Current Debates on Variability in Child Welfare Decision-Making: A Selected Literature Review

Abstract

This article considers selected drivers of decision variability in child welfare decision-making and explores current debates in relation to these drivers. Covering the related influences of national orientation, risk and responsibility, inequality and poverty, evidence-based practice, constructions of abuse and its causes, domestic violence and cognitive processes, it discusses the literature in regards to how each of these influences decision variability. It situates these debates in relation to the ethical issue of variability and the equity issues that variability raises. I propose that despite the ecological complexity that drives decision variability, that improving internal (within-country) decision consistency is still a valid goal. It may be that the use of annotated case examples, kind learning systems, and continued commitments to the social justice issues of inequality and individualisation can contribute to this goal.

1. Decision Variability in Child Welfare

Decisions made in response to the problems facing families have considerable consequences in child welfare social work. The frameworks informing such decisions are essentially frameworks of social meaning and, as such, are highly influenced by the social contexts within which they occur. This context is far from monolithic, as many factors combine in a complex and contingent web of influences that converge and conflict. Thus, decisions are the outcome of highly situated categorisations of meaning sourced from available discourses in the national, political, organisational, professional, theoretical, technological, relational, material and personal environments. Such diverse and malleable social interpretations create a very complex context, and one outcome of this complexity is variability in decision outcomes. Drawing on a decision ecology approach, this article is a selected literature review of current research on several decision elements across the ecological spectrum. It discusses how they contribute to decision variability and why this matters [1,2]. It is not a systematic review in that it is not the result of a specific literature search and code technique; rather, the influences have been selected broadly based on current debates common in Anglophone contexts for their “reach” across the ecological spectrum. I argue that between-country variability is inevitable; nevertheless, it presents a major opportunity for critical examination of the impacts of various constructions of abuse, risk and family and their interactions with political contexts and policy orientations. In addition to this, within-country variability also exists. This represents an ethical problem, as general consistency at decision points across similar cases represents the universalist duties inherent in human rights and duty-based ethics.

This article outlines the occurrence of decision variability, considers the ethical questions such variability provokes, then discusses a selected range of known contributors to variable decision outcomes. Inevitably, this leads to political considerations. Decision variability in child welfare social work ultimately reflects inherent tensions in the normative versus radical traditions of social work, conflicting methods of risk assessments and perceptions, differing attitudes on child welfare outcomes and epistemological conflicts about the nature of child abuse [3,4]. All of these aspects—ecological, ethical, psychological and political—are reflected in decision outcomes that can vary widely in terms of access to services, recommendations for statutory assessment and decisions to remove children and place them in foster care. Why does such variability exist, even within countries that have the same policy orientation to child welfare [5]? Is it a problem, and for whom? If it is, can it be resolved? This article addresses these questions. It begins by introducing decision-making concepts, outlines variability and discusses the ethical consequence of this. It then covers a range of selected issues impacting variability broadly organized ranging from the macro to the micro: policy orientations, risk, inequality and poverty, evidence-based practice, constructions of abuse and its causes and cognitive processes.

2. Decision-Making: Rational and Consistent?

Decision-making is not necessarily rational. Kemshall ([6], p. 216) suggests the rational actor paradigm, “… fails to acknowledge the importance of macrosociological and historical forces on the one hand, and non-rational, individual forces on the other”. She argues instead that all decision makers are “situated”, that their decisions are contingent upon the meaning making resources and priorities of their social context. Many studies seek to elucidate the complexities and lack of consistency in child welfare decision-making in practice. Such studies illustrate decisions at a range of points on the continuum from notification to a child protection service, substantiation, decisions to assess, to intervene, to remove and to return children. A range of studies show that many countries have variable placement or substantiation decision outcomes in response to the same or similar cases [1,7,8,9,10,11]. These differences have variously been related to ethnicity, risk perceptions, professional background, poverty, worker attitudes towards removal, the specific site office and family wishes, while others note that the presence of alternative care resources is also implicated in variable decisions [12,13,14,15,16,17].

For example, Forkby and Hojer [18] found that decisions to place children in residential care were shaped by several factors. Firstly, it was viewed by practitioners as a last resort option, and as the outcomes of it were unknown, this resulted in a “damned if you do, damned if you don’t” position for the social worker. Removal as containing significant risk of harm has been shown elsewhere to result in a “weighing up of harms” by social workers, rather than a simple assumption of removal equates to safety position. This is especially so if the underpinning cultural preference is for family maintenance rather than removal [19]. Rossi et al. [20] compared the decisions of experts (academics, agency leaders, etc., n = 27) and frontline workers (n = 103) in response to real case summaries. They found that although all emphasised the same characteristics in the cases, these did not lead to uniform or consistent decisions. Instead, there was significant variability between both experts and workers spread across the spectrum from: close the case, offer ordinary services, offer family preservation services or remove. Experts were slightly more weighted towards the family preservation options, and the most influential case characteristic for removal was a prior record of abuse. Characteristics of the decision makers, such as gender or age, had little effect on decision outcomes [20].

Britner and Mossler [21] compared different professionals’ responses to the same case using vignettes. These included judges, guardians, court advocates, social workers and mental health workers. They found that the professional discipline determined the decision outcome more than the content of the complaint. Social workers and mental health providers relied on information about the severity and pattern of abuse, information about services offered in the past and parental responses to those services. Judges and guardians ad litem tended to emphasise information about “the likelihood of a reoccurrence of abuse and the child’s ability to recount the abuse, whereas CASA (court advocate) volunteers rely on information about the stability of the family” ([21], p. 317). Arad-Davidson et al. [9] asked social workers about real decisions they had made via a questionnaire. They questioned them at two points in time and found that 21% of decisions, mostly to remove, had not been implemented in the six-month time period between the questionnaires. The reasons given were mostly either parental or child objections, particularly if the child was older or the worker was more experienced; however, all children whose mothers were drug or alcohol addicts were removed.

Arad-Davison and BenBenishty [3] in another decision-making study of 200 social workers in Israel examined case worker attitudes via a questionnaire and compared them to responses to a vignette. Workers’ attitudes towards issues in child welfare (removal from the home of a children at risk, the ability of alternative care to foster children’s development, the optimal duration of alternative care and parent’s and children’s participation in the intervention recommendation) were compared with their risk assessments and intervention recommendations. They found that workers could be organized into “pro-removal” and “anti-removal” groups, via the questionnaire. They found that the former pro-removal group rated risk more highly and was more likely to recommend removal in response to the vignette family. They found that mother and child wishes had no effect on decisions and that social worker variables (such as experience) had no effect on which attitudinal group into which they could be categorised. Jergeby and Soydan [22] found that when comparing social workers’ responses to a vignette between and within countries, that even within countries, there was no evidence of standardised responses to the same situation.

In Aotearoa/NZ, the rate of notification to substantiation ranges from 5% to 48%, depending on the location of the site office [23], and a recent qualitative workforce review of child protection services noted the difficulties of establishing a clear and consistent threshold between NGO family support and statutory child protection intervention [24]. This review concluded that despite the assistance of assessment frameworks, decision-making protocols and legislation, that “…these sources were open for interpretation when determining whether or not the needs of a child or young person met the threshold for Child, Youth and Family assessment and intervention” ([24], p. 34). Also noted are the various impacts on social worker’s decision-making in a complex context, including public perceptions of the social worker role, the nature and duration of the actual concerns for the child, the practitioner’s own skill level and experience, the interface with other agencies providing services in local areas and the clarity of organisational messages and drivers ([24], p. 49).

Thus, variability in decision outcomes reflects a range of influences across the ecological spectrum, far beyond individual characteristics [16]. These include site locations, professional discipline, attitudes toward family maintenance, underpinning beliefs about the possible harm of foster care and the organizational contexts of practice. These aspects highlight the situated actor Kemshall [6] speaks of in contrast to the rational actor, as these factors highlight the contingent and socially-influenced aspects of decision-making in real-life contexts.

The Ethical Consequences of Variability

Thus, decision outcomes are variable and influenced by cultural variations, some of which are to be expected. As child abuse is ultimately socially constructed, it is unsurprising that reactions to it are also variable in line with differing national contexts [4,25]. However, comparing why variations exist between countries is valuable. It draws our attention to connections between the macro and micro realms and enables critical examination of the taken for granted [26,27]. Examining internal variability is even more important. Internal variability in decision outcomes represents an ethical difficulty that is not easily resolved. Platt and Turney [28] offer a cogent critique of attempts to standardise threshold decisions on several counts: firstly, that the complex decision-making context involves numerous factors beyond the actual case characteristics or agency decision-making tool; secondly, that the worker’s own sense-making processes contain numerous aspects to do with cognitive processes, intuition emotion and values; thirdly, that neither linear, rational decision-making nor the use of heuristics and biases as “defective” are realistic appraisals of real-life decision-making. They also point out that the various types of abuse and varied circumstances it occurs in make comparing them in order to make a distinct “line in the sand” impossible. However, all of these critiques are directed at the difficulties of standardising decisions in terms of the real-world context, but do not address the ethical issues inherent in variability. Certainly, descriptive theories of decision-making should be complex and situated, rather than linear and “rational”, yet pointing this out does not resolve the ethical issues raised by decision variability.

I contend that some kind of consistency in decision-making in regards to child protection decision-making should be attempted, however imperfectly, despite complexity. Why is consistency important ethically? As a statutory agency, the level of protection extended to children and the level of respect accorded to the parental rights of parents should not intersect at a line so malleable as to make the response of a statutory social work unreliable and arbitrary. If the macro factors (orientation) are held equal and the case facts are similar, deontological ethics (that is, that the same duties are accorded to all clients based on their existence as humans with intrinsic value) would suggest that outcomes should be consistent [29]. Thus, either a child does or does not reach a threshold for out-of-home care or access to a therapeutic service. The home is or is not safe enough to warrant return. Decisions such as these should have some level of transferability within a nation state, as an expression of the commitment to maintaining both the duty of care to protect children equally and to respect the autonomy and guardianship rights of parents. Otherwise, the law is arbitrary.

However, close studies of substantiation and other decision points do not conform to this expectation. As explained above and below, this issue is evident in many countries and represents several contributing influences. However, first, is this issue the ethical problem I have posed it as? It may be that variability, even within countries, is not quite the ethical problem I have posited. It could be argued that the research methods and ethical theories relied on to reach such a conclusion are inadequate. For example, most decision-making research is quantitative research limited to variables that are measurable and able to be manipulated, for example recorded case data or factors derived from factorial surveys using case vignettes. Many real-life variables that impact on decisions are not captured in either types of data, for example the in-depth, contingent and case-specific aspects, such as the social worker-client relationship, the reactions of both parents and children to the social worker, the perceived trustworthiness and caring attitude of the social worker toward the client and compliance. These variables are all found in qualitative studies to impact case decisions, albeit indirectly [30,31,32]. From an ethics of care perspective, some decision variability may not necessarily be negative, as such variability may reflect the type of relationship a worker and client family have. According to an ethics of care perspective (as opposed to a “duty” perspective), ethical decisions should not be determined solely by identifying one’s duties, but also by the context-dependent nature of the case and the relational or caring obligations that derive from it [33,34,35,36].

An ethical decision in child protection, therefore, could be considered not only one that reflects the weighing up of universal duties to protect children and respect the self-determination rights of parents or the costs and benefits of possible courses of action (traditional justice approaches), but one that is contingent upon the nature of obligations derived from a caring and supportive relationship with the family. Thus, it may be that some variability reflects the different relational qualities in a case that are not captured by research methods comparing the surface facts of a case, but are ethically defensible. This conclusion may have some merit, yet it is not a complete resolution, as the fairness question remains, as do concerns regarding transparency. However, it does lend us a more nuanced understanding of the ethical component of this issue and leads beyond the more straightforward question of what drives decision outcome variability, to when is decision variability ethically justified and when is it not.

3. Policy Orientations

Therefore, what drives these decision differences? One driver in between-country differences is the policy orientation of the nation state. At the macro level, differences abound in the historical development of services; for example, Duffy and Collins [26] note that the historical contexts of the U.S. and Northern Ireland have led to differing emphases on legal, cultural and political aspects of child welfare. Gilbert et al. [5] outline the differences between the policy orientations of nation states, deriving three main types: child protection, child welfare and child focussed. Others have examined the interrelationship between these orientations and decision reasoning, noting that the first leads to a focus on specific acts of abuse, the second locates the cause of problems in family needs, while the final one can result in a ‘child-centric’ practice that protects children’s individual rights and wellbeing, but may downplay family relationships [37,38,39,40,41].

Policy orientations can be described as broad “flavour differences” in overall policy directions [5,37,38,39,42]. Within a child welfare orientation, the best interests of the child are closely linked to the interests of the family as a whole, and assessments include strengths, as well as difficulties. Its purpose is preventive, and the focus within this approach is to “…create those material and social conditions within which all children are given sufficient opportunities to reach their full potential” ([43], p. 2; [44]). Thus, a child welfare orientation offers a broad-based prevention policy framework, based on a long-term understanding of epidemiology, rather than focussing on a single risky event [45].

Gilbert et al. [5] conversely characterise a “child protection” orientation as one that frames problems in individualistic and moralistic ways, directs legalistic and investigatory intervention types, promotes adversarial state-parent relationships and results in the use of mostly involuntary out-of-home placement. Fargion [43] adds to this description, stating that child protection orientations define children’s best interests narrowly in terms of children’s protection, reify ‘abuse’ as something objectively apprehendable and utilise standardised assessment tools. Significantly, she claims this approach “treats difficulties as signals of risks” ([43], p. 2). A further emerging direction is a “child focussed” orientation, one that directs policy focussed on promoting the rights and wellbeing of individual children; however, a significant downside of this orientation is that children can become viewed as entirely separate from the context of their social relationships [41].

These three orientations thus shape policy frameworks, but the application of these orientations is not straightforward, with “mixed” orientations increasingly common [5,39]. Kriz and Skivenes [46] provide an example of this contrast, stating that “the American and English welfare systems provide social assistance to the most destitute, with a child welfare system that focuses on risks for children. This type of welfare system stands in contrast to the Norwegian social democratic welfare state, which provides universal services related to health and social welfare, many of which are aimed at families, and with a child welfare system that focuses on children’s needs as well as risks” ([47], p. 1867).

Orientation frameworks shape decision-making by setting out the preferred aims and functions of a child protection system within national contexts. Variability often relates to clashes between these orientations, not only between nation states, but also within them. Kriz and Skivenes [47] for example, found that people in different national contexts, namely California, England and Norway, evaluated risk in the same case vignette differently, with Californians evaluating risk the lowest, then England and Norway. They found that the child welfare orientation in Norway could be attributed to an increase in perceptions of risk by Norwegian social workers, as the threshold for intervention is much lower when prevention is framed as a family support service provision aimed at preventing out-of-home placement. Of note were significant differences in the perceptions of which factors created risk for the children involved, as well as interesting findings that showed that the least regulated environment, Norway, produced the most homogenous types of reasoning around risk, while the other two contexts were more internally heterogeneous. They argue that this reflects not only different risk assessment tools, but also the policy orientations and type of welfare state of each nation. Other studies have also tracked the congruence between policy orientations and decision-making, for example Khoo et al.’s study of Canadian and Swedish approaches to child abuse and Skivenes and Stenberg’s study of social worker responses to domestic violence [38,47,48,49]. Internal variability can also reflect these broad frameworks, as the study by Arad-Davidson and Benbenishty (described below) shows that internal variation along these same lines (child welfare/child protection orientations) impacts on decision variability about out-of-home care.

Another aspect of decision variability is the disjuncture between official policy frameworks and how social work is actually practised. As Buckley argues, the two can be quite disparate, with real practice differing so markedly from espoused policy guidelines, that she questions whether it is an “ungovernable enterprise” [50]. Kriz and Skivenes [51] compared the ways that social workers in Norway, the U.K. and England responded as ‘street level bureaucrats’ to their policy contexts, that is, they managed the demands made of them by official policy and legal frameworks given institutional and financial limitations. This study highlighted that although many countries share similar commitments to the principles of the best interests and/or wellbeing of the child, family preservation, permanence and safety, that nevertheless, the degree to which governments focus on individual principles differs, and importantly, how social workers deal with the sometimes conflicting nature of those principles is quite a different matter. Such a high level of discretion can contribute to differences in decision outcomes. Pendulum swings in reaction to media reports can also shape practitioner responses in a manner that may diverge from official policies, as public opinion that swings from criticising too much removal to criticising family preservation attempts can drive decisions based on fear of public opinion, rather than in line with stated policy [12,52].

Orientations, Risk and Neo-Liberalism

Such studies highlight the culturally-defined nature of responses to child protection issues across national contexts, patterned by policy frameworks. These orientations clearly have some influence on across-country variability and internal variability as more countries adopt mixed orientations. Mixed orientations can be seen in conflicting legal imperatives, such as family preservation and children’s “best interests”, but other influences fuelling the fragmentation of ways of approaching decision-making may be the influence of neo-liberal ideologies and the impact of risk thinking. Risk hyper-sensitivity is particularly likely within a “risk society”. Beck’s [53] analysis of a risk society proposed that risk is increasingly perceived as ubiquitous, as people are more aware of risks than ever before. Whatsmore, risk is viewed as widespread and potentially catastrophic, and simultaneously, people are expected to guard against all possible risks to themselves [54]. Within a neo-liberal political system, this expectation is heightened. Neo-liberal ideologies focus broadly on reducing the role of government, individualising social problems and “responsibilising” both professionals and individuals [55,56]. This results in the spreading of responsibility for child welfare across NGOs and statutory services, common in neo-liberal environments focussed on economic efficiency and devolution of responsibility. As rates of referral to child protection services increase dramatically, intense pressure on child protection systems have led many countries to significantly reconstruct their service systems in these ways, particularly in the wake of the fiscal constraints imposed by the worldwide recession. Such changes commonly focus on spreading the responsibility for child abuse responses beyond statutory agencies and introducing new technologies in attempts to make practice responses more systematised and targeted to those most at risk [57,58]. The state’s role moves from being a “social state” to a “facilitating state”, merely a coordinator of services, rather than a body responsible for their existence, functioning or resourcing [59]. This spreading of responsibility among multiple agencies may multiply the various logics and beliefs relating to the correct or proper role of statutory social work with children, as agencies have their own cultural beliefs, practices and tools to assess families. It may also produce an uneven impact on the practitioner’s level of risk aversion, as an increasingly accountable workforce reacts to the pressure to manage risks to the self in different ways, with some workers complying and others resisting [60]. Within such a system, the possibility is that social workers become focussed on lowering their own risk of future accusations of poor practice, rather than building durable and humane solutions for children and their families. Decision-making becomes framed as avoidance of risk rather than promotion of wellbeing or self-determination.

Further, the professional knowledge and relationship focus of social workers are downplayed within a context driven by managerial and technical discourses, as are the radical traditions of social work [61]. One effect of this is that “this framing of social problems encourages those professions working on social problems to adopt an ‘inward-looking’ perspective that minimizes the connections between structural change and the manifestation of individual problems” ([62], p. 1023). Another is that targeting of “programmes” is considered the only solution to abuse, and universal, supportive services are eroded. Abuse itself becomes considered as an individual problem, to which the correct and proper response is the correction of parenting practices, rather than considering the ways that parent’s interactions with their children are shaped by their material and social resources. It can be reasonably concluded that where a social worker is positioned along the radical-normative continuum will drive decision variability in responses to risk in the context of neo-liberal states.

4. Risk and Safety in a Risk Society

Thus, how risk is conceptualised is another major driver of decision-making in Anglophone countries [6,63]. Indeed, risk assessment and its associated “risk logics” have become the focus of child protection systems nearly worldwide and, with it, an economic-rationalist assumption that risks are measurable, predictable and avoidable [57,64]. Considering risk assessment highlights two debates in decision-making: how risk assessment tools are derived and used and whether they include strengths or safety perspectives. Houston [65] states differences in ways of conceptualizing risk reflect objectivist, subjectivist and critical meta-theoretical paradigms, each with their own epistemology, ontology and axiology in relation to risk. One expression of an objectivist approach to risk, for example, has been in the development of risk factor science and related risk assessment tools. A concern about risk and a perceived need to identify and reduce risk has led to increasing research into risk factors, using longitudinal and quasi-experimental methods, to measure the effect sizes of different variables. These have been incorporated into risk assessment tools, often termed “actuarial” approaches to risk assessment work. While providing valuable information about patterns across populations and successful in some studies to predict future abuse, on their own, such methods have significant limitations for decision-making [65,66].

Limitations of decision tools exist for two related reasons—the method of quantification and the way they are used in the practice of decision-reasoning. Firstly, in terms of the production of risk factors via statistical correlations, they are neither “necessary nor sufficient” to predict abuse in the future [67], thus producing a high rate of “false positives” [68]. Secondly humans as decision makers are notoriously imperfect at incorporating accurate assessments of probability and effect sizes into decision-making [69]. France et al. [70] note this in social work when they state that while statisticians might understand the level of surety or otherwise are able to be infer it from statistical correlation, nevertheless, risk factors are often treated as if they can give certainty about future outcomes in practice [70]. Gillingham [12] and others have noted that practitioners may subvert the intention of structured decision-making tools to get the outcome they desire or use the language of risk to legitimate decisions they have already made on other grounds [71]. However, if used correctly, actuarial approaches can be combined with consensus or discretionary approaches to produce a more complete method of risk assessment than one or the other would alone, with some research demonstrating that actuarial approaches are more accurate in terms of the actual risk of future notification than professional discretion approaches only [72]. However, using them alone to decrease decision variability is undesirable, as many other factors in addition to future risk must be considered.

Either way, in addition to actuarial knowledge, case-specific knowledge, a relationship with the family, understanding how risks change over time, theoretical underpinnings and ethical commitments are required to be able to make decisions related to particular families [72,73]. This is particularly true given that estimations of future harm are just one aspect of decision-making about children’s lives. While preferred within modernist economic rationalist paradigms that attempt to predict and control uncertain outcomes, the use of risk-factor approaches can fuel practice dominated by risk aversion and forensic investigatory-type approaches [64,74]. The development of strengths and safety-oriented perspectives have attempted to counter this deficit approach to risk assessment, arguing that searching for risk factors alone is neither accurate, fair, nor useful in a context of unequal power relations and where the ongoing relationship is the best hope for actual change in the family [75,76,77]. These approaches attempt to resist risk saturation by actively searching for and including aspects related to client safety and strengths. Initial research notes the impact on decision-making of the Signs of Safety approach as producing knowledge about clients in ways that retain moral subject positions for parental clients to occupy. This contributes to shared power and parental engagement, creates the opportunities for personal change and leads to decisions that are likely to be less risk averse [78]. In terms of decision variability, which tools are specifically adopted by different agencies is therefore a significant driver of differences in decision outcomes, as those using traditional risk assessment formats and those utilising strengths or safety-oriented tools may reach very different conclusions about the level of risk and, thus, the need for intervention. Safety-oriented tools are also more likely to open up different pathways for intervention, as they focus on creating an intensely collaborative relationship with families and utilising their own networks and resources, whereas traditional risk assessment methods, used in the context of targeting and resource efficiency, tend to use the assessment to determine who should receive evidence-based standardised programmes delivered by a third party.

5. Inequality and Poverty

This leads us to consider the role of poverty in family lives, as how the impact of poverty is constructed by decision-makers is another major driver of differences in decision outcomes. An individualised focus can lead to a narrow evaluation of quantifiable and individualised risks, obscuring a major concern with the pervasive over-representation of those most socio-economically disadvantaged in child protection systems. An enduring correlation between poverty and abuse, while not deterministic, nevertheless raises questions about the relationship between poverty and abuse in terms of causality and ethical questions about child welfare intervention as a reflection of inequality [41,79,80,81,82]. Paxson and Waldfogel [83] in a large study found that an increase from 10% to 15% of children living in extreme poverty was associated with a 22% increase in child abuse. Many others have found similar results [84]. However, interpreting these findings in decision-making contexts requires careful examination: “The experiences of those trying to parent in a profoundly unequal society are not interrogated rigorously enough in current responses, with causation and correlation confused in a highly abstract language that renders real people and their lives invisible and/or unintelligible” ([41], p. 5).

In terms of decision-making, the connections between poverty, inequality and abuse are important for practitioners to grasp. However, how does one incorporate them in decision-reasoning? There is some evidence that social workers can become de-sensitised to poverty as an integral and seriously damaging aspect of client’s lives [85]. Other studies have found that poverty made no impact on social worker’s decision-making, although other markers of poverty, such as substandard housing, did [14]. While Stokes and Schmidt [14] point out that social workers should be considering the impact of poverty on the ability of families to care for children, particularly from an anti-oppressive standpoint that recognises the impact of macro-structural conditions on families’ lives, Moraes et al.’s study [86], while investigating a similar question and with similar findings, came to quite a different conclusion. Moraes et al. [86] used a representative sample drawn from the Canadian Incidence Study of reported child abuse and neglect. They found that over five indicators (case substantiation, provision of ongoing child welfare services, referrals to child and family support programmes, out-of-home placement, applications to child welfare court and police involvement), poverty had an extremely minimal impact (6%) on the variance in decision outcomes. Nor did a poverty index of the family’s circumstances, although, again, living in unsafe housing (perhaps a proxy for poverty) had some impact on substantiation of abuse, particularly in cases where there was no physical injury. They conclude that there is no evidence in this study that social workers act in a ‘discriminatory’ manner in relation to families living in poverty and posit this as a positive finding.

As Keddell [87] notes, this contrasting interpretation of findings in relation to poverty in decision-reasoning is indicative of the intensely contestable nature of the definition of the social work role and the tensions practitioners must navigate: one must be aware of the oppressive nature of poverty and its impacts on parenting, but at the same time, resist judgemental and negative assumptions about parents in poverty. This dichotomy reflects the difference between reflection and critical reflection: the former encourages self-examination with a view toward eradicating discriminatory stereotypes, while critical reflection moves beyond this to include structural awareness of social problems in one’s assessment of a family, so as to recognise the interconnections between family practices and structural conditions [88]. These two types of reflection have significant implications for decisions where poverty and child welfare concerns converge, as decisions reflect, at least partly, the practitioner’s evaluation of the culpability for, or source of, the problems facing the family. Without at least some inclusion of structural impacts, culpability lies only with parents. Decision variability may reflect these tensions.

Despite the radical and critical traditions implicit in social work education, political changes in many countries (particularly Australia, New Zealand and the U.K.) towards the individualisation of social problems in both welfare reform and child protection systems reconfigurations constantly cast social work as a micro-level profession sent to maintain control and regulation of the masses, rather than advocate for them [89]. Marston and McDonald [62] bemoan the de-politicisation of social work, arguing that recent neo-liberal developments of policy in many Western countries have resulted in a significant reconstruction of the social work role and, with it, the replacement of “…sociological and political-economy approaches to problems like poverty, and have instead opted for behavioural-economic understandings of human behaviour. The combined effect of these changes is to cast doubt on the knowledge and actions of social workers as political actors, particularly those social workers directly engaged in work that seeks to redress social injustice and to influence public policy” ([63], p. 1023). Interestingly, Moraes et al. also draw a link between attitudes toward poverty and decision reasoning, hypothesising that in “…nations where poverty is viewed as primarily an individual responsibility and income inequality is pronounced, child welfare measures may be more intrusive among poor than non-poor families. On the other hand, in nations where poverty is viewed largely a societal responsibility and income inequality is more moderate, child welfare professionals may be more egalitarian in their decision-making” ([87], p. 167). Thus, the challenge for social workers in increasingly neo-liberal climes is to incorporate poverty in decision-making beyond simply eliminating negative stereotypes towards poor families, but must continue to remain resistant to pervasive social discourses that cast the poor as primarily responsible for their own demise [90].

6. Evidence-Based Practice

Likewise, while including known risk factors in decision processes is important background knowledge, current exhortations to make “better use of research” can result in a deterministic use of risk factors or an inappropriate weighting of them in decision reasoning. Many effect sizes are tiny or the relationships between factors indeterminate. For example, Appleyard et al. [91] found in their study examining the relationship between childhood abuse, adult substance abuse and victimisation of one’s own children that there were traceable relationships, but the effect size was small (0.26/0.19). Studies of the impact of young parental age at first birth, a commonly relied on risk factor, actually show that effects are mixed. Some show a clear correlation with maltreatment, but these effects are also influenced by lower economic status, lack of social support and high stress levels [92]. This means that the age at first birth may not be as influential as first thought, or, all of the above factors combine for the effect. However, when research is translated for practitioners, it is often oversimplified, so that the nuances involved are lost [41]. Thus, the use of research in practice can be used inappropriately, as when over-simplification is combined with natural tendencies to over-estimate effect sizes, the result can be too much emphasis on negative attributes [92]. These issues led Munro et al. [67] to conclude: “The repeated exhortations to child protection services to predict and prevent maltreatment, and the associated blame when they fail to do so, should be replaced with more modest expectations” ([68], p. 70). Thus, while understanding broad population patterns in relation to risk factors is certainly important background knowledge, its use in ways that assume they confer certainty in any specific situation is misguided and may contribute to variability [93].

7. Constructions of Abuse and Its Causes

Alongside these macro influences of national orientation and changing conceptualisations of risk and evidence, other changes are also occurring in regards to child abuse definitions and understandings of the causes and consequences of abuse. Definitions of abuse are played out at the micro level, as the idiosyncratic nature of family circumstances and behaviour is often difficult to fit into formal definitions of abuse, or if it does, deciding at what point or threshold it is ‘bad enough’ for intervention is a slippery fish. As noted, child abuse is a socially-constructed phenomenon in terms of its meanings, definitions and social responses [94]. This is reflected in the changing definition of abuse for differing purposes, such as for research, practice or legal purposes [95], differences in social worker’s responses to the same cases, intense public debate as to the boundary line between physical punishment and abuse and beliefs about what constitutes reasonable care or “good enough” parenting [15,96,97,98]. Modern definitions of abuse cover a variety of types, and establishing their parameters in real-life situations remains difficult.

Considering definitions and constructions of abuse leads to a serious consideration of the causes of abuse. Many system’s responses, drawing on a neo-liberal paradigm, propose that abuse is caused by psychological or criminal tendencies and, thus, prescribe systems of surveillance, control and sanction as ways of reducing and preventing abuse [99]. These changes, without responding to a wider range of abuse causes, risk factors and effects of abuse, remain partial, only targeting one aspect of abuse prevention (surveillance and prediction) and only the most severe types of abuse. Such responses assume that definitions of abuse are obvious, straightforward and its solutions are self-evident [100]. In turn, decision-making itself is conceived of as correctly “recognising” abuse and categorising a family as “abusive” or “non-abusive”, rather than the more difficult holistic assessment of child and family needs, opinions and wishes.

Risk factors, as discussed, cannot predict abuse with any accuracy. Profiling of the most serious offenders can result in effective prevention of secondary abuse and sometimes primary. However, the vast majority of families referred to child abuse services, as well as those for whom abuse is substantiated, have much more complex, diffuse and chronic issues, with strong relationships to structural conditions, rather than the innate pathology of individuals [67,101,102,103]. What is more, child deaths are extremely rare events that are not predictable, even when considering a high-risk population [93]. In the use of the predictive risk model in Aotearoa/New Zealand that tracked babies from birth over their first five years, the top two deciles of risk were 66% accurate; in other words, it also identified 34% of those deciles as a ‘high risk’ population that did not go on to become abusive [104]. Pritchard and colleagues found that those who killed children had several identifiable salient features, such as having a history of violence and being not biologically related to the children they killed. They conclude that: “The juxtaposed results indicate that the assailants’ problems are essentially psycho-criminological, especially violence, rather than socio-economic, although poverty worsens most situations” ([101], p. 1403). However, such cases make up a tiny proportion of those referred to child protection services, even of the substantiated cases. In Aotearoa/New Zealand in 2012, for example, emotional abuse and neglect made up around 80% of substantiated cases, with the remainder physical and sexual abuse [105]. However, substantiated cases (18,595 distinct children) were, in turn, a tiny proportion of the total of 148,659 notifications (of which 57,766 were police notifications for domestic violence) [24,105]. In Canada, 85% of referrals to their child protection service can be categorised as chronic needs, while the remainder are urgent protection cases [106]. Thus, practice responses attempt to respond to both high-end perpetrators, who may well have individual psychological risk markers for serious physical or sexual abuse, as well as the much more common diffuse, chronic situations of a combination of mental illness, poverty, drug and alcohol addiction, discrimination and domestic violence [24,107]. Based on an assumption that the former type of abuse is what most abuse “is” (influenced strongly by media representations), there is a substantial mismatch between systems and services and the actual phenomena of child abuse [52]. Wolfe et al. [107] state the problem thus: “Child abuse is an event, not a uniform disorder, and therefore it is necessary to consider multiple causes that interact unpredictably. Notwithstanding the critical role of the adult offender, child abuse is rarely caused by a single risk factor. Although risk signs and indicators are present, it is still very difficult to predict who will become abusive and who will not—child abuse may emerge in any given family if the ‘right’ conditions exist. These causal conditions stem largely from the interaction of child, familial, and cultural influences, but it is not possible to predict with precision when they will occur” ([108], p. 35).

The variety of psychological and social work theories, as well as theories of human development, constructions of childhood and what might constitute children’s best interests also impact on decisions about what counts as abuse and harm across the lifespan [108]. For example, conflicts arise when considering the line between punishment and physical abuse, which despite being criminalized in a number of countries, remains a contested areas as to what amount of force, with what intention, to what age child, over what period of time and combined with what other family factors constitute abuse; and if so, is it to the level of harm such that statutory intervention is required [3]? Likewise, the impact on neglect, particularly of poverty and mental illness, can make a finding of neglect difficult if the parent is considered not in reasonable control over the contributing circumstances. Each of these (definitions of neglect and physical abuse) may have considerable cultural dimensions, with child rearing practices differing markedly across ethnic and cultural groups. This leads to a consideration of the way that a variety of types of abuse impacts on decision variability. It is only in recent times that the vast array of harms to children have been categorized as “abuse”: neglect, physical abuse, sexual abuse and emotional abuse. Some of these are clearer than others in terms of the degree of social consensus on harm, including the research evidence. For example, Chan et al. [109] found that despite an assumption that professionals would have some kind of common, professional view on abuse and its consequences, it was found instead that their attitudes towards child abuse reflected their cultural backgrounds rather than a professional consensus. Such differences in the interpretation of child abuse illustrate the contested nature of child abuse. This must refract through the varied decisions of social workers.

Domestic Violence as the New Child Abuse

A further major recent change in the construction of the child abuse definition has been the inclusion of domestic violence as a type of emotional abuse of children [110]. Increasingly, convincing research on the harm to children from exposure to or witnessing of domestic violence has led to its transformation from an issue primarily affecting adult victims, most commonly women, to also having child victims [111,112,113,114]. There is also evidence that in addition to exposure, domestic violence also has a high co-occurrence with direct abuse of children, with some estimates between 30%–60% [115]. This has led to interesting developments in the responses of child protection systems. In Aotearoa/New Zealand, for example, police callouts for domestic violence where children were present prompted a now-mandatory referral to child protection services. Due to this change, one third of families referred to child protection services each year come directly from Police (see above). As in other countries, this represents a challenge to child protection services, where the reconstruction of domestic violence as a new type of child abuse leaves a service more familiar with traditional types of abuse ill-prepared to intervene effectively [110,116,117], with either excessive criminalization or child removal having significant drawbacks [118,119,120].

Considerable collaboration across services is needed to respond in ways that effectively protect children from harm (and women) [121,122]. However as Hester [121] notes, this is difficult when child protection systems and women’s protection systems operate within very different paradigms, leading to conflict about the most urgent problem faced by families and, thus, how to intervene successfully [119]. Simply educating child protection workers about domestic violence, while a good start, is insufficient in establishing an effective and systematic response to this phenomenon, where both the values and knowledge of child protection and women’s advocacy services are needed to be included, combined with case-specific consultation in order for collaboration to be successful [123,124]. This issue presents considerable tensions for social workers, for whom the definitions and conceptualisation of domestic violence presents challenges for establishing decision consistency. How it should be coded (as emotional abuse or neglect), if it should be considered a risk to children at all and who is assigned responsibility for protecting children from it, all impact on decisions in relation to it. Skivenes and Stenberg [49] compared responses to two vignettes, one a neglect case and one a domestic violence case. They note that in response to a domestic violence case, that responses between the U.S. and the U.K. were similarly concerned with risk, and this may be due to the U.K. swinging towards a child protection orientation; although, in another type of vignette offered (neglect), they found much more of an emphasis in the U.K. on a family service or child welfare-type approach. They contend that this is due to the type of case and that this may show the impact of evidence-informed practice, as knowledge about the effect of domestic violence has a more established cross-national consensus than other types [49]. However, they found that despite the uniformity of views on domestic violence as constituting a risk factor, that marked heterogeneity again appeared in the reports of what action social workers would actually take in response to this risk.

Another issue in responding to domestic violence in case decision-making relates to who is perceived as responsible for it. As other research shows, where parents are viewed as culpable, more intrusive decisions follow. A variety of studies show that women are perceived as “bad mothers” when they are faced with domestic violence in the child protection system, and this results in a deficit mothering discourse that blames them for their predicament. In terms of decision-making, this can result in an ultimatum approach from child protection services of requiring women to separate from their partners or remove the children [125]. Where women are viewed as both responsible, yet holding little real control over the risk presented by their partner to both themselves and their children, intrusive decisions become constructed as justified [126]. In a study examining the ways child protection workers attributed blame in domestic violence cases [127] found that the presence of domestic violence increased worker’s perception of risk to children. Every construct they used in a factorial vignette approach showed that workers were more likely to view women as more culpable for exposing their children to harm than their partner, even though in every case, their male partner was the violence perpetrator. Thus, understanding how domestic violence is affecting decision variability requires an understanding of the ways blame, culpability and risk are understood in relation to domestic violence, as well as how child protection systems interact with other services in particular locations.

8. Cognitive Processes and Group Decision-Making

Finally, the variable ways in which all of these factors are used in decision-making highlight the known cognitive problems with decision-making. For example, the tendency of humans to make decisions based on the most vivid or recent piece of information [128], the lack of the computational ability of humans to consider and weigh up all relevant information, the effect of “frames” (the implicit contextual parameters of decisions) satisficing—taking the best available option rather than the ideal—and the use of heuristics and biases, particularly the tendency to have a confirmation bias. Many of these dynamics can lead to an overconfidence in human judgement, particularly when people are forced to make decisions quickly, or the “latent conditions” under which they are forced to operate lead to negative decision “shapers” such as incomplete information and high uncertainty, systemic feedback errors, improper drivers (such as technical tasks rewarded, rather than family outcomes) or a lack of training [69,92,129,130]. However, slower consideration of relevant issues, combined with education about cognitive biases, can help reduce their impact [92,131]. Others argue that heuristics (short-hand rules of thumb derived from practice experience) are not necessarily problematic, but the complex use of short cuts in response to recognised patterns that are often correct when used in their specific environmental contexts. It is this proposition on which proponents of ecological rationality and naturalistic decision-making essentially rely [28,132,133]. Hogarth [134] extends these ideas to consider heuristics as an aspect of the intuitive components of decision-making, which together with analytic reasoning, make up decision processes [73]. He argues that the learning of “correct” heuristics as one part of developing sound intuitive reasoning must take place in “kind”, as opposed to “wicked” learning environments. A “kind” environment is one that “…provides timely feedback and allows the tacit or experiential system to shape practitioner responses. In contrast, ‘wicked’ learning environments provide misleading, or no, feedback resulting in unfounded practitioner confidence” ([134], p. 4).

In terms of decision-making variability, therefore, a reliance on either individual practitioners or the immediate ecological context presents challenges for national consistency in decision-making, as individual differences in perceptions, as well as site differences and their varieties of latent conditions, frames and whether there is a kind or wicked immediate environment for the learning of intuitive heuristics (very similar to “practice wisdom” concepts in social work), may produce inconsistency between locations. A child in one town, with its particular mix of available resources and site office systems and cultures, should not be afforded a different level of protection than the same child in similar circumstances in a different location, nor should the rights of parents be impinged upon differently. For example, one reason for the interaction between inequalities and child protection involvement may be the lack of supportive family resources in poorer neighbourhoods that leave few choices open to both NGO and statutory social workers than to use legal intervention; an understandable response and one entirely consistent with naturalistic decision-making, but not an equitable one. Establishing a fairer response to threshold decisions would highlight the increased need in some neighbourhoods for more family support services, rather than involve statutory services that are unwarranted and can be harmful (in terms of intrusive and stressful interventions and the uncertain outcomes of foster care) [135,136,137]. Furthermore, while many social work processes use groups to attempt to include a range of stakeholders, the influence of “group think” can also be negative, wherein group process can lead to a decision that reflects a desire for consensus rather than the best or most accurate decision. Current studies tend to ignore these cognitive processing issues, with many practitioners forced by working conditions to make decisions based on scanty information, under time pressure and in unstructured group decision-making contexts [24]. This impacts on decision variability, as without considered decisions and well-structured systems that support them, the internal variables of biases and values, the impact of a “wicked” learning environment, or “group think”, can become primary decision drivers, leading to increased variability.

9. Conclusions

This conceptual article discusses selected research as it applies to decision-making variability in the child protection realm. There are some omissions, as to discuss decision-reasoning is to discuss the entirety of the social context, and this is impossible to cover in an article. Nevertheless, this article raises many issues currently driving both between-country and within-country decision variation in response to child abuse and neglect. It highlights factors ranging from the macro to the micro, showing that orientations shape both external and internal variability. They contribute to variations in the ways that the causes of abuse are framed and highlight that child abuse remains subject to intense contestation as to its nature, consequences and best interventions. The impact of responsibilisation in neo-liberal environments, constructions of risk, responses to poverty and inequality, evidence-based practice, contestable views on abuse definitions and its causes and cognitive and systemic decision-making processes present intense challenges to decision consistency. I argue that despite this complex ecological context, finding ways to establish realistic decision systems and direct guidance is more desirable than ever. While concepts from either the “situated actor” of sociology or the naturalistic decision-maker from cognitive psychology are both useful ways to describe how decisions are made in context, neither offers guidance on how decisions should be made. To simply say that the environmental and cognitive issues at play are too complex brushes over the social justice issues at stake. One idea for developing realistic decision guides may be to develop complex case exemplars with annotation to highlight decision principles that would apply at each decision point and provide such exemplars for each type of child abuse, for each age of child. Attention to systems to ensure “kind” learning environments that enhance the development of sound professional discretion for practitioners, as well as logical decision pathways embedded in organisational structures are also imperative. Continuing to emphasise critical reflection in social work education and practice is another important aspect of increasing consistency in decision-making, as is addressing the numerous organizational systems, cultural and processual elements of decision-making. However, in addition to these micro-meso solutions, settling the macro framing of child abuse and its ideological dimensions is crucial to improving decision variability. This must include the ongoing commitment of social work to address political and macro issues of access to resources and the harmful effects of neo-liberal individualisation. Ideally, this commitment would result in establishing decision-reasoning processes that acknowledge influences from all levels of the ecological approach on problem causation, as well as the ethical commitments that social work espouses.