a superior, like the headteacher, feeling that from that moment on they would not be responsible for what was decided regarding the potential referral: I referred it to the child safeguarding team, which is the agency that should take the subsequent steps (Respondent 53) I referred it to the school management team, who would report it (Respondent 248) I left the case in the hands of the child safeguarding and management team (Respondent 161) b. Reporting to a colleague. Other responses (n = 6) reflected that participants chose to talk it over with the child's main teacher, and then leave the decision for referral up to them: I referred it to the child's main teacher (Respondent 7) I did not refer the case personally to social services … I referred it to the child's main teacher (…) (Respondent 228) c. The case had already been referred. Some respondents said that they were warned by other members of the school staff that an external service was already aware of the case (n = 7). This made them decide that there was no need to make a referral outside school. Social services were already aware of the case (Respondent 266) The cases were already being treated by social services (Respondent 430) They were children that were already being monitored by an external agency (Respondent 167) d. Other people would do the referral. Some participants either knew or trusted that someone else would make the referral (n = 11), which discouraged them from pursuing it themselves: Other people would take care of it (Respondent 288) Other members of the school staff with greater responsibility would do it in my place (Respondent 209) Most of them said that the referral was made after talking to their colleagues or the child safeguarding team (n = 8): People in the management team did it (Respondent 233) The girl's main teacher made it [the referral] (Respondent 182) e. We solved the problem within the school framework. Answers classified under this subcategory (n = 9) implied that the school proposed a solution to the problem that was considered as a sufficient substitute for reporting, like sharing their concerns with the child's family or having an interview with the primary caregivers. After performing this action, they decided not to report the case elsewhere: It was solved within the school (Respondent 291) We talked with the family (Respondent 188) The school was able to mediate and find a solution (Respondent 75) f. My superiors did not support my initiative. A few respondents (n = 4) explicitly stated that their superiors or child safeguarding team had prevented them from making the referral: Faced with my suspicions, the assessment and application of the maltreatment protocol was not pursued because the management team blocked it (participant 351) The management team did not support me and I did not dare to make it [the referral] on my own (Respondent 291) Some participants even reported that they had had a hard time pursuing or coping with the disagreements they had with the management team or other school staff members: I did not want to deal with my colleagues, superiors or the institution to which I belong and I consider them to be part of the maltreatment, since they do not look after the children's well-being (…) (Respondent 443) After following the protocol and even going beyond my role to comply with it, the management and child safeguarding team did not consider it necessary to inform other agencies (Respondent 418). Figure 2 aims to summarize the pathway followed from initial suspicions to referral, according to the results for this first category.

Thought that suspicion should satisfy certain criteria
The second most frequent reason for not making a referral (25.5%, n = 37) was the belief that the suspicions should satisfy certain criteria in order for a referral to be made to an agency outside the school. a. Certainty. A substantial number of respondents (n = 28) said they had not referred the case outside of school because they were not certain about the victimization actually occurring: I did not have enough signs to be sure of it, it was more a feeling than a certainty (Respondent 175) I was not sure that it was really happening (Respondent 426) I was not completely sure (Respondent 374) Figure 2. Path of suspected cases referred only within the school.
Note. Percentages are calculated based on the answers in each category (n = 64, because some answers were included in 2 subcategories, for example, participant 248).
Grey area represents the school framework.
White arrows represent extremes of the path, that is, ways in which the suspicion enters the circuit and leaves it.
Black arrows represent paths within the school framework, among staff inside the school.
Dotted arrow represents a path that is exclusively from a superior or the safeguarding office.
b. Severity. Another group of respondents (n = 6) stated that they had not made a referral because they felt the potential case was not severe enough to be dealt with by an external agency: I did not consider it was that severe (Respondent 47) The case was never severe enough to report it (Respondent 165) I was never able to detect a severe case (Respondent 68)

Inconsistencies
Quite a few responses (9.7%, n = 14) were not consistent with the answers given to previous items. That is, some respondents reported having had at least one suspicion of victimization during their career but never made a referral. However, when asked the open-ended question about reasons why they did not make a referral, they said that they had never had any suspicions.

Felt it was not their responsibility
Another group of respondents (9%, n = 13) indicated that they did not feel that communicating their suspicion about potential victimization cases to external agencies was their responsibility: It is not my role at school (Respondent 82) I did not think it was my responsibility (Respondent 293) One respondent specified that they felt they were not entitled to make a referral outside of school by themselves I am not entitled to [make a referral] (participant 383).

Lack of knowledge
Some of the answers (8.3%, n = 12) showed that respondents had decided to make a referral but did not know where or to whom the referral should be made.
To whom should we report our suspicions? We talked it over at the school, without knowing that it was victimization (Respondent 32) We did not know to whom [to address the referral] or the protocol. We tried the hospital (Respondent 168) Because of lack of awareness (Respondent 315) Journal of Interpersonal Violence

Other reasons
There were some answers (5.5%, n = 8) that we could not classify into any of the other categories. Some respondents took action by themselves and decided not to report their suspicions afterwards, such as: Some other answers implied broader issues, such as fear of the consequences or feeling that making a referral would enhance social reactions they did not want to deal with: High social controversy (Respondent 444)

Comparison of Respondents According to the Reasons for Not Making a Referral
To assess our third research question, that is, whether participants giving different reasons for not reporting suspicions differed in terms of their sociodemographic characteristics or level of knowledge, in Table 3 we present the descriptive statistics of respondents in each of the categories described. We only used the first category for each respondent, to avoid repeating information about individuals. We can see that respondents giving answers for the category "felt it was not their responsibility" were the youngest respondents on average. Those with answers that belonged to the category "lack of knowledge" were those with the fewest years of experience. On the contrary, respondents giving answers in the category "decided not to report outside once reported within the school" were those with the highest mean age and years of experience. Regarding gender or school level in each category, the distribution was similar to the distribution of the sample (refer to Table 1). None of the differences was found to be statistically significant. Figure 3 shows the average level of knowledge about victimization for respondents in each category. When analyzing knowledge about victimization, respondents in the category "inconsistencies" displayed the highest level of knowledge and those with answers in the category "other" displayed the lowest level of knowledge.  Table 3. Distribution of Respondents According to the Main Category Assigned to their Answer.
Note. Comparisons were made with chi-squared test for frequencies and Kruskal-Wallis test for quantitative variables.
In the same figure, it can be seen that the level of knowledge regarding detection was relatively even across all categories. Interestingly, in terms of knowledge about reporting, respondents that fell into the category "lack of knowledge" displayed the highest level of knowledge on average and those that belonged to the category "felt it was not their responsibility" displayed the lowest.

Discussion
The main aim of this research was to try to determine how many suspected cases of child victimization go unreported by schools, why and which variables might explain this. We also aimed to compare participants who gave different reasons for not reporting in terms of their sociodemographic variables or level of knowledge. In order to use the most suitable methodological approach to address each of these questions, we combined qualitative and quantitative techniques that allowed us to provide new and enriching insights into existing research.
Regarding the first research question (i.e., the extent to which school staff report cases of potential victimization to external agencies), most of the respondents in our sample had suspected at least once that a student under their care might be being victimized. The good news is that the proportion of suspicions of potential victimization cases reported found was slightly higher than in previous studies (Dinehart & Kenny, 2015;Greco et al., 2017). This could mean that awareness about the need for early reporting is actually increasing in school staff. Another piece of good news is that most respondents that followed a reported case thought this decision had benefited the child or youth involved. This is important information as it overturns some myths regarding the potential negative effect of social services interventions (Walsh & Jones, 2016).
Nevertheless, the majority of the suspected cases still go unreported, even in the context of this research in which teachers are obliged to refer any suspicion to the corresponding authorities. When analyzing the reasons that school staff gave to explain this behavior in order to answer our second research question, some of the participants mentioned reasons already cited in previous studies, such as the lack of knowledge (Falkiner et al., 2017;Jenkins & Palmer, 2012), or the belief that one needs certainty before making a referral (Walsh & Jones, 2016). Thus, our research adds evidence to the existence and persistence of these barriers. However, other reasons that have not been mentioned by previous studies also emerged.
Our study showed that, despite efforts to increase awareness around the need for early reporting by schools (Gilbert et al., 2009), some respondents still feel that this is not their responsibility. They felt they were not entitled to make a referral, or that only people with specific roles in schools could make this decision, or that it was not their responsibility (Alazri & Hanna, 2020). Interventions aimed at increasing early reporting of these cases should take this finding into account and include ways to empower school staff members to act on these suspicions.
Our results revealed other reasons why suspected victimization cases were not reported, including that once the respondents had shared their concerns with someone within the school framework, they decided not to pursue the referral any further. This situation has been reported in previous studies (Alazri & Hanna, 2020), but elucidating the particular dynamics that occur in the relationship among different members of the school when sharing these concerns was only possible through qualitative strategies. Figure 2 shows that a suspicion tends to be communicated first either to a colleague (who is usually the main teacher of the student potentially at risk) or to a superior or child safeguarding team. When the safeguarding team, which has the most responsibility for deciding to make a referral, does not support the suspicion, the information does not reach any agency outside school, despite the legal framework that requires adults to communicate these situations. In addition, in some of these disagreements, the respondents felt that their superior or safeguarding team was deliberately allowing the victimization to keep occurring. It is important to increase awareness among school staff that they are entitled (and, in this context, mandated) to make a referral to the corresponding agency as individuals, even when their superiors do not agree with this procedure. Of course, a unilateral disclosure to a third party without the agreement of the school may be harder to deal with, as issues of confidence may arise (Bryant & Baldwin, 2010). However, school staff must rely on their compliance with the legal and ethical framework to pursue the referral if they consider that someone might be in danger.
Another common pathway shown in Figure 3 is that the school team proposes an action in order to address the potential victimization. Believing that a single (and according to the descriptions analyzed, quite simple) action will be sufficient to solve a complex problem such as students potentially exposed to violence entails many dangers, especially if it prevents children suspected to be at risk from reaching a source of help (Wekerle, 2013). But it could also discourage victims from asking for help (McElvaney et al., 2014), if the consequences of the intervention proposed by the school staff (e.g., talking with the family) causes the risk of further victimization for the children.
The remaining pathways lead to making a referral to an external agency. However, the individual who first suspected the victimization may always add information to both an already reported case or a case that will be reported by someone else from school. In this sense, it is important to strongly encourage all school staff members to be familiar with the protocol (Bryant & Baldwin, 2010) and to get involved in the referral procedure as much as they can. More fluid communication between schools and agencies in charge of child protection would also be useful (Nohilly, 2019). Simultaneously, school staff should also offer help directly to the victim, as evidence suggests that victims would like more support from schools (Buckley et al., 2007). Finally, in a context in which school staff is mandated to report suspicions about children at risk, it is important to recall that governmental agencies are responsible for ensuring the access of this population to the current protocols and conventions.
Finding that a considerable number of school staff members answered inconsistently might be related to the discrepancies about what constitutes a reasonable suspicion (Crowell & Levi, 2012). Further research using a similar approach may shed light on how school staff members classify a child as potentially at risk.
Lastly, there were also some unexpected results regarding our third research question, that is, the contrast between participants with different reasons for reporting in terms of their sociodemographic variables and level of knowledge. For instance, it was interesting that respondents in the category "lack of knowledge" were found to score the highest in terms of knowledge about reporting. This seems to suggest that even though knowledge is accurate, other factors may discourage school staff members from making a referral (Nohilly, 2019). Ways of coping with insecurity or perceptions of low self-efficacy need to be found to encourage the early reporting of students suspected to be at risk of victimization. In terms of the validity of our categorization, it is also important to point out that participants alleging a lack of knowledge were the least experienced, and respondents who felt that it was not their responsibility were the ones with the lowest levels of reporting knowledge.
Even though our study followed rigorous guidelines on both quantitative (Lantz, 2013) and qualitative analysis (Stenius et al., 2017) and responded to the call for more studies using these techniques (Alazri & Hanna, 2020), some limitations are worth considering. First, it must be borne in mind that the generalizability of qualitative findings depends on the context in which a phenomenon takes place, according to the proximal similarity model (Collingridge & Gantt, 2019). This is why we provided a detailed description of the situation in which this study took place, based on the model's gradient of similarities criterion (i.e., time, place, people, and setting), so that future research may use this information to assess to what extent the findings are likely to be replicated in another context. Second, the fact that the respondents were from different regions and were recruited using different procedures may hide an effect of the context. However, although the samples came from two different geographical areas and the recruitment procedures Journal of Interpersonal Violence differed, no differences were found in terms of sociodemographic characteristics when comparing subjects in the different groups. Third, despite the similarities between our sample and school staff populations, the fact that the respondents were mostly teachers may have limited the views that were recorded in our study. Research including greater representation of other types of school workers may reach complementary conclusions. Finally, probably the most important limitation is that we did not record details about the family or socioeconomic background of children suspected to be at risk, an aspect that might also influence the reasons for not reporting. Our study focused on school staff variables so it is important to consider that excluding students' characteristics like age or ethnicity might have hidden effects in reporting. Future studies might include a consideration of the influence of these characteristics on the decision not to report.

Conclusions
A considerable amount of suspected cases of child or adolescent victimization detected at schools is not being reported to external agencies. Most school staff that reported and followed up their suspicions perceived that the intervention was beneficial for the potential victim. School staff gave diverse reasons for failing to report a suspected case, such as being discouraged to pursue referral once their concerns had been shared with other members of the school staff, not feeling entitled to make a referral on their own, not knowing how to proceed or thinking that their report had to fulfill certain criteria (i.e., certainty, seriousness). All school staff should be encouraged to become familiar with the reporting requirements and procedures, as well as to act on suspicions of students at risk of victimization, even if they feel unsure. Being aware of the internal school dynamics that might discourage staff from reporting and addressing misbeliefs could be an efficient way to promote the use of early reporting as a source of help.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors gratefully acknowledge the support from Agency for Management of University and Research Grants and by the European Social Fund [Case file: 2018FI_B 00638] and the school staff members who actively participated in this research.