Factors associated with the decision to investigate child protective services referrals: A systematic review

Correspondence Jeri L. Damman, PhD, MSc, Lecturer, Department of Social Work and Social Care, University of Sussex, Brighton BN1 9RH, UK. Email: j.l.damman@sussex.ac.uk Abstract Limited resources for child protection create challenging decision situations for child protective services (CPS) workers at the point of intake. A body of research has examined the factors associated with worker decisions and processes using a variety of methodological approaches to gain knowledge on decision-making. However, few attempts have been made to systematically review this literature. As part of a larger project on decision-making at intake, this systematic review addressed the question of the factors associated with worker decisions to investigate alleged maltreatment referrals. Quantitative studies that examined factors associated with screening decisions in CPS practice settings were included in the review. Database and other search methods were used to identify research published in English over a 35-year period (1980–2015). Of 1,147 identified sources, 18 studies were selected for full data extraction. The studies were conducted in the United States, Canada, and Sweden and varied in methodological quality. Most studies examined case factors with few studies examining other domains. To inform CPS policy and practice, additional research is needed to examine the relationships between decision-making factors and case outcomes. Greater attention needs to be given to the organizational and external factors that influence decision-making.

making has become an expansive body of work providing important knowledge across decision points, including screening referrals at intake for further investigation. As part of a larger study on decisionmaking, this systematic review addressed the question of the factors associated with worker decisions to investigate alleged reports of child maltreatment in quantitative studies over a 35-year period. This paper provides overviews of the CPS decision-making process at the point of referral and the approach to the empirical study of decisions at this point in the child welfare continuum. The Decision-Making Ecology (DME; Baumann, Kern, & Fluke, 1997;Baumann, Dalgleish, Fluke, & Kern, 2011) is introduced as the review's conceptual framework, followed by the systematic review. The paper concludes with a discussion of findings and implications for practice and research.

| BACKGROUND
When a report of alleged maltreatment is made to a child protection agency, the typical process that CPS workers engage in involves the collection of available information about the alleged victim, the perpetrator of abuse/neglect, and the presenting concern. The decision to assign the referral for investigation (i.e., the screening decision) is generally made based on whether concerns meet definitional or threshold criteria as described by policy to warrant investigation (Child Welfare Information Gateway [CWIG], 2016). If a referral meets the criteria and is 'screened in', an assessment of risk then determines the response priority (CWIG, 2017;Steen & Duran, 2014). Reports may be 'screened out' at this stage if referral information is insufficient or if another authority is deemed most appropriate to respond to concerns (Children's Bureau, 2019). After screening in a report, an investigation is undertaken, and subsequent decisions are made by CPS including the decision to substantiate or remove the child to a place of safety.
Where a differential, or alternative, response procedure is in place at intake, referrals that do not meet statutory definitions or thresholds may be offered voluntary services as an alternative to being screened out (Children's Bureau, 2019). Screened-in referrals may be diverted through alternative response procedures when a worker determines that voluntary services are safe and appropriate, or they may be diverted from services when a worker determines that engagement with the agency is no longer warranted (Morly & Kaplan, 2011).
CPS agencies typically manage a high number of referrals with a smaller proportion resulting in investigation though geographical variation exists (Gilbert et al., 2009). In the United States, for example, an Screening decisions present particular challenges for workers, such as managing a high volume of referrals containing limited, vague, or conflicting information (Cross et al., 2014;Saltiel, 2015). Intake workers must also filter referrals within relatively tight timescales and determine if they meet the required maltreatment definitions. Additionally, screening for the likelihood of maltreatment to correctly determine if the investigation is warranted can be challenging given the broad definitions of maltreatment, safeguarding, or significant harm and the potentially complex and changeable circumstances of families (Saltiel, 2015). System pressures and intense public scrutiny present further challenges for workers attempting to ensure the correct decision is made (Munro, 1996). Finally, the use of differential response has additional implications for defining screening decision thresholds (CWIG, 2014).

| CURRENT APPROACHES TO CHILD WELFARE DECISION-MAKING
Child welfare practice and research related to decision-making has historically focussed on building knowledge to improve the accuracy of worker decisions. The two major approaches that have been applied to achieve this goal can be described as technical and cognitive. Technical approaches, borne from the broader trends towards more technocratic and managerial practices in child welfare (Lees, Meyer, & Rafferty, 2011), focus on identifying variables or factors in a CPS referral that predict the most accurate decision. This approach values consistency and standardization acquired through the use of instruments and tools to navigate and support worker decisions. Cognitive approaches focus less on factors and more on the process of decision-making and the use of professional judgments. This approach values knowledge and expertise acquired through experience and the importance of judgement in dynamically complex, real-life situations.

| Technical approaches to decision-making
Increased attention to CPS screening decisions has intensified interest in technical approaches to improve decision-making (Cull, Rzepnicki, O'Day, & Epstein, 2013). Influenced by assessment tools in psychology, the field of child welfare embraced the notion of assessment tools to support worker judgements in the 1980s (Shlonsky & Wagner, 2005). By the early 1990s, formal risk assessment instruments and frameworks (Department of Health, 2000;Hughes & Rycusa, 2006) were under development in child protection systems internationally.
More recently, the use of technical approaches in the screening process has become increasingly standardized (Akin, McDonald, & Tullis, 2010) with the narrowing of models used by the field, and the increasing popularity of Structured Decision Making ® (Brown & Packard, 2012;Children's Bureau, 2019). Interest has also grown in the use of predictive analytics and predictive risk modelling (London Councils, no date, n.d.; Packard, 2016) with the use of algorithms to inform screening decisions at intake. For example, in the United States, predictive analytics were featured heavily in recommendations by the Commission to Eliminate Child Abuse and Neglect Fatalities (2016)) to prevent the most severe forms of maltreatment.

| Cognitive approaches to decision-making
In contrast to the technical approaches described, the Munro Review in England emphasized the importance of professional judgements over prescriptive practices and the importance of reflection to strengthen expertise in the field (Munro, 2011). This movement towards enhanced expertise is also evident in the United States as states are increasingly adopting centralized intake units (Children's Bureau, 2019) to enhance expertise in referral intake and screening and address variation in screening practices.
Alongside these developments, efforts have been underway to better understand how workers make effective decisions, partly to inform the design of strategies to cultivate expertise among decisionmakers. Work by Klein, 1998 andKlein et al. (1993) on the naturalistic decision-making framework and its recognition-primed decisionmaking model (RPD) has been recently applied to child welfare decision-making (Platt & Turney, 2014;Whittaker, 2018). According to RPD, decision-making is a process that relies on past experience and includes the use of intuition to evaluate situations, mental simulation to predict the future or explain the past, metaphor to draw on and compare experiences, and storytelling to consolidate experiences (Klein, 1998). ShadowBox ® training, which is designed to facilitate expertise in decision-making through the use of scenarios and reflective rationale, has been piloted with CPS workers (Newsome & Klein, 2017;Newsome, Wright, Klein, Flory, & Baker, 2015;Whittaker, 2018). Early findings suggest differences in the ways in which decisions are made between experienced and novice decisionmakers, with an overreliance on procedure among the latter group.
Such differences in cognitive processes are important given the context of CPS workplaces where turnover is high, and reliance on procedures is high in efforts to standardize decision-making processes.
Cognitive approaches to decision-making also contribute to a more nuanced and critical understanding of child welfare decision-making.
Opportunities for reflection on how and why decisions are made provide insight into current conceptualisations of child maltreatment and awareness of the ethical dilemmas associated with decision-making.

| CONCEPTUAL FRAMEWORK FOR CHILD WELFARE DECISION-MAKING
The DME is a useful organizing framework as it integrates both technical and cognitive approaches through its focus on factors (DME framework) and cognitive processes (General Assessment and Decision Making [GADM] model). The DME model was first presented in the late 1990s based on early screening research . In the DME framework, case, decision-maker, organizational, and external factors are hypothesized to inform decisions and their associated outcomes ( Figure 1). The DME later incorporated the GADM model, to include the cognitive process of decision-making (Baumann et al., 2011). Together, the DME and the GADM provide a framework for understanding how human decisions are made in child F I G U R E 1 Decision-making ecology framework [Colour figure can be viewed at wileyonlinelibrary.com] welfare. The DME '… takes human error as the starting point for understanding decision-making and suggests that decisions need to be understood within their context' (Baumann et al., 2011, p. 4).
The categorization of technical and cognitive approaches provides a useful structure to begin examining the expansive and diverse body of research on child welfare decision-making. Research on child welfare decision-making typically addresses decision points and processes related to the maltreatment response (e.g., investigation, substantiation, and service referral), child placement (e.g., removal and foster care), and case outcomes (e.g., permanency and parental rights termination), with an evidence base more developed at some decision points than others. A recent systematic review was conducted by the National Institute for Health and Care Excellence (2017) in England to identify research published in 2000-2016 on effective tools in recognizing maltreatment at the screening decision point. This latter review identified only three studies, all of which were poor quality, and highlighted a general lack of robust research to inform screening practices. Another systematic review by Lauritzen, Vis, and Fossum (2018) summarized the factors associated with multiple CPS decisions (i.e., decisions to investigate, dismiss, substantiate, and refer families for services) from studies published since 2005 that employed a broad array of methodologies. Over the past 35 years, several studies have been undertaken to build knowledge to inform technical approaches to decision-making at the point of CPS intake. However, to date, no systematic reviews have specifically examined factors associated with worker decisions to investigate referrals of alleged maltreatment. The current review explores factors identified in quantitative studies of CPS decision-making. A future, yet to be completed, review of qualitative studies will address the cognitive processes associated with decision-making. This review sought to contribute knowledge of the methodological approaches and findings that have been generated in relevant studies over time. For purposes of this systematic review, we use the DME as a conceptual framework to organize and analyse factors associated with CPS screening decisions.

| METHODS
The study methodology was guided by approaches described in Bronson and Davis (2012) and Littell, Corcoran, and Pillai (2008) and informed by guidance established by the Cochrane Collaboration (2012) and Campbell Collaboration (2014). The research team developed a protocol consistent with PRISMA standards (Moher et al., 2015), which is available upon request.

| Eligibility criteria
Eligible studies included the quantitative examinations of worker decisions to refer alleged maltreatment reports for investigation using data drawn from actual (vs. hypothetical) decision-making settings. Vignettes, or other hypothetical scenarios were excluded.
Although vignette methodologies offer important contributions to the literature that should not be overlooked, they have been found to have limited transferability to actual decision-making (e.g., Evans et al., 2015). Only sources published in English were included in the review due to resource constraints. Studies included in the review spanned a 35-year period from January 1980 to January 2015. This time period was selected due to potential challenges in accessing studies prior to 1980.

| Information sources
The search strategy included bibliographic database searches, a grey literature search, hand searches of selected journals, and forward and backward reviews of selected sources. Additional sources were identi-

| Search terms
The terms used to search the bibliographic databases included 'child' and 'decision', as well as four terms related to maltreatment, three terms related to research, and 11 terms related to screening ( Figure 2). This strategy was further developed to include selected Boolean operators, which were then adapted for each database.

| Data management and selection process
Bibliographic software EndNote X7 was used to collect, de-duplicate, organize, and manage references and full text documents. After data collection was completed, a two-stage screening process was undertaken to identify relevant sources for study inclusion. The first stage involved the application of study inclusion criteria to the review of titles and abstracts. The second stage involved applying the same criteria to full text documents selected for inclusion as result of the title and abstract review process. Each phase of the screening process was conducted independently by two raters. Discrepancies were resolved by a third rater. The number of discrepancies identified at each phase was minimal, indicating a high level of interrater reliability.

| Data extraction
Sources that met study inclusion criteria were subject to data extraction using a form developed by the research team. Information was systematically collected about study sample characteristics, worker characteristics, study methods, and study findings using the DME framework. The study quality was assessed by the research team using established criteria that evaluated the domains of study design, sampling, and internal and external validity. Two raters independently extracted data from the majority of selected sources. Due to staff changes at the end of the study period, a single rater extracted data from the final set of sources. Raters entered data into a single database for comparison and subsequent analysis. No discrepancies were identified in the data collection process. Data extraction included an assessment of quality based on internal and external validity.

| FINDINGS
After de-duplication, 1,147 sources were identified for inclusion in the review. Of these, 121 were included for further screening after the title and abstract review. Of the 121 sources reviewed, 22 were subject to full data extraction. Of these 22 sources, two were excluded as they contained duplicate findings, and two others were excluded based on the study criteria. Figure
The majority of studies explored case factors related to the referral, child, caregiver, and family. The remaining factor domains received less attention with organizational factors examined in seven studies and decision-maker and external factors examined in only one study.
Many studies examined CPS reports for all maltreatment types, though Carlson (1988) and Silva (2011) examined decision-making for only physical abuse and child sexual abuse cases, respectively. Three studies collected data on all maltreatment types and then conducted separate analyses for each form of maltreatment (Gilbert et al., 1997;Karski, 1999;Schwab et al., 1997). Most studies employed multivariate statistical analysis and one study used bivariate analysis (Carlson, 1988).
Final sample sizes ranged from 100 to 2,905 children or child referrals, with the exception of Steen and Duran (2014) where U.S. states (n = 44) were the unit of analysis. Four studies used single sites (Carlson, 1988;Gilbert et al., 1997;Karski, 1999;Silva, 2011), and the remaining studies involved multiple sites, ranging from two to 12.

| Major findings
Across studies, 51% to 68% of reports were screened in for investigation. Statistically significant (P < 0.05) factors associated with the decision to investigate and the decision not to investigate were extracted from the 18 studies, organized by maltreatment type, and categorized  Schwab et al. (1997) Identifies variables that predict, and patterns in, decision-making

| Case factors
Factors in the case domain included those related to the referral, child, caregiver, and family. Referral factors (Table 2) included information on the reporter, CPS report, alleged perpetrator, and nature of allegation. Child factors (Table 3) included number of victims, demographics, and child characteristics. Caregiver factors (Table 3) included caregiver characteristics and parental difficulties. Family factors (Table 3) included family characteristics and major life events.

| Referral
Reports from mandated reporters (Wells et al., 1995) and professionals including schools (Giovannoni, 1987), law enforcement (Giovannoni, 1987;Schwab et al., 1997), doctors (Schwab et al., 1997), and other authorities (Östberg, 2014) were associated with the decision to investigate. In the case of nonprofessional reports, referrals by neighbours were significant in two studies (Giovannoni, 1987;Wells et al., 1995). Across studies, parent reports (nonperpetrating parent, parent) were significant as both a factor associated with screening in (Schwab et al., 1997) and screening out (Hutchison, 1988(Hutchison, , 1989Wells et al., 1995) reports. Anonymous reports (Wells et al., 1995) and witnesses to physical abuse reports (Karski, 1999) were also significant in decisions to investigate, but reports from relatives were significantly related to decisions to not investigate (Wells et al., 1995). Report characteristics significantly related to decisions to investigate included the number of reports received (Giovannoni, 1987), reports received in a single day (Hutchison, 1988), and the completeness of the report (Gryzlak et al., 2005;Johnson et al., 2002;Wells et al., 1995).
Maltreatment reports involving a known perpetrator (Wells et al., 1995) were more likely to be investigated than those with unknown perpetrators. A known perpetrator, or knowing the alleged perpetrator's full name, was also associated with the decision to investigate sexual abuse reports (Silva, 2011), as were reports with adults, females, family members, or parent figures as alleged perpetrators (Silva, 2011). The perpetrator having a relationship with the child was also associated with the decision to investigate physical abuse reports (Carlson, 1988) and, in the case of mothers, neglect reports (Gilbert et al., 1997). Access to the alleged child victim was also a significant factor in decisions to investigate all maltreatment types (Carlson, 1988;Schwab et al., 1997;Silva, 2011) including, in sexual abuse reports, if the perpetrator lived with the child or if access was unknown (Silva, 2011). Reports with perpetrators who were unknown (Silva, 2011;Wells et al., 1995), a third party, or aged 12 to 17 years were significantly more likely to be screened out (Silva, 2011).
The presence of an injury (Hutchison, 1988(Hutchison, , 1989 or a suspicious injury (Carlson, 1988) including an alleged (Carlson, 1988) or even suspected/unknown injury (Johnson et al., 2002) were factors associated with the decision to investigate. The type of injury (Wells et al., 1995) including whether there were visible signs (Carlson, 1988), a face or head injury (Schwab et al., 1997), or an injury of a 5-to 9-year old (Gilbert et al., 1997;Karski, 1999) were also factors significantly related to decisions to investigate. Both injuries that were severe (Gryzlak et al., 2005;Wells et al., 1995) and minor/other injuries or physical conditions (Wells et al., 1995) were significant factors in the decision to investigate. In physical abuse and neglect reports, direct evidence was also a significant factor for screening in reports for investigation (Gilbert et al., 1997;Karski, 1999).
Maltreatment type was also a salient factor: neglect, physical abuse, sexual abuse, and emotional abuse were significantly related to decisions to investigate (Gilbert et al., 1997;Gryzlak et al., 2005;Hutchison, 1988Hutchison, , 1989Jones, 1996;Karski, 1999;Östberg, 2014;Wells et al., 1995). However, in the case of sexual abuse reports where more precise measures were used, reports described as 'rape' or 'assault' were associated with decisions not to investigate (Silva, 2011).

| Child
Case factors relevant to the alleged child victim included number of victims, demographics, and characteristics. More than one child, or multiple victims, was associated with decisions to investigate across maltreatment types (Wells et al., 1995), and neglect specifically (Schwab et al., 1997). A child's younger age was also relevant to screening decisions with children less than 2 years (Gryzlak et al., 2005;Wells et al., 1995), less than 6 years from certain referral sources (Johnson et al., 2002), less than 7 years in cases of neglect (all types) and physical abuse (Schwab et al., 1997), or between the age of 2 to 12 years compared with teens (Gryzlak et al., 2005) more likely to receive an investigative response. Findings related to gender were mixed with female victims more likely to be investigated in some studies (Gilbert et al., 1997;Östberg, 2014) and to be screened out in others (Gryzlak et al., 2005;Wells et al., 1995). Ethnicity was significant (Wells et al., 1995) for decisions to investigate in two studies using the Children's Bureau Screening Study data, but dependent on site when tested as an interaction (Gryzlak et al., 2005).
Additional child characteristics included a child being unable to protect themselves (neglect or physical abuse), being fearful (physical abuse), and the child's physical condition (medical neglect). Child characteristics associated with screening out were a behaviour disorder (Giovannoni, 1987) or, in reports of physical abuse, hostile or aggressive behaviour (Schwab et al., 1997). However, school problems were associated with the decision to investigate (Giovannoni, 1987).

| Caregiver
Case factors relevant to the caregiver included caregiver characteris- Note. 1: Carlson (1988); 2: Gilbert et al. (1997); 3: Giovannoni (1987); 4: Gryzlak et al. (2005); 5: Hutchison (1988); 6: Hutchison (1989); 7: Jagannathan and Camasso (2013); 8: Johnson et al. (2002); 9: Jones (1996); 10: Karski (1999) Note. 1: Carlson (1988); 2: Gilbert et al. (1997); 3: Giovannoni (1987); 4: Gryzlak et al. (2005); 5: Hutchison (1988); 6: Hutchison (1989); 7: Jagannathan and Camasso (2013); 8: Johnson et al. (2002); 9: Jones (1996); 10: Karski (1999) of a learning disability (McDaniel, 2003;McDaniel & Slack, 2005) were associated with the decision to investigate. Referrals indicating drug use in neglect or physical abuse cases, or emotional problems in cases of refusal to accept parental responsibility, were also associated with the investigation decision (Gilbert et al., 1997). However, a caregiver's experience of abuse as a child was significantly associated with the decision not to investigate (McDaniel, 2003;McDaniel & Slack, 2005). Schwab et al. (1997) identified a number of parenting difficulties associated with the decision to investigate: inappropriate discipline (in emotional and physical abuse cases), a negative view of child (in medical neglect, and cases of caregiver refusal to accept parental responsibility), rejection by caretaker (in emotional abuse cases), name calling by the caretaker (in cases of emotional abuse), caregiver aggressive behaviour/anger towards child (in emotional abuse cases), and caretaker views of the child as a burden or unwanted (in abandonment cases, and cases where a caregiver refuses to accept parental responsibility). A caregiver not seeking medical treatment was also significant in reports of emotional abuse. Whereas Schwab et al. (1997) found inability to cope as a significant factor associated with the decision to screen out reports of physical abuse, Östberg (2014) found that parents' need for relief (from the child) was a significant factor in decisions to assign a case for investigation.

| Family
Case factors relevant to the family included family characteristics and major life events. Factors such as multiple children (McDaniel, 2003;McDaniel & Slack, 2005), or a large number of siblings in reports of sexual abuse (Silva, 2011), were associated with decisions to investigate. Reports with unknown household types were significantly more likely to be screened out (Wells et al., 1995). Open cases (Hutchison, 1988(Hutchison, , 1989Silva, 2011) or prior CPS involvement were factors consistently associated with the decision to investigate across multiple studies, as were prior reports (Carlson, 1988;Wells et al., 1995), incidents (Silva, 2011), investigations (Gilbert et al., 1997;Karski, 1999), or substantiation (Schwab et al., 1997). Receipt of cash assistance (AFDC, and later TANF), currently (Gilbert et al., 1997;Karski, 1999), and in the prior 3 months (McDaniel, 2003;McDaniel & Slack, 2005), was also associated with the decision to investigate. In the case of neglect reports, AFDC receipt was significantly associated with the investigation decision but, in physical abuse cases, only when there was a child aged 0 to 4 years (Gilbert et al., 1997;Karski, 1999).
Reports were less likely to be screened in for investigation when cash assistance had ended in the month prior, or where the family of concern had a higher income, or when an income was greater than $12,500 (McDaniel, 2003;McDaniel & Slack, 2005). Other family factors associated with the decision to investigate included inadequate housing (Giovannoni, 1987), unspecified environmental factors (Jones, 1996), or a lack of child custody issues or arrangements (Gilbert et al., 1997;Karski, 1999). Major life events associated with decisions to investigate ranged from a birth, school suspension, a family move, to arrest (McDaniel, 2003;McDaniel & Slack, 2005) while a convicted criminal offence was associated with the decision to not investigate in cases of physical abuse (Schwab et al., 1997).

| Decision-maker, organizational, and external factors
Studies had a more limited focus on factors within the remaining DME domains of decision-maker, organizational, and external factors (Table 4). Decision-maker factors focussed on worker beliefs. Organizational factors focussed on site differences, workforce, and supervision factors. External factors addressed were population and child maltreatment trends and community relationships.

| Decision-maker
In a study that examined worker beliefs in relation to screening decisions, the belief that CPS should investigate every incident of a child being at risk of harm and the belief that the decision to investigate is irrespective of the availability of community resources were both associated with decisions to investigate (Wells et al., 2004).

| Organizational
At the organizational level, site level differences were associated with screening decisions after controlling for other factors (Gryzlak et al., 2005;Johnson et al., 2002;Wells et al., 1995;Wells et al., 2004). Decision-making from site to site varied by injury type, report completeness, age of youngest child (Johnson et al., 2002), ethnicity, and when no harm was reported (Hutchison, 1989). Site differences in the time between contact and decision varied by site, and site structure was also relevant, with integrated rather than specialized units more likely to investigate (Östberg, 2014). In addition to site differences, agency specific practices were related to the decision to investigate including consultation with a manager prior to decision-making in sexual abuse reports (Silva, 2011). Increases in CPS staff were also associated with increased investigation rates (Jagannathan & Camasso, 2013).

| Study limitations
Studies included in the review were of variable and, in some cases, poor quality. Studies were also primarily from the United States so the relevance of findings to other countries may be limited given variations in practice, policy, and cultural context. The methodological diversity of the studies reviewed limited our ability to apply summary statistical techniques such as meta-analysis, thereby limiting the generalizability of the results. Studies prior to 1980 were also excluded. It is possible that earlier grey literature sources may have been missed.
It is also possible that our review suffers from biases associated with history as report volumes, child welfare policies, demographics,  (Carlson, 1988;Giovannoni, 1987;Gryzlak et al., 2005) due to small size and limited power to detect significant results (Steen & Duran, 2014). Finally, the presentation of study methods and findings were limited in some studies (Jagannathan & Camasso, 2013;Karski, 1999;Östberg, 2014), which limited the study team's evaluation of methodological quality.

| Implications
The studies reviewed sought to identify characteristics associated with decision-making at the point of CPS intake to improve child wel- Future research on factors associated with screening would also benefit from exploring the interactions, groupings, combined influence of the variables identified using advanced statistical methods. Such approaches would offer a more integrated understanding of these influential factors within and between DME factor domains and provide insight into how decision processes may vary depending on the type of maltreatment alleged. Finally, none of the studies examined the accuracy of the decisions that were made and future research on decisionmaking effectiveness is needed. This includes both the extent to which decisions accurately screen child protection referrals and decisions influence child and family outcomes. Relatedly, future research needs to also consider how the role of supervision or decision-making in teams contributes to effective decision-making.
This study also provides preliminary practice implications given the gaps in knowledge on factors associated with screening decisions.
Developing expertise in decision-making requires the use of reflective and critical thinking to consider factors individually and collectively.
CPS workers need to have feedback on their decisions and opportunities for reflection to consider how factors inform their own decisionmaking and the practice or empirical evidence base to support these factors in the context of indicators, risk, and protection. For example, Keddell (2014, p. 931) recognizes the social justice issues that may exist for families involved with CPS, such as those living in poverty, and the pivotal role of decision-making in determining the CPS respond to family difficulties. Keddell argues for attending to not only how decisions are made. but also developing a focus on what decisions should be made For example, a screening decision when a lowincome family lacks access to community supports or services to prevent crisis. Factors identified in this study raise questions about how workers are using a social justice lens when screening CPS referrals. Keddell (2014) suggests decision-making principles at each decision point to retain this lens and the use of critical reflection. Critical reflections by workers of such factors identified in this review can ensure workers maintain a focus on the social work profession's person-in-environment perspective and attend to the organizational and external context (social, political, economic, and otherwise) of decision-making. For example, opportunities for workers to reflect on the influence of DME external factors such as inequality, organizational budget constraints, political ideology, or media influence on their individual decision-making practices are important to achieve both effective and just decision-making practices.
Findings from this study also highlight the presence of procedural factors within the DME organizational domain, including the completeness of referral, the name of the alleged perpetrator provided, and the number of reports received about an incident. Although sufficient information is essential, it is important that decision-making is not procedurally led. Early work by Newsome et al. (2015) on novice and expert decision-makers in child welfare involving tests of the RPD referred to earlier and in the context of CPS decision-making has identified two types of decision-makers: (a) inquisitive and critical thinkers that rely on pattern recognition and mental simulation (Investigators) and (b) novices that rely on procedural knowledge outlined in structured assessment tools to justify decisions (Proceduralists) (Newsome et al., 2015). The demands and pressures on CPS workers have contributed to a prioritization of procedural factors; yet reflective spaces and feedback loops are important to strengthen pattern recognition and critical thinking over time.

| CONCLUSION
This systematic review was a first step in examining the extensive body of work on child welfare decision-making with a focus on screening decisions and the factors associated with them across DME domains. In addition to identifying important research and practice implications, this review highlights the limited knowledge that exists in the field on particular decision points, despite the wealth of literature on the topic more broadly. This review also highlights the importance of attending to factors across DME domains in recognition of the less visible but equally influential factors that inform decisionmaking. Further reviews are needed to organize and synthesize information to prompt decision-making practice advances and future research on effective and just decision-making.