1.1. Background

Substantial growth in the proportion of community-dwelling older adults has been observed globally as a result of lower birth rates and increased longevity [1]. As the ‘baby boomer’ generation matures, the proportion of adults aged 65 years and older is expected to double between 2019 and 2050 [1]. In Australia, the proportion of the population aged 65 years and older, which was 16% in 2020 [2], is expected to reach 19% by 2050 [3], and those aged 85 years and older (2% in 2020) [2] are expected to reach four percent [3].

While it is not certain whether the increase in the number of adults over 65 years old in Victoria, Australia will equate to an actual increase in homicide incidence, there is little to indicate these numbers will decrease. One reason for this is that despite trends for decline in the general homicide rate, the rate for older adults has changed little over time [4]. In fact, trends in Australia for older adults have been reported as either slightly increasing [5] or stable [6], rather than decreasing, which has also been observed more recently in the United States [7,8]. This suggests a likely increase in older adult homicide events even without a notable change in rate. Adding to this, the most recent report on homicide in Australia noted a 16% increase in homicide on the previous year and the highest since the 2005–06 reporting period [9]. Contemporary issues that may contribute to a future with an increase in older adult homicides include inter-generational familial stressors, such as financial issues, caregiver stress, and increasing mental illness in the community [10].

With this increase in the older population, it is expected that there will be a concomitant increase in the frequency of older adult homicide. Research that specifically investigates the factors contributing to homicide among older adults is limited [4,10,11] despite an increasing volume of research describing the prevalence, characteristics and prevention of elder abuse and neglect [12–14]. According to the Global Status Report on Violence Prevention [15], key global data on homicide lacks the specificity required for research into this age group.

The first meta-analysis of international older adult homicide research described a pooled prevalence of older adult homicide of 2.02 per 100,000 population, and that older adult homicide victims were predominantly killed in their home, by either a family member or a person that was a stranger to them, and during either an argument or a criminal activity against them [4]. Compared to younger adult victims, older victims were significantly more likely to be female, die at the home location, by an offender that was a stranger to them, and during a criminal activity against them. Conversely, older adults were significantly less likely to be killed during an argument, or with the use of a firearm [4]. A recent review by Rogers and Storey [11] identified an emerging typology of the older female victim killed by someone known to her in her own home.

These contemporary reviews highlight the need for further evidence on the distribution and determinants of homicide amongst community-dwelling older adults. Furthering this research domain requires a deeper understanding of factors associated with the fatal incident that are potentially preventable. These include whether homicide results from ongoing abuse or neglect, an understudied area [11], and whether there were individual factors related to deceased or the offender such as presence of alcohol and/or other drugs, psychological and medical history and other sociodemographic variables [10,11,13,16].

The nature of adult homicide is heterogeneous and typologies are mostly characterised by the deceased-offender relationship [17,18], with sub-types commonly derived from the homicide motive [11,17,18]. Older adult homicides predominantly occur in the process of criminal activity or during an argument [4,11,19,20], though further data on the individuals involved and the homicide incident are lacking. This affects our understanding of what contributed to the event and how it might have been prevented. Offenders of older adult homicide are frequently known to the older adult victim [4], and familial offenders are not necessarily intimate partners [10]; for example their adult children [21], of which motives and contributing interpersonal factors are not well documented [10,22]. While much work has been done to specify intimate partner homicide (IPH) typologies [23,24], research into older adult IPH has done little more beyond the categorisation of ‘mercy killing’ [25], and deserves greater attention. Importantly, while older adult homicide victims are frequently killed by someone known to them [4,11], the proportion that are killed by their intimate partner is small when compared with younger adults [26,27].

1.2. The social-ecological model and older adult homicide

Violence is recognised as a complex social and public health problem [28]. Public health conceptual frameworks have been increasingly applied in interpersonal violence research and policy to identify and evaluate evidence-based primary, secondary and tertiary prevention strategies [29]. One such conceptual framework, the public health approach states that once a problem is defined, risk and protective factors are then identified so that appropriate prevention efforts can be developed, assessed for effectiveness and implemented [30].

Risk and protective factors can be examined using another public health framework, the social-ecological model (SEM), first introduced by Bronfenbrenner [31]. As with the public health approach, the SEM has been applied to strengthen the understanding of factors associated with and protective against interpersonal violence [18,30–32]. While Bronfenbrenner depicted the nested ecological model components as micro-, meso-, exo- and macrosystems, public health practitioners, and more specifically violence prevention researchers, have reframed the nested categories as individual, interpersonal, community, and societal [30,33]. In a modification appropriate to better understanding homicide, the incident location, mechanism of injury, weapon, injury severity, and time of day can be presented as a stand-alone ‘incident’ category that intersects the individual and interpersonal categories (Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Modified social-ecological model–older adult homicide.

https://doi.org/10.1371/journal.pone.0292837.g001

Individual level factors known to be associated with homicide among older adults include: sex; race and social isolation for the deceased; mental and physical health, and historical exposure to violence of both the offender and the deceased; historical drug and alcohol abuse or presence of alcohol in the offender at the time of the homicide event; and previous contact with the legal system or suicide threats by the offender [10,11,21]. In their study of older adult homicide-suicide, Malphurs found offenders had a low documented history of receiving psychiatric care (5%), which contrasted to the high proportion reported as displaying mental illness (51.9%) [34].

Across adults of all ages, mental illness, history of substance abuse [35], and a criminal record for intimate partner violence (IPV) may increase a person’s chance of committing homicide [36]. Previous criminal history, however, is not always present in IPH [37], and a history of criminal charges does not definitively indicate that a homicide will take place [38]. Being a migrant or an ethnic minority is also associated with a higher likelihood of IPH victimisation [39,40].

At the interpersonal level, factors associated with older adult homicide include: the deceased-offender relationship; motive; financial dependence of the offender on the deceased; physical dependence of the deceased on the offender; and for intimate partners, recent or impending separation; history of domestic violence/discord; and the age difference between the deceased and offender [10,22,27]. Malphurs [34] found homicide-suicide perpetrators were three times more likely to be a caregiver than those who died by suicide alone. The study also identified that spouses, followed by other family members, were the primary offenders of homicide-suicide, with physical or verbal marital discord noted for 19% of incidents, and 29% of homicide-suicides being related to an impending divorce.

It is not known whether established risk factors for IPH in adults in general, such as recent separation, access to weapons, a history of actual or threatened violence [35,39], or pre-migration trauma [39], apply to the same extent in older adult homicide. Being married rather than cohabiting was an identified protective factor for IPH across adult age groups [27].

A recent bereavement or change in life circumstance in the deceased has also been noted in an analysis of domestic homicide reviews in the UK [10], as well as dependence on the deceased by the offender and a caring role within the relationship dyad [41].

Incident level factors include: weapon used; location and time of day of incident; and the severity of injury, for example the use of excessive force [42]. A recent meta-analysis of older adult homicide found firearms to be used significantly less in older adults compared to younger adult homicide victims, and the homicide location to be significantly more often at home [4]. The mechanism is most frequently by sharp object, firearm or blunt trauma [11] and differs significantly to younger adult homicide [4]. In addition to the established concept that older adults are more susceptible to injury and a poorer recovery from physical assault [43], are studies concluding that attacks on older adults can involve a use of excessive force [42,43]. Additional empirical evidence is required to confirm excessive force is consistently more present than in younger counterparts.

Community level factors include geographic area and socio-economic variables such as household structure [44]. Older adult homicide rates vary by geographic location, for example prevalence is higher in Miami, Florida, compared to Chicago or Houston [26], and is higher for the US overall compared to other high-income Organisation for Economic Co-operation and Development countries [45]. In their analysis of culture and context, Weaver and colleagues [44] found that lower religious adherence and higher robbery rates at the community level were associated with higher prevalence of older adult homicide [44]. Adult IPV victims residing in rural locations have also been identified to have greater risk of homicide [39,46]. Other frequently examined community level factors in the US in particular, are the association of gun ownership and relevant state-based legislation on fatal violence [45,47].

Societal level factors include governmental and other cultural influences [48]. Exploring these societal structures is beyond the scope of the current study. Studies directed at the societal or structural level usually involve comparing large geographic areas according to the prevalence of homicide or another outcome of interest with known or potential factors at that level, focussing less on the individual, interpersonal, incident and community level factors [13,44,48].

Through the lens of the modified social-ecological model, there are still relevant details at the individual-, interpersonal- and incident-level that have not been described for older adults. Specific areas that require a greater understanding include potential risk factors for offenders [4,11], for example history of violence or mental illness or contact with the criminal justice system. This information is seldom reported in older adult homicide research, with the exception of research describing offender-only data collected via the criminal justice system [16].

Data that provides information about the interaction between the deceased older adult and the offender is also crucial, but is not usually explained beyond the relationship [49,50]. While understanding if there was a history of interpersonal violence is well recognised in intimate partner violence research [51], it is not frequently available in older adult homicide research. The same is true for history of recent separation, and age difference between the deceased and offender in partner homicide.

The incident circumstances also require a greater than basic description of weapon, location and motive [43,50], for example the use of alcohol and other drugs proximal to the event.

1.3. Medico-legal death investigation

In Australia, as in many other developed countries, unexpected and unnatural deaths, including suspected homicides, are required to be reported to a Coroner. Following a criminal investigation, the Coroner may perform additional investigations before they make a finding on: the deceased’s identity, cause of death and, in some cases, relevant proximate and historical circumstances leading up to the death [52]. This routinely collected information is recognised as both a reliable and valuable source for research [53].

Studies that have relied on administrative databases with uniformly collected data tend to provide a limited level of detail [43]. In the area of domestic homicide, there have been some promising studies from the United Kingdom containing a greater level of granularity [10,21], though these are derived from text-based documents that are usually qualitative in presentation and not prepared in a uniform manner.

Many studies examining homicide, particularly from offender-based data only, include cases where an offender has been identified and successfully prosecuted [16,54]. The use of coronial data permits inclusion of cases regardless of whether an offender was identified or there was a conviction.

In addition, domestic or family violence death reviews have been established in many Australian jurisdictions [52,55], and have led to the development of data systems to support their work. The first of these data systems, the Victorian Homicide Register (VHR), was established at the Coroners Court of Victoria (CCOV) by the Coroners Prevention Unit in 2014 to support family violence death reviews as part of their Victorian Systemic Review of Family Violence Deaths [52]. Coded and free text information is recorded for both the deceased and offender(s) and includes: socio-demographic characteristics; deceased-offender relationship; physical and mental health history (including alcohol and other drug use); criminal history; service contact history; and interpersonal factors, including any history of domestic violence perpetration or victimisation. This detailed and rich data makes the VHR a unique and useful information source to examine homicide at the individual, interpersonal, incident and community levels as set out by the modified SEM (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Individual, interpersonal, incident and community level variables associated with adult homicide and availability in the Victorian Homicide Register.

https://doi.org/10.1371/journal.pone.0292837.t001

1.4. Comparison with younger adult homicides

Use of comparisons groups for exploring the factors associated with homicide in older adults is conceptually and logistically challenging. One approach is comparison with non-fatal and serious injury of older adults within the geographic region. The comparison of older non-fatal and fatal assault has been performed in the US [43,49,82], and these highlight substantive methodological issues. One of these issues is that data sources differ due to the original purpose of information collection. Specifically, data sourced from surveys or the healthcare system include different variables to data generated from the criminal justice process, with further disparity in variable definition and record-keeping reliability [43,57].

Another issue is that homicide is reported at a high level of accuracy compared to assaults which may be reported either to the police or measured through hospital presentations [83]. The latter may only represent a fraction of the actual number of assaults in the population for multiple reasons which are difficult to reconcile [49].

An alternative approach to comparison of older adult homicides is of populations in different geographic regions. This also problematic because of variance in methods for collecting data within each jurisdiction, and the inability to control for community-level factors that might influence the homicide rate. Examples include legislation, political environments or other socio-cultural factors [11,44].

Comparisons between older and younger adults is a commonly used data analysis method in the international homicide research literature [4,11,57]. It also facilitates hypothesis generation and is reasonable when the limitations are acknowledged. This approach also recognises that, an optimal comparison study population may not be available, not logistically feasible to obtain, or of a such a low frequency count to negate robust statistical analyses.

There are benefits to developing policy and prevention initiatives by determining whether the underlying factors for homicide between young and older populations are similar or different. At closer inspection, individual studies have identified that rather than being an intimate partner, the relationship is more likely to be intergenerational, for example parent-child or grandparent-grandchild [10,21] The recent meta-analysis reported that older homicide victims were significantly more frequently female, killed by a stranger in their own home and were less often killed with a firearm than their younger counterparts [4]. These established factors recognise older adults as a discrete group.

It is important to specify the factors unique to older adults as well as those shared with other age and sex cohorts in the population. In their evaluation of the state of older adult homicide research, Addington [50] identified the need to expand on the descriptive research, including increasing understanding of the diversity among victims, offender characteristics, in-depth analyses, and identifying risk and protective factors.

There is value in comparing older and younger adult homicides from within the same society in the Australian context. An increased recognition and response to family violence in Australia over the last decade has resulted in strategies focused on younger people (usually women with children) who experience violence from men they have been in an intimate or familial relationship with [84]. While the focus of this response is reasonable given the public health burden, the strategies may not address the factors and circumstances unique to older adults. Specific examples where the scope is not specific to older adults include firearm restrictions [85], and intimate partner homicide risk assessment tools [51].

1.5. Research questions

As outlined above, the VHR offers a rich and unique data source in which to provide a detailed description of the nature of homicide against the older adult. The modified social ecological model is a useful framework for examining these known and potential risk factors at the individual, interpersonal, incident and community levels (Table 1). A comparison between older adult and younger adult victims will also serve to contextualise the older adult homicide situation. The following research questions will be addressed: 1. What are the individual (victim and offender), interpersonal, incident and community level factors associated with older compared to younger adult homicide? 2. Are the key factors associated with homicide among older adults in Victoria the same as for younger adults? 3. What are the characteristics of older adult homicide typologies based on the deceased-offender relationship? 4. Are the key factors associated with homicide among older adults in Victoria the same across deceased-offender relationship type? 5. What are the key factors associated with homicide perpetrated by a family member against older adults in Victoria? 6. Are the key factors associated with homicide perpetrated by a family member against older adults in Victoria the same as for younger adults?

1.6. Aims

This study will examine the epidemiology of homicides among community-dwelling older adults (65+ years) compared to younger adults (18‐64 years) in the Australian state of Victoria during the period 2000 to 2015, using data from the VHR. The specific aims are to: 1. Compare the annual frequency and rate per 100,000 population of homicide amongst community dwelling older and younger adults in Victoria;

2. Compare the individual, interpersonal, incident, and community level factors associated with specific older adult homicide typologies; and

3. Compare the individual, interpersonal, incident, and community level factors associated with family homicide among community-dwelling older and younger adults in Victoria identified from Aim 1.

2.1. Study design

The study will be a jurisdictional population-based retrospective analysis of consecutive homicides of community-dwelling older adults in Victoria, Australia for the period 2001 to 2015.The proposed study protocol to our knowledge, provides the first descriptive comparison of familial homicide between younger and older adults through the lens of the social-ecological model providing detail at the individual-, relationship-,incident- and community-level. It is also the first study to compare the rates for this specific time period and geographic location, contributing to the response to calls for more international empirical data in this area of research [7,50]. The study has institutional human research ethics committee approval and the initial data extraction has taken place. At the time of manuscript submission data had not been processed or coded beyond what exists following data collection and entry, and no analysis had taken place.

2.2. Ethics

The project was approved by the Victorian Department of Justice and Regulation Human Research Ethics Committee on 1 June 2017 (renewed 10 December 2020; JHREC Reference CF/20/17503), endorsed by the State Coroner on 18 April 2017 (renewed 30 October 2020; CCOV RC Reference 376) and the Monash University Human Research Ethics committee (Project 29051, registered 17 May 2021).

2.3. Setting

The research setting is Victoria, a state of Australia, that in 2016 had a population of 5,926,624, 16% of which (922,603) was aged 65 years and older [87]. In 2021, there were 16 homicide incidents in Victoria involving and older victim aged 55 years and over and the rate for this group was 0.9 per 100, 000 population [88]. Victoria has seen a yearly increase in offender incidents of crimes against the person, that includes homicide and serious assault, with 49,000 recorded incidents for the period ending March 2022, up from 43,380 in 2018 [89].

The study will include adult homicide deaths reported to a coroner in Victoria between 1 January 2001 (the first year for which VHR data is available) and 31 December 2015, where the coroner’s investigation has been finalised. The time period was selected because coroners’ investigations in homicide investigations can take some years to finalise—the coroner by legal convention waits until proceedings in other jurisdictions (such as criminal trials and appeals) are concluded before commencing the coronial investigation—and 2015 was determined to be the most recent year for which most or all homicide investigations will be finalised at data collection.

2.4. Data source

The primary data source for the study is the VHR, a purpose-built database developed and maintained by the Coroners Prevention Unit, CCOV. The VHR is a Microsoft Power Platform database comprising coded, structured and unstructured free text entered following a review of the information generated for the Coroners’ investigations and Judges’ sentencing remarks (where present) for homicide and suspected homicide deaths. The VHR contains 242 variables comprising a core and enhanced dataset coded from Victorian coronial records, specifically: police report of death to the Coroner; post-mortem medical and scientific reports (e.g., autopsy and toxicology reports); the coronial brief; the Coroners’ finding; sentencing remarks from any criminal proceedings; and family violence death review reports (if present). Cases classified as family violence are coded with the additional enhanced dataset that includes specific data on history of family violence perpetration or victimisation for the deceased and offender.

Specially trained project officers at the CCOV maintain the VHR, where entries undergo regular and rigorous quality assurance, specifically full and partial validity checking, and random full-record entry by a second coder, to confirm inter-coder reliability and resolve disparities through consensus agreement.

An advantage of the VHR is the broad inclusion criteria on initial screening to facilitate the inclusion of all potential homicides in the database, regardless of whether an offender is identified, which is updated once the investigation is completed by the Coroner (see Table 2 for detailed definitions). This reduces the likelihood of selection bias by ensuring homicide cases are included in the dataset, regardless of criminal conviction status, and that it is kept up-to-date as new information arises.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Definitions of terms in this protocol.

https://doi.org/10.1371/journal.pone.0292837.t002

The initial case identification process for the VHR is rigorous in that it includes known and suspected homicides at the onset of police investigation. A comparative study found that the National Homicide Monitoring Program, which uses a similar case identification method, yielded a higher number of homicides over a 12 year period than the causes of death data from the Australian Bureau of Statistics [93]. Case identification for the VHR is also more robust than other international homicide research that use voluntary reporting of homicides. For example, Chu identified that supplemental homicide reports may be missing 8% of homicide cases [43].

2.5. Case identification

Cases will be identified using the VHR where the homicide incident occurred between 1 January 2001 and 31 December 2015, and the Coroners’ investigation has been completed by 31 July 2017. Cases will be identified by an experienced member of the Coroners Prevention Unit (JD) against the inclusion criteria using a query of the VHR and exported to a Microsoft Excel spreadsheet.

2.6. Eligibility criteria

While the inclusion age for older adult homicide research can vary between populations and geographic locations, the predominant cut-off is 65 years [11,57]. This also suits the Australian context as it is the general administrative cut-off used to describe older adults [90].

Fig 2 shows the case identification process. Cases will be included if: the incident and death occurred in Victoria during study period of 1 January 2001 and 31 December 2015; the Coroners’ investigation and subsequent data entry is complete at time of data extraction; the deceased was aged 18 years or older at the time of the fatal incident; and the coroner determined that the death was the result of homicide.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Case Identification diagram: Adult homicides in Victoria, Australia 2001–2015.

^aSource: Victoria Police Crime Statistics.

https://doi.org/10.1371/journal.pone.0292837.g002

Cases will be excluded if the incident occurred outside the Australian state of Victoria or the deceased was residing in an institution (e.g., nursing home, prison, mental health facility, hospital, detention centre) at the time of the incident.

2.7. Sample size

There are 758 eligible homicides for the study period, 63 older adult homicide victims (aged 65 years and older) and 695 younger adults (aged 18–64 years) to serve as a comparison group. Numbers will be verified during the data processing and descriptive analysis process. All of the eligible older adult homicides have been completely coded across the VHR domains relating to inclusion, sociodemographic, identity, context, mental illness, mental treatment and service contacts (described below and at Table 1). All of the younger adult homicides (n = 695) have been coded with core data suitable for summary comparison with older adult homicide victims (Table 1). This includes: age, sex, relationship, local government area (LGA) of incident, incident location type, mechanism of injury, use of excessive force (overkill) or presence of alcohol and drugs in the deceased and details of intimate partner relationship status, including intention to or recent separation and relationship duration.

2.8. Data collection/variables

The primary outcome measure for the study is the incidence of homicide in older adults. The secondary outcomes include the nature of older adult homicide (the individual-, interpersonal-, incident- and community-level factors described below); the nature of older adult homicide typologies based on the deceased-offender relationship (intimate and familial, acquaintance, stranger or unknown); and the comparison of older adult homicide with younger adult homicide. The variables to be extracted from the VHR for each homicide case of victims aged 65 years and older (and for a subset of younger adult victims) are shown in detail at Table 1, along with relevant research evidence and availability within the VHR for deceased, offender, older and younger adults.

Individual level variables include: demographic characteristics, substance use or exposure to violence; mental or physical illness diagnoses and treatment, and service contacts for both the deceased and offender (Table 1). The following new individual level variables will be generated from the existing data: Age group in ten-year brackets.

Interpersonal level variables include: the deceased-offender relationship, motive, and family violence history. New interpersonal level variables will include: Age difference between deceased and offender, motive (collapsed), and homicide-suicide.

Incident level variables include: location description (i.e. victim’s home), Overkill (excessive physical force), toxicology, injury mechanism, and justice outcomes. New incident level variables will include: season, day of week.

Community level variables includes LGA and new variables will include: Socioeconomic Index for Areas (SEIFA) Index of Relative Socioeconomic Disadvantage ranking and Australian Statistical Geography Standard remoteness structure for both deceased and offender residence and for the incident location.

2.9. Data de-identification

In accordance with coronial legislation, some of the information that this study seeks to examine includes records that state the identity of the deceased person. The records cannot therefore be accessed in a de-identified manner. The data extracted for this study will, however, be recorded in a de-identified manner. This will be done by removing any names, street addresses, day and month of birth, names of hospitals or other service providers, and free text fields used for annotation. Researchers will only access the VHR onsite at the CCOV and any data extracts will be stored without potentially identifiable information (as described above).

2.10. Data management

Data files will be accessed via password protected computers, and saved on secure drives at the Department of Forensic Medicine’s Southbank location according to Monash University and Victorian Institute of Forensic Medicine (which are co-located) IT security protocols.

Each data file will be password protected and only the researchers listed in this application will have access to the password(s). Investigators will work in adherence to Monash University’s Information Technology Policies and Procedures. Study data will be destroyed after a retention period of seven years.

2.11. Data analysis

All data analysis will be conducted using IBM SPSS Statistics for Windows, version 28.0 (IBM Corp., Armonk, N.Y., USA). The study is exploratory in nature so any findings regarding similarities or differences in the outcome measures are considered important. The importance of the results from chi-square and regression analyses will be determined by a p-value less than 0.05, and the magnitude and uncertainty in the estimated effect size.

2.12. Bias

The initial screening process for the VHR is broad and so facilitates inclusion of all potential homicides and reduces the likelihood of selection bias because deaths with no known offender or without convictions can still be included.

The larger size of the reference group (aged 18–64) means that there will be differences in margins of error between the older and younger adult homicide groups. Due to the higher prevalence of homicide in the younger adult group, this is unavoidable.

2.13. Dissemination

Results from the analysis will be submitted for peer-reviewed publication, presented at conferences, and form part of BK’s PhD thesis. The publication of aggregated data will help practitioners, policy makers and other key stakeholders (including Coroners) that are concerned with the health and injury prevention of older Victorians, to make decisions to prevent future intentional‐cause deaths in older Victorians. Study findings will also be disseminated via relevant media networks on publication to ensure a greater audience for the research beyond the scientific community.