The chaos of ACES

ACEs are an attempt to identify a set of traumatic conditions experienced before the age of eighteen, and to trace the ‘score’ of events in a simple causal manner through to alleged long-term biological damage to physical and mental health that these early experiences are purported to create. Findings from studies using ACEs are regarded as ‘hard’ data for policy and decision-making. However, while statistical methods and evidence have an important role to play in policy-making they have certain important limitations. Jerrim and de Vries (Reference Jerrim and de Vries2017) discuss a wide array of uncertainties that should be made transparent, including weak measures, measurement error, missing data, and statistical significance. These are all evident in ACEs studies, as we shall show below (see Hartas, Reference Hartas2019). In contrast to an acknowledgement of its provisional nature, the evidence as deployed by the ‘ACE movement’ advocates and by policy-makers is promoted as scientifically definitive, e.g. http://www.barnardos.org.uk/news/press_releases.htm?ref=129366.

For rigorous tracing of causal inputs through to effects, ACEs need to be a clearly defined set of experiences. Yet, they are chaotic in encompassing a shifting range of possible abuses and dysfunctions, and show inconsistencies in claims about severity, timing and duration. For example, in standard ACEs inventories (e.g. Felitti et al., Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards, Koss and Marks1998; Bellis et al., Reference Bellis, Lowey, Leckenby, Hughes and Harrison2014) the boundaries between common family circumstances and abnormal experiences become blurred. A ‘yes’ answer to ‘were your parents ever separated or divorced?’ constitutes an ACE whether it was amicable or adversarial, or occurred before the respondent was born, when a toddler, or a teenager. Similarly, the ACE criterion ‘living with anyone who was depressed, mentally ill or suicidal’ takes no account of who this is, severity or duration. ACE advocates are casting the net ever more widely to include more situations, including parental disability, mothers’ health, lack of routine, inter-parental conflict, moving home, and violence involving a sibling or peer (e.g. Harold and Sellers, Reference Harold and Sellers2018). The implication is that all these different experiences and the variety of combinations of them are comparable, underpinned by a common mechanism.

There is further chaos in the methodologies adopted by ACEs studies. There are retrospective studies based on people’s recollections, and prospective longitudinal designs subject to the specificities of the temporal period from which they start (Reuben et al., Reference Reuben, Moffit, Caspi, Belsky, Harrington, Schroeder, Hogan, Ramrakha, Poulton and Danese2016), as well as different sources of information and assessment. Whatever their methodology, most of the putative ACEs have in common their narrow remit of consideration: that is, the ‘household’, and in particular, parent/s and child. There is no attention to the influence of subsequent experiences in ameliorating or exacerbating their effects (see Coyne’s (Reference Coyne2017) discussion of Horwitz et al., 2001). The concept and measurement of ACEs does not capture confounding contextual issues that are beyond parental control and that can harm people emotionally and physically, such as being subject to racism/Islamophobia in a context where recorded ‘hate crimes’ are rising in the UK (O’Neill, Reference O’Neill2017). They do not extend to contextual factors, including wider family and friendship networks, school experiences, neighbourhood circumstances, the provision of formal and informal support services, and the broader socio-political regime (Petrie et al., Reference Petrie, Ayrton and Tinson2018).

ACEs then are an example of the methodological notion of a ‘chaotic concept’ (Sayer, Reference Sayer2010) – an abstraction that both conflates issues and divides indivisible processes with the consequence of hiding or ignoring the essential features shaping an outcome, a point to which we return later. ACE studies can provide no indication of how best to intervene, cannot point to whether or not an intervention, of what type and when, works. Indeed as Kelly-Irving and Delpierre (Reference Kelly-Irving and Delpierre2019) point out, they cannot and should not be used to predict individuals at risk. So, let us briefly examine the evidence from within the paradigm.

Sound and fury? How do the ACES measure up?

Despite the conceptual chaos, we have noted the bold and portentous claims made by policy makers, clinicians and practitioners in relation to ACEs. Intergenerational disadvantage and reducing the demands on health care systems caused by ‘multimorbidities’ are currently policy priorities. ACEs form part of a biological mode of explanation that has become common-place over the last decade.

The explanatory story is simplified into persuasive tropes through a variety of intermediaries with various mechanisms at work. For example, civil servants are often faced with an over-saturation of evidence. Their sense of personal efficacy, and hope of career advancement, is afforded when the government of the day accepts their proposals as policy. Thus, a good deal of a civil servant’s day is spent in discussion and argument with others within the state machine and crafting persuasive stories is central (Stevens, Reference Stevens2011). Broer and Pickersgill (Reference Broer and Pickersgill2015: 59) describe how opinion formers in Scotland use neuroscience to give ‘epistemic authority’ to policies which they felt were politically and morally right, engaging self-consciously in ‘pragmatic reductionism’ to render the science into simplified form, ‘packaged’ to persuade. There is a strong lineage in the vocabularies associated with ACEs, which can be traced to the Harvard Center on the Developing Child and its long-standing collaboration with the FrameWorks Institute, a communications company. This was focused on the production of

The production of a ‘core story’ required the development of a new vocabulary to develop ‘powerful frame cues’ (Shonkoff and Bales, Reference Shonkoff and Bales2011: 20) in the form of metaphors and values. Expert knowledge was thus recast through folk understandings, to show people what they think they already know. The act of translation by the FrameWorks collaboration, in the US context, involved challenging a dominant cultural notion that childhood adversity was something to be overcome by rugged individualism and self-reliance (Bales, Reference Bales2004). This, by necessity, had to involve emphasising the ‘damage’ and ‘toxicity’ of suboptimal childhoods. ACEs are currently described on the Harvard site as follows:

We note that the site shows a welcome shift in tone and now explicitly refers to environmental factors and social deprivation as part of the ACE story, but the narratives of biological damage, dose and toxicity remain, albeit modified by caveats.

In the UK context, these stories enter an increasingly interventionist child protection system (inter alia, Featherstone et al., Reference Featherstone, Gupta, Morris and White2018) which has shown itself to be enthusiastic about transferring the notion of future harm into the more coercive of the State’s activities – the compulsory removal of children from their families. Over the decades, precarious and fragile child development has featured in the legitimating narratives of the system. We have arguably reached a high-water mark of precautionary removals. In the first English study focused on newborn infants (under one week of age) in the family justice system, Broadhurst et al. (Reference Broadhurst, Alrouh, Mason, Ward, Holmes, Ryan and Bowyer2018) note that care proceedings were issued on 1,039 newborn babies in 2007/8. By 2016/17, the number was more than double at 2,447. Whilst the report does not explore reasons for the increases, it seems highly likely that the need to nip biological damage before it buds is a significant part of the story.

We have noted that the ACE concept as invoked in policy and practice is chaotic. At one level, it seems benign enough – the imperative to pay attention to people’s social histories when they misbehave in school, or present with mental health difficulties, or end up in trouble with the police makes a lot of sense. In this iteration, no measurement or scoring is necessary; we simply need sensitive professional practice. But the ACE agenda borrows an actuarial vocabulary. Researchers make strong claims about dose and invite targeting by ACE score. Even if the response to screening is simply more questioning, this has the potential to drag more families and children into state surveillance as we have noted above. With such consequential possible sequelae, it is vital that the evidence is carefully appraised; and frequently it is not. In the remainder of this section, we will interrogate two examples of the primary research, in order to examine effect sizes, and inter alia to bring out some of the contradictions and confusions that arise within the paradigm itself; we seek also to show something of the ideological direction of travel. One paper is from the US, published in the American Journal of Preventive Medicine (Wade et al., Reference Wade, Becker, Bevans, Ford and Forrest2017); the other derives from the UK, published in the Journal of Public Health (Bellis et al., Reference Bellis, Lowey, Leckenby, Hughes and Harrison2014).

Prompted by the high level of intercorrelation amongst ACEs (i.e. that individuals with one adversity tend to have others), the aim of the former study was to boil down the standard eleven-item (five categories of abuse, six of household stress) Behaviour Risk Factor Surveillance System (BRFSS) ACE measure to a shorter instrument. Results are based on a retrospective survey of over 70k adults, and in common with the norm, exposure to 4 or more ACEs was adopted as the criterion for elevated risk of ‘chronic health problems’ (Wade et al., Reference Wade, Becker, Bevans, Ford and Forrest2017: 2). The study identified ‘emotional abuse’ (being sworn at, insulted or put down by parents) as the ‘adversity’ most commonly ‘endorsed’ (a suggestive term) in the ‘abuse’ category for individuals with four or more ACEs, and equivalently ‘household alcoholism’ (living with a problem-drinker or alcoholic) in the ‘household stressor’ category. Interestingly, parental divorce/separation was not considered for inclusion ‘as these experiences are protective for some children’ (p. 4).

The main finding of the study was that the two-item questionnaire correlated very highly with the eleven-item measure, and was just as predictive of poor health outcomes, leading the authors to advocate the two item instrument as ‘ideal for the rapid identification of individuals with significant childhood adversity’. In terms of effect sizes, the results are also of interest. These are expressed as odds ratios. To pick one example, the odds of heavy drinking (defined as more than fifteen alcoholic beverages per week for men, and eight for women) are approximately 50 per cent greater for individuals reporting four or more ACEs. At one level, this sounds quite a salient result, but how large an effect is this? Using the most widely used method to quantify the magnitude of an effect (Cohen’s d), this just squeaks in as a small effect (d = .2). A more intuitive concept in a clinical context is that of the Number Needed to Treat (NNT), which indicates the number of individuals who would have to receive some form of intervention to ‘cure’ one additional case; here NNT is eight. In general, the effect sizes (odds ratios) across a broad range of health outcomes range are less than two (small effect) for most outcomes (thirteen out of sixteen).

Turning to the UK study (Bellis et al., Reference Bellis, Lowey, Leckenby, Hughes and Harrison2014), this is also a retrospective survey (approximately 1500 adults) of health and other outcomes, using the eleven-item BRFSS scale (although with some differences in definitions). The study begins by again drawing attention to the high levels of intercorrelation between the experience of different types of adversity. Regarding outcomes, it is notable that those in the categories of sexual behaviour, mental health, substance use and violence/criminal justice are by far the most seriously affected. Of the twenty-six specific outcomes assessed, nine of the ‘top ten’ are in these four domains, with use of cannabis the most commonly associated with four or more ACEs, followed by regular smoking. For comparison with the US study, ‘heavy’ consumption of alcohol (consuming more than six drinks, once or more per week) comes out fourth; the risk of this increases from 7.5 per cent in the sample for respondents reporting zero ACEs to 30 per cent for those with four or more. Bellis et al. do not give odds ratios in their analysis of outcomes, although they are liberally used elsewhere in their report. A rough translation of the increases in risk they report into odds ratio is possible though. For alcohol consumption, the odds ratio comes out at around five, a much higher figure than that reported by Wade et al., yielding a Cohen’s d of approximately 0.9 (large effect). In general, the effect sizes are much higher than those of the US study, with the majority falling into the intermediate or large effect categories. Neither of the two papers discusses effect sizes, and we are not able to comment other than to draw attention to such disparities.

In the detail of Bellis et al., Reference Bellis, Lowey, Leckenby, Hughes and Harrison2014, some intriguing anomalies appear. For example, whereas there is a greater propensity for obesity in the four or more group, it appears that this group also exercise more and have roughly half the risk of type II diabetes. No effects of ACEs are found for educational outcomes, although respondents are more likely to describe themselves as unemployed (or long-term sick). Bellis et al. also report that ‘having 4+ ACEs was strongly associated with higher deprivation’. Nonetheless, they go to considerable lengths to discount the importance of this link. Their concluding discussion opens as follows ‘Independent of relationships with deprivation, increasing ACE counts are strongly related to adverse outcomes through the life course’ (p. 88), reiterating this point on the following page ‘Although ACEs are more likely to occur in poorer communities, independent of deprivation, ACE counts correlate with worse… outcomes over the life course’. Such statements can be misleading; they imply deprivation to be unimportant. Far from it: they only mean that making a statistical correction for deprivation (based on an assumed, and implausible, linear model) does not cancel out the effect of early adversity.

The validity of the ACE framework is assumed to lie in its specification of risk to children, but as Finkelhor (Reference Finkelhor2017) notes, the indicators were contingent to the original study. There has been no rigorous scientific review to establish ACEs as the strongest predictors of poor life outcomes and little acknowledgement that mediating factors are crucial in explaining the established correlational associations. We know that there is a strong correlation between events labelled ACEs and the experience of family poverty. For example, roughly half of all children in lone parent households in the UK live in poverty, around twice as many as in couple families (DWP, 2017). Relationship breakdown is both a cause and effect of disadvantage and hardship, with mothers and children at most risk of falling into persistent poverty (Stock et al., Reference Stock, Corlyon, Castellanos Serrano and Gieve2014). A similar association is evident with regard to mental illness, in that low-income groups and their families are considerably more likely to suffer poor mental health, while parents suffering mental health difficulties are at much greater risk of experiencing economic hardship (Boardman et al., Reference Boardman, Dogra and Hindley2015). Likewise, the link between deprivation and child abuse / neglect is major and corroborated right across the developed world (Bywaters, Reference Bywaters2015) as is the association between domestic violence and poverty (Fahmy et al., Reference Fahmy, Williamson and and Pantazis2016)

We conclude this technical interlude with two final points on commonalities between the two exhibits above. First, there is the rhetorical effect of quantitative evidence that we noted earlier; the tables of statistics provide the lustre of science, but even the statistically informed reader must work hard to make a sound judgment and is left with unanswered technical questions. Second, neither paper seriously takes account of the limitations of retrospective studies compared with prospective ones; in particular, the operation of attributional biases, working either to mitigate or to exculpate failings and adversities in people’s lives, by blaming others or circumstances. Such biases are well-established features of human behaviour, attested by a plethora of research in social psychology; failing to take such biases into account is a serious weakness of the ACE research paradigm. With such a degree of contestability in the ACE framework, why is it so popular and why is policy so silent on addressing self-evident environmental adversities?

Bodies, brains and the outside world: inconvenient materialities

We have noted that it is reasonable to propose that children experiencing high numbers of ACEs are those most likely to be managing the effects of hardship. This is an assumption made by most ACE advocates. Yet, the irrefutable and long-recognised relationship between child poverty, poor health, lower educational attainment and reduced life expectancy is concealed by alleged ACE pathways. More significantly, as we have noted, ACE driven concerns to protect children largely promote behaviour change solutions. ‘ACE awareness’ relies on early identification, family intervention, resilience building and other therapeutic style approaches, despite their often weak and contradictory evidence base (Wastell and White, Reference Wastell and White2017). In contrast, the benefits of providing material support to poor families are clear and well-established, underlining the primary influence of poverty on later life chances. The positive effects of providing cash handouts to struggling families across the developing world is among the best evidenced of all anti-poverty strategies, with studies (many of them RCTs) demonstrating transformative effects on health, nutrition, school attendance and cognitive development (Hagen-Zanker et al., Reference Hagen-Zanker, Bastagli, Harman, Barca, Sturge and Schmidt2016).

The absence of poverty in the ACE framework does not stem from a neutral scientific calculation, nor accidental omission. Rather it is decentred by design. Poverty is separated out from other childhood adversities and reframed as a symptom of a damaged brain and body. For example, Public Health Wales claim that children experiencing ACEs are ‘ultimately less likely to be a productive member of society’ (Public Health Wales, 2015), while Public Health Scotland conclude that ACEs have a clear role in explaining inequality (Couper and Mackie, Reference Couper and Mackie2016). From this perspective, poverty is viewed merely as a symptom of dysfunctional development. Thus, the solution is perceived to lie not in raising household incomes but in breaking intergenerational ‘cycles of deprivation’. According to the ACE model, problems reside in the quality of the individual rather than the lack of resources available to them. Regardless of all the authoritative-sounding references to neuro-biological pathways this remains a value-laden position.

This strand of reasoning has a long and dubious pedigree, stretching back at least as far as the nineteenth century (Gillies et al., Reference Gillies, Edwards and Horsley2017; Wastell and White, Reference Wastell and White2017). The strategic detachment of deprivation from socio-economic and structural conditions and the emphasis on ‘intergenerational transmission’ of inferior culture and/or biology has featured in various guises over the years. But as Gordon (Reference Gordon2011: unpaginated) pointed out in evidence to the Social Mobility and Child Poverty Policy Review:

The ACE agenda provides yet another persuasive-sounding explanation. It promises finally and authoritatively to locate the seeds of dysfunction in the brains and bodies of poor children. Far from pioneering new ground, contemporary efforts to optimise biological development in children tread familiar ground. In the early twentieth century the Eugenics Society were similarly concerned to reposition disadvantage as a public health issue and were likewise fixated on ‘problem families’ and their biological consequences (Lambert, Reference Lambert2016). While attention may have shifted from genetic to social inheritance, a eugenic logic of predicting and preventing abnormality and weakness remains. Just as in the past, the desired ends are presented as benevolent and self-evidently progressive, justifying increasingly authoritarian means. Some ACE proponents are already calling for legally enforceable ‘parental competence evaluations’ (Willems, Reference Willems2012). We ask, where might this logically end?

Driving the ACE movement fervour is the spectre of long lasting physical and mental damage inflicted in childhood by careless or already damaged parents. ‘The first years last forever’, state the ACE advocates but this is by no means an established fact (inter alia, Bruer, Reference Bruer1999; Wastell and White, Reference Wastell and White2017). The adversities listed in the ACE framework are remarkably commonly experienced in the general population. In the original US ACE study two-thirds of the (relatively privileged) participants experienced at least one ACE, while in the UK the figure is estimated at 50 per cent (Bellis et al., Reference Bellis, Hughes, Hardcastle, Ashton, Ford, Quigg and Davies2017). Yet the vast majority appear to have overcome their adversities to live fulfilling lives (ONS, 2018). As Rutter et al.’s (Reference Rutter, Sonuga-Barke, Beckett, Castle, Kreppner, Kumsta, Schlotz, Stevens and Bell2010) research on resilience ascertained, human development is naturally adaptive to stress given the right conditions. Masten (Reference Masten2015) has described this as ‘ordinary magic’; the everyday processing that equips children to manage the setbacks and challenges that are integral to human life. If we were as fragile as the ACE proponents claim, it is hard to imagine how humanity could ever have survived its own traumatic history.