From Caring to Killing: A Typology of Homicides and Homicide–Suicides Perpetrated by Caregivers

Abstract

In the news media, there are regular reports of family caregivers killing the people for whom they care, but scholarly research on this phenomenon is fragmented, and there has been little effort to predict or prevent future deaths. The aim of this study was to develop a typology of caregiver-perpetrated homicides that could provide a framework for more rigorous research and targeted responses in policy and practice. Ideal Type Analysis was applied to sixty-four homicides and homicide–suicides perpetrated by family caregivers in England and Wales between January 2015 and December 2019. The cases clustered into seven clear types: Ending Suffering; Genuine Burden of Care; Pre-existing Mental Illness; Neglect; Exploitation; Caregiver as Victim of Domestic Violence, Abuse or Coercive Control; and Caregiver as Perpetrator of Domestic Violence, Abuse, or Coercive Control. Each type was characterised by a distinct motive, context, or course of events leading to the homicide. This is the first typology of homicides and homicide–suicides perpetrated by caregivers. The Caregiver-Perpetrated Homicide Typology challenges previous claims that caregiver-perpetrated homicides are isolated events and provides a framework for the development of evidence-based prediction and prevention initiatives.

1. Introduction

Caregivers (also known as unpaid carers, informal carers, family carers, or family caregivers) are people who provide essential care for family members or friends with long-term illnesses or disabilities (NHS England 2025; APA 2023). In the UK alone, more than five million people identify themselves as caregivers (ONS 2023), and more than one million of those provide at least 50 h of care per week (ONS 2023). As a group, their unpaid labour is estimated to save the British government GBP 162 billion per year (Petrillo and Bennett 2021). Similar statistics have been reported for caregivers globally (Reinhard et al. 2019; Fast et al. 2024; Carers Australia 2020).

Although caregivers make a significant social and economic contribution, the role can take a considerable toll on their own wellbeing. More than six decades of international research have shown that caregivers experience higher than average rates of physical and mental illness, social isolation, and financial distress (Adelman et al. 2014; Pinquart and Sorensen 2003; Brimblecombe and Farias 2022; Cooper 2021). A growing body of research has also shown that caregivers contemplate suicide at higher-than-average rates, with some also contemplating homicide (killing the person for whom they care) or homicide–suicide (O’Dwyer et al. 2021; Whitehead et al. 2012; Cooke et al. 1998; O’Dwyer et al. 2016). It also clear—from media reports, grey literature, case reports, and public vigils—that some caregivers act on those homicidal thoughts and kill the person for whom they are caring (see, for example: Disability Day of Mourning; Catlin 2003; Cina et al. 1996; Mukhida 2007). This is a global phenomenon, with deaths by homicide and homicide–suicide reported in numerous countries including the UK, USA, Canada, Australia, Japan, Singapore, South Africa, Uganda, Russia, and Belgium.

Despite this, there has been very little rigorous research on caregiver-perpetrated homicides and homicide–suicides (Cohen 2019). Where caregiving has been considered, it has frequently been as a subset of other domestic homicides (e.g., Oya et al. 2024; Malphurs and Cohen 2002; Benbow et al. 2019; Salari 2007; Malphurs et al. 2001) or within a narrow definition of ‘caregiver’ that excludes non-familial relationships or caregivers not living with the care recipient (e.g., Malphurs and Cohen 2002; Stoneman et al. 2022). Some studies have also excluded homicide–suicides where the caregiver survived the suicide attempt (e.g., Malphurs and Cohen 2002), excluded homicides where the caregiver acted in self-defence (e.g., Lucardie and Sobsey 2005), or failed to distinguish between professional and family/friend caregivers (e.g., Karch and Nunn 2011). In addition, the majority of studies have focused on specific types of caregivers—such as older people, parents of disabled children, or caregivers without their own physical or mental health problems (e.g., Ohrui et al. 2005; Coorg and Tournay 2013; Salari 2007; Bourget et al. 2010; MacPherson et al. 2020; Roberto et al. 2013; Malphurs et al. 2001; Jung et al. 2020; Benbow et al. 2019)—or on specific illnesses or disabilities (e.g., Lucardie and Sobsey 2005; Guan et al. 2022; Brown 2012; Cina et al. 1996). As a result, the existing evidence has impeded recognition of caregiver-perpetrated homicides as a distinct phenomenon and enabled only fragmented understandings of that phenomenon (Frederick et al. 2019). For brevity, the term homicide will be used throughout the remainder of this paper to refer to both homicides and homicide–suicides, unless a distinction needs to be made.

Many studies have also failed to take a systematic approach to the identification of cases, often drawing on media reports or high-profile cases as their only source of data (e.g., Ohrui et al. 2005; Coorg and Tournay 2013; Roberto et al. 2013; Lucardie and Sobsey 2005; Cina et al. 1996; Catlin 2003; Guan et al. 2022). With nearly 30% of homicides not reported in the media, and media reports frequently inaccurate or lacking key information, studies that rely exclusively on data from media reports may be underestimating the number of caregiver-perpetrated homicides and misrepresenting the contexts in which they occur (Malphurs and Cohen 2002; Roberto et al. 2013; Coorg and Tournay 2013; Canetto and Hollenshead 2000; Guan et al. 2022).

The lack of rigorous, high-quality research has led to unsubstantiated claims that caregiver-perpetrated homicides are rare, isolated tragedies that cannot be predicted or prevented and do not warrant further investigation (Boxall 2018; Devon and Cornwall Police 2016; Coorg and Tournay 2013; MacPherson et al. 2020; Brown 2012). This is despite the fact that a substantial proportion of domestic homicides involve a caregiving relationship—9% of all homicides in the US, 31% of homicide–suicides perpetrated by older adults in the US, and 70% of all homicides in Japan—and clear evidence of risk factors and warning signs in many reported cases (Schwab-Reese et al. 2021; Malphurs and Cohen 2002; Oya et al. 2024; Catlin 2003; MacPherson et al. 2020).

The study, prediction, and prevention of domestic homicides is generally guided by taxonomies or typologies that provide a framework for distinguishing between homicides based on motivation, method, relationship, or risk factors (e.g., Resnick 1969; Hanzlick and Koponen 1994; Marzuk et al. 1992; Jung et al. 2020; Guileyardo et al. 1999; Biron and Reynald 2015). Although there have been suggestions of different types of caregiver-perpetrated homicides (e.g., Karch and Nunn 2011; Frederick et al. 2019; Jung et al. 2020), none of the existing homicide typologies are specific to caregiving and, even where they mention caregiving, they do not adequately capture the diversity and complexity of caregiver-perpetrated homicides.

The development of a specific typology for caregiver-perpetrated homicides would provide a framework for more rigorous research (which could, in turn, facilitate a more comprehensive understanding of the phenomenon) and facilitate targeted prediction and prevention initiatives based on the distinct features of each type of homicide (Malphurs et al. 2001; Malphurs and Cohen 2002; Frederick et al. 2019). As Karch and Nunn (2011) note, “treating these incidents with a common approach” will not be sufficient to prevent further deaths (p. 152).

The aim of this study was to develop a typology of caregiver-perpetrated homicides. In order to achieve this aim, our objectives were to: (1) identify all caregiver-perpetrated homicides in England and Wales in a five-year period; (2) use Ideal Type Analysis to develop a typology; and (3) describe the key characteristics of cases within each type. Additional objectives and their associated analyses will be reported elsewhere.

2. Materials and Methods

2.1. Ethical Framework

This research comprised a secondary analysis of publicly available data, so ethical approval was not required. These data do, however, reflect the experiences of real people whose lives have been, and continue to be, profoundly affected by homicide. In being exposed to these data, there is also a risk of vicarious trauma for the research team (Cullen et al. 2021; Micanovic et al. 2020). With that in mind, we drew on feminist care ethics (Tronto 1998; Brannelly and Barnes 2022; O’Dwyer et al. 2018) and the slow movement (Berg and Seeber 2017; Gallagher 2020) to create a culture of care in and around this project that included caring for the data and caring for each other. To care for the data, we: stored and shared data securely; used multiple sources of data for each case; used the real names of caregivers and care recipients, rather than pseudonyms or case numbers, to retain their humanity and ours; prioritised rigour over speed; and engaged in an iterative analysis process that kept us in constant contact with the data. To care for each other, only four members of the team were exposed to all the cases. These researchers met face-to-face whenever possible, journaled reflexively, and debriefed regularly, with the disclosure of emotional responses to the data considered an act of authenticity and solidarity rather than weakness or incompetence (Micanovic et al. 2020; Gallagher 2020). We also encouraged and supported all team members to take time away from the project; arranged work schedules to avoid excessive periods of time immersed in the data; actively promoted and modelled self-care; and provided access to free, confidential telephone counselling.

2.2. Inclusion/Exclusion Criteria and Definitions

Homicides and homicide–suicides perpetrated by caregivers, against family members or friends with long-term disabilities or illnesses, in England and Wales, between January 2015 and December 2019, were included.

A caregiver was defined as (NHS England 2025; APA 2023):

i.

A family member (parent, partner, child, grandchild, sibling) or close friend. This could include former or ex-partners, de facto partners, step-children, step-parents, foster or kinship families, and housemates, lodgers or neighbours.

WHO WAS

ii.

Identified in official reports as providing care for a person with a long-term illness or disability

OR

iii.

Was likely to be providing care on the basis of cohabitation with a person with a long-term illness or disability

OR

iv.

Was likely to have been providing care on the basis of cohabitation prior to institutionalisation (e.g., an elderly person caring for a spouse with dementia in their own home before the spouse was admitted to a long-term care facility1).

A long-term illness or disability was defined as (CDC 2024; Equality Act 2010):

i.

A physical or mental condition lasting or expected to last more than 12 months.

THAT

ii.

Required (or was likely to require based on typical symptoms) support to maintain independence, wellbeing, or quality of life

AND

iii.

Was identified in reports by relatives or professionals.

For children to be included, it had to be clear that the illness or disability required care beyond what would be age-appropriate in the absence of that condition. The definition of a mental condition encompassed—but was not limited to—personality disorders and addiction or substance abuse problems (WHO 2022, 2025).

A homicide was defined as a death caused or enabled by the action or omission of another person. This definition of homicide goes beyond legal definitions of murder and manslaughter in order to facilitate the inclusion of cases that were not prosecuted or were convicted of lesser charges (such as assisting suicide). As Jung et al. (2020) explain, “each country and its legal jurisdiction has a unique system of cultural, social, and legal background that comes into play when judging … homicide, only examining the ‘sentenced homicide’ or ‘criminal homicide’ cases would possibly distort the reality” (p. 2). A homicide–suicide was defined as a homicide followed by the suicide of the perpetrator. No limit was placed on the time between the homicide and the suicide, and homicides followed by attempted suicides were included.

England and Wales were selected as the sites for this study because the law mandates that every homicide, homicide–suicide, or suspicious death in these countries is subject to a statutory review, with the findings of those reviews to be made public (unless publishing them would endanger a vulnerable child or adult)2. A five-year period was selected to ensure a sufficient number of cases for analysis, with the specific dates chosen to avoid the onset of the COVID-19 pandemic (when caregiving roles and responsibilities changed in response to lockdowns and temporary service closures) and increase the likelihood of criminal proceedings, statutory reviews, and inquests being completed.

Cases were excluded if: the perpetrator was a paid care worker, a health professional, or housemate/lodger without a specified caregiving role; a criminal proceeding or coronial inquest was still underway; the illness or disability was assumed or implied in media reports without confirmation from family or relevant professionals; the victim was an older adult without a long-term illness or disability; the victim was a child without a long-term illness or disability; or, the caregiving role was a result of a fabricated or induced illness (formerly known as Munchausen Syndrome by Proxy; Glaser 2020).

2.3. Identification of Cases and Data Collection

Potentially relevant cases were first identified through the statutory review reports, which are published on Local Authority3 websites and online at the National Case Review Repository (https://learning.nspcc.org.uk/case-reviews/national-case-review-repository/, accessed on 25 April 2025), the Safeguarding Adult Reviews Library (https://nationalnetwork.org.uk/search.html, accessed on 25 April 2025), and the NHS England website (https://www.england.nhs.uk/publications/reviews-and-reports/invest-reports/, accessed on 25 April 2025). To ensure that all relevant reviews had been identified, each of the 339 Local Authorities in England and Wales was then contacted and asked to review the list of cases identified in their county/borough and either confirm that the list was complete or provide any reviews that had not been published. As an additional check, the date and location filters on the Disability Day of Mourning website (https://disability-memorial.org/, accessed on 25 April 2025) were used to identify relevant deaths of disabled people in England and Wales between 2015 and 2019.

Once potential cases were identified through these means, additional information about each case was collected from media reports, sentencing remarks4, Court of Appeal documents, and Prevention of Future Death Reports (issued by coroners5). Media reports were identified using Nexis and Google searches for the perpetrator and victim’s name (note: although statutory review reports are frequently anonymised, key details—such as date of death, location, and ages of caregiver and care recipient—allow them to be matched to media reports and court records). Court of Appeal documents were identified via the National Archives website (https://caselaw.nationalarchives.gov.uk, accessed on 25 April 2025). Sentencing remarks and Prevention of Future Death Reports were identified via the Courts and Tribunals Judiciary website (https://www.judiciary.uk/, accessed on 25 April 2025). (Note: we have used the terms ‘caregiver’ and ‘caregiving’ throughout this paper because they are the most commonly used terms in the international literature. In the UK, the terms ‘unpaid carer’ and ‘caring’ are preferred and so were used—along with other relevant terms for disabilities and illnesses—in our search for cases and additional information).

A case file was created for each potential case, including all statutory reviews, all media reports, any available sentencing remarks, and any available Court of Appeal or inquest documents. Each case file was reviewed in its entirety by two members of the research team (SOD, LH) to check if the case met our inclusion criteria and, for cases in which a homicide was followed by a suicide or attempted suicide, that there was a clear link between the two events.

2.4. Ideal Type Analysis

Ideal Type Analysis is grounded in Weber’s (1978, 2012) notion that observed reality can inform the development of conceptual representations (ideal types) of a given social phenomenon and that, in turn, those concepts can be used to deepen understanding of the reality of that phenomenon. As Stapley et al. (2022) explain, “we can regard ideal types as generalisations or mental representations of a social phenomenon that will never be identical with reality, but which will help make reality understandable” (p. 2). Ideal Type Analysis is a valuable method for the analysis of new or under-studied topics because it provides a framework against which future instances of the phenomenon can be compared and contrasted, and offers a heuristic to guide research, policy, and practice (Weber 1978, 2012; Swedberg 2017; Gerhardt 1994). Although Ideal Type Analysis has its roots in history and sociology, it is a flexible method that can be adapted for use with different forms of data and across research disciplines (e.g., Stapley et al. 2022; Gerhardt 1994; Stuhr and Wachholz 2001).

Given the topic of our research, it is important to clarify that “the word ‘ideal’ in an Ideal Type Analysis context does not mean ‘best’, but rather it refers, philosophically, to an idea” (Stapley et al. 2022, p. 2). In choosing this method of analysis, we were not implying that some homicides are more defensible or desirable (i.e., more ideal) than others (or indeed that any homicide is defensible or desirable). Instead, we were seeking to identify, compare, and contrast the features of different types (‘ideas’) of caregiver-perpetrated homicides, in order to facilitate a greater understanding of, and more nuanced responses to, these complex events.

Stapley et al.’s (2022) approach to Ideal Type Analysis was adapted for use in this study. Their approach involves seven steps: familiarisation with the dataset, writing case reconstructions; constructing the ideal types; identifying optimal cases; forming the ideal type descriptions; checking credibility; and making comparisons. We adhered to these steps, with the exception of writing case reconstructions. Given the complex topic, and the tertiary nature of much of the data (i.e., the facts of a case had already been summarised, synthesised, or interpreted by others), it was decided that creating short narrative summaries of each case risked introducing further bias, and so case files were used in their entirety. We also decided against using the term ‘optimal cases’. Ideal Type Analysis has traditionally sought to identify an ‘optimal’ or ‘pure’ case for each type—i.e., a case that is most representative of that type (Gerhardt 1994). Given the topic of our research, however, we have referred to these as archetypal cases, to avoid the implication that there is an optimal way of killing a disabled person or that certain motives are morally pure. A detailed description of the analysis undertaken at each step is provided in

Table 1. Ideal Type Analysis stages and activities.

Familiarisation	Four members of the team (SOD, LH, CB, and RG) read all the included case files and independently made notes on the key features of each case, discrepancies between different data sources within each case file, and their reflections on the similarities and differences between cases.
Constructing Ideal Types Forming Ideal Type Descriptions Identifying Archetypal Cases	Three meetings were held, with each meeting lasting approximately four hours. Four members of the research team (SOD, LH, CB, and RG) were involved at this stage. Meeting One (August 2023): SOD, LH, CB, and RG shared their notes and reflections on each case and, working collaboratively, agreed on seven preliminary types. The approach here and through the remainder of the analysis was informed by Weber’s notion of ‘mental experiments’, in which each aspect of a type is “tested separately with a view on whether or not it is indispensable for the development of the phenomenon under investigation” (Gerhardt 1994, p. 88), such that the final description of each type includes only those elements that are absolutely necessary for understanding that type. Between Meeting One and Meeting Two: SOD, LH, CB, and RG independently assigned each case to one of the preliminary types and made explanatory notes if they had difficulty assigning a particular case. They also made notes about any issues or inconsistencies with the types themselves that had emerged as a result of attempting to assign the cases. Meeting Two (August 2023): SOD, LH, CB, and RG reviewed and discussed their assignment of cases to the preliminary types. For n = 49 cases, this discussion resulted in agreement on the assignment. For n = 17 cases, agreement could not be reached, or team members felt another reading of the case files was required to make a decision. These 17 cases were referred to as Queries. Between Meeting Two and Meeting Three: SOD, LH, CB, and RG re-read the case files for the Queries and tried again to assign them to a type. Also, between Meeting Two and Meeting Three, each member of the team independently wrote a description of each type, outlining the unique elements of that type and identifying a potential archetypal case. Meeting Three (October 2023): SOD, LH, CB and RG reviewed the Queries, with consensus considered to have been reached when at least three of the four team members agreed on the assignment of a case. In reviewing and assigning Queries to types, the team drew on Weber’s concepts of adequacy on the level of meaning and causal adequacy (Swedberg 2017) and considered two questions: (1) What was the primary motivation or driving force behind the homicide? and (2) Would the homicide have occurred without this? (For example, in cases assigned to the Neglect type, would the victim have been killed in a more active manner by the perpetrator if they had not died of neglect?). Where the answer to the second question was ‘No’, the team felt confident in assigning cases to that type. During the review of Queries, two cases were excluded (one because there was insufficient evidence of a caregiving role and one because there was insufficient evidence of a disability or long-term illness). Of the remaining fifteen Queries, eleven remained in the same type as originally proposed in Meeting Two and four were assigned to a different type. The Meeting Three discussions also led to an additional four cases (whose type was initially agreed in Meeting Two) being assigned to a different type. Following these discussions, each team member shared their description of the types and their potential archetypal case for each one. Type names and descriptions were then refined until consensus was reached. In refining our descriptions of the types, we were informed by Weber’s notion of ‘interpretive adequacy’ and sought to ensure that each description included the course of events or actions that were essential for meaningfully distinguishing one type from another (Gerhardt 1994).
Checking Credibility	The refined types, their descriptions, and a selection of cases were given to two members of the research team who had not been involved in the original analysis (GMT, DS). GMT and DS worked independently and for each case read the entire case file, assigned the case to a type, and provided a written rationale for their decision. Where they were unable to assign a case to any of the types, they indicated this and provided an explanation for their decision. In selecting cases for the credibility check, SOD and LH chose two cases per type (n = 14 cases or approximately 20% of the total sample), with one being an archetypal case and the other selected randomly from the list of cases originally assigned to that type. GMT and DS were not told how cases had been assigned by the other members of the team until they had completed the credibility check. GMT and DS were able to assign all fourteen cases to a type. For seven (50%) of the cases used in the credibility check, there was complete agreement (that is, GMT and DS’s assignment to type matched each other and matched the wider research team). For four additional cases, either GMT or DS’s assignment matched that of the wider research team (with the other person’s notes indicating that they had vacillated between two types, one of which was the type that case was assigned to by the wider research team) and this was taken as majority agreement. For three cases, neither GMT nor DS had assigned the case to the same type as the wider research team. Meeting Four (December 2024): SOD, LH, CB and RG met to explore whether those three cases needed to be reviewed again or if the type definitions needed to be refined to ensure all cases could be reliably assigned to a single type. As a result of this meeting, two of the three cases remained in the type to which the wider team had assigned them, and one case was assigned to a different type.
Making Comparisons	To enable comparison across cases, data were extracted from the case files and entered into a matrix created in Microsoft Excel. Each case was one row in the matrix, with columns capturing the type to which the case had been assigned and key characteristics of the cases (including but not limited to: age; gender; ethnicity; relationship; care recipient’s illness or disability; nature of care provided; caregiver’s illness or disability (if any); social deprivation; squalor; past suicidal or homicidal ideation or attempts; prior convictions; view of the caregiving role; homicide method; homicide location; prior violence, coercive control, or neglect; engagement with services; statutory review recommendations; the nature and extent of media reporting; and criminal charges, convictions, and sentences. Where there were discrepancies in a case file (e.g., a media report and a statutory review gave different ages for the caregiver), these were noted in the matrix. Once the matrix was populated, the cases within each type were reviewed as a group and frequencies calculated for each characteristic. Because the ratio of cases to types was small, it was not possible to compare characteristics quantitatively across the types, so a qualitative comparison was conducted.

.

3. Results

3.1. Cases

Ninety-one potential cases were identified. On initial review, 25 of those did not meet our inclusion criteria and were excluded. Two additional cases were excluded during the Construction of Types. The 64 remaining cases were included in the Ideal Type Analysis. Forty-seven of these were homicides, ten were homicide–suicides, and seven were homicides followed by an attempted suicide. Among the cases that involved a homicide–suicide or a homicide followed by an attempted suicide, all but two of the (attempted) suicides occurred at the same time as, or immediately after, the homicide. In one Neglect case, the suicide attempt occurred on the day the neglect was discovered by authorities, but before the care recipient had actually died. In one Caregiver as Perpetrator of DVA/CC case, the features of the case meant it was not possible to determine exactly how long after the homicide the suicide attempt occurred, but there was clear evidence that the events were connected.

3.2. Types

The analysis identified seven types of caregiver-perpetrated homicides: Ending Suffering; Genuine Burden of Care; Pre-existing Mental Illness; Neglect; Exploitation; Caregiver as Victim of Domestic Violence, Abuse, or Coercive Control (DVA/CC)6; and Caregiver as Perpetrator of DVA/CC. These types are described in Section 3.3 below (with the number of cases in each type provided in parentheses), followed by the archetypal cases in Section 3.4 and a narrative summary of death and demographic characteristics by type in Section 3.5.

The term ‘care recipient’ has been used in the type descriptions to refer to the person with the long-term illness or disability. We have deliberately avoided using the terms ‘victim’ and ‘perpetrator’ when describing the homicides, in order to keep the focus on the caregiving relationship and avoid confusion with the types that involve DVA/CC (see below). The term ‘Genuine’ has been included in the Genuine Burden of Care type to distinguish these cases (in which there was clear evidence that the caregiver was under significant strain) from cases in other types where the caregiver (without evidence of such strain) claimed burden as part of their legal defence.

3.3. Type Descriptions

3.3.1. Ending Suffering (n = 8)

At the time of the homicide, the care recipient was experiencing pain, suffering, or distress as a result of their illness or disability, and the caregiver believed their actions would end that suffering. This type included—but was not limited to—cases where the care recipient had specifically asked the caregiver to help them die. For cases of homicide–suicide (or homicide followed by an attempted suicide) to be included in this type, there had to be clear evidence that ending the suffering of the care recipient was the primary motive for the homicide.

3.3.2. Genuine Burden of Care (n = 15)

At the time of the homicide, the caregiver was completely overwhelmed by the physical, emotional, social, and/or financial demands of the caregiving role (with alternative care arrangements either unavailable or considered unacceptable by the caregiver and/or care recipient). The homicide generally occurred in response to a crisis (including a recent move into a long-term care facility) or a sudden increase in care demands. Up until the homicide, cases in this type were characterised by good provision of care, generally good engagement with services, and help-seeking. For cases of homicide–suicide (or homicide followed by an attempted suicide) to be included in this type, there had to be clear evidence that a reprieve from the burden of care was the primary motive for the homicide.

3.3.3. Pre-Existing Mental Illness (n = 9)

The caregiver had a mental illness that is typically characterised by symptoms such as delusions or paranoia. The homicide occurred when the caregiver was experiencing psychosis or had poorly managed symptoms (including non-compliance with medication or as a result of intoxication). Although some caregivers with pre-existing mental illnesses also experienced a genuine burden of care, to be included in this type it had to be clear that the homicide was not a response to that burden. This type included—but was not limited to—cases where the caregiver was experiencing acute symptoms of dementia, such as delusions or delirium.

3.3.4. Neglect (n = 8)

The care recipient died because the caregiver failed to provide the level of care sufficient to ensure their survival. Cases in this type were characterised by profound neglect over an extended period of time (usually several years), as well as a marked lack of engagement with (or complete disconnection from) services and extreme social isolation. In these cases, the poor standard of living generally pre-dated the caregiving role (i.e., the onset of the illness/disability). Although some caregivers who neglected care recipients also experienced a genuine burden of care, to be included in this type it had to be clear that the neglect could not be attributed solely to the burden of care.

3.3.5. Caregiver as Victim of DVA/CC (n = 5)

Prior to the homicide, the relationship was characterised by a long-standing pattern of violence, abuse, or coercive control perpetrated by the care recipient against the caregiver. The homicide occurred when the caregiver ‘snapped’ or acted to protect themselves from further abuse.

3.3.6. Caregiver as Perpetrator of DVA/CC (n = 14)

Prior to the homicide, the relationship was characterised by a long-standing pattern of violence, abuse, or coercive control perpetrated by the caregiver against the care recipient. The homicide was a continuation or escalation of that abuse. For cases to be included in this type, it had to be clear that the abuse could not reasonably be attributed to the challenges of the caregiving role or the care recipient’s symptoms. Cases where a child under 16 was killed in the context of violence and abuse were included in this type7. Cases of neglect were also included in this type if there was clear evidence that the neglect was an extension of long-standing coercive control.

3.3.7. Exploitation (n = 5)

Prior to the homicide, the caregiver initiated a relationship with the care recipient to gain access to the money, goods, services, or housing the care recipient received as a result of their disability or illness. The homicide was a continuation or escalation of that exploitation, or a response to the possibility of the exploitation being ‘found out’. Although cases in this type often featured violence, abuse, or coercive control prior to the death, the relationship (and therefore the history of abuse) was generally short-term and unknown to agencies (or known to agencies, but with insufficient time to act).

3.4. Archetypes

A brief account of the archetypal case for each type is provided in

Table 2. Archetypal cases by type ^a.

Ending Suffering

Paula Meadows, 85, had dementia and was cared for by her husband, Tony Meadows, 84. In media reports, Tony was described by neighbours as taking very good care of Paula prior to her death. According to media reports, the coronial inquest was told that Paula had been in “absolute agony” in the weeks preceding her death due to back pain and that Tony had planned the deaths for several days. The couple died by asphyxiation in a homicide–suicide and the coroner described the homicide as an act of compassion. Their daughter was quoted as saying, “In my opinion daddy killed mummy because her life was dreadful.”

Genuine Burden of Care

Tristan Barrass, 14, and Blake Barrass, 13, were two of their parents’ six children. They had diagnoses of Attention-Deficit Disorder/Attention-Deficit Hyperactivity Disorder (ADD/ADHD) and speech and language difficulties. One of them also had a history of suicidal thoughts and self-harm. Two of the other four children also had special educational needs and disabilities. Their parents were Sarah Barass, 35, and Brandon Machin, 39. Although he did not live with Sarah and the children, and was not known to services as their father, Brandon was involved in his family’s life. The statutory review notes that professionals viewed Sarah’s parenting as “overwhelmingly positive and she was seen as a competent, caring, and articulate parent who supported and fought hard for her children’s access to appropriate support and help. She frequently self-referred to support services”. Despite this, there were difficulties in securing formal diagnoses, a lack of timeliness and consistency in the provision of support by some agencies, and failure to provide ‘whole family’ support (with the children being supported by multiple agencies with limited coordination and poor or delayed information sharing). Sarah also had difficulty gettobtaining information and support for managing her sons’ problematic sexual behaviours (which had escalated in the period preceding the deaths and included a report, on the day of the deaths, of a sexual assault against a child outside the family). According to media reports of the criminal proceedings, Sarah had once texted a friend, “I’ve thought of every possible solution to this mess. Mass murder, putting them all in care, checking into the local nut house. I love my kids too much to kill them, I can’t put them into care for the same reason”. Sarah and Brandon gave Tristan and Blake an overdose of medication, then later strangled and suffocated them. Sarah and Brandon pleaded guilty to two counts of murder, conspiracy to murder all six children, and five counts of attempted murder. They were both sentenced to a minimum term of 35 years’ imprisonment.

Pre-existing Mental Illness

Priscilla Edward, 78, had dementia and lived alone. Priscilla was cared for by several family members, including her daughter Regina Edwards, 52, who was described in the statutory review as spending “long periods with her during the day.” Regina had a history of mental health problems, including a diagnosis of schizoaffective disorder, and had previously been detained in the psychiatric unit of a hospital for 10 years after stabbing one of her children. Although Regina received psychiatric and psychological care in the years following her release from hospital, her compliance with medication was described in the statutory review as “fluctuating” and she was not seen by a psychiatrist in the five months preceding the death. On the weekend of the death, Regina stayed overnight with Priscilla and the statutory review notes that “she had left all her medication at home so took some of her mother’s old risperidone tablets, and … later admitted she had not been taking her own medication regularly before this.” The statutory review notes that on the weekend of the death, Regina thought her mother “looked different, really old and witchlike, and she thought her mother had long bony claws like a witch.” The statutory review further notes that Regina “had thought for many years that there was some generational issue with witches in the family” and said after the death that she “remembered thinking that her mother had to die to end the curse”. Regina strangled Priscilla. Regina pleaded guilty to manslaughter by diminished responsibility and was sentenced to life imprisonment with a minimum term of 10 years to be served in a secure hospital unit.

Neglect

Jordan Burling, 18, had speech and language delays, poor attention, and difficulty interacting with others (possibly caused by Fragile X Syndrome b). He was neglected by his mother (Dawn Cranston, 43), grandmother (Denise Cranston, 68), and sister (Abigail Burling, 23) and died from bronchopneumonia caused by malnutrition, immobility, and infected pressure ulcers. Expert witnesses at the criminal trial estimated that he had been severely malnourished for many months, possibly years, and that the pressure ulcers would have taken weeks, if not months, to develop. In sentencing remarks, the judge noted that “this was not a deprived household in materials terms … although the house was full of clutter you all had mobile phones, laptop or tablet computers, and a great deal of other equipment. The house was well stocked with food.” The judge also noted that although Dawn Cranston likely had a dissociative disorder it “in no way excuses your failure to take the simple step of calling a doctor much sooner”. Neglect was evident throughout Jordan’s childhood—including severely decayed teeth, poor hygiene, and recurrent headlice infestations—and the family repeatedly failed to take Jordan to appointments scheduled by health, social care, and education professionals. He was permanently removed from school by his mother at age 11. The statutory review notes that Jordan was “unknown to any services and virtually invisible to his local neighbourhood” for the three years preceding his death, but that during this time “other members of the family were continuing to use health services” and “at none of these appointments was [Jordan’s] condition raised or advice sought about his health.” Dawn and Denise were both found guilty of manslaughter and sentenced to four and three years’ imprisonment, respectively. Abigail was found guilty of causing or allowing a vulnerable adult’s death and sentenced to 18 months’ imprisonment.

Caregiver as Victim of DVA/CC

Julie Collier, 55, had a long-term alcohol addiction and serious alcohol-related illnesses, including cirrhosis of the liver. She also had mobility problems and poor vision, necessitating the use of a wheelchair or scooter outside the home. Her husband, Christopher Collier, 52, also had an alcohol addiction and alcohol-related illnesses. In the years preceding the homicide, he had disclosed to multiple professionals that his wife was physically abusing him, noting that he was scared of her and worried he might snap and harm her. He also told professionals that he was struggling to meet her care needs but could not leave the relationship because no-one else would care for her. Although there were multiple occasions when Julie presented to her GP or other health professionals with bruising, it could not be determined if these were sustained while drunk (GP records suggested some injuries were consistent with falling while drunk) or were the result of abuse by Christopher (either as the primary perpetrator or as a victim retaliating). Julie had a history of injuring herself while drunk and had physically assaulted a previous partner. After a night of heavy drinking, Julie hit Christopher, who retaliated by punching her, and she fell to the floor, possibly hitting her head. She died from a subdural haemorrhage caused by blunt force trauma. Christopher was charged with murder, pleaded guilty to manslaughter, and was sentenced to four years’ imprisonment.

Caregiver as Perpetrator of DVA/CC

Sian Blake, 43, had Motor Neurone Disease (MND) and lived with her partner, Arthur Simpson-Kent, 49, and their two children (aged 8 and 4). Although Sian only received a formal diagnosis of MND a few days before her death, her symptoms had been worsening and interfering with activities of daily living and employment for up to two years. At the time of diagnosis, she was told that the progressive nature of the condition meant she had between one and three years to live. During sentencing, the judge described Arthur as having “the difficult task of caring for Sian and their children as her health deteriorated.” The statutory review also noted that the children had taken on significant caregiving roles. Based on reports from her family and friends, the statutory review describes Arthur’s relationship to Sian and the children as controlling, noting that he isolated them from her family, limited her job opportunities, and strictly controlled and monitored their food intake. On the day of the death, Arthur had learned that Sian was planning to move back into her mother’s home with their sons. Sian and her children died after being beaten and stabbed. Arthur concealed the bodies, lied about Sian’s whereabouts, and fled to Ghana. Upon arrest, Arthur claimed he and Sian had made a suicide pact because of her illness, but there was no evidence to support this. Arthur was extradited to the UK, pleaded guilty to murder, and was given a whole life order (life imprisonment with no possibility of early release).

Exploitation

Adrian Munday, 51, was deprived of oxygen at birth and, in early adulthood, developed a drug addiction and serious mental health problems (most likely schizophrenia). His family described him as “particularly vulnerable” and substantial involvement from multiple agencies was required to help him live independently. A few weeks prior to his death, Adrian met Stuart Hodgkin, 40. Hodgkin was on probation and had 39 previous convictions for offences including arson, assault, and theft. The statutory review notes that Stuart “moved into Adrian’s accommodation and exploited him for money and possessions”. He also slept in Adrian’s bed, with the statutory review noting that “Adrian appeared to be sleeping on the floor”. Hodgkin identified himself to members of the public as Adrian’s caregiver. Stuart beat Adrian to death and then set his body on fire in what the judge described as “a determined effort to dispose of or disfigure his body.” Stuart was found guilty of murder and sentenced to life imprisonment with a minimum term of 20 years.

^a The archetypal case should not be used to infer that all cases in that type were the same age, gender, relationship, or illness/disability. ^b Fragile X Syndrome is “the leading inherited form of intellectual disability and autism spectrum disorder, and patients can present with severe behavioural alterations, including hyperactivity, impulsivity and anxiety, in addition to poor language development and seizures” (Hagerman et al. 2017, p. 1).

. After careful consideration, we decided that these accounts should not be anonymised. We justified this decision on the basis that (1) all the information used in this analysis is publicly available and (2) typologies developed through Ideal Type Analysis require validation. Although it may be distressing for surviving family and friends to see our description of these cases, their feedback on how the cases have been classified could help to validate the typology and refine it for further use in policy and practice.

3.5. Death and Demographic Characteristics by Type

The death and demographic characteristics of cases by type are detailed in

Table 3. Death and demographic characteristics by Type (n = 64 cases).

	Homicide Form	Methods Note: Number not provided because many cases used multiple methods	Location	Care Recipient Gender (n = 65 Care Recipients)	Caregiver Gender (n = 68 Caregivers)	Care Recipient Age (n = 65 Care Recipients)	Caregiver Age (n = 68 Caregivers)	Caregiver’s Relationship to Care Recipient	Care Recipient Illness or Disability Note: Number not provided because most cases had multiple conditions	Caregiver with Illness or Disability
Ending Suffering (n = 8 cases)	Homicide (n = 4) Homicide–Suicide (n = 3) Homicide–Attempted Suicide (n = 1)	Suffocation Overdose a Shooting Throwing b Strangulation Poisoning c	Home (n = 6) Long-Term Care Facility (n = 1)	Female (n = 6) Male (n = 2)	Female (Transgender) (n = 1) Male (n = 7)	Older Adult e (n = 8)	Adult (n = 3) Older Adult (n = 5)	Partner (n = 5) Adult Child (n = 3)	Chronic Pain Dementia Heart Condition Mental Illness Mobility Problems Multiple Sclerosis Parkinson’s Disease Pulmonary Disease Stroke Traumatic Brain Injury	Yes (n = 4) No/Not Known (n = 4)
Genuine Burden of Care (n = 15 cases)	Homicide (n = 6) Homicide–Suicide (n = 7) Homicide–Attempted Suicide (n = 2)	Beating d Strangulation Suffocation Overdose Stabbing Cutting Shooting Arson Throwing	Home (n = 14) Home on a visit from the Long-Term Care Facility (n = 1)	Female (n = 12) Male (n = 4) Note: Two care recipients (both male) killed in one case.	Female (n = 3) Male (n = 13) Note: Two caregivers (one male, one female) deemed responsible for the deaths in one case.	Child f (n = 3) Adult g (n = 3) Older Adult (n = 10) Note: Two care recipients (both children) killed in one case.	Adult (n = 5) Older Adult (n = 10) Note: Two caregivers (both adults) deemed responsible for the deaths in one case.	Partner (n = 10) Adult Child (n = 1) Parent h (n = 5) Note: Two caregivers (both parents) deemed responsible for the deaths in one case.	ADD/ADHD Arthritis Autism Cerebral Palsy Chronic Obstructive Pulmonary Disease (COPD) Diabetes Dementia Emphysema Epilepsy Global Developmental Delay Heart Condition Learning Disability Mild Cognitive Impairment Mental Illness Mobility Problems Quadriplegia Rubenstein–Taybi Syndrome Speech and Language Disorder	Yes (n = 12) No/Not Known (n = 3)
Pre-existing Mental Illness (n = 9 cases)	Homicide (n = 7) Homicide–Attempted Suicide (n = 2)	Stabbing Beating Decapitation Strangling Shooting Overdose Smothering	Home (n = 9) Long-Term Care Facility (n = 1)	Female (n = 6) Male (n = 1)	Female (n = 3) Male (n = 6)	Adult (n = 1) Older Adult (n = 8)	Adult (n = 6) Older Adult (n = 3)	Partner (n = 2) Adult Child (n = 7)	Addiction Arthritis Cancer Chronic Pain Dementia Diabetes Hearing Impairment Heart Condition Mental Illness Mobility Problems Paraplegia Spinal Stenosis Stroke	Yes (n = 9)
Neglect (n = 8 cases)	Homicide (n = 7) Homicide–Attempted Suicide (n = 1)	Failure to: Feed Clean Toilet Turn Move after fall Treat infections or sores	Home (n = 8)	Female (n = 5) Male (n = 3)	Female (n = 7) Male (n = 5) Note: Two caregivers (both male) deemed responsible for the death in one case. Three caregivers (all female) deemed responsible for the death in one case. Two caregivers (one male, one female) deemed responsible for the death in one case.	Adult (n = 3) Older Adult (n = 5)	Adult (n = 8) Older Adult (n = 4) Note: Two caregivers (one adult, one older adult) deemed responsible for the death in one case. Three caregivers (two adults, one older adult) deemed responsible for the death in one case. Two caregivers (one adult, one older adult) deemed responsible for the death in one case.	Partner (n = 2) Adult Child (n = 4) Parent (n = 3) Sibling (n = 2) Grandparent (n = 1) Note: Two caregivers (one parent, one sibling) deemed responsible for the death in one case. Three caregivers (one parent, one grandparent, one sibling) deemed responsible for the death in one case. Two caregivers (one former spouse, one adult child) deemed responsible for the death in one case.	Addiction Arthritis Dementia Developmental delays Devic’s Disease Down Syndrome Emphysema Liver Disease Mobility Problems Traumatic Brain Injury	Yes (n = 5) No/Not Known (n = 3)
Caregiver as Victim of DVA/CC (n = 5 cases)	Homicide (n = 5)	Stabbing Smothering	Home (n = 4) Elsewhere (n = 1)	Female (n = 3) Men (n = 2)	Female (n = 2) Male (n = 3)	Adult (n = 4) Older Adult (n = 1)	Adult (n = 3) Older Adult (n = 2)	Partner (n = 4) Adult Child (n = 1) Parent (n = 1)	Addiction Autism Cancer Chronic Pain Cirrhosis Learning Disability Mental Illness Mobility Problems Traumatic Brain Injury Vision Impairment	Yes (n=4) No/Not Known (n=1)
Caregiver as Perpetrator of DVA/CC (n = 14 cases)	Homicide (n = 13) Homicide–Attempted Suicide (n = 1)	Beating Stabbing Throwing Strangulation Drowning Neglect (as extension of CC) Shooting	Home (n = 13) Elsewhere (n = 1)	Female (n = 11) Male (n = 3)	Female (n = 1) Male (n = 13)	Child (n = 1) Adult (n = 9) Older Adult (n = 4)	Adult (n = 11) Older Adult (n = 3)	Partner (n = 9) Parent (n = 1) Friend (n = 2) Lodger (n = 1)	Addiction Amputation Arthritis Cancer Chronic Pain Diabetes Dementia Fibromyalgia Global Developmental Delay Heart Condition Hemiplegia Kidney Disease Learning Disability Mental Illness Mobility Problems Morbid Obesity MND Multiple Sclerosis Traumatic Brain Injury Vision Impairment	Yes (n = 9) No/Not Known (n = 5)
Exploitation (n = 5 cases)	Homicide (n = 5)	Beating Arson	Home (n = 4) Elsewhere (n = 1)	Female (n = 1) Male (n = 4)	Female (n = 2) Male (n = 3)	Adult (n = 3) Older Adult (n = 2)	Adult (n = 5)	Partner (n = 1) Lodger (n = 3) Neighbour (n = 1)	Addiction Acquired Brain Injury COPD Diabetes Epilepsy Heart Condition Learning Disability Mental Illness Stroke Vision Impairment	Yes (n = 4) No/Not Known (n = 1)

^a Prescription or over-the-counter medications. ^b Including throwing from a height, throwing across a room, or throwing down a flight of stairs. ^c Rat poison. ^d Including kicking, punching, hitting with a blunt object. ^e Aged 65 years or over. ^f Aged under 18 years. ^g Aged 18 to 64 years. ^h Including de facto step-parents.

. Homicides and homicide–suicides (or homicides followed by an attempted suicide) were evident in the Ending Suffering, Genuine Burden of Care, Pre-existing Mental Illness, Neglect, and Caregiver as Perpetrator of DVA/CC types. The Caregiver as Victim of DVA/CCC and Exploitation types featured homicides only. In the Neglect and Genuine Burden of Care types only, there were cases where multiple caregivers were deemed responsible for the care recipient’s death. In one Genuine Burden of Care case, two care recipients died.

Across all types except Neglect, the majority of homicides were violent (e.g., stabbing, beating, strangulation) and did not involve a firearm, but instead involved the caregiver’s hands or household objects such as dressing gown cords, walking sticks, electrical cables, and kitchen knives. In the Exploitation and Genuine Burden of Care types only, there were cases of homicide by arson.

Across all types except Neglect, the majority of caregivers were male. Across all types, the majority of care recipients were female.

In the Ending Suffering, Genuine Burden of Care, Pre-existing Mental Illness, and Neglect types, the majority of care recipients were older adults (aged over 65). Children as care recipients were only evident in the Genuine Burden of Care and Caregiver as Perpetrator of DVA/CC types (although in one Neglect case, the care recipient died at the age of 18 as a result of neglect throughout childhood). All the caregivers were adults or older adults.

Exploitation was the only type where the majority of relationships were not familial. In all other types, the majority of caregivers were either the adult child, parent, or partner of the care recipient.

Across all types, the majority of homicides occurred in the home, and the caregiver and care recipient were (or had recently been) cohabiting. In the Ending Suffering and Genuine Burden Care types, some homicides occurred in a long-term care facility (e.g., nursing home or care home) or when the care recipient had been taken out of the facility for a visit to the family home.

There were no obvious patterns within or across types in terms of the care recipients’ illnesses or disabilities, although it is notable that in many cases (across all types), the care recipient had multiple long-term conditions. In many cases (across all types), the caregiver also had an illness or disability, although in the Genuine Burden of Care type official reports frequently suggested that this was a consequence of the caregiving role.

4. Discussion

To our knowledge, this is the first typology of homicides perpetrated by caregivers. The sixty-four cases clustered into seven clear types, each characterised by a distinct motive, context, or course of events leading to the homicide. This challenges previous claims that caregiver-perpetrated homicides are isolated events and highlights opportunities for prediction and prevention.

The Ending Suffering, Genuine Burden of Care, Pre-existing Mental Illness, Neglect, and Caregiver as Perpetrator of DVA/CC types align with the findings of previous studies, where suffering, burden, mental illness, DVA/CC, and neglect were reported as factors contributing to caregiver-perpetrated homicides (Karch and Nunn 2011; Brown 2012; Bourget et al. 2010; Frederick et al. 2019; Stoneman et al. 2022). The current study goes beyond previous work, however, in recognising these as unique types of caregiver-perpetrated homicides, with distinct characteristics. The Ideal Type Analysis also identified two additional types—Exploitation and Caregiver as Victim of DVA/CC—that have, to date, been largely overlooked.

DVA/CC was the driving force in only two of the types. This is consistent with previous studies, where the majority of caregiver-perpetrated homicides were found to have no history of DVA/CC prior to the homicide (Bourget et al. 2010; Karch and Nunn 2011; Schwab-Reese et al. 2021). Where DVA/CC was present, the typology distinguished between abuse perpetrated by the caregiver and abuse perpetrated by the care recipient. None of the previous studies of caregiver-perpetrated homicides have considered the latter, which may be due to the difficulty of distinguishing between the perpetrator and the victim of DVA/CC in relationships where the victim retaliates or the lack of research more generally on the experiences of DVA/CC victims who become caregivers for the perpetrators of that abuse (Kernsmith 2005; Isham et al. 2017; MacPherson et al. 2020). Although disabled people are statistically more likely to be victims of DVA/CC (Dammeyer and Chapman 2018; Breiding and Armour 2015; Krnjacki et al. 2016), prediction and prevention of the Caregiver as Victim type will depend on policy makers and frontline professionals recognising that disabled and chronically ill people can also be perpetrators of abuse. Without this, abuse against caregivers may go undetected and hinder the timely provision of support that could prevent homicides of this type.

In separating cases with DVA/CC as the driving force from cases with other driving forces, the typology also necessitates more nuanced thinking about the motivation and risk factors for caregiver-perpetrated homicides. In particular, it requires researchers, policy makers, and frontline professionals to recognise that (1) a caregiver who provides good quality care and engages well with services can still resort to homicide when overwhelmed by the caregiving role or the suffering of the care recipient, and (2) for these caregivers, homicide may seem like the ultimate act of care. Consequently, while detecting and preventing DVA/CC is necessary, in the absence of other initiatives it will not be sufficient to prevent all types of caregiver-perpetrated homicide (Karch and Nunn 2011).

Relatedly, mental illness was the driving force in only one type. As Frederick et al. (2019) note, for some caregivers, homicide or homicide–suicide is “a rational decision albeit one that is unpalatable” (p. 323). This view is supported by studies that have identified thoughts of suicide and homicide in caregivers without mental illness (O’Dwyer et al. 2021; O’Dwyer et al. 2024). So while identifying and treating mental illness (including addiction) in caregivers may help to prevent some caregiver-perpetrated homicides, additional action will be required to predict and prevent homicides that are driven by a desire to end suffering or escape the burden of care (O’Dwyer et al. 2016; Karch and Nunn 2011).

The terms ‘mercy killing’ or ‘altruistic motives’ have frequently been used to describe caregiver-perpetrated homicides that occur in the absence of mental illness or DVA/CC (Brown 2012; Malphurs and Cohen 2002; Roberto et al. 2013; Lucardie and Sobsey 2005; Salari 2007). As Canetto and Hollenshead (2000) note though, when it comes to caregivers and care recipients, “it can be difficult to ascertain who was feeling the most pain and who wanted to be relieved of it” (p. 95). In distinguishing between Ending Suffering and Genuine Burden of Care, the typology goes some way to addressing this problem and providing more precise language with which to describe and understand these homicides. It also aligns with research that has shown some caregivers contemplate homicide–suicide because they want to end their own life but cannot bear to leave the care recipient behind, while others contemplate homicide–suicide because they want to end the life of the care recipient but cannot bear to be left alone (or face the criminal consequences of their actions) (O’Dwyer et al. 2016; Kim et al. 2019; Nakigudde et al. 2016)8. Of course the concept of ending suffering is still morally ambiguous, and homicides—no matter how compassionate the motive—are still crimes (Brown 2012; Malphurs and Cohen 2002; Roberto et al. 2013; Lucardie and Sobsey 2005; Salari 2007). The typology, therefore, should not be used to endorse or pardon homicides of this kind, but onlyrather to facilitate a more nuanced approach to understanding and addressing them.

Exploitation has been given only limited consideration in previous research on caregiver-perpetrated homicides (see Benbow et al. 2019; Roberto et al. 2013; Schwab-Reese et al. 2021). This may be due to the narrow definition of caregiver used in previous studies and/or because surviving family members disputed the fact that care was provided by housemates, friends, or neighbours (Benbow et al. 2019). The presence of this type (along with the DVA/CC types) challenges the view that relationships which include the material provision of care are caring relationships. The notion that care is only ever provided in relationships characterised by love, devotion, and kindness is pervasive, with care recipients often described by policy makers and professionals as “loved ones”, despite clear evidence that they are not all loved (Weicht 2008; Robertson 2023). And, as the Genuine Burden of Care type demonstrates, even when care recipients are loved, love may not be enough to mitigate the overwhelming demands of the caregiving role (Weicht 2008; Robertson 2023; O’Dwyer et al. 2016). Acknowledging that (a) not all caregivers are family members and (b) caregiving relationships can be violent, coercive, or exploitative, may be an important step towards predicting and preventing caregiver-perpetrated homicides of the Exploitation (and DVA/CC) types.

Although cases of neglect have been identified in previous studies of caregiver-perpetrated homicides and homicide–suicides (Roberto et al. 2013; Karch and Nunn 2011; Lucardie and Sobsey 2005; Frederick et al. 2019), the typology offers more insight into these deaths. In particular (and consistent with the broader literature on neglect), cases in the Neglect type were characterised by extreme social isolation and disconnection from formal services (Brandon et al. 2014; Ansello and O’Neill 2010). Although caregivers who are disconnected from services may be harder to identify and support, the passive nature of these homicides means there is considerable time in which to prevent them. Given that a poor standard of living (often characterised by squalor or hoarding) generally pre-dated the onset of the caregiving role in this type, consideration should be given to who is fit to take on a caregiving role, and homicide prevention may ultimately depend on removing vulnerable people from the care of those deemed unfit.

A wide range of illnesses/disabilities and caregiving relationships were evident across types. By focusing on particular sub-groups of caregivers or care recipients, previous studies have inadvertently implied that certain relationships, ages, or illnesses/disabilities pose a greater risk for homicide. The typology does not support this and, instead, highlighting the importance of recognising caregivers as a population in their own right. This may be particularly important for informing policy (Brown 2012; Frederick et al. 2019; Ohrui et al. 2005). In England, for example, caregivers are not currently recognised as a priority group in the national Suicide Prevention Strategy (DHSC 2023) nor in the Serious Violence Strategy (Home Office 2018), and the risk of homicide was not mentioned in the national Carers Action Plan (which expired in 2020 and, despite calls from advocacy groups, has not yet been replaced with a new plan) (DHSC 2018). Similar oversights are also evident in relevant policies internationally. Although our finding of approximately 13 caregiver-perpetrated homicides per year in England and Wales supports claims that caregiver-perpetrated homicides are rare, they are not rare enough to justify this lack of recognition. Guided by the typology, future research must take a more holistic approach to the study of caregiver-perpetrated homicides in order to facilitate evidence-based policy that could help prevent these deaths.

The use of violent methods by caregivers—particularly in the Ending Suffering and Genuine Burden of Care types—may seem counter-intuitive, but it is consistent with previous studies (Ohrui et al. 2005; Coorg and Tournay 2013; Malphurs et al. 2001; Karch and Nunn 2011; Lucardie and Sobsey 2005; Jung et al. 2020). It could be an issue of gender (with the majority of caregivers who perpetrated homicide being male) (Canetto and Hollenshead 2000; Garcia-Moreno et al. 2015), a lack of access to alternative means (such as controlled drugs), or the limited efficacy of less violent means (in several cases smothering or medicating was followed by more violent methods). Although restricting access to means is a common recommendation for the prevention of caregiver-perpetrated homicides—particularly in places with easy access to firearms (Malphurs et al. 2001; Cheung et al. 2016; Eliason 2009)—most of the cases in the current study used common household items, for which restriction is impossible. A focus on improving support for caregivers, removing caregiving responsibilities from those who are ill-suited to the role, and tackling male violence more generally, may be more effective than means restriction in preventing homicides perpetrated by caregivers.

4.1. Strengths and Limitations

This study is the first of its kind. With a systematic approach for identifying cases, a rigorous method for establishing types, multiple sources of data for each case, and no limits on the nature of the caregiving relationship or the illness/disability, it overcomes the fragmented nature of previous research and offers a tangible framework to guide future research, policy, and practice in the UK and globally.

It does, however, have some limitations. Although statutory reviews and their publication are a legal requirement in England and Wales—and so should have been a reliable means of identifying all relevant cases—for one-third of cases the review was not available (either because it had not been published or the local authority did not respond to our request to supply it). In addition, 82 local authorities failed to respond to our request for confirmation of the cases identified in their counties/boroughs, so it is possible some relevant cases were not included. The study also relied on publicly available data, and there was considerable variation in how much information was available about each case from media reports, statutory reviews, sentencing remarks, and coronial reports. Access to complete information would have enabled more in-depth descriptions of the types, including more detailed demographic profiles (including ethnicity, education, and occupation, which were not reported consistently enough to include reliably in this paper). To that end, we endorse calls made by other scholars for routine collection and reporting of disability and caregiving data in official homicide statistics (Frederick et al. 2019; Malphurs and Cohen 2002; Cohen 2019; Guan et al. 2022) and add our own call for mandatory (and enforced) publication of all sentencing remarks, coronial reports, and statutory reviews.

The Ideal Type Analysis was based on cases in England and Wales only, over a five-year period, so the typology will need to be validated in other countries and across other time periods. Although we are confident that the types have global applicability, validation studies in countries outside the UK could determine whether there is a need to add more types, or modify the description of existing types, to capture culturally specific motives or contexts (for example, the ritual killings of disabled people that occur in some African nations; Taylor et al. 2019; Bayat 2015). In addition, with sixty-four cases spread across seven types, our cell sizes were too small for statistical comparisons of the demographic characteristics. Larger samples collected over longer time periods or across multiple countries could help to overcome this in future studies and provide greater insight into demographic risk factors for each type.

Stapley et al.’s (2022) method for Ideal Type Analysis stipulates that each case must fit into only one type. Consistent with this, and informed by Weber’s concepts of mental experiments and adequacy (see Table 1), we focused on the driving force or primary motivation for each homicide when identifying the types and assigning cases to them. There is a risk, however, that this leads policy makers and frontline professionals to assume these are the only factors contributing to homicides in each type. Caregiver-perpetrated homicides are highly complex and a range of intersecting factors can contribute to these deaths (Frederick et al. 2019). For example, some cases of Ending Suffering or Genuine Burden of Care in this study also featured mental illness or DVA/CC. So although the typology offers an important conceptual lens through which to understand and address caregiver-perpetrated homicides as a phenomenon, more research is required to understand how complexity at the level of the individual case can be addressed in policy and practice (Swedberg 2017).

4.2. Future Directions

Despite these limitations, this study represents a significant step-change in research on caregiver-perpetrated homicides, both in the UK and globally. It will no longer be “necessary to borrow from what is known” about other domestic homicides (Brown 2012, p. 7), because the typology provides a framework for dedicated research on caregiver-perpetrated homicides that can be used to inform evidence-based prediction and prevention.

In particular, the typology could be used to guide more in-depth research on the risk and protective factors for caregiver-perpetrated homicides. Although the description of each type offers some direction for prediction and prevention, a deeper understanding of the social, economic, and political contexts in which the different types of homicides take place could facilitate more targeted action in policy and practice (Frederick et al. 2019; Karch and Nunn 2011; Malphurs et al. 2001). The typology could also be used to guide research on the criminal justice responses to caregiver-perpetrated homicides and to examine media coverage of these events. Previous research has highlighted a lack of consistency in charges and sentencing when caregiver-perpetrated homicides are prosecuted (Jung et al. 2020; Canetto and Hollenshead 2000; Catlin 2003; Brown 2012; Cohen 2019; Benbow et al. 2019), but no systematic research has been undertaken. There have also been suggestions that overly detailed media reporting of caregiver-perpetrated homicides can give caregivers ideas for homicide methods that ‘work’ (O’Dwyer et al. 2016), while sensationalist reporting may make it harder for caregivers who are contemplating homicide to seek help. The typology could provide a lens through which to promote evidence-based charging and sentencing and develop guidelines for safer media reporting.

There is also a need for research on the lived experience of caregiver-perpetrated homicides (Canetto and Hollenshead 2000). This study drew on secondary data, which often obscured the voices of those closest to the cases. Qualitative research involving caregivers who have perpetrated homicides, surviving families, and the legal, health, and social care professionals who have worked on these cases could further understanding of each type of caregiver-perpetrated homicide. Alongside this, research on the experiences of caregivers who have attempted homicide (or care recipients who have survived an attempted homicide) would strengthen understanding of the factors that prevent caregiver-perpetrated homicides. As Biron and Reynald (2015) note, “it should be self-evident that information from serious non-fatal (or near fatal) cases is vital in preventing future deaths: the question ‘why did this [person] survive’ is equally as important as the question ‘why did this [person] die?’” (p. 13; [child] in the original).

5. Conclusions

Caregiver-perpetrated homicides are not isolated events unworthy of investigation. They are a distinct phenomenon that requires dedicated research to inform nuanced prediction and prevention initiatives. As populations age and governments invest less in formal health and care services, the need for—and pressure on—caregivers will increase (Milne and Larkin 2023). As a result, the number of caregiver-perpetrated homicides and homicide–suicides is also likely to rise (Ohrui et al. 2005). When a caregiver kills there is usually a public outcry, but as Brown (2012) notes, “when the rhetoric dies down, it seems as if [those] who died, lived lives that consisted of a never-ending struggle but succeeded in attracting attention only when it was too late to help” (p. 7). The Caregiver-Perpetrated Homicide Typology presented here provides a clear framework for research and a starting point for targeted action in policy and practice. It is our fervent hope that it will draw attention to the very real risk of homicide in caregiving relationships while there is still time to help.