Intimate Partner Homicide in England and Wales 2011–2013: Pathways to Prediction from Multi-agency Domestic Homicide Reviews

Research QuestionWhat pathways to more accurate prediction of intimate partner homicide (IPH) can be found by reviewing two years of official Domestic Homicide Reviews in England and Wales?DataThis study conducted a detailed review of investigative source material, police database information and the official independent author reviews of the 188 cases of intimate partner homicide recorded in England and Wales between April 2011 and March 2013.MethodsDescriptive analytical techniques were used to explore the prevalence of various characteristics of victims, offenders and relationships in these cases, with special attention given to offender suicide ideation as a precursor to the crimes.FindingsOffenders in these cases were 86% male, with high rates of both chronic substance abuse (61%) and prior reported offending (50%) against their homicide victim. The most disproportionately prevalent characteristic appears to be that 40% of the male offenders were known by someone, but often not to police, as suffering suicidal ideation, self-harm or attempted suicides. The prevalence of that marker, while not measureable in the general population, is over four times higher than the pre-offence police indications of suicidal tendencies across 80 domestic homicides in Leicestershire (Button et al., 2017).ConclusionsIt is plausible that many more intimate partner homicides might be accurately predicted, and perhaps prevented, with more public investment in obtaining data on suicidal indicators and more proactive treatment of domestic abuse offenders known to suffer suicidal tendencies.


Introduction
While reported domestic abuse (DA) has remained at persistently high levels over the past decade, its relationship to domestic homicide remains difficult to unravel. Recent UK research (Bland and Ariel 2015) suggests that there may be little relationship between reporting of DA to agencies and subsequent homicide, making prediction and prevention a difficult task. The decline in overall homicide in England and Wales over the past decade has not been matched by declines in intimate partner homicide (IPH), which now make up one in five of all homicides (See Fig. 1) The gender of IPH victims has remained relatively stable at 70% female (Osborne et al. 2012). Studies have also shown IPH is the type of homicide where women come closest to matching the prevalence of men as offenders. In the US for each 100 men that kill their partner, 75 women do the same (Wilson and Daly 1992). This is not reflected in other types of homicide and intimate partner relationships are where women most commonly kill (Peterson 1999). This characteristic suggests a dynamic that sets it apart from all other victim and perpetrator relationships.
Victim mental health within the context of IPH is lacking in substantial research but the psychological profile of those that kill has been subject to vigorous discussion. It has been identified that passive aggressiveness, paranoia and depression are most common within the personalities of such perpetrators (Aldridge and Browne 2003). Campbell et al. (2007) found in her US study that 29% of female victims and between 13 and 28% of perpetrators had a prior history of mental health problems. Another element of mind-set is that of suicidal ideation or attempts prior to or in the course of committing an IPH. Homicide followed by suicide during IPH in the US constitutes some 74% of homicides that are followed by suicide across all categories (Campbell et al. 2007). This pattern has been established on an international basis indicating a consistent characteristic of IPH across differing culture and geography (Starzomski and Nussbaum 2000). Despite the high prevalence of domestic callouts to police, IPH fortunately constitutes a tiny fraction of all domestic cases. Yet that ratio creates a huge difficulty in arriving at any reliable method of risk assessment and prediction. As research in Thames Valley by Thornton (Thornton 2011, 2017 and in Dorset by Chalkley (2015;Chalkley and Strang 2017) have shown, tools such as the Domestic Abuse Stalking and Harassment instrument (DASH) have failed to predict most cases of domestic homicide.

Research Question
Given the difficulty of predicting statistically rare events, it is useful to start by comparing the characteristics of those events to a larger population of people who are thought to be at risk. While precise estimates of risk factors are beyond the scope of this methodological approach, exploring the homicide cases may provide pathways to prediction by future studies. The key question for the present study is what characteristics of these homicides might have been identified prior to their occurrence, so that future homicides could be prevented with more accurate prediction? Answering this question requires that unexpectedly large concentrations of certain variables be identified, even without precise knowledge of their prevalence in the non-homicidal population of intimate partners. Any and all factors which seem to be unusually prevalent in these cases may be candidates for further research as pathways to prediction.

Data and Method
The data for this study is derived from the review of original case file material collected by the investigating police force, the Police National Computer database (PNC) and available Domestic Homicide Reviews (DHRs) of all cases recorded within the Home Office Homicide Index in the 24-month period 1st of April 2011-31st March 2013, that led to charge of homicide under Common Law involving murder or voluntary manslaughter. The Index contains data submitted to the Home Office by police in each instance of homicide and details every homicide in England and Wales.
Domestic Homicide Reviews The DHR documents are authored by a professional with extensive experience in public sector service, based on a comprehensive review of all material held by agencies engaged with either of the parties involved in the homicide event. As seen in Fig. 2, the completion rate of DHRs by the time of this study in 2015 was low: only 52.1% of those commissioned were available for review as part of this study. This is despite the statutory obligation for completion and expected publication following such events. A further 6.5% had been commissioned but not yet completed despite at least two years passing since the homicides had occurred. Some were unavailable as too sensitive due to ongoing legal issues or other features of the case.
Relationships included in this dataset were heterosexual only, involving murder or manslaughter committed by a current or previous spouse, common law partner, girlfriend, boyfriend or anyone of the opposite sex with whom the perpetrator had an intimate relationship. A stringent evidential test applied by the Crown Prosecution Service ensured there was a reasonable prospect of conviction of the suspected perpetrator beyond reasonable doubt. The data set also includes cases where a suspect committed suicide prior to charge. A total of 188 offences satisfied the data criteria and were available to be included in the dataset.
In addition to the case summary document compiled by the police investigation team and presented to the Crown Prosecution Service, the DHR authors had access to material held by all agencies across the public sector where the individuals involved were known. They also had the reports submitted to the Coroner for inquest purposes and antecedent statements of those closest to the victim and offender. The reviews provide a comprehensive summary of the information established by the author, independent of public agencies. It documents what was known to agencies that had involvement with the victim or perpetrator, seeking retrospective learning from each case.
The findings are presented here primarily in relation to offender characteristics. The Master's thesis (Bridger 2015) of which the current study is a summary provides more detail and a larger range of variables examined. The present study reports the findings from that thesis that showed the greatest prevalence of factors known or knowable prior to the homicide.

Findings
Gender The dataset contained 162 (86%) female and 26 (14%) male victims. These figures are reversed for perpetrators as all the relationships were heterosexual in nature.

Fig. 2 Completion of DHR within cohort
Ethnicity There was a high frequency of white North-European victims in the cohort at 79% of all cases, with much lower rates of black and Asian (Indian sub-continent) victims evident, both at 9% and the remaining 'other' contributing 3%. All male victims in the cohort were of white ethnicity. Perpetrators showed similar results but with greater contribution by minority groups; 72% were white North-European, 12% black, 12% Asian (Indian sub-continent) and remaining 'other' contributing 4%. All female perpetrators in the cohort were of white ethnicity.
Occupation and Employment As shown in Fig. 3, perpetrators show very high concentrations in unskilled employment (28.9%) or are unemployed (45.5%). All non-earning categories collectively contribute 57.8% of the perpetrators in the cohort.
Substance Abuse and Intoxication Analysis of the perpetrators (Table 1) established that 41.1% were substance abusers, with alcohol as the main contributor (33.7%). When further disaggregated by gender, it shows that 39.9% of male perpetrators were substance abusers compared to 52% of female perpetrators.
In the perpetrator cohort, only 21.2% were intoxicated by alcohol at the time of the offence and a further 28.3% were described by witnesses as being intoxicated, a combined total of 49.8%. When split by gender, male perpetrators were intoxicated (as shown through toxicology) in 22.6% of cases with a further 27% said to be intoxicated by witness evidence. With females, these figures were 15% and 50% respectively.
Mental Health Perpetrators of both genders had a combined prevalence of 31.3% with a diagnosis of mental illness, and 26.9% had depression. When considered by gender, rates were established as 32.5% of male perpetrators generally, with a 27.4% contribution by depression, and 24% female perpetrators, all of which included depression but some had a multiple diagnosis. Suicidal Ideation Pre-offence suicide ideation and attempts had been present in 38.6% of perpetrators, with 40.3% of male and 28.0% of female perpetrators. In most of those cases, however, there is no evidence that these facts were known to police prior to the homicide. Post-offence suicide attempts occurred in 33.4% of cases with a 'success' rate for the perpetrator of 24.2% of those attempts.
Previous Abuse of the Victim In 36.4% of cases, there was a PNC recorded previous crime against the victim by the homicide perpetrator. The vast majority consisted of physical assaults with only 3.2% being attributed to other crime types. When considered by gender, 50% of the male victims had a crime recorded compared to 34.2% of female victims. A disclosure or allegation of domestic crime had been made by the victim to a friend, family member or public agency in 60.8% of cases. Prior Criminality Victims had at least one previous criminal conviction in 26.6% of cases though only 2.1% possessed a conviction for an assault of grievous bodily harm or more severe; 16.2% of victims had a prior arrest for a domestic abuse crime. When considered by gender, it was found 73.1% of the male victims had some kind of previous conviction and 61.5% had a prior DA crime arrest, compared to 19.1% and 8.4% respectively for female victims. Table 2 shows the frequency of convictions recorded against perpetrators. Over half (55.3%) had at least one previous criminal conviction before the homicide. It was found that female perpetrators had a higher percentage of previous convictions at 61.5% compared with 54.3% for male perpetrators. The national average for possession of a criminal conviction is 33% for men and 9% for women (Ministry of Justice 2010). When the sample of this study is compared, it is apparent that rates of criminality are much higher.
There were no repeat cases of homicide in the previous convictions reviewed and levels of previous violence convictions were very similar between genders. Male perpetrators had elevated levels of property, drugs and weapons convictions whereas female perpetrators were elevated with regard to public order convictions. Over half (54.3%) of perpetrators had a previous arrest for a DA crime recorded on PNC and 20.9% had a prior physical assault offence against a different partner recorded.

Relationship and Cohabitation
The distribution for each relationship category is shown in Fig. 4. Whilst 40.9% of all homicides involved a couple who were legally married, the highest proportion of offences occurred between couples that were cohabiting at the time of the homicide incident, for a total of 55.3% of all couples.
Length of Relationship Analysis of the length of the relationships within the homicide cases established the largest single concentration existed in the fifteen plus years' category (21.8%) ( Table 3).

Fig. 4 Percentage distribution of relationship categorisation across homicides
Pregnancy and Children Pregnancy at the time of the homicide was the case in only four instances in this cohort: this represented 2.5% of all of the homicides involving a female victim. The rate of pregnancy within the sample was only 4.8% of the female victims aged between 15 and 44 years, considerably lower than the national rate for women in the same age range, which was 7.8% (ONS 2013). Thus, pregnancy does not seem to elevate the risk of domestic homicide.
Pre-offence Behaviour of Perpetrator Behaviours of the perpetrator in the lead up to the homicides revealed a number of important findings, with results displayed in Table 4. In 19.7% of the homicides, there was a disclosure made by the perpetrator detailing an intention to cause the victim serious harm or death, and in 24.3%, the perpetrator told others that they intended to kill themselves.
There was a general lack of offender preparation for the homicides: sourcing equipment was done by only 30%, only 8% made plans for disposal of the victim's body, 1.7% made plans for escape from the country, and preparation of alibi was made by only 15%. Most notable was the prevalence of harassment or stalking behaviour by the perpetrator towards the victim in the lead up to the offence. This did not occur in any of the cases where there was a female perpetrator but was present in 38.1% of cases of male perpetrators. Rank-Ordered Prevalence Looking across the highly prevalent characteristics, it is useful to rank them in terms of prevalence among male offenders, who comprised 86% of the murderers. Table 5 displays that ranking for all factors present in over one-third of the offenders.

Conclusions
Of all of the characteristics ranked in Table 5, suicide ideation appears to be the most over-represented in relation to the general population. Chronic substance abuse, cohabitation and even prior crime against the victim are so widespread and prevalent in the population generally that they would massively over-predict domestic homicide. Suicidal ideation or attempts, however, appear to be much rarer in the population. Thus, a 40% rate of suicidal indication among the male offenders may be the most useful of any of these characteristics in distinguishing people who are much more likely to kill their partners than other offenders. While we have no reliable estimate of the prevalence of suicide ideation or attempts in the general population, it can be noted that the male suicide rate for the UK in 2012 (ONS 2014) was only 18.2 per 100,000 population. In percentage terms, that is approximately two-one-hundredths of 1%. Thus, the male offender, pre-offence suicidal ideation prevalence among perpetrators of IPH is 2000 times higher than the general population male suicide rate at the time of the homicides in this study. Even if this estimate is adjusted for the multiple years in which the suicidal ideation or behaviour may have been expressed by the perpetrators in this study, the use of suicidal tendencies as a marker for domestic homicide would seem to be a valuable pathway to prediction.
The 40% finding for suicidal offenders also suggests how much of this information police are not receiving before an IPH. In an analysis of 80 domestic homicides in Leicestershire over 1997, Button et al., (2017 found that 9% of them had a suicide or self-harm marker in the Crime and Intelligence System (CIS), which has since been closed. Extrapolating from the present study's findings, approximately 77% of the Leicestershire homicide offenders who had had suicidal tendencies were unknown by the police to have suffered that problem prior to committing a homicide as Button et al., (2017) defined the crime (including attempts). Thus, the case for police investing in resources to identify more cases of overlap between suicidal tendencies and domestic abuse seems to be great. If ever there was a case for partnership policing rather than silo policing, this would seem to be a prime example.
Policy Implications The most important implication of these findings is that IPH cannot be predicted from police records alone. Current assessment techniques require, at minimum, that a prior act of abuse (before a homicide) become known to professionals. Yet in nearly 40% of cases that opportunity never presented itself. Suicidal ideation becoming known from any source, and perhaps other information from victims or their families, may be absolutely necessary to improve prediction of IPH.
This work does not provide a definitive list of what those predictive factors are. But it is the first predictive analysis in the UK that has been able to include variables that were not known to police prior to the homicide. This means that the Domestic Homicide Reviews could provide a valuable ongoing source of analysis and improvement in both prediction and prevention.
The findings from the DHRs suggest important areas for consideration by professionals in terms of the level of completion of the reviews, the need for commonality across public agencies when considering risk, and the value of information sharing whenever the opportunity arises to allow for decisions to be made with the best knowledge available. The third of those elements has developed through the introduction of Multi Agency Safeguarding Hubs (MASH) in many force areas, ensuring that information sharing occurs across safeguarding agencies earlier and deeper than ever before.
Research Limitations This work details the factors present in IPH on a national basis over a two-year period, albeit with a low completion rate of the DHRs. This research cannot say what factors are actually predictive of IPH. A matched cohort of non-fatal situations analysed over a similar period may be helpful in allowing for comparison and predictive modelling. To date, this has not been done with a case control research design beyond a single constabulary level.
It is notable in this analysis that many offences occurred with no apparent buildup of violence in relationships that outwardly show little if any signs of abuse. These offences mean opportunities for prevention and pursuant investigation are extremely difficult, particularly when considering motive or trying to prove deliberate act. There is limited in-depth research in this respect but this dataset and source material could be used to explore greater detail for further contribution to understanding.
The study featured analysis of rarely used material from the official reviews compiled by independent authors in the aftermath of such homicide events. Key themes such as failure to share information, insufficient training of professionals, inconsistency in risk assessment and the poor management of complex needs of the perpetrators of DA were identified as the most regularly occurring allowing lessons to be learned by professionals in the field. Completion rates and availability of DHRs were found to be much lower than anticipated considering the statutory obligation of the review.
This study further demonstrates that homicide is not exclusive to relationships where violence increases over time before murder occurs as the final transaction. Whilst antecedent violence can be an important factor, many murderous acts occur for reasons that current theory and techniques for prediction do not encompass. This work may be seen as the first step in a programme of research that incorporates data not reported to police. If combined with the opportunities afforded by technology to analyse large volumes of data, such research could lead to greater understanding of IPH and allow professionals to better protect future generations of intimate partnerships.