Content uploaded by Colin Pritchard
Author content
All content in this area was uploaded by Colin Pritchard on Aug 18, 2015
Content may be subject to copyright.
Who Kills Children? Re-Examining the
Evidence
Colin Pritchard*, Jill Davey, and Richard Williams
Colin Pritchard has been Research Professor in Psychiatric Social Work at the Centre for Social
Work and Social Policy at Bournemouth University since 2003. Recent publications in the
British Journal of Cancer and British Journal of Neuro-Surgery illustrate the value of a
social work and social policy perspective in medicine and a forthcoming study of female
youth suicide in two Catholic continents in international social work highlights the Centre’s
commitment to ‘giving a voice to the unheard’. Richard Williams joined the university as a
senior lecturer in 2007 after an innovative thirty-year career in educational social work,
leading to his book Breaking the Cycle of Educational Alienation, his recent publication
with Colin comparing five-year cohorts of former looked after children with excluded-from-
school adolescents as young men, highlighted his continuing link with the field and the value
of an evidenced-based approach in social work practice. Jill Davey’s social work career spans
twenty-five years, with her holding strategic and managerial positions in child protection and
an advisor’s role with the DfES, including a number of Serious Case Reviews and published
in child-care social work. She joined the university as a senior lecturer eight years ago, was
involved in developing the Post-Qualifying Child Care Award and currently is programme
lead for the BA/MA in Social Work.
*Correspondence to Professor Colin Pritchard, Centre for Social Work and Social Policy,
School of Health and Social Care, Bournemouth University, Royal London House,
Christchurch Rd, Bournemouth, BH1 3LT, UK. E-mail: cpritchard@bournemouth.ac.uk
Abstract
Violent children’s deaths have become a surrogate indicator of effective child protection
but can those who kill children be better identified? A decade-long study of child homi-
cide assailants (population of 2.5 million) is re-examined in the context of nineteen
Western nations’ child mortality rates and child-abuse-related deaths, correlated with
four international measures of relative poverty, focusing on income inequality. Child
mortality rates of the nineteen countries were ranked and correlated with levels of
poverty. Child mortality and poverty strongly correlated but, unexpectedly, child-
abuse-related deaths did not. Child homicide assailants are extremely rare, but three dis-
tinct within-family assailant categories can be identified: mentally ill parents, mothers
with a child on the Child Protection Register and men with previous convictions for vio-
lence. Mentally ill parents were the most frequent assailants, but violent men killed over
five times the rate of mentally ill parents. The juxtaposed results indicate that the assai-
lants’ problems are essentially psycho-criminological, especially violence, rather than
socio-economic, although poverty worsens most situations. Despite the dangers of
#The Author 2012. Published by Oxford University Press on behalf of
The British Association of Social Workers. All rights reserved.
British Journal of Social Work (2012) 1–36
doi:10.1093/bjsw/bcs051
British Journal of Social Work Advance Access published May 3, 2012
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
‘false positives’, children’s services need to give greater weighting to the child protec-
tion– psychiatric– violence interface to assist front line staff in improving risk assessment
and contribute to reducing the impact that parental mental illness can have on the child.
Keywords: Child protection, psychiatric, criminological, risk assessment
Accepted: March 2012
Introduction
Child protection policy has been driven by high-profile tragedies, from
Dennis O’Neil in the 1940s, to Maria Colwell in the early 1970s and, in the
2000s, by the deaths of Victoria Climbie
´and Peter Connolly. This has led
to blaming the child protection services by media and politicians, so that
the success of child protection has become equated to the avoidance of
such tragedies (Greenland, 1987;HMSO, 1995;NSPCC, 2002;Pritchard
and Williams, 2010). This has led to a bureaucratic defensiveness (Munro,
2011) to avoid ‘worst-case scenarios’. This study will show that, with better
identification of key factors found amongst the very rare fatal child assailants,
risk assessment could be improved, possibly reducing front line staff anxieties.
This re-analysis of the evidence of ‘who killed children’ was prompted by
seeing the picture of Baby Peter Connelly that led to the humane question:
‘How could anybody be so persistently cruel to such a child?’
The authors have struggled with some of the implications of the answers
because they challenge some vitally important social work values, such as
‘acceptance’ of the person. It is argued that social workers must not be
fearful of making evidenced-based judgements because of fear of being
considered ‘judgemental’.
Hard data on child homicide assailants are re-analysed. These data
had previously identified levels of risk of the different type of assailants
(Pritchard, 2004;Pritchard and Sayers, 2008), instead of asking what was
common to the assailants.
In a recent summary of twenty-five years’ research, Professor Baron-
Cohen (2011) asked a similar question about genocide: ‘How could any
human being treat another person in such a way?’ We believe our conclu-
sions will converge in our exploration of trying to better understand
‘Who kills children?’.
Context of child neglect, abuse and poverty
There is considerable evidence that relative poverty is associated with many
negative psycho-social outcomes, such as poorer education, employment
opportunities, greater crime and delinquency (Feinstein et al., 2007;
Page 2 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Wilkinson and Pickett, 2009). Relative poverty is also associated with child
neglect—a feature found in every Western country (Sidebotham et al., 2001;
Blakely et al., 2003;Weissman et al., 2003;Moore, 2005;Mulvaney et al.,
2009). Moreover, poverty is also strongly associated with child mortality,
as, in most Western countries, poorer children die at a disproportionately
higher rate than affluent children (Judge and Benzeval, 1993;UNICEF,
2001;Blakely et al., 2003;Singh and Kogan, 2007;Kitsantas, 2008;Freeman-
tle et al., 2009;Parslow et al., 2009;Anderson and Thomas, 2010). This dis-
advantage persists into adulthood (Mackenbach et al., 2003;Parslow et al.,
2009), giving a further twist to the cycle of intergenerational disadvantage
(Williams and Pritchard, 2006;Feinstein et al., 2007;Conroy et al., 2010).
While there is debate about definitions of poverty (Laderichi et al., 2003;
UNMDG Task Force, 2009), most of the above studies were concerned with
relative poverty, as they came from the unequivocally richest countries in
the world and not the absolute poverty found in the developing world
(Gordon et al., 2000).
One feature that needs to be discussed is the association between socio-
economic disadvantage and mental illness (Dohrenwend and Dohrenwend,
1969). Modern, integrative research highlights the epigenetic interaction,
namely the constant interplay of nurture and nature, highlighting the preva-
lence of psycho-social pathology with poverty (Kahn et al., 2000;Costello
et al., 2003;Conger and Donnellan, 2007).
Hudson (2005, 2007) explored the perennial issue of whether poverty
causes, or is a result of, mental disorder, as there is evidence that supports
the theory of drift (Eaton et al., 2001), which suggests that mental illness
makes people more vulnerable to socio-economic difficulties and conse-
quently more often gravitate into poverty (Saraceno et al., 2005;Hudson,
2005, 2007). Furthermore, Turkheimer et al. (2003) demonstrate the
scientific base of a bio-psycho-social interactive model, where the socio-
economic and psychological environment influences hereditability, impact-
ing in a measurable way upon personal psychopathology (South and
Krueger, 2011). With regard to Child-Abuse-Related Deaths (CARD),
Baird et al. (1999) have shown that ‘neglect’ is measurably different from
abuse, although sometimes they overlap at the extreme. If there is a con-
tinuum from neglect to abuse to the extreme of abuse, the death of a
child, as suggested by a number of authorities (Kempe and Kempe, 1978;
NSPCC, 2002), then it would be expected that the statistical association
of poverty and ‘ordinary’ all-causes-of-death (ACD) would strengthen
and increase along a continuum of probable or actual CARD.
The ‘probable/possible’ abuse-related death categories are the WHO
(2008) mortality categories of ‘Ill-Defined Signs and Symptoms’ (IDSS) (in-
cluding sudden infant death syndrome (SIDS)) and those containing a
degree of violence, namely as ‘Accidents and Adverse Events’ (AAE)
and ‘Undetermined Deaths’ (UnD), which some believe are the most
likely source of under-reported CARD (Newlands and Emery, 1991;
Who Kill Children? Page 3 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Creighton, 1993;Newton and Vandeven, 2006). At the end of the con-
tinuum are confirmed homicides, unequivocally child abuse-related.
This paper challenges the implicit assumption that CARD are primarily
related to socio-economic disadvantage and along a continuum of neglect to
severe abuse to violent death. It is argued that key factors to be found
amongst CARD assailants are essentially psychological, rather than
social, although there is often an overlap. If we can more accurately identify
potential assailants, we can also exclude people as likely assailants, assisting
to more accurately consider the reality of ‘worst-case scenarios’, thus hope-
fully reducing defensive bureaucratisation (Munro, 2011).
To do this, we re-examine macro and micro data and ask four research
questions:
Macro:
1 Will international comparisons of ACD, namely total child mortality (birth to
fourteen), be statistically related to relative poverty?
2 Will CARD be more strongly linked to relative poverty than for ACD of chil-
dren (birth to fourteen)?
3 With respect to relative poverty, will homicides of adults be statistically
correlated?
Micro:
4 Will there be significantly different levels of risk found amongst categories of
a decade of known child homicide assailants?
Methodology
The main themes are enunciated to show the flow of a four-stage argument.
Macro analysis: as context in which to re-analyse a regional study of a
decade of child homicides:
1 Four measures of ‘relative poverty’ will be correlated with the various types of
child mortality.
2 Total children’s (birth to fourteen) deaths are explored along a continuum
from possible to actual CARD.
3 The relative rarity of extreme violence in the UK of adults and children will be
examined. The above three stages will be a context for the micro re-analysis
of a decade of known CARD assailants.
Micro analysis:
4 Re-evaluates a decade of actual cases of child homicide assailants from a total
regional population equivalent to 4 per cent of the UK population, to deter-
mine any differential risks of ‘killing a child’.
Page 4 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Macro (international)-level analysis
Stage 1: Relative poverty (income inequality) and child mortality
The poverty dimension is explored as a context in which to analyse different
mortalities.
Measuring relative poverty: There are a number of scales indicating ‘rela-
tive poverty’ based upon slightly different but often overlapping sets of cri-
teria and four are explored, as Burkhauser et al. (1994) found marked
differences between the scales, but argued for the older and most common-
ly used, the OECD Gina coefficient (ILO, 1998). This measures the distri-
bution of all incomes within an economy and calculates the variation of
people deviating from a theoretically perfect equal distribution. The
authors, however, prefer the more recent work of Wilkinson and Picket
(2009) as most suitable for our purposes; it uses World Bank data to
measure nation-specific levels of relative poverty, namely ‘income inequal-
ity’, which is the gap between the top and bottom 20 per cent of incomes.
Wilson and Picket have been criticised for what we consider a strength,
as, unlike the Gina, which includes all incomes in a total population, they
focus upon the bottom and top fifths to calculate a ratio and, like a
Likert scale, highlight, not blur, the end-points, as can happen with Gina
(d’Ercole and Fo
¨ster, 2005). Moreover, ‘income inequality’ includes more
factors than the other three scales and has demonstrated internationally
that income inequality is statistically related to a range of negative educa-
tional, psycho-social and health outcomes. A third scale comes from the
USA and focuses upon relative fiscal and ethnic-related ‘child poverty’ as
a proportion of such children in the general population, based upon US
definitions of relative poverty (USDHSS, 2006). The most recent scale is
the UNICEF (2010), whose measures are based upon three semi-global
factors: material, educational and health well-being of children. However,
it only reports on sixteen of the nineteen countries to be reviewed. As
these scales have different bases, all four scales will be used to analyse
the mortality data. This then becomes an inter-rator reliability test by cor-
relating each scale with each other and with the death rates found in nine-
teen Western countries. It is believed that this is the first time that all four
poverty scales have been used together to explore child mortality.
Stage 2: Mortality categories
Children’s (birth to fourteen) mortality data are drawn from the WHO
(2008) data and will be reported in rates per million (pm) rather than
actual numbers, to enable comparisons to be made between countries of
differently sized populations. These rates can then be placed in rank
order of highest to lowest and correlated with income inequalities and
the other poverty scales.
Who Kill Children? Page 5 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
The core rationale for examining total child mortality comes from the
UNICEF (2001) statement that ‘child mortality is an indicator of how
well a nation meets the needs of its children’, which implicitly implies
that relative failure is a form of neglect (Pritchard and Williams, 2010).
ACD are total child mortality, expressed in rates per million (pm) for
infant (under one); small child (one to four) and child (five to fourteen)
deaths from which the ACD (birth to fourteen) rate is calculated. As
infants are the most vulnerable to early adverse events and have long
been known to have the highest rates of confirmed homicides (Kempe
and Kempe, 1978;WHO, 2008), the focus will be upon infant (under one)
and children’s (birth to fourteen) mortality rates.
The mortality analysis is for the latest three-year averages and all cat-
egories are classified in the latest International Category of Deaths, the
10th edition (WHO, 1992).
Possible or actual child-abuse-related deaths (CARD): It has been argued
that some under-reported CARD might be inaccurately recorded ending in
SIDS, AAE or the UnD categories, especially the UnD (Creighton, 1993;
Emery, 1993;NSPCC, 2002). This is possible, but only a confirmed homi-
cide can be an unequivocal abuse-related death; however, these ‘possible’
CARD are analysed to counter any argument of under-reporting:
1Ill-Defined Signs and Symptoms (IDSS): The IDSS category does not include
any known violence and is coded R95-99 (International Classification of
Disease (ICD); WHO, 1992). This is includes where there is a death that is
‘a sudden death of unknown causes, excluding sudden cardiac death’ (R96)
and ‘unattended deaths’, where ‘the body was found but no cause could be
discovered (R96-7)’, and finally ‘other ill-defined and unspecified causes of
mortality’ (R99) (WHO, 1992). Crucially, IDSS includes SIDS, coded R95.
Although SIDS had been thought of as possible child abuse-related (New-
lands and Emery, 1991;Emery, 1993;Newton and Vandeven, 2006), later re-
search produced a complexity of findings ranging from poverty,
environmental pollution, possible neurological abnormality, unidentified
virus, premature births, parents who smoke, etc. (e.g. Brookman and
Nolan, 2006;Pharaoh and Platt, 2007;Matturi et al., 2008;Campbell et al.,
2008;Woodruff et al., 2008). It seems evident, therefore, that there are inter-
active factors, probably linked to relative poverty (Turkheimer et al., 2003;
Campbell et al., 2008;Kitsantas, 2008).
2Accident and adverse events (AAE): AAE deaths have a degree of violence
(WHO, 1992), although ‘suspicious road crashes’ are separately reported in
the UnD category (coded Y32; WHO, 1992).
3Undetermined Deaths (UnD): UnD are classified because it was ‘not possible
for the medical or legal authorities to determine whether it was accident, self-
harm or assault’ (WHO, 1992, p. 1095). It includes ‘poisoning ...(ranging
from drugs to vapours and gases) ...hanging ...suffocation ...drowning
Page 6 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
and submersion ...’; but, in each case, ‘intent could not be determined’
(WHO, 1992, p 1095). Thus, a UnD of a infant (under one year) and possibly
a small child (one to four years) might be suspicious, whereas, in an older
child (five to fourteen), these might have been self-inflicted, but sometimes
coroners give an open verdict to protect grieving families and such
deaths are subsequently categorised as undetermined (Linsley et al., 2001;
Stanistreet et al., 2001;Pritchard and Hansen, 2005).
Earlier research on UnD and possible links to abuse suggested that between
10 and 50 per cent of all UnD of under 5s were possibly abuse-related
(Newlands et al., 1991;Emery, 1993). Therefore, to avoid the critique
that we might be missing under-reported abuse deaths, we take an approxi-
mate mid-point of the above assumption (10 – 50 per cent) and combine
confirmed homicide rates with 33 per cent of infant and small child (one
to four) UnD as an estimate of the likely maximum CARD. Interestingly,
a recent Department of Education report combined homicides and un-
determined ‘whether accidentally or purposely inflicted’ together (Sidebo-
tham et al., 2011) in an attempt to determine the maximum of actual or
possible abuse-related deaths. However, with the closer scrutiny that
child deaths receive today (Bennett et al., 2006;Hochstadt, 2006;Jenny
and Isaac, 2006), it seems unlikely that the majority of ill-defined and un-
determined deaths will be abuse-related, but they are included to avoid
under-estimating the extremes of the problem of abuse.
It should be remembered, however, that undetermined and ill-defined
deaths have been associated with miscarriages of justice (Pritchard, 2004)
and this negative publicity has inhibited paediatricians from reporting
suspected abuse, undermining the collaboration of children’s services
(Brookman and Nolan, 2006;King et al., 2006;Williams, 2007), whilst,
equally, we need to be extremely cautious lest we inadvertently add distress
to grieving parents.
4Confirmed Homicides: Deaths are so categorised when there is known to be
an assailant and is the epitome of an abuse-related death.
Statistics
Micro analysis: When using statistics to determine patterns emerging from
the behaviour of small groups of people, it might be argued that the results
should be considered as indicative rather than definitive statistics. The key
is to focus upon what the data mean rather than the absolute numbers,
hence the use of non-parametric statistics in the field of psychology,
social work and education (Guildford, 1978). This is especially important
when dealing with events of real statistical rarity, such as child homicides.
For example, in the UK during 2004–06, the average numbers of confirmed
child (birth to fourteen) homicides and ‘possible’ abuse-related deaths aver-
aged twenty-two, from a population nearly eleven million (WHO, 2008;
Who Kill Children? Page 7 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Pritchard and Williams, 2010), whilst, in a 2009 national cohort of children
aged from birth to seventeen years old, there were seventy deaths possibly
abuse-related, calculated from combining all confirmed homicides and all
undetermined deaths and deaths ‘not yet classified’, whereas police-recorded
deaths as homicides (birth to seventeen years) were just fifty-two
(Sidebotham et al., 2011), indicating the problem of definitive statistics
over time. Moreover, with such small numbers nationally, a multiple
killing of say three or four children in one family quickly distorts the
annual rates. Consequently, a decade of regional child homicide assailants
is, in a statistical sense, a very ‘special sample’ but as stable a population as
can be expected in this problematic field.
As will be seen, the assailants could be designated into three mutually ex-
clusive categories and we wish to determine which of the three groups pro-
portionally killed more often, thus the use of the non-parametric
Chi-square test. This test is a goodness-of-fit test, determining between
the expected frequency and the observed frequency and the degree to
which the results might have occurred by chance (Guildford, 1978;Kanji,
2006). The expected frequency is derived from the proportions in the
samples under scrutiny and what the numbers would be if nothing other
than chance had occurred. The observed frequency is what actually hap-
pened and the difference between the expected and observed is the prob-
ability (p) that something other than chance was operating, from which
degree of statistical significance is calculated; the lowest level of
significance is usually 5 per cent, or expressed as p,0.05. The Chi-square
was also used in the Department of Education’s analysis of Serious
Case Reviews (SCRs), also with relatively small numbers, to
differentiate between family size and numbers of SCRs per annum
(Sidebotham et al., 2011).
Another approach might be described as a commonsense approach to
statistics when dealing with such rare events from within large populations.
These are simple comparative ratios based upon numbers of deaths from
within a well-defined population. For example, out of a five-year cohort
of 215 former excluded-from-school adolescent males now aged sixteen
to twenty-four, three were subsequently found to have murdered, that is
1.4 per cent of this ‘special’ population. However, at a national level, to ran-
domly find one murderer of same age range would require 78,000 males,
that is 0.0013 per cent of the ‘general’ population (Pritchard and Williams,
2010). From the same ‘special cohort’ cohort, two of the 215 young men
committed suicide, that is 0.9 per cent, which is considerably higher than
the rate found amongst mentally ill people at approximately 0.1 per cent
(Pritchard and King, 2004). Moreover, to find at least one suicide in this
age group by chance would require 19,000 men from the general population
(i.e. 0.005 per cent). Hence, to find five statistically rare violent deaths, in
2.3 per cent of this small cohort, or a rate of 23,256 per million (pm), is re-
markably high when compared against the highest UK cause of death in this
Page 8 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
age group—road deaths, at 168 pm (WHO, 2008). This demonstrates the
need to keep in mind the nature and rarity of the events, rather than
numbers per se. So, when examining the micro data, the child homicides
will also be expressed as a ratio of one assailant from within the total of a
well-defined population and compared against the general population by
age and gender of the actual assailants.
Macro analysis: For the analysis of the death rates and relative poverty in
the countries reviewed, we cannot use the Pearson product-moment coeffi-
cient, which can only be used when ‘the two variables X and Y are mea-
sured on continuous metric scales’, so we use the classic the Spearman
Rank Order (Rho) correlation, ‘when the numbers of pairs or N (the vari-
ables) is less than 30. It is even more conveniently applied when the data is
already in terms of rank orders rather than in terms of interval measure-
ments’ (Guildford, 1978, pp 304–5). Thus, each country’s death rates are
given an ordinal rank and juxtaposed against the poverty scale, which can
also be ranked. As rank order correlations do not include all information,
it requires a higher correlation than the Pearson to demonstrate any statis-
tical significance.
Each of the five WHO mortality categories will be correlated with the
Spearman Rank Order (Rho) correlation to test any association with rela-
tive poverty, namely income inequalities (Wilkinson and Picket, 2009) and
three other scales (ILO, 1998;USDHSS, 2006;UNICEF, 2010), to deter-
mine which, if any, has the best statistical fit with the various mortalities.
This statistical approach has been successfully used in a number of recent
cohort and international studies in the fields of medicine and social work
(Hansard, 2011;Pritchard and Williams, 2009,2010,2011;Pritchard et al.,
2010, 2011; Pritchard and Hickish, 2011;Pritchard and Wallace, 2011).
Tables of all the mortality categories from nineteen Western countries
will show the latest available three-year average rates (2004–06). A few
countries’ latest available rates come from earlier years and these are indi-
cated in the tables, such as Italy 2001 – 03 and USA 2003 – 05. However,
Belgium and Denmark are excluded from the study, as their latest data
were 1997 and 2001, respectively, is too great a time gap for comparison
with the other countries.
Stage 3: Rarity of serious violence: context for CARD
To highlight the extreme rarity of those who kill children, we examine the
level of serious violence in the UK population.
Data on violent offenders are given as numbers and in ratios of deaths in
the general population (Home Office, 2009) based upon the assailants’ age.
This is a commonsense approach, as both child and adult homicides in the
UK are very rare and far fewer than children’s road deaths, for example
(Pritchard, 2002;WHO, 2008).
Who Kill Children? Page 9 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Stage 4: Micro: re-analysing decades of regional child homicides and
suicides
The original research, at the police’s invitation, was a regional two-year
cohort study of child sex abusers (CSA) and a decade of confirmed child
homicides in 4 per cent of the UK population (Pritchard, 2004) based
upon national police records. In social science terms, this is a very large
sample, covering a total population of a region, but identifying subgroups
within that population.
Suicide rates came from a confidential regional suicide register, enabling
us to determine the suicide rates of child abuse victims and perpetrators
(Pritchard, 2004;Pritchard and King, 2004). Despite the almost inevitable
overlapping of substance abuse, including alcoholism, it was possible to
ascribe the assailants to three distinct within-family assailant categories
and one extra-family category.
Within-family: The first is mentally ill mothers (MIM) and mentally ill
fathers (MIF), based on records of being treated by a consultant psychiatrist
and/or suicide, or a severe attempt (Pritchard, 2004), and second are mothers
whose children were then on the Child Protection Register (CPR), identified
from Social Service Department records, and ‘step-fathers’, namely males
who were in a parental role to the child, and who had a previous conviction
for violence, identified through national police records.
The number of people who belonged to these categories in the regional
general population (2.4 million) was estimated, matched against the ages
of our actual assailants in the region (men aged nineteen to forty-two and
women aged twenty to thirty-four) from which a homicide rate per
million was calculated for each category of assailant.
The numbers of mentally ill people matching the age of the assailants was
based upon national estimates (Meltzer et al., 1995;Jenkins et al., 1998)of
mental illness in the general population, yielding 13,419 men and 8,746
women. The numbers of men with known convictions for violence came
from police records of 901 men and the numbers of mothers of children on
the CPR (723) are known from social service records. Thus, a murder rate
could be calculated for each category based on the number of actual assailants.
In the initial analysis, there were twenty-seven assailants, who, over a
decade, killed thirty-three children, of whom twenty-two were within-family
assailants, the most frequent being eight MIM (30 per cent) aged twenty to
thirty-four and four MIF (15 per cent) aged nineteen to forty-two years.
In retrospect, it was realised that the high ‘mental ill health’ background
of the assailants was an unwelcome result, fearing it added stigma to men-
tally ill people (Morgan and Bhugra, 2010), and this inadvertently led the
researchers away from focusing upon the commonality of assailants to
examining differential risk levels (Pritchard and Bagley, 2001;Pritchard,
2004;Pritchard and Sayers, 2008). We now appreciate that this was a
mistake!
Page 10 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
The focus should have been upon what the assailants had in common and,
whilst risk levels are important, we failed to appreciate the practice impli-
cations of what was discovered.
Other within-family assailants were ‘step-fathers’, namely men in father
roles to the child, with previous convictions for violence (15 per cent), and
six mothers linked to the Child Protection Register (22 per cent).
The five extra-family assailants had more convictions for non-sex
offences than sexual crimes, as well as being convicted of a sex offence
against children, plus one serious conviction for violence. Consequently,
they were designated Violent-Multi-Criminal-Child-Sex-Abusers
(VMCCSA) (Pritchard, 2004). As this group were outside the family,
they could considered as the responsibility of not the children’s services,
but rather the criminal justice services because of their unequivocal crimin-
ality. This group are reported upon elsewhere (Pritchard and Sayers, 2008)
and little else will be said about them.
The initial epidemiological rate to determine differential risk levels of
assailants found that MIM killed at a rate of 91 pm, whilst mothers with
children on the Child Protection Register killed at a rate of 830 pm and
the violent ‘step-fathers’ at 498 pm per annum.
The extra-family assailants, with their violence and multi-criminality,
were the most dangerous, as these men killed at a rate of 5,102 pm, more
than fifty times the most frequent within-family assailant, the MIM
(Pritchard, 2004).
This study corrects the original error by focusing upon what the assailants
have in common. This approach differentiates between the assailants and a
more accurate picture emerges of those who pose the greatest physical
threat to children.
The Chi-square test is used to examine the differentials of within-family
assailants, as we know the number people in the assailant’s category within
the regional population, enabling us to measure the difference between
expected and observed numbers of assailants—a statistic used in other
studies that focused upon the rarity of homicides, especially in the UK
(Grubin, 1994;Liem and Pridemore, 2012).
The re-analysis is necessary because it was recognised that four of the six
mothers of children on the Child Protection Register had been jointly con-
victed with their male partners because it was not possible to prove whom
the main culprit was. These cases were before the offence of familial homi-
cide was introduced in the 2004 Domestic Violence Crime and Victims Act,
legislation designed to resolve the dilemma of jointly charged assailants
(Drakeford and Butler, 2010). The jointly convicted mothers’ sentences
were between five and seven years, but their violent male partners received
life sentences. This indicated whom the court thought was the primary as-
sailant and justifies our decision to exclude the four mothers as primary
homicide assailants. This changes the risk levels considerably and can
provide a more accurate weighting for future risk assessments.
Who Kill Children? Page 11 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Results
The analysis begins with relative poverty measures as a context for the
national child morality rates.
Macro: poverty–child neglect interface
Table 1lists the income inequality of the nineteen countries, led by the
USA, whose top 20 per cent of incomes were 8.5 times that of the bottom
20 per cent. The UK is third, with a ratio of 7.2 times, whilst Sweden,
Norway, Finland and Japan were less than four times.
Each of the four ‘inequality scales’ significantly and positively correlated
with each other, but the income inequality scale (Wilkinson and Pickett,
2009) has the largest correlation with the Gina, which is the most estab-
lished inequality measure.
Children’s (birth to fourteen) mortality
In parentheses, ACD, IDSS and UnD mortalities are all considerably lower
than before, so, in every country, despite their relative ranking with each
other, the death rates are far lower than the rates in the 1970s (Pritchard
and Williams, 2010).
Table 1 Income inequality (top times bottom 20 per cent rate), Gina coefficient, USBC and UNICEF
relative poverty measures
Country
W and P income
inequality Rank and Gina
USBC Child
Poverty Rank %
UNICEF Child
Global Score
1 USA 8.5 2. 0.381 1. 20.6 1
2 Portugal 8.0 1. 0.385 3. 16.6 6
3UK 7.2 45. 0.335 13. 10.1 4
4 Australia 7.0 11. 0.301 11. 11.8 n/a
5 New Zealand 6.8 4¼. 0.335 8. 15.0 n/a
6 Italy 6.7 3. 0.352 6. 15.5 2
7 Greece 6.2 7¼. 0.321 10. 13.2 3
8 Ireland 6.1 6. 0.328 4¼. 16.3 13
9 Switzerland 5.7 14¼. 0.276 14. 9.4 14
10 Canada 5.6 10. 0.317 7. 15.1 9
11 France 5.6 13. 0.281 15. 7.6 8
12 Spain 5.6 9. 0.319 2. 17.3 5
13 Netherlands 5.3 16. 0.271 12. 11.5 15
14 Germany 5.2 12. 0.298 4¼. 16.3 7
15 Austria 4.8 18. 0.265 15. 6.2 10
16 Sweden 4.0 19. 0.234 19. 4.0 11
17 Norway 3.9 14¼. 0.276 17. 4.6 12
18 Finland 3.7 17. 0.269 18. 4.2 16
19 Japan 3.4 7¼. 0.321 9. 13.7 N/a
Page 12 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
All-Causes-of-Death (ACD)
Table 2shows the total mortality rates in the nineteen countries of children
(birth to fourteen) and baby (under one year).
Infant ACD was highest in the USA (at 6,839 pm) and New Zealand
(5,855 pm), with England & Wales fifth highest (at 5,046 pm). The lowest
were Sweden (2,900 pm) and Japan (2,862 pm).
Focusing on all children’s (birth to fourteen) total mortality, this was led
by the USA (at 2,430 pm), New Zealand (2,105 pm), Canada (1,877 pm)
and England & Wales fifth (at 1,795 pm). The lowest four counties, with
half the rate of the USA, were Japan (1,073 pm), Sweden (1,075 pm),
Finland (1,181 pm) and Norway (1,200 pm).
Ill-Defined Signs and Symptoms (IDSS)
Table 3shows that New Zealand (at 853 pm) and the USA (828 pm) had the
highest infant IDSS rate; England & Wales (at 443 pm) are ranked eighth.
The lowest countries were Italy and Greece (at 180 and 86 pm,
respectively).
Accidents and Adverse Events (AAE)
Table 4lists children’s (birth to fourteen) AAE deaths. New Zealand
(332 pm) and the USA (253 pm) had the highest rates. England & Wales
Table 2 All-cause deaths of infants (,1), small child (1– 4), child (5– 14) and children (0– 14) rates per
million of population 2004 –06 ranked by infant rates and rank income inequalities
Country and years if not 2004– 06 Infant (,1) Small child (1 – 4) Children (0 –14)
1 USA 2003–05 6,839 699 2,530
2 New Zealand 2002 –04 5,855 301 2,105
3 Canada 2002–04 5,287 220 1,877
4 Portugal 2001–03 5,206 356 1,945
5 England & Wales 5,046 309 1,795
6 Australia 2001–03 4,924 256 1,767
7 Netherlands 4,575 229 1,638
8 Italy 2001–03 4,318 200 1,773
9 Switzerland 2003–05 4,248 211 1,525
10 Ireland 4,183 183 1,494
11 Austria 4,086 195 1,466
12 Germany 3,928 201 1,634
13 Greece 3,847 193 1,392
14 France 2003– 05 3,817 214 1,381
15 Spain 2003– 05 3,719 237 1,363
16 Norway 2003– 05 3,273 205 1,200
17 Finland 3,255 224 1,181
18 Sweden 2003– 05 2,900 203 1,075
19 Japan 2,732 251 1,073
Who Kill Children? Page 13 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
(at 58 pm) were ranked thirteenth; Ireland (22 pm) and Sweden (36 pm)
were the lowest.
Undetermined Deaths (UnD)
Table 5lists UnD, led by Switzerland’s infant rates (at 37 pm) and
England & Wales (29 pm), but, as thirteen countries had rates of less
Table 3 Ill-defined and signs and symptoms of infant (,1) and small child (1 – 4) rates per million,
ranked by infant rates
Country and years 2004– 06 Infant (,1) Small child (1–4)
1 New Zealand 2002 –04 853 13
2 USA 2003–05 828 15
3 Germany 600 17
4 France 2003–05 553 22
5 Australia 2001–03 532 17
6 Canada 2002–04 466 19
7 Switzerland 2003–05 450 22
8 England & Wales 443 8
9 Ireland 384 20
10 Sweden 2003–05 339 9
11 Portugal 2001 –03 331 30
12 Austria 326 4
13 Norway 2003–05 293 16
14 Finland 263 5
15 Japan 262 12
16 Spain 2003–05 227 11
17 Netherlands 217 17
18 Italy 2001–03 135 9
19 Greece 2003–05 86 4%
Table 4 Accidents and adverse events rates per million ranked by infant rates
Country for 2004– 06 or other years Infant (,1) Small child (1 – 4) Children (0 – 14)
1 New Zealand 2002 –04 332 115 170
2 USA 2003–05 253 106 141
3 Portugal 2001–03 198 86 119
4 Japan 146 55 77
5 Australia 2001–03 127 89 85
6 France 2003–05 91 60 61
7 Canada 2002–04 86 58 63
8 Spain 2003–05 79 57 59
9 Switzerland 2003–05 73 21 34
10 Italy 2001–03 71 33 47
11 Finland 60 56 57
12 Germany 59 42 42
13 England & Wales 58 30 37
14 Austria 55 47 44
15 Norway 2003–05 53 53 48
16 Netherlands 50 40 38
17 Greece 49 50 49
18 Sweden 2003–05 36 46 48
19 Ireland 22 29 27
Page 14 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
than 10 pm per annum, this highlights the rarity of UnD. For example,
translating the undetermined rates of England & Wales into percentages,
UnD were 0.56 per cent of all total baby deaths.
Homicides—children
Confirmed CARD (i.e. homicides) are shown in Table 6and, again,
the USA tops the table (79 pm for baby and 37 pm for all children (birth
to fourteen)), whilst England & Wales, who used to be fourth highest
(Pritchard and Williams, 2010), are now sixteenth (with an infant rate of
5 pm, and 3 pm for children (birth to fourteen)). Infant homicides were
0.097 per cent of all total infant deaths, statistically very rare, with most
countries having rates of less 20 pm.
Homicides and 33 per cent infant and small child (birth to four)
undetermined deaths
To account for possible unrecognised homicides, Table 7lists infant and
small child abuse deaths (i.e. confirmed homicides) plus a third of undeter-
mined rates for the birth-to-four groups.
Led by the USA at 87 pm for infant rates and 57 pm for infant and small
child (birth to fourteen), fourteen countries have combined rates of less
than 20 pm; England & Wales (at 15 pm) were equal thirteenth, again high-
lighting just how statistically rare homicide of children in the majority of
Western countries is.
Table 5 ‘Undetermined’ deaths infant (,1) and small child (1–4) rates per million ranked by infant
rates
Country, years and rank order Infant ( ,1) Small child (1–4)
1 Switzerland 2003–05 37 7
2 England & Wales 2004– 06 29 8
3¼USA 2003–05 23 3
3¼Germany 23 2
5 Portugal 2001–03 15 22
6 Japan 14 3
7 Netherlands 12 3
8 Canada 2002–04 9 2
9¼Finland 6 4
9¼Ireland 6 3
11¼France 2003 –05 5 1
11¼Sweden 2003 –05 5 1
13¼Australia 2001 –03 1 0
13¼Italy 2001 –03 1 1
17¼Austria 0 0
17¼New Zealand 2002 – 04 0 0
17¼Greece 0 0
17¼Norway 2003 –05 0 0
17¼Spain 2003 –05 0 1
Who Kill Children? Page 15 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Adult homicides (Table 6)
The third column in Table 6lists adult homicide rates in all countries. The highest
three adult homicides were the USA, Finland and Portugal, whose infant homi-
cides ranked first, thirteenth and twelfth, respectively. The lowest three infant
ranks were England & Wales, Spain and Ireland, whose adult rates ranked
Table 6 Homicides of infant (,1), and children (0– 14) rates per million ranked by infant rates and
adult rates
Country, years and infant rank Infant (,1) Children (0– 14) All adult homicides Adult rank
1 USA 2003–05 79 37 66 1
2 New Zealand 2002 –04 54 26 13 6¼
3 Austria 2004–06 38 16 7 14¼
4 Australia 2001–03 34 15 16 4
5 Germany 30 12 6 16
6 Netherlands 27 12 10 10 ¼
7 Canada 2002–04 25 13 14 5
8 Norway 2003–05 24 9 5 17¼
9 Switzerland 2003–05 23 23 9 12
10 Japan 20 10 5 17¼
11 France 2003–05 17 8 7 14¼
12 Finland 13 10 19 2
13 Portugal 2001 –03 9 5 17 3
14 Sweden 2003–05 7 4 10 10¼
15¼Italy 2001 –03 6 4 11 8
15¼Greece 6 3 10 10¼
17 England & Wales 5 3 3 19
18 Spain 2003–05 4 3 13 6¼
19 Ireland 0 1 8 13
Infant and adult homicides Rho ¼+0.1665 n.sig; children and adult homicides Rho ¼+0.2895 n.sig.
Table 7 Combined homicide and 33 per cent of undetermined infant and small child rates per million
Country, years and rank order Infant (,1) Small child (1 – 4) Infant and small child
1 USA 2003–05 87 26 57
2 New Zealand 2002 –04 54 9 32
3 Germany 38 5 22
4 Austria 38 5 22
5 Australia 2001–03 35 7 21
6 Switzerland 2003–05 35 13 24
7 Netherlands 31 7 19
8 Canada 2002–04 28 9 19
9 Japan 25 11 18
10 Norway 2003–05 24 3 14
11 France 2003–05 19 5 12
12¼Finland 15 5 10
125England & Wales 15 5 10
14 Portugal 2001 –03 14 6 10
15 Sweden 2003–05 9 2 6
16 Italy 2001–03 7 3 5
17 Greece 6 1 4
18 Spain 2003–05 4 1 3
19 Ireland 2 1 2
Page 16 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
eighteenth, sixth and equal-thirteenth. There was no significant correlation, with
Spearman correlations of +0.1665, showing the markedly different dynamics in
adult and infant homicides in the majority of countries reviewed.
Correlating mortality and income inequalities
Table 8provides the rank order of all the mortalities and the relative
poverty scales. Only the relationship between income inequalities scales
and mortality will be discussed, as all four poverty scales significantly cor-
related with each other. Crucially, the income inequality scale of Wilkinson
and Pickett (2009) had the highest correlation with the Gina coefficient
(Rho ¼+0.8669, p,0.001). Moreover, on each of the child mortality cor-
relations, the income inequality scale had the highest correlation of the four
scales, indicating that it was the most sensitive in detecting relative poverty
and its impact upon children.
A major finding was that the six countries with the widest income inequal-
ity occupied the five highest death rates, whilst the four countries with the
narrowest relative poverty also had the lowest four infant death rates.
Starting with ACD along a continuum to confirmed CARD, it was found
that income inequality with infant and children’s total deaths were
highly significantly correlated (Rho ¼+0.7694 and +0.7763, respectively;
p,0.001), indicating that child mortality and relative poverty were
linked, although correlation does not necessarily mean cause.
Table 8 Correlating Western world countries mortality and relative poverty scales
Deaths and income inequality with Rho W&P Rho Gina Rho USCP Rho UNICEF
ACD Child (0– 14) +0.7763 +0.5039 +0.4601 +0.3706
,0.001 ,0.025 ,0.025 ,0.05
Ill-Defined Signs and Symptoms +0.3614 +0.1269 +0.1469 +0.0117
,0.1t n.sig n.sig n.sig
Infants and Small Child Accidents and Adverse Events +0.5325 +0.4716 +0.4071 +0.2823
,0.025 ,0.025 ,0.05 n.sig
Children Accidents and Adverse Event +0.5273 +0.5199 +0.3491 +0.3971
,0.01 ,0.025 ,0.01t ,0.1t
Infant and Small Child Undetermined deaths +0.2990 +0.1071 +0.1119 –0.2466
n.sig n.sig n.sig n.sig
Infant homicides +0.2345 – 0.0159 – 0.0939 – 0.1360
n.sig n.sig n.sig n.sig
Children’s homicides +0.0149 – 0.2858 – 0.0444 – 0.1419
n.sig n.sig n.sig n.sig
Infant and Small Child (0 –4) Homicides and 33% UnD +0.0993 – 0.1406 –0.0358 +0.2056
n.sig n.sig n.sig n.sig
Adult homicides +0.3800 +0.3702 –0.1695 +0.1290
n.sig n.sig n.sig n.sig
Inter-correlation of Scales
WP n/a+0.8669 +0.5578 +0.6911
GINA coefficient n/a+0.9022 +0.7566
USCP n/a+0.6184
UNICEF n/a
Who Kill Children? Page 17 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Infant IDSS were not significantly correlated, although there was a stat-
istical trend with the income inequality (Rho ¼+0.3614, p,0.1) but with
none of the other scales (Rho ¼+0.4483). However, there were significant
links with AAE deaths (Rho ¼+0.5325, p,0.25) and with lower correla-
tions for the Gina and USCP but not with the UNICEF scale.
There were no significant correlations with any of the poverty scales and
either infant or small child (birth to four) undetermined deaths
(Rho ¼+0.2990) or infant and children’s homicides (Rho ¼+0.0149);
indeed, there were negative but not significant correlations with children’s
homicides and infant and small child (birth to four) homicides plus a
third of UnD with the Gina, USC and UNICEF poverty scales.
These results demonstrate a positive correlation between poverty and overall
child mortality at a national level, and some link, as might be expected, with
AAE but no correlation between poverty and any of the violent child deaths.
Adult homicide and inequality
Adult homicides and income inequality had the highest correlation
(Rho ¼+0.3800), which fell short of statistical significance. The USCP
and the UNICEF scales both found non-significant negative correlations,
showing that, at national levels, adult murders are different from children’s
homicides, especially of children under five years.
Relative rarity of severe violence (Table 9)
Rates in the general population: adults
The rarity of severe violence in the wider community in England & Wales is
based upon levels of current sentences for violent offences (Home Office, 2009).
In 2008, there were 1,000,000 convictions of ‘violence-against-the-person’
in England & Wales. However, in terms of severity, most were relatively
minor; only 44,000 (4.4 per cent) victims required medical attention
Table 9 Ratios of crime to England & Wales’ population indicating statistical rarity of extreme vio-
lence (confirmed homicides 2004–06)
Category
Population
(millions)
Number of
deaths Ratios
Children’s homicides (0– 14) 9.528 10 1: 952,802
Children’s homicides +33% infant
and small child UnD
9.528 22 1: 433,092
All other homicides 53.728 117 1: 459,214
Serious violent offenders if all homicides 21,320 127 1: 168
Males (19– 42): re children 8.962m 10 1: 896,259
Re all homicide 127 1: 70,569
Page 18 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
(Home Office, 2009). This is not to discount the distress to individuals, but it
places the figures in a comparative context in relation to serious violence.
The latest average (2004–06) number of all confirmed homicides in
England & Wales (WHO, 2008) was 127 (or 3 pm or 0.0003 per cent).
Even in the USA, which has long had the highest homicide rate in
Western countries, whilst there was an average of 17,887 annual homicides,
equivalent to 61 pm, this is only 0.0061 per cent of their general population
(WHO, 2008).
If we assumed that the latest average annual homicide assailants in
England & Wales (127) were of the age of the regional child assailants
(nineteen to forty-two years), then, of the 8,962,258 of this age in the
general population, it would require 70,569 male assailants to randomly
find one murderer.
Rates of prison inmates
Probably the greatest concentration of people with chaotic lives is the
82,000 men currently in prison (Home Office, 2009); 21,320, or 26 per
cent, had sentences for serious violence, namely ABH to murder, including
sex crimes. If we assumed these 21,320 seriously violent offenders were re-
sponsible for the total Anglo-Welsh 127 homicides, this would give a ratio of
1:168, which is 400 times the rate of the nineteen-to-forty-two-year-old men
in the general population. This stresses the relative rarity of the seriously
violent offender, even from amongst the most ‘disturbed’ population, but
give emphasises to the weighting that should be linked to previous convic-
tions when undertaking risk assessments.
Rarity of extreme violence to children (birth to fourteen)
The national average infant and children’s homicide rates were 5 and 3 pm,
respectively (Pritchard and Williams, 2010), and, even when combining
birth to four-year-old homicides with a third of UnD from within a popula-
tion of 3.120 million children, this would yield a ratio of only one in 141,812
infants.
This evidence might seem to challenge the experience of front line practi-
tioners, who frequently come across hostile, aggressive and abusive beha-
viours. However, it is important to recognise that, where actual extreme
violence exists, it highlights the exceptional nature of that person’s behav-
iour. This is demonstrated in estimates that 1 per cent of all A&E admissions
are possibly abuse-related (Woodman et al.,2008), which would yield 4,144
children aged birth to four years in England (Department of Health, 2011).
This equates to eighty children per week, or eleven children per day, but,
when contrasted against the maximum estimated twenty-two CARD of all
Who Kill Children? Page 19 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
children in 2004–06, it is a ratio of one abuse-related death to 188 possible
abuse-related A&E admissions of under four-year-olds, and is indicative of
the very special nature of people who actually kill children.
Decade of child homicide assailants: identifying the commonalities
(Table 10)
Within-family assailants
In the re-analysis of the Wessex child homicides, the most frequent assai-
lants were MIM (eight), who killed at a rate equivalent to 91 pm. The
four MIF killed at a rate of 30 pm, thus, overall, 67 per cent of the eighteen
within-family assailants were mentally ill parents.
We excluded the four mothers with a child on the Child Protection Regis-
ter from the analysis, because they had been jointly charged with their male
partner and the court did not directly identify the primary assailant, as each
carer had blamed the other. This left two mothers with a child on the Child
Protection Register who were the primary assailants and they killed at a
rate of 227 pm, more than two-and-a-half times the most frequent assailant,
the MIM.
However, the four male assailants who were not biological parents all had
previous convictions for serious violence and their assailant rate was
498 pm, more than five times the rate of the MIM and more than sixteen
times that of the MIF.
Translating these rates into equivalent ratios per annum, over the decade,
one in 33,557 MIF killed, one in 10,933 MIM, one in 3,615 mothers with a
child on the Child Protection Register and one in 2,007 ‘step-fathers’ with
previous convictions for violence.
When the rates are translated into ratios for men and women of the same
age in the general population, one in 1,493,710 men aged nineteen to
Table 10 Who kills children (0–14) numbers primary assailant ratios to population and rates per
million, per annum in a decade of Wessex and annual England & Wales homicides
Category and numbers Population Ratio to population
1 p.a.
Rates pm per annum
General population
Men (19– 42) (6 pa) 8,962,258 1: 1,493,710 0.67 pm
Women (20 –34) (4 pa) 5,252,575 1: 1,313,144 0.76 pm
Wessex
Mentally ill parents* 13,419 1: 33,557 pa 30 pm
Fathers 4 (0.4 pa) 8,746 1: 10,933 pa 91 pm
Mothers 8 (0.8 pa)
Mother CPR 2 (0.2pa) 723 1: 3,615 pa 277 pm
Previous violence
SFPVC 4 (0.4 pa) 803 1: 2007 pa 498 pm
VMCCSA 5 (0.5 pa) 98 1: 196 pa 5,102 pm
Page 20 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
forty-two killed a child (or 0.67 pm) and one in 1,313,144 women aged
twenty to thirty-four killed a child. Thus, the ‘step-father’ with previous vio-
lence killed more than 740 times more than the same-aged male in the
general population. The MIM killed more than 120 times more than
women in the general population.
With such small numbers of within-family assailants, however, can we
meaningfully compare between the three categories, to determine levels
of dangerousness? We believe we can.
Table 11 shows the results of the Chi-square test and gives the expected
frequency of assailants, based upon the proportions of perpetrators from
within the different categories. It is shown that it was the step-fathers
with previous convictions who had a significantly higher assailant rate
than would have been expected (x
2
¼26.2793, 3 d/f, p,0.01), indicating
that, proportionately, these men were far more dangerous than mentally
ill parents of either sex.
The finding that the violent men were the most dangerous is given further
support when looking at the extra-family assailants. All were men with
VMCCSA records, who killed at a rate of 5,102 pm, more than fifty times
the most frequent assailant, and, over a decade, annually one in 196
killed a child.
Study’s limitations
The study has a number of limitations; the first is that it might be thought to
contain two separate studies, namely the macro analysis of child mortalities
and relative poverty in the nineteen Western nations and the micro re-
analysis of the decade of child homicides. However, we believe that bringing
the two together strengthens the argument about the special nature and cen-
trality of the psycho-criminality of the child homicide assailants. For, whilst
the finding of a strong statistical link between total child mortality and
poverty was to be expected, the lack of a statistical link to violence-related
child deaths at a national level was unexpected but suggests that factors
other than poverty were operating within this statistically very rare group.
The third limitation was the exclusion of the jointly charged mothers as
co-equal assailants in the micro study, although this appeared justified by
Table 11 Comparing within-family assailants from the regional population
Categories of
assailant
Assailant’s category
in population
Expected numbers
of assailants
Observed numbers
of assailants
Mentally ill mothers 8,746 6.6 8
Mentally ill fathers 13,419 10.2 4
Mothers CPR 723 0.5 2
Violent ‘step-fathers’ 901 0.68 4
x
2
¼Differentiating within-family Wessex assailants; x
2
¼26.7279, 3 d/f, p,0.001.
Who Kill Children? Page 21 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
the court’s subsequent sentences (Drakeford and Butler, 2010). Fourth is
the small numbers of assailants, but, as Grubin (1994) states, homicides,
even at national levels, are ‘special’ and small samples of twenty can
provide valuable information in dealing with such rare statistical phenom-
ena a recent European-wide study of homicide confirms (Liem and
Pridemore, 2012).
Fifth, by focusing only upon child fatalities, we might be underestimating
the level of extreme physical violence, as not all such events will end fatally.
Consequently, a commonsense approach helps to resolve this by looking at
differential ratios of assailants along a continuum of general population
compared with assailant categories. This approach was utilised in the
NIHR programme that analysed more than sixty-six studies concerned
with children admitted to A&E for a physical injury (Woodman et al.,
2008). They showed that current UK assessment models were not very re-
liable but that the younger the child, especially under four years, the more
likely that subsequent abuse was discovered and it was estimated that 1 per
cent all A&E admissions were probably abuse-related (Woodman et al.,
2008). Taking this estimate to the 2006–07 A&E admissions in England
of more than 400,000 children (birth to four) (Department of Health,
2011), this would equate to 4,134 children thought to be possibly physically
abused, which again highlights just how relatively rarely possible
abuse-related events result in a fatality. Conversely, caution is needed
lest we over-include undetermined or ill-defined deaths and inadvertently
stigmatise innocent parents (Pritchard, 2004) and this shows how difficult
it is, if not impossible, to gain any absolute precision in a field in which
many of the actors will be seeking to hide their possibly abusive behaviour.
Our reading of the Woodman et al. (2008) critique of current assessments
is that the various tools focused mainly on the child and not primarily their
carers, the most likely assailants.
It is acknowledged that the use of aggregated national data cannot tell us
about individuals and there is the danger of the ecological fallacy of projecting
national results onto individuals (Bland, 2000), which is discussed by Bland in
detail (pp. 42–3), which is also a problem for the recent review of SCRs
(Sidebotham et al., 2011), as different categorisations throw up different
rates. For example, assailants who kill children under seven years old are
markedly different from those who kill older children and, in the Wessex
study, the pre-seven-year-old victims were all killed by parents, whereas
those older than eight-and-a-half had an extra-family assailant (Pritchard,
2004)—a finding noted in other European studies (Liem and Pridemore, 2012).
Nonetheless, it is acknowledged that this Wessex indicative study, based
upon individuals from population-level data, cannot be regarded as conclu-
sive, but rather is hypothesis-generating that requires replication and
further research, if possible on bigger samples and based upon common cri-
teria. However, the juxtaposition of micro and macro statistical results
helps to overcome the possible ecological fallacy (Bland, 2000), as the
Page 22 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
results point in the same direction: the child protection–psychiatric–crim-
inological interface.
Finally, there is the danger of false positives. Although this study high-
lights the greatly increased risk posed by a male with a history of serious
violence, compared with men in the general population, it must be recog-
nised that the vast majority of males with a history of serious violence do
not go on to kill, either children or adults. An even stronger qualification
must be made in relation to people with mental illness, but it will be
argued that these two factors should be given far more weight than previ-
ously in any risk-assessment tool.
The challenge facing the profession in seeking to predict harm, in order to
protect the most vulnerable, will mean rethinking our approach to current
risk-assessment tools.
Discussion
Main finding
The four research questions considered can be answered positively. First, it
was possible to differentiate between the known assailants and highlight
that, despite the majority of within-family assailants being mentally ill, it
was the men with previous convictions for violence, both within-family
and extra-family assailants, who posed the far greater comparative risk.
Second, using four different scales to measure relative poverty, all were
strongly correlated with total children’s (birth to fourteen) mortality, the
strongest correlation being the income inequality scale (Wilkinson and
Picket, 2009).
Third, whilst there was a small, significant correlation with poverty and
AAE deaths, there were NO significant links with ill-defined deaths, or
UnD, theoretically the most likely source of ‘hidden’ abuse deaths, or homi-
cide, nor the combined UnD and homicides.
Fourth, there was no statistically significant correlation between adult
homicides and any of the relative poverty measures. Whilst some clinical
studies have shown a link between inequality and adult murder (Daly and
Wilson, 2001;Shaw et al., 2005;Awofeso, 2008;Birken et al., 2009;Liem
and Pridemore, 2012), the Gartner (1990) study from the USA, however,
found that this was not true for US child murderers, supporting these and
other findings that the dynamics of child and adult homicides are different
(Bennett et al., 2006;Cavanagh et al., 2007;Liem and Koenraadt, 2008).
Lessons from evolutionary psychology
New research on child homicide comes from the field of evolutionary
psychology, as Daly and Wilson (1988,2001,2002,2006) suggested that
men in the role of step-fathers, not having a genetic investment in their step-
Who Kill Children? Page 23 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
children, are more likely to be overtly aggressive; this is supported by other
research (Weekes-Shackelford and Shackleford, 2004) that identified a
pattern similar to that found amongst other primates, which fits our findings
on the violent ‘step-fathers’. Recent neuroscience research indirectly sup-
ports Daly and Wilson, as it was found that blood fathers bond with their
newborn child when they share the first six months of child rearing
(Toney, 1983). Furthermore, neuroscience and endocrinology research
shows that different levels of neurotransmitters and testosterone are
lowered in new biological fathers sharing the child rearing, apparently influ-
encing the new father’s bonding with the child by reducing potential male
aggression (Seifritz et al., 2003;Grey et al., 2006;Feldman et al., 2007,
2010;Gordon et al., 2010). This appears to be an intrinsically protective
mechanism that is not ‘switched’ on with non-biological fathers who
come later into the family. This highlights the constant interaction of the
bio-psycho-social of human behaviour, shown in the above neuroscience-
type studies that should move us away from the sterile
nurture-versus-nature debate. Thus, the case for insisting on adequately in-
cluding all father figures in assessing child protection risks is given further
emphasis. This is supported by various studies of SCRs (Brandon et al.,
2008;Ofsted, 2010) and urged by Munro (2011).
Conversely, a non-violent, non-abusing step-father joining a family can
be very positive for the child and help to raise the family out of poverty
(Boggess, 1998;Musick and Meier, 2010), which is a further reminder of
the danger of misconstruing trends and of possible false positives in any
assessment.
Daly and Wilson (2001,2002,2006) expanded their ideas about ex-
tremely violent behaviour, including homicide in which victim and assail-
ant were unrelated, with the focus on male dominance in situations in
which males were threatened, especially with scarce resources. They
bring together their evolutionary psychological explanation with the
sociological, in an ingenious study linking income inequality and homi-
cide in North America (Daly and Wilson, 2001). However, whilst this
profile matches our ‘step-fathers’ result and their associated links with
socio-economic disadvantage and multi-criminal backgrounds (Zagar
et al., 2009), it does not fit the mentally ill biological parents who
killed. This highlights the extent of the psychopathology of a parent
killing their children, often triggered in delusional hallucinatory states,
where the parent kills ‘in love’ to protect their child from some delusion-
al impending tragedy (Pritchard, 2004,2010,2012;Liem and Pridemore,
2012). We must never forget that these tragedies are fortunately ex-
tremely rare and that, whilst mentally ill parents are the most frequent
assailants, they are proportionately far less dangerous than men with pre-
vious convictions of serious violence.
Page 24 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
The child protection, criminological and psychiatric interface
This study highlights the fact that people who kill children are likely to have
a history of serious violence or, and to a lesser but significant degree, mental
illness. In effect, we have rediscovered the psychiatric –criminological inter-
face with child protection and, although very rare, mentally ill or personal-
ity disordered parents can be a threat to children, especially if the child is
included in some delusional or hallucinatory state. Furthermore, it is recog-
nised that mothers with schizophrenia, as well as those with a severe person-
ality disorder, sometimes have long-standing difficulties, especially if
compounded by substance abuse problems (Howard et al., 2003;Salmon
et al., 2004;Wan et al., 2007;Elkin et al., 2009;Schechter and Willheim,
2009), whilst there is long-standing evidence of a psychiatric link to the
homicide of children, albeit statistically rare (Falkov, 1996;Friedman
et al., 2005;Dil et al., 2008;Kim et al., 2008;Liem and Koenraadt, 2008;
West et al., 2009). It is suggested that there is a reluctance to acknowledge
this issue, lest further stigma is added to the mentally ill (Pritchard, 2004,
2010). Yet, it can and must be stressed that the vast majority of mentally
ill parents do not harm their children and that, given appropriate and ad-
equate services, they can be maintained safely in the community. Indeed,
we would argue that the presence of children whose parents have a psych-
osis, severe depression or an anti-social personality should be a trigger and
an opportunity to ensure optimal integrated psychiatric care for the whole
family that can make such a difference to patients’ lives (Fenton and
Schooler, 2000;Liberman and Glick, 2004;Pritchard, 2006;Hides et al.,
2010;Gleeson et al., 2010).
But what of the men of violence, who are a greater risk to children than
the mentally ill? Here, we draw upon Baron-Cohen’s (2011) synthesising
work as he defines, with impressive neuroscience-based evidence, that
such men can be designated as having severe personality disorders, demon-
strating what he described and measures as men with ‘zero empathy’.
As described in the Diagnostic Statistical Manual (APA, 2010), people
with severe personality disorders have long-standing characteristics and
relative persistent behaviour, which, if it includes violence, often starts at
a relatively young age. This fits the designation of personality disorder
(APA, 2010) or ‘psychopath’ explored in Baron-Cohen’s (2011) integrated
psycho-social, neuroscience and endocrinological research study.
Furthermore, criminologists have long known that the most consistent in-
dicator of future violence is previous violence (Farrington and Loeber,
2000;Langevin, 2003;Thornton et al., 2003;Urbaniok et al., 2006;Collins,
2008;Rettenberger et al., 2010;Rossegger et al., 2011a,2011b).
It is appreciated that such a bald statement runs counter to a key social
work value of being non-judgemental and might be thought to be against
natural justice because a person should not be condemned more than
Who Kill Children? Page 25 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
once for the earlier offence. However, being afraid of appearing to be
judgemental should not stop us making evidence-based judgements and,
along a chain of vulnerability, the potential child victim must take
precedence.
External validation for the psychiatric and criminological child
protection interface
Comparative research is always stronger if there is a control or comparison
group. The recent seminal study of Brandon et al. (2008) on SCRs is a very
reasonable ‘control’ group for our sample, as she concentrated upon a
group of forty cases, which involved thirty-four child deaths compared to
the original Wessex cohort of thirty-three deaths. In Brandon et al.’s inten-
sive cohort, 53 per cent of cases involved male domestic violence and 63 per
cent of parents were either currently suffering from or previously had suf-
fered from a mental disorder, which gives some support to our results,
which are similar. Moreover, the recent Ofsted (2010) review of SCR
again found similar levels of violence and recent or current mental illness
in the parent carers. Both SCR studies reflect the important findings of
Cavanagh et al. (2007), who also highlighted the risk that men with previous
violence pose.
It may be asked, given the relative prevalence of the psychiatric and vio-
lence dimensions noted in Brandon et al. (2008) and Ofsted (2010), why
these features were not given greater prominence in the reports. In all prob-
ability, it was because their focus was upon procedures and processes within
the SCRs, rather than on the ‘content’ of the family situations.
Interestingly, a new study of Serious Case Reviews in thirteen counties
also found similar proportions of domestic violence and mental illness in
the families as the above SCR studies. The research used Baird et al.’s
(1999) actuarial-based assessment tool that was able to accurately identify
more than 80 per cent of children at risk of re-abuse, who were re-abused
(Wood, 2012). However, the risk-assessment tool did not differentiate
between those children who were subsequently killed and those who
lived, apparently because it did not weight the violence and psychiatric
components.
Consequently, despite our small sample, there is a very strong case for
substantially weighting ahistory of serious violence and/or mental disorder
in any future risk assessment, whilst at the same time always being aware of
false positives.
The key to identifying potential assailants is to recognise that it is their
psycho-psychiatric situation, rather than their social position per se,
though poverty invariably makes a difficult situation worse. The greatest
weighting should therefore be given to men with histories of serious
Page 26 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
violence and, in practical terms, living with children who are not their bio-
logical children.
Policy and practice implications
In practice, the Framework for Assessment for Children in Need and Their
Families (Department of Health, 2000) continues to be used by child-care
social workers as a key risk-assessment tool, although there is a degree of
ambivalence about its effectiveness (Brandon et al., 2006;White et al.,
2009), and it is now criticised by Munro (2011). It is essentially an ecological
approach that gives equal consideration to the child’s developmental needs,
parenting capacity and the family and environmental factors (Brandon
et al., 2008). This is where we emphasise the need to give weighting to
the significant question ‘Has any of the adults a history of serious violence
or a mental health problem?’, and recognise the issues of recidivism in both
violent and sexual offenders (Thornton et al., 2003;Collins, 2008;Gibbon
et al., 2010;Rettenberger et al., 2010). We argue for hard but evidence-
based case-specific judgements to be made that give greater recognition
of the child protection–psychiatric – criminological interface. Of course, im-
provement in inter-agency and interdisciplinary collaboration would be
required; for example, adult psychiatrists need to consider ‘child protection’
with regard to any patient with a severe personality or psychotic disorder,
and ask themselves whether there is a child involved, whether there is vio-
lence present and what the implications are.
However, social workers in the field complain that, despite their knowing
families for months, the psychiatrist’s ‘judgement’ after but a relatively
brief contact often outweighs their views, with the danger of over-
medicalising the situation (Pritchard, 2006). In effect, two over-stressed ser-
vices are in danger of entering the ‘blame game’ and having too narrow a
focus, whereas a new partnership is needed based upon recognition of
mutual contributions.
Equally, the social worker should enquire whether there is a history of
serious violence and/or a mental health problem and act accordingly.
It is appreciated that many psychiatrists might be wary of further adding
stigma that may appear to be a threat to their patients’ maintaining their
children at home. Yet, in cases of mental illness, if there is good integrated
treatment, most situations can be safely and effectively managed (e.g. Fenton
and Schooler, 2000;Pritchard, 2006,2012;Gleeson et al., 2010;Hides et al.,
2010). However, as yet, we are less sanguine about severe personality disor-
ders (Gibbon et al., 2010) or what Baron-Cohen might describe as the
‘zero-empathetics’.
In the context of serious violence, social workers carrying out child pro-
tection risk assessments need to have access to data relating to serious crim-
inality to make better-informed risk assessments, especially to ensure
Who Kill Children? Page 27 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
adequate exploration of ‘father figures’ in the child’s family (Munro, 2011).
There should be a requirement for social workers trained in child protection
to also have training to ensure their capacity to understand the impact of
violent criminality or mental illness upon the child.
Nonetheless, we have to warn about misconstruing the child protection –
psychiatric axis lest there be too many false positives, as we need less, not
more defensive bureaucracy that overly fears the ‘worst-case scenario’
that so often inhibits front line staff.
Finally, it is hoped that, in re-examining the evidence of ‘who kills chil-
dren’, this paper may contribute to the understanding of the interface
between child protection, psychiatry and violent criminality and move
towards a better, more effective, more efficient assessment of extreme
risk to a child.
Acknowledgements
We are indebted to Sister Beryl Pritchard for her diligent archival work and
to Professor Peter Thomas, Professor of Health Care Statistics and Epi-
demiology, Bournemouth University, for his helpful statistical advice.
References
APA (2010) Statistical Diagnostic Manual, 4th edn, Washington, American Psychological
Association.
Anderson, L. and Thomas, B. (2010) ‘Inequalities in premature mortality in Britain: Ob-
servational study from 1921 to 2007’, British Medical Journal,341, p. c3639.
Awofeso, N. (2008) ‘Poverty, murder and respect: Unlikely to be linked in the UK’,
British Medical Journal,337, p. 911.
Baird, C., Wagner, D., Healy, T. and Johnson, K. (1999) ‘Risk assessment in child pro-
tective services: Consensus and actuarial model reliability’, Child Welfare,78, pp.
723–48.
Baron-Cohen, S. (2011) Zero Degrees of Empathy: A New Theory of Human Cruelty,
London, Allen Lane.
Bennett, M. D., Hall, J., Frazier, L., Jr, Patel, N. and Shaw, K. (2006) ‘Homicide of chil-
dren aged 0–4 years, 2003 – 04: Results from the National Violent Death Reporting
System’, Injury Prevention,12(Suppl. 2), pp. ii39–ii43.
Birken, C. S., Parkin, P. C., To, T. and Macarthur, C. (2009) ‘Neighbourhood socio-
economic status and homicides among children in urban Canada’, Pediatrics,123,
pp. 815–19.
Blakely, T., Atkinson, J., Kiro, C., Blakelock, A. and D’Souza, A. (2003) ‘Child mortality,
socio-economic position and one-parent families: Independent associations and vari-
ation by age and cause of death’, International Journal of Epidemiology,32, pp. 410 –
18.
Bland, M. (2000) An Introduction to Medical Statistics, 3rd edn, Oxford, Oxford Univer-
sity Press.
Page 28 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Boggess, S. (1998) ‘Family structure, economic status and educational attainment’,
Journal of Population and Economics,11, pp. 205–22.
Brandon, M., Belderson, P., Watson, C., Howe, D. and Black, J. (2008) ‘Analysing child
deaths and serious injuries through abuse and neglect: What can we learn? A bi-
annual analysis of serious case reviews 2003–2005’, Department of Children,
Schools and Families, RB023 Research.
Brandon, M., Howe, A., Dagley, V., Salter, C., Warren, C. and Black, J. (2006) Evaluat-
ing the Common Assessment Framework and Lead Professional Guidance and Imple-
mentation 2005–06, London, Department for Education and Skills.
Brookman, F. and Nolan, J. (2006) ‘The dark figure of infanticide in England and Wales:
Complexities of diagnosis’, Journal of Interpersonal Violence,21, pp. 869–89.
Burkhauser, R. V., Smeeding, T. M. and Merz, J. (1994) Relative Inequality and Poverty
in Germany and the United States Using Alternative Equivalence Scales, Discussion
Paper No. 13, Forschungsinstitut Freie Berufe Universitat Luneburg.
Campbell, M. J., Hall, D., Stephenson, T. and Madan, J. (2008) ‘Recurrence rates for
SIDS: The importance of risk stratification’, Archives of Childhood Diseases,93,
pp. 936–9.
Cavanagh, K., Dobash, R. and Dobash, R. P. (2007) ‘The murder of children by fathers in
the context of child abuse’, Child Abuse and Neglect,31, pp. 731 –46.
Collins, R. (2008) Violence: A Micro-Sociological Theory, Princeton, Princeton Univer-
sity Press.
Conger, R. D. and Donnellan, M. B. (2007) ‘An interactionist perspective on the socio-
economic context of human development’, Annual Review of Psychology,58, pp.
114–27.
Conroy, K., Sandel, M. and Zuckerman, B. (2010) ‘Poverty grown up: How childhood
socioeconomic status impacts adult health’, Journal of Development Behavioural
Paediatrics,31, pp. 154–60.
Costello, E. J., Compton, S. N., Keeler, G. and Angold, A. (2003) ‘Relationships between
poverty and psychopathology: A natural experiment’, Journal of the American
Medical Association,290, pp. 2023–9.
Creighton, J. (1993) ‘Children’s homicide: An exchange’, British Journal of Social Work,
23(4), pp. 643–4.
d’Ercole, M. and Fo
¨rster, M. (2005) Income Distribution and Poverty in OECD Countries
in the Second Half of the 1990s, OECD Social, Employment and Migration Working
Papers, Paris, OECD.
Daly, M. and Wilson, M. (1988) ‘Evolutionary social psychology and family homicide’,
Science,242, pp. 519–24.
Daly, M. and Wilson, M. (2001) ‘Income inequality and homicide rates in Canada and the
United States’, Canadian Journal of Criminology,4. pp. 219–36.
Daly, M. and Wilson, M. (2002) ‘Murderous parents’, International Journal of Therapy
and Compensatory Criminology,April, pp. 123–43.
Daly, M. and Wilson, M. (2006) ‘Stepfather involvement and adolescent well-being: Do
mothers and non-resident fathers matter?’, Journal of Family Issues,September, pp.
1191–213.
Department of Health (2000) Framework for the Assessment of Children and Need and
Their Families,London, The Stationery Office.
Department of Health (2011) Trends in Children and Young People’s Care: Emergency
Admission Statistics 1996/97–2006 – 07, London, Department of Health.
Who Kill Children? Page 29 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Dil, L. M., Doreleijers, T. A. and Schoevers, R. (2008) ‘Filicide, psychiatric disorders in
parents who murder their children: A literature review’, Tidschirie Psychaitrica,50,
pp. 263–72.
Dohrenwend, B. P. and Dohrenwend, B. S. (1969) Social Status and Psychological Dis-
order: A Causal Inquiry, New York, Wiley.
Drakeford, M. and Butler, I. (2010) ‘Familial homicide act and social work’, British
Journal of Social Work,40, pp. 1419–33.
Eaton, W. M., Mutaner, C., Bovasso, G. and Smith, C. (2001) ‘The role of inter- and intra-
generational mobility, government assistance and work environment’, Journal of
Health and Social Behaviour,42, pp. 277–94.
Elkin, A., Gilburt, H., Slade, M., Lloyd-Evans, B. and Howard, L. M. (2009) ‘A national
survey of psychiatric mother and baby units’, Psychiatric Services,60, pp. 629–33.
Emery, J. L. (1993) ‘Child abuse, sudden infant death syndrome and unexpected infant
death’, American Journal of Diseases of Childhood,147, pp. 1097–100.
Falkov, A. (1996) Fatal Child Abuse and Parental Psychiatric Disorders, ACPC
Series, Report No. 1, London, Department of Health Study of Working Together,
Part 8.
Farrington, D. P. and Loeber, R. (2000) ‘Epidemiology of juvenile violence’, Child and
Adolescent Psychiatric Clinics of North America,9, pp. 733–48.
Feinstein, L., Hearn, P., Renton, Z. and Abrahams, C. (2007) Reducing Inequalities: Re-
leasing the Talents of All, London, Institute of Education.
Feldman, G., Hayes, A., Kumar, S., Greeson, J. and Laurenceau, J. P. (2007) ‘Mindful-
ness and emotion regulation: The development and initial validation of the Cognitive
and Affective Mindfulness Scale-Revised (CAMS-R)’, Journal of Psychopathology
and Behavioural Assessment,29, pp. 177–90.
Feldman, R., Gordon, I., Schneiderman, I. and Weisman, O. (2010) ‘Natural variations in
maternal and paternal care are associated with systematic changes on oxytocin follow-
ing parent-infant contact’, Psychoneuro-endrocinology,35, pp. 1133–41.
Fenton, W. S. and Schooler, N. R. (2000) ‘Evidence-based psychosocial treatment for
schizophrenia’, Schizophrenia Bulletin,26, pp. 1–3.
Freemantle, N., Wood, J., Griffin, C., Gill, P., Calvert, M. J. and MacArthur, C. (2009)
‘What factors predict differences in infant and perinatal mortality in primary care
trusts in England? A prognostic model’, British Medical Journal,339, p. 2892.
Friedman, S. H., Hrouda, D. R., Holden, C. and Resnick, P. J. (2005) ‘Child murder com-
mitted by severely mentally ill mothers: An examination of mothers found not guilty
by reason of insanity’, Journal of Forensic Science,50, pp. 1466–71.
Gartner, R. (1990) ‘The victims of homicide: A temporal and cross-national review’,
American Sociological Review,55, pp. 92–106.
Gibbon, S., Duggan, C., Stoffers, J., Hubbard, N. and Lieb, K. (2010) ‘Psychological
interventions for antisocial personality disorders’, Cochrane Database Systematic
Review, CD007668.
Gleeson, J. F., Cotton, S. M., Alvarez-Jimenez, M., Wade, D. and McGorry, P. D. (2010)
‘Family outcomes from a randomised control trial of relapse prevention therapy in
first-episode psychosis’, Journal of Clinical Psychiatry,71, pp. 475–83.
Gordon, D., Levitas, R., Pantazis, C., Payne, S. and Townsend, P. (2000) Poverty and
Social Exclusion in Britain, York, Joseph Rowntree Foundation.
Gordon, I., Zagoory-Sharon, O., Leckman, J. and Feldman, R. (2010) ‘Prolactin, oxyto-
cin and the development of paternal behaviour across the first six months of father-
hood’, Hormones and Behaviour,58, pp. 513–18.
Page 30 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Greenland, C. (1987) Preventing Child Abuse and Neglect Deaths, London, Tavistock
Publications.
Grey, P. B., Yang, C. F. and Pope, H. G. (2006) ‘Fathers have lower salivary testosterone
levels than unmarried men and married non-fathers in Beijing, China’, Proceedings of
Biological Science,273, pp. 333–9.
Grubin, D. (1994) ‘Sexual murder’, British Journal of Psychiatry,165, pp. 624 –9.
Guildford, J. P. (1978) Fundamental Statistics in Psychology and Education, London,
McGraw-Hill.
Hansard (2011) Health: Question ‘To ask Her Majesty’s Government what action will
they take in response to Professor Colin Pritchard’s study published recently in the
British Journal of Cancer?’, House of Lords, 733.240 Tuesday 13 December, pp.
1159–62.
Hides, L., Samet, S. and Lubman, D. L. (2010) ‘Cognitive behaviour therapy (CBT) for
the treatment of co-occurring depression and substance abuse: Current evidence and
future research’, Drug and Alcohol Review,29, pp. 508 –17.
HMSO (1995) Child Protection: Messages from Research, London, HMSO.
Hochstadt, N. J. (2006) ‘Child death review teams: A vital component of child protec-
tion’, Child Welfare,85, pp. 653–70.
Home Office (2009) Prison Statistics, London, Home Office.
Howard, L., Shah, N., Salmon, M. and Appleby, L. (2003) ‘Predictors of social services
supervision of babies of mothers with mental illness after admission to a psychiatric
mother and baby unit’, Social Psychiatry and Psychiatric Epidemiology,38, pp.
450–5.
Hudson, CG (2005) Socioeconomic status and mental illness: Test of the social causation
and selection hypothesis. American Journal of Orthopsychiatry,75, pp. 3 –18.
ILO (1998) Gina Coefficient Index, Geneva, OECD, International Labour Organisation.
Jenkins, B., Bebbington, P., Brugha, D., Farrrell, M., Lewis, G. and Meltzer, H. (1998)
‘British psychiatric morbidity surveys’, British Journal of Psychiatry,173, pp. 4 –7.
Jenny, C. and Isaac, R. (2006) ‘The relation between child death and child maltreatment’,
Archives of Diseases of Childhood,91, pp. 265–9.
Judge, K. and Benzeval, M. (1993) ‘Health inequalities: New concern about children of
single mothers’, British Medical Journal,306, pp. 677–80.
Kahn, R., Wise, P., Kennedy, B. and Kawachi, I. (2000) ‘State income in-equality, house-
hold income and maternal mental and physical health: Cross sectional national
survey’, British Medical Journal,321, pp. 1311–15.
Kanji, G. K. (2006) 100 Statistical Tests, 3rd edn, Sage Publications
Kempe, C. and Kempe, H. (1978) Child Abuse, Chicago, University of Chicago Press.
Kim, J. H., Choi, S. S. and Ha, K. (2008) ‘A closer look at depression in mothers who kill
their children: Is it unipolar or bipolar depression?’, Journal of Clinical Psychiatry,69,
pp. 1625–31.
Kim, S. W. and Pridemore, W. A. (2005) ‘Poverty, socio-economic change, institutional
anomie and homicide’, Social Science Quarterly,86, pp. 1377–98.
King, W. K., Kiesel, E. and Simon, H. K. (2006) ‘Child abuse fatalities: Are we missing
opportunities for intervention?’, Paediatric Emergency Care,22, pp. 211–14.
Kitsantas, P. (2008) ‘Ethnic differences in infant mortality by causes of death’, Journal of
Perinatology,28, pp. 573–9.
Laderichi, C. R., Saith, R. and Stewart, F. (2003) ‘Does it matter that we do not agree on
definitions of poverty? A comparison of 4 approaches’, Oxford Development Studies,
31, pp. 253–74.
Who Kill Children? Page 31 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Langevin, R. (2003) ‘A study of psychosocial characteristics of sex killers: Can we iden-
tify them before it is too late?’, International Journal of Offender Therapy and Com-
parative Criminology,47, pp. 366–82.
Liberman, R. P. and Glick, J. D. (2004) ‘Drug and psychosocial curricula for
psychiatry residents for treatment of schizophrenia: Part 1’, Psychiatric Services,55,
pp. 1217–19.
Liem, M. and Koenraadt, F. (2008) ‘Familicide: A comparison with spousal and child
homicide by mentally disordered perpetrators’, Criminal Behaviour and Mental
Health,18, pp. 306–18.
Liem, M. and Pridemore, W. (2012) Sourcebook of European Homicide Research,
New York, Springer.
Linsley, K. R., Schaparia, K. and Kelly, T. P. (2001) ‘Open verdicts versus suicide: Im-
portance to research’, British Journal of Psychiatry,178, pp. 465–8.
Mackenbach, J. P., Bos, V., Andersen, O., Cardano, M., Costa, G. and Kunst, A. E.
(2003) ‘Widening socio-economic inequalities in mortality in six Western European
countries’, International Journal of Epidemiology,32, pp. 838–9.
Matturi, I., Mauri, M., Elena-Ferreo, M. and Lavezzi, A. (2008) ‘Unexpected
peri-natal loss versus SIDS: A common neuropathy entity’, Open Neurology,2, pp.
45–50.
Meltzer, H., Gill, B., Pettiigrew, M. and Hinds, K. (1995) The Prevalence of Psychiatric
Morbidity among Adults Living in Private Households, London, HMSO.
Moore, A. (2005) ‘Changing patterns of childhood mortality in Wolverhampton’,
Archives of Diseases of Childhood,90, pp. 687–91.
Morgan, C. and Bhugra, D. (2010) Principles of Social Psychiatry, 2nd edn, Oxford,
Wiley-Blackwell.
Mulvaney, C., Kendrick, D., Towner, E., Brussoni, M., Hayes, M. and Ward, H. (2009)
‘Fatal and non-fatal injuries in England 1995 –2004: Time trends and inequalities by
age, sex and area deprivation’, Journal of Public Health,31, pp. 154 –61.
Munro, E. (2010) ‘Learning to reduce risk in child protection’, British Journal of Social
Work,40, pp. 1135–51.
Musick, K. and Meier, A. (2010) ‘Are both parents always better than one? Parental con-
flict and young adult well-being’, Social Science Research,39, pp. 814 –30.
Newlands, M., Adamson, E., Ghulam, S., Saleh, M. and Emery, J. L. (1991) ‘Differences
in post-perinatal infants deaths between North and Southern Derbyshire’, Archives
Diseases of Childhood,66, pp. 1421–6.
Newlands, M. and Emery, J. L. (1991) ‘Child abuse and cot deaths’, Child Abuse and
Neglect,15, pp. 275–8.
Newton, A. W. and Vandeven, A. M. (2006) ‘Unexplained infant and child
death: A review of SIDS, sudden unexplained deaths and child maltreatment
fatalities including shaken baby syndrome’, Current Opinion in Paediatrics,18, pp.
196–200.
NSPCC (2002) Campaigning on Child Abuse Deaths, London, NSPCC.
Ofsted (2010) Learning Lessons from Serious Case Reviews 2009–2010, available online
at tp://www.ofsted.gov.uk/content/download/11643/136464/file/Learning.pdf.
Parslow, R. C., Tasker, R. C., Draper, E. S., Parry, G. J. and Jones, S. (2009) ‘Paediatric
intensive care network: Epidemiology of critically ill children in England and Wales:
Incidence, mortality, deprivation and ethnicity’, Archives of Diseases of Childhood,
94, pp. 210–15.
Page 32 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Pharaoh, P. O. and Platt, M. J. (2007) ‘Sudden infant death syndrome in twins and single-
tons’, Twins Research and Human Genetics,10, p. 644 –8.
Pritchard, C. (1992) ‘Children’s homicide as an indicator of effective child protection: A
comparative study of Western European statistics’, British Journal of Social Work,23,
pp. 645–52.
Pritchard, C. (2002) ‘Children’s homicide and road deaths in England and Wales and the
USA: An international comparison 1974–1997’, British Journal Social Work,32, pp.
495–502.
Pritchard, C. (2004) The Child Abusers: Research and Controversy, Buckingham, Open
University Press.
Pritchard, C. (2006) Mental Health Social Work: An Evidenced-Based Approach,
London, Routledge.
Pritchard, C. (2010) ‘The psychiatric–child protection interface: Research to inform
practice’, in Morgan, C. and Bhugra, D. (eds), Principles of Social Psychiatry, 2nd
edn, Oxford, Wiley-Blackwell, pp. 483–95.
Pritchard, C. (2012) Family Violence in Europe: Child Homicide and intimate partner
violence, linking the crime and psychiatric interface. In Liem, M. and Pridemore
(eds) Sourcebook of European Homicide Research, New York, Springer, 75– 93.
Pritchard, C. and Bagley, C. (2001) ‘Suicide and murder in child murderers and child
sexual abusers’, Journal of Forensic Psychiatry,12, pp. 269–86.
Pritchard, C. and Hansen, L. (2005) ‘Adolescent and youth suicide in England and Wales:
An international comparison’, International Journal of Adolescent Medicine Health,
17, pp. 239–53.
Pritchard, C. and Hickish, T. (2011) ‘Comparing cancer mortality and GDP health ex-
penditure in England and Wales with other major developed countries 1979 –2006’,
British Journal of Cancer,105, pp. 1788–994.
Pritchard, C. and King, E. (2004) ‘Comparing child-sex-abuse-related and
mental-disorder-related suicide in a six year cohort of regional suicides: The import-
ance of the child protection– psychiatric interface’, British Journal Social Work,34,
pp. 181–97.
Pritchard, C. and Sayers, T. (2008) ‘Exploring potential extra-family child homicide assai-
lants in the UK and estimating their homicide rate: Perception of risk and the need for
debate’, British Journal of Social Work,38, pp. 290–307.
Pritchard, C. and Wallace, M. S. (2011) ‘Comparing the USA, UK and 17 Western coun-
tries efficiency and effectiveness in reducing mortality’, Journal of Royal Society of
Medicine Shorts,2.60, pp. 1–10.
Pritchard, C. and Williams, R. (2009) ‘Does social work make a difference? A controlled
study of former looked-after-children and excluded-from-school adolescents now
men aged 16–24: Subsequent offences, being victims of crime and suicide’, Journal
of Social Work,9, pp. 285–307.
Pritchard, C. and Williams, R. (2010) ‘Comparing possible child-abuse-related-deaths in
England and Wales with the major developed countries 1974–2006’, British Journal of
Social Work,40, pp. 1700–18.
Pritchard, C. and Williams, R. (2011) ‘Poverty and child (0– 14 years) mortality in the
USA and other Western countries as an indicator of “how well a country meets the
needs of its children” (UNICEF)’, International Journal of Adolescent Medicine
and Health,23, pp. 251–5.
Who Kill Children? Page 33 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Pritchard, C., Brackstone, J. and MacFie, J. (2010) ‘Adverse event and patient safety in
the operating theatre: Perspectives of 549 surgeons’, Annals of Royal College of Sur-
geons,(Supplement) 92, pp. 1–4.
Pritchard, C., Cox, M., Foulkes, L. and Lindsay, K. (2011a) ‘The patient’s voice in neuro-
surgery: Psycho-socio-economic benefits of a patient-designed versus standard service
following treatment for a subarachnoid haemorrhage’, Social Care and Neuro-
Disability,2, pp. 80–96.
Pritchard, C., Lindsay, K., Cox, M. and Foulkes, L. (2011b) ‘Re-evaluating the National
Subarachnoid Haemorrhage study (2006) from a patient-related-outcome-measure
perspective: Comparing fiscal outcomes of treatment-as-usual with an enhanced
service’, British Journal of Neurosurgery,25, pp. 376–83.
Pritchard, C (2012) Family violence in Europe: Child homicide and intimate partner vio-
lence, linking the crime and psychiatric interface. In M. Liem and R. Pridemoer (eds).
Source of European Homicide Research, New York, Springer, pp. 75–93.
Rettenberger, M., Matthes, A., Boer, D. P. and Eher, R. (2010) ‘Prospective actuarial
risk assessment: A comparison of five risk assessment instruments in different
sexual sub-types’, International Journal of Offender Therapy and Comparative Crim-
inology,54, pp. 169–86.
Rossegger, A., Endrass, J., Urbaniok, F., Vetter, S. and Maercker, A. (2011a) ‘From
victim to offender: Characteristics of sexually abused violent and sexual offenders’,
Nervenartz,24, Epub ahead of publication.
Rossegger, A., Laubacher, A., Moskvitin, K. and Endrass, J. (2011b) ‘Risk
assessment instruments in repeat offending: The usefulness of FOTRES’, Internation-
al Journal of Offender Therapy and Comparative Criminology,11, Epub ahead of
publication.
Salmon, M. P., Abel, K., Webb, R., Warburton, A. L. and Appleby, L. (2004) ‘A national
audit of joint mother and baby admissions to UK psychiatric hospitals: An overview
of findings’, Archives of Women’s Mental Health,7, pp. 65–70.
Saraceno, B., Levav, I. and Kohn, R. (2005) ‘The public mental health significance of re-
search on socio-economic factors in schizophrenia and major depression’, World
Psychiatry,4, pp. 181–5.
Schechter, D. S. and Willheim, E. (2009) ‘Disturbance of attachment and parental psy-
chopathology in early childhood’, Child Adolescent Psychiatry Clinics of North
America,18, pp. 665–86.
Seifritz, E., Esposito, F., Neuhoffm, J., Luthi, A. and Di Salle, F. (2003) ‘Differential sex-
independent amygdala response to infant crying and laughing in parents and non-
parents’, Biological Psychiatry,54, pp. 1367–75.
Shaw, M., Tunstall, H. and Dorling, D. (2005) ‘Increasing inequalities in risk of murder in
Britain: Trends in the demographic and spatial distribution of murder 1981–2000’,
Healthy Places,11, pp. 45–54.
Sidebotham, P. and Golding, J. and the ALSPAC Study (2001) ‘Child maltreatment in
the “Children of the Nineties”: A longitudinal study of parental risk factors’, Child
Abuse and Neglect,25, pp. 1177 –2000.
Sidebotham, P., Bailey, J., Belderson, P. and Brandon, M. (2011) Fatal child maltreat-
ment in England 2005–2009. Child Abuse and Neglect.35, pp. 299 – 306.
Singh, T. P. and Kogan, M. D. (2007) ‘Widening socio-economic disparities in US child-
hood mortality 1969–2000’, American Journal Public Health,97, pp. 1658 – 65.
Page 34 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
South, S. C. and Krueger, R. F. (2011) ‘Genetic and environmental influences on inter-
nalising psychopathology vary as a function of economic status’, Psychological Medi-
cine,41, p. 107.
Stanistreet, D., Taylor, S., Jeffrey, V. and Gabby, M. (2001) ‘Accident or suicide? Predic-
tors of coroners decisions in suicide and accident verdicts’, Medicine, Science and
Law,41, pp. 111–15.
Thornton, D., Mann, R., Webster, S., Blud, L. and Erickson, M. (2003) ‘Distinguishing
and combing risk for sexual and violent recidivism’, Annals of New York Academy
of Sciences,989, pp. 225–46.
Toney, L. (1983) ‘The effects of holding the newborn at delivery on parental bonding’,
Nursing Research,32, pp. 16–19.
Turkheimer, E., Haley, A., Waldron, M., D’Onofrio, B. and Gottesman, I. (2003) ‘Socio-
economic status modifies heritability of IQ in young children’, Psychological Science,
14, pp. 623–8.
UNICEF (2001) Child Deaths by Injury in Rich Nations, Florence, Innocenti Research
Centre, Report No. 2, United Nations Children Fund.
UNICEF (2010) Global Study of Child Poverty and Disparities. New York, UNICEF.
UNMDG Task Force (2009) Millennium Development Goals: A Report, New York,
United Nations.
Urbaniok, F., Rossegger, A. and Endrass, J. (2006) ‘Can high-risk offenders be reliably
identified? A follow-up study on dangerous offenders in Switzerland released from
prison for legal reasons’, Swiss Medical Weekly,136, pp. 761–8.
USDHSS (2006) Federal Poverty Level. US Dept of Health & Human Services, Wash-
ington DC.
Wan, M. W., Warburton, A. L., Appleby, L. and Abel, K. M. (2007) ‘Mother and baby
unit admissions: Feasibility study examining child outcomes 4–6 years on’, Australian
and New Zealand Journal of Psychiatry,41, pp. 150–6.
Weekes-Shackelford, V. A. and Shackleford, T. K. (2004) ‘Methods of filicide: Steppar-
ents and genetic parents kill differently’, Violence and Victims,19, pp. 75–81.
Weissman, A. M., Jogerst, G. J. and Dawson, J. D. L. (2003) ‘Community characteristics
associated with child abuse’, Child Abuse and Neglect,27, pp. 1145–59.
West, S. G., Friedman, S. H. and Resnick, P. J. (2009) ‘Fathers who kill their children: An
analysis of the literature’, Journal of Forensic Science,54, pp. 185 –8.
White, S., Hall, C. and Peckover, S. (2009) ‘The descriptive tyranny of the Common As-
sessment Framework: Technologies of categorisation and professional practice in
child welfare’, British Journal of Social Work,39, pp. 1197–217.
WHO (1992) International Classification of Diseases, 10th edn, Geneva, World Health
Organisation.
WHO (2008) Registered Deaths by Cause, Age and Gender: Table 1, available online at
www.who.whosis.
Wilkinson, R. and Pickett, K. (2009) The Spirit Level, London, Allen.
Williams, C. F. (2007) ‘The UK General Medical Council fails child protection’, Paedi-
atrics,119, pp. 800–2.
Williams, R. and Pritchard, C. (2006) Breaking the Cycle of Educational Alienation,
Buckingham, Open University Press.
Wood, H. (2012) ‘Assessing the feasibility of using an actuarial risk assessment tool to
identify risk in child protection cases’, unpublished Ph.D. thesis, School of Health
and Social Care, Bournemouth University.
Who Kill Children? Page 35 of 36
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
Woodman, J., Pitt, M., Wentz, R., Taylor, B. and Gilbert, R. E. (2008) ‘Performance
screening tests for child physical abuse in accident and emergency departments’,
Health Technology Assessment,12, pp. 1–95.
Woodruff, T. J., Darrow, L. and Parker, J. D. (2008) ‘Air pollution and postnatal infant
mortality in the USA 1992– 2002’, Environmental Health Perspectives,116, pp. 110 –15.
Zagar, R. J., Isbell, S. A., Busch, K. and Hughes, J. R. (2009) ‘An empirical theory of the
development of homicide within individuals’, Psychological Report,104, pp. 199–245.
Page 36 of 36 Colin Pritchard et al.
by guest on May 7, 2012http://bjsw.oxfordjournals.org/Downloaded from
