A small constellation: risk factors informing police perceptions of domestic abuse

ABSTRACT

Police in the United States (US) and the United Kingdom (UK) now routinely use risk assessment tools to identify common risk factors for re-abuse and lethality when responding to domestic abuse. Nevertheless, little is known about the extent to which officers understand and perceive the importance of factors commonly included on risk assessment tools for predicting future abuse. This study attempts to shed some light into this area of research by exploring the responses of 720 British and American police officers to questions regarding how important and how essential various risk factors are for evaluating the level of risk or harm a victim of domestic abuse may face in the future. Findings indicated that British and American officers were largely in agreement about a small constellation of risk factors that they considered integral to the risk assessment process: using or threatening to use a weapon; strangulation; physical assault resulting in injury and escalation of abuse. The results revealed that officers’ country of employment, rather than their demographic characteristics or experience policing domestic abuse, was a particularly influential predictor of their perceptions, and that both the situational context and the victim's perception about risk are important in domestic abuse risk assessment.

KEY WORDS:

• Domestic abuse
• risk assessment
• US
• UK

Introduction

The police response to domestic abuse has been in the spotlight of scholars across multiple disciplines for more than 30 years. Empirical research has not only identified weaknesses in how police deal with this crime, it has also provided the impetus for dramatic changes in the policy structures and recommended practices of police officers worldwide. Perhaps, the most well-known example is the widespread adoption of pro- and mandatory arrest policies following field experiments testing the benefits of arrest in the 1980s (see Sherman 1992). Twenty years later, a body of research identifying the risk factors for re-victimisation prompted some police agencies to implement the use of risk checklists at the initial response to domestic abuse calls (Robinson 2006, National Policing Improvement Agency 2008, Campbell et al. 2009). Police in the United States (US) and the United Kingdom (UK) now routinely use such tools to identify common risk factors for re-abuse and lethality. A recent survey of 358 US law enforcement agencies found that 42% use ‘structured risk’ or lethality assessments to determine the level of danger to the victim (Police Executive Research Forum 2015). Another study indicated tools such as the Lethality Assessment Program (LAP) are in use in 32 US states (Messing et al. 2015). Since 2009, all UK police forces are expected to use the Domestic Abuse, Stalking, and Honour Based Violence (DASH) risk assessment tool in cases of domestic abuse, as recommended by the Association of Chief Police Officers (Robinson 2010). Thus, it is increasingly expected that police officers will be able to draw upon a professional understanding and awareness of the level of risk inherent in a wide variety of domestic abuse situations in order to respond effectively.

These changes to the way police respond to domestic abuse underscore the idea of ‘risk’ informing the way the police relate to members of society and illustrate how police officers now act as ‘risk communicators’ in their everyday work (Ericson and Haggerty 1997). Taking a ‘risk-led’ approach to domestic abuse is becoming routine in many countries, yet there are also inherent and possibly unsurmountable challenges to police work being informed by their perceptions of ‘risk’ (Hoyle 2007, Robinson 2010, Bland and Ariel 2015). Most pertinent to the current study are practitioner-based challenges such as uneven training and expertise with regard to the practice of risk assessment. This was recently highlighted in the inspection of UK police forces by Her Majesty's Inspectorate of the Constabulary (HMIC), which found a range of different and inconsistent risk-based practices across forces, combined with the troubling assessment that officers generally had ‘low levels of appreciation as to why [risk assessment] is important, and a poor understanding as to what officers should do beyond the completion of the DASH form’ (HMIC 2014, p. 70). Not only are there discrepancies in how risk is assessed across police agencies, but the meaning attached to risk factors can vary between police and other service providers coming into contact with domestic abuse. For example, ‘relationship separation’ is treated as a heightened risk factor by the police, whereas it is often seen as the goal of intervention by social workers in child protection cases (Stanley and Humphreys 2014). Understanding the meaning attached to certain risk factors by relevant practitioners and how they make judgements about risk can be considered a necessary precondition to successful multi-agency partnerships working to reduce domestic abuse, as well as help better understand police responses to domestic abuse (Trujillo and Ross 2008).

The current study explores perspectives of police officers in the UK and the US about ‘risk’ in domestic abuse cases, a topic that has received surprisingly scant attention by researchers, given the scale of changes described above. Furthermore, cross-national comparative research on this topic is absent, precluding an understanding of how societal and cultural factors might shape officers’ understanding of domestic abuse risk. For example, although the overall rates of domestic abuse in the US and UK are comparable, the rate of intimate partner homicide is much higher in the US (Maxwell and Robinson 2013), in large part due to the use of firearms, particularly handguns, which is a distinctive characteristic of American homicide (Robinson and Maxwell 2016). Thus, weapons as a signifier for producing potentially lethal harm are more likely to inform police perceptions of risk in the US. Conversely, given the UK government's recent definitional change of domestic abuse to include coercive control (see Stark 2007), we might expect British officers to see the relevance of a broader array of risk factors, going beyond physical assault (Robinson et al. forthcoming). To address this gap in the literature, the current study adopts a comparative approach, drawing upon a sample of police officers from the US and UK, to examine three key questions: (a) How do officers rate the importance of various risk factors, and which do they consider to be essential for evaluating risk?; (b) How are police perceptions about the importance of specific factors informed by their prior training and experience dealing with domestic abuse? and (c) What do officers’ comments reveal about their perceptions of risk factors in cases of domestic abuse?

Prior research

There is a vast literature on risk factors and risk assessment related to both the prediction and prevention of violence, dating from the 1970s (Douglas and Kropp 2002, Andrade et al. 2009). More recently, a subset of this literature has focused on domestic abuse specifically, with researchers identifying common risk factors correlated with re-victimisation (a catchall term that covers a spectrum of negative outcomes, ranging from a repeat occurrence of any type to more severe injurious violence and homicide; for a review, see Messing and Thaller 2015). Thus, a range of empirical research crossing interdisciplinary boundaries and using a variety of measures has informed the development of risk assessment tools for domestic abuse, many of which are now widely used in both frontline operational settings (e.g. LAP in the US, DASH in the UK, etc.) and clinical settings (e.g. SARA for correctional officers in the UK and Canada; see Kropp et al. 1995, Dutton and Kropp 2000). Although ample research underpins the composition of these tools (for a review, see Bennett Cattaneo and Goodman 2005) and has established to varying degrees their validity and reliability (Messing and Thaller 2013), far fewer studies examine practitioners’ perceptions of these tools and the relative importance they attach to the various risk factors contained within them.

Extant research on how practitioners evaluate risk in individual cases tends to be based on general violence risk assessment with practitioners working in clinical settings, such as psychologists (e.g. Heilbrun et al. 2000), rather than practitioners using these tools ‘in the field’, such as police officers. Available research in this vein has identified a filtering process whereby field practitioners rely upon a subset of factors to judge risk, with situation-specific information being particularly influential. For example, research in Australia showed that despite collecting extensive information using a structured risk assessment tool, in practice, police officers’ judgements of risk were largely based upon a subset of information: the victim's level of fear, the escalation of incidents, the number of prior incidents and the involvement of drugs and alcohol in the current situation to make a decision about risk (Trujillo and Ross 2008). The researchers also cautioned that risk assessment tools are unlikely to capture every factor influencing officers’ judgements of risk, and that research in this area is still limited. Robinson and Howarth's (2012) study examining the risk assessments of British advocates working with victims of domestic abuse found that risk factors including the escalation of violence, use of weapons, stalking and significant injuries were important for judging risk. A study of victim advocates in the US lends further support to the idea that the severity of physical violence and drug/alcohol abuse are especially significant situational factors (Bennett Cattaneo 2007). Finally, studies have demonstrated that victim-related factors such as fear of the perpetrator and/or fear of further injury are recognised as important by both police (Belfrage and Strand 2008, Trujillo and Ross 2008) and victim advocates (Robinson and Howarth 2012).

Other research has focused on implementation challenges associated with risk-led approaches to policing domestic abuse. Canadian research found that police officers recognised the value in conducting risk assessments in cases of domestic abuse but felt that a lack of understanding of the rationale and theory underpinning risk assessment hindered its effectiveness ‘on the ground’, leading to a recommendation for specialised training prior to implementation (Blaney 2010). Similarly, interviews with police officers in New Zealand revealed that most officers were positive about the idea of risk assessment, but in practice implementation depended upon ‘ … the variables at any given scene [that] shape officers’ decisions about where, when and how the risk assessment process will occur’ (Grant and Rowe 2011, p. 61). Overall, extant research reminds us that police officers create schemas or ‘working rules’ to provide a framework for interpreting events, people and situations which will inevitably influence their frontline responses (Stalans and Finn 1995, Robinson, 2000, Stroshine et al. 2008), and that police schemas are increasingly informed by notions of ‘risk’. The current study aims to gain a better understanding of officers’ mind-sets when responding to domestic abuse, by focusing on how they understand and engage in risk assessment, in order to identify ways to improve police decision-making at these calls.

Methodology

Sample and data

The sample and data for this study stem from a larger cross-sectional survey in which officers in one US agency and one UK agency were administered an online, anonymous survey to explore how they respond to and perceive domestic abuse. Both agencies are large, with more than 5000 employees. At the time of this study, however, the US agency was divided into more territorial divisions, had one centrally located domestic abuse unit and served an urban area comprising roughly 1.5 million residents representing a racially/ethnically diverse resident population (approximately 40% of the population was Black or of another minority race or ethnicity). The UK agency, on the other hand, had fewer territorial divisions, with a domestic abuse unit serving each unit. The UK agency also served a mix of rural, small towns and small cities whose resident population was 1.3 million and was far less racially/ethnically diverse in its resident population (15% of residents were Black or of another minority race or ethnicity) than its US counterpart. The two agencies under study both use formal risk assessment instruments; the UK officers use the DASH and the US officers use the LAP. In both jurisdictions, responding officers are expected to complete the assessment.

Because this study was aimed at exploring the perceptions of officers most likely to respond to domestic abuse incidents, the targeted sample for this study was all patrol officers/police constables and sergeants employed in the two agencies from February through April 2015 (n = 2437 and 1025 officers in the UK and US samples, respectively). Officers were sent an email via an internal, organisational distribution list requesting their participation in a voluntary, anonymous online survey regarding domestic abuse. Of the officers who received the emailed request, 810 clicked on the link and began the survey; however, some officers were excluded if their country of employment could not be determined (n = 33) or if they did not wish to participate in the survey or answer any questions (n = 4). Thus, 11% (n = 265) of the UK sample and 50% (n = 508) of the US sample remained, resulting in a total of 773 officers. The final sample size was 720 officers, which was determined after removing all cases for which no valid response was provided for each of the outcomes under study (n = 53). Put another way, officers were only removed from the sample if they failed to answer every risk assessment question; if the officer responded to at least one question regarding risk, he/she remained in the sample. Sample descriptive statistics for these officers are provided in Table 1.

Table 1. Sample descriptive statistics.

	Total		United States		United Kingdom
	Mean	SD	Mean	SD	Mean	SD
United States	.68	.47	–	–	–	–
Male	.85	.35	.93	.25	.69	.46
Age	38.18	7.73	36.69	7.39	41.30	7.50
Minority	.18	.39	.26	.44	.03	.17
College degree	.38	.49	.38	.49	.38	.49
Total months at current agency	126.81	74.94	115.06	60.98	152.19	93.82
Domestic abuse training	.96	.19	.98	.15	.94	.24
Domestic abuse specialist	.09	.29	.05	.23	.17	.38
Large domestic abuse caseload	.30	.46	.34	.47	.23	.42

Note: Missing data account for 8% or less for each of these variables.

Measures

The primary measures of interest for this study were 20 distinct risk factors for domestic abuse. Because the participating officers were from the US and UK, we amalgamated the items from the two most widely used risk assessment tools in these countries (the LAP and the DASH) to form a list of risk factors. This was then piloted with a small sample of officers in both countries, who did not suggest any revisions to the list. Appendix A includes the original risk factors from the LAP and DASH alongside our list of 20 items.

The decision to use an amalgamated list was taken to ensure the survey was applicable to both British and American officers, and that they were responding to the same questions. If we had administered only LAP-based questions to US officers and only DASH-based questions to UK officers, this would have precluded making robust comparisons across samples, one of the main aims of our study. Furthermore, US officers would not have been provided the opportunity to give their views about many risk factors which are not included on the LAP (which contains 11 items). Because the DASH is longer (27 items), we expect our list was more likely to contain risk factors familiar to UK officers, even if differently worded or presented in combined form. The LAP contains a higher proportion of items relating to weapons (2 of 11) than does the DASH (1 of 27); ‘access to firearms’ is the only item on our list that is derived from the LAP but not the DASH. Overall, the DASH is clearly longer and has the potential to obtain more information about a wider spectrum of issues that are relevant for identifying domestic abuse risk.

Officers were instructed to consider domestic abuse in general and then rate each risk factor in terms of the officer's opinion of its importance for evaluating the level of risk or harm the victim may face in the future. Responses were based on a 7-point Likert scale ranging from ‘not at all important’ to ‘extremely important’ (not at all important; very unimportant; somewhat unimportant; neither important nor unimportant; somewhat important; very important; extremely important). For this study, the response categories were collapsed such that ‘very important’ or ‘extremely important’ were coded as ‘1’ and all others were coded as ‘0.’ Officers were then asked to choose only 5 of the 20 listed risk factors they viewed as essential for evaluating risk (1 = essential ; 0 = not essential), in order to identify how officers might focus on certain factors more than others, even when many are deemed important. Although it is possible that limiting the number of essential factors to five might have prevented some officers from listing all of the factors they perceive to be essential, we felt that forcing officers to select a subset of essential factors more accurately reflects the decision-making of field practitioners, in line with extant research discussed previously.

Five main independent variables were included to account for officers’ prior training and experiences responding to domestic abuse. Because officers in different countries use different risk assessment tools, are guided by different laws and may have different experiences responding to domestic abuse, we included country of employment (1 = USA; 0 = UK). Tenure at current agency refers to the number of months the officer has been employed by the current agency; it does not necessarily reflect total law enforcement experience, as some officers may have been employed at other law enforcement agencies prior to their current employment. Domestic abuse training refers to whether the officer had ever received dedicated training about domestic abuse in their entire law enforcement career (1 = yes; 0 = no). Domestic abuse specialist is a dichotomous variable that reflects whether the officer ever held a specialist role related to domestic abuse (e.g. sergeant in a domestic violence unit) (1 = yes; 0 = no). Large domestic abuse caseload refers to whether over half of the officer's caseload is domestic abuse-related (1 = yes; 0 = no).

Finally, officer demographic characteristics were included as control measures. Male (1 = yes; 0 = no), age (continuous measure), minority (1 = Black or another minority race or ethnicity; 0 = White) and college degree (1 = yes; 0 = no) were included in each model.

Analytic strategies

The survey lends itself to analysis of both quantitative and qualitative data. Given the relative infancy of this line of research, much of the quantitative analyses for this study are descriptive in nature. First, we calculated the percentage of officers who reported that each respective risk factor was either very important or extremely important for evaluating future risk. Then, we calculated the percentage of officers who viewed each risk factor as essential for determining future risk. Both computations were completed for the entire pooled sample and then separately by country; chi-square tests were computed to assess whether significant differences existed across the two countries. Finally, logistic regression analyses were conducted for each of the 20 risk factors in order to examine the factors which help predict perceptions of future risk (very important/extremely important).

An analysis of the qualitative data from an open-ended question was also conducted to complement the quantitative data analysis. Directly after officers evaluated the importance of the 20 risk factors, they were asked whether there were ‘any comments you wish to make about these risk factors (e.g. terminology, wording, omissions, etc.)?’ It is important to note that officers were not prompted to comment on specific types of risk factors (e.g. victim-centred; violence-centred). The open-ended question was left intentionally broad so that officers could comment generally on risk factors and assessment. The third author conducted open, axial coding to analyse the data (Miles and Huberman 1994). First, codes were developed based on participants’ words (Charmaz 2006). Then, after comparing the codes through multiple passes, the codes were collapsed into major categories (Charmaz 2006).

Results

Quantitative data

Table 2 presents the results of the risk factors identified by officers as being either very important or extremely important for evaluating the level of risk a victim may face in the future. Overall, regardless of country, it was notable that officers considered the majority of risk factors to be either very or extremely important for determining future risk; 16 of the risk factors were designated as such from at least half of the pooled sample. The five risk factors that the largest percentage of officers viewed as very or extremely important were: using a weapon or threatening to use a weapon (∼96%), strangulation (∼96%), sexual assault (∼93%), physical assault resulting in injury (∼92%) and escalation of abuse (∼91%). Crucially, our findings indicate that British and American officers were in agreement about a small constellation of risk factors which they considered to be the most important to consider when evaluating risk. Similarly, there was consensus around the least important risk factors, with officers in both countries least likely to perceive that having a blended family (∼11%) or being unemployed (∼20%) was very or extremely important for determining risk.

Table 2. Percentage of officers who perceive each risk factor to be very/extremely important.

Risk factors	Total		United States		United Kingdom
	%	Rank	%	Rank	%	Rank
Escalation of abuse	91.1	5	88.4**	5	96.9	4 (tie)
Relationship breakdown/separation/divorce	47.84	17	47.0	16	49.6	18
Using a weapon or threatening to use a weapon	95.8	1 (tie)	94.1**	2	99.6	1
Making threats to kill	87.9	7	86.1*	7	91.7	8
Strangulation	95.8	1 (tie)	94.5**	1	98.7	2
Jealously/controlling behaviours	75.4	12	69.6**	11	87.7	10
Unemployment	19.5	19	21.6*	19	14.9	19
Threats of suicide	67.4	14	67.5	13	67.1	15
Blended family (step-children)	11.1	20	10.2	20	13.2	20
Spying/stalking/harassing/threatening messages	77.2	10	71.8**	10	89.0	9
Access to a firearm	77.6	9	69.0**	12	96.1	7
Victim's level of fear	68.3	13	59.9**	14	86.4	11
Physical assault resulting in injury	91.9	4	89.6**	4	96.9	4 (tie)
Conflict over child contact	47.78	18	43.3**	17	57.5	17
Victim pregnant or recently had a baby	50.7	16	40.3**	18	73.1	14
Sexual assault	93.0	3	90.8**	3	97.8	3
Mistreatment of an animal or family pet	59.5	15	58.5	15	61.8	16
Drug/alcohol abuse	76.0	11	76.2	8	75.4	13
Bail/injunction/protection order	77.9	8	74.4**	9	85.5	12
Criminal history for domestic or sexual violence	89.6	6	86.4**	6	96.5	6

Note: **p ≤ .01; *p ≤ .05 indicate between-country differences in the percentage of officers who reported each risk factor to be very/extremely important.

Although the rank order of risk factors was similar for both the American and British officers, for 15 of the 20 risk factors, significant differences emerged in the raw percentages of officers who reported them to be very/extremely important. Specifically, a greater percentage of British than American officers perceived 14 of the 20 risk factors to be very or extremely important for determining risk. This pattern reveals a tendency of British officers towards a heightened evaluation of the importance of a greater number of risk factors relative to their American counterparts. The only instance where a greater percentage of American officers perceived the risk factor to be very/extremely important was unemployment. The five risk factors that received a statistically equivalent amount of support were relationship breakdown/separation/divorce (∼47% in the US; ∼50% in the UK); threats of suicide (roughly 67% in both countries); blended family (∼10% in the US; ∼13% in the UK); mistreatment of an animal or family pet (∼59% in the US; ∼62% in the UK) and drug/alcohol abuse (∼76% in the US; ∼75% in the UK).

After assessing the importance of each risk factor, officers were asked to choose which 5 of the 20 risk factors they considered to be essential for evaluating risk (Table 3). Overall, the following five risk factors received the most support from the pooled sample: using a weapon or threatening to use a weapon (∼67%), strangulation (∼64%), escalation of abuse (∼53%), physical assault resulting in injury (∼51%) and making threats to kill (∼43%). Consistent with our previous findings, British and American officers were largely in agreement about a subset of risk factors which they considered to be essential when evaluating risk. Similar to earlier analyses, very few officers viewed unemployment (∼1%) and having a blended family (<1%) as essential for evaluating risk.

Table 3. Percentage of officers who perceive each risk factor to be one of the top five Essential risk factors.

Risk factors	Total		United States		United Kingdom
	%	Rank	%	Rank	%	Rank
Escalation of abuse	52.9	3	46.8**	5	66.2	1
Relationship breakdown/separation/divorce	8.3	14	8.3	13	8.3	15
Using a weapon or threatening to use a weapon	66.7	1	71.5**	1	56.1	3
Making threats to kill	43.1	5	48.6**	4	31.1	7
Strangulation	64.4	2	65.2	2	62.7	2
Jealousy/controlling behaviours	29.3	8	28.7	7	30.7	8
Unemployment	1.1	19	1.6	18	0.0	19
Threats of suicide	18.2	12	22.6**	12	8.8	14
Blended family (step-children)	0.1	20	0.2	20	0.0	19
Spying/stalking/harassing/threatening messages	25.7	10	24.0	10	29.4	9
Access to a firearm	27.2	9	27.2	9	27.2	10 (tie)
Victim's level of fear	13.1	13	6.5**	14	27.2	10 (tie)
Physical assault resulting in injury	51.0	4	54.3**	3	43.9	4
Conflict over child contact	3.5	17	3.1	16	4.4	17
Victim pregnant or recently had a baby	4.3	15 (tie)	1.0**	19	11.4	12
Sexual assault	31.7	6	30.7	6	33.8	6
Mistreatment of an animal or family pet	4.3	15 (tie)	4.5	15	4.0	18
Drug/alcohol abuse	18.8	11	23.2**	11	9.2	13
Bail/injunction/protection order	3.2	18	2.0**	17	5.7	16
Criminal history for domestic or sexual violence	31.4	7	27.9**	8	39.0	5

Note: **p ≤ .01; *p ≤ .05 indicate between-country differences in the percentage of officers who reported each risk factor to be essential.

Significant between-country differences in officer perceptions of whether a particular risk factor was essential for evaluating risk existed for 10 of the 20 risk factors. Unlike earlier analyses which indicated that British officers were more likely to judge a greater number of risk factors as very/extremely important, when officers were asked to choose which risk factors were essential, it became more difficult to discern a pattern when comparing the responses of American to British officers. British officers were more likely to view the following five of the risk factors as essential: escalation of abuse, victim's level of fear, victim is pregnant/recently had a baby, bail/injunction/protection order and criminal history for domestic or sexual violence. American officers were more likely to view using a weapon or threatening to use a weapon, making threats to kill, threats of suicide, physical assault resulting in injury and drug/alcohol abuse as essential for evaluating risk.

Table 4 presents the results of the 20 logistic regression analyses predicting the likelihood of officers judging a risk factor to be very or extremely important for evaluating the level of risk or harm the victim may face in the future. The most notable finding across the models was the importance of officers’ country of employment. Similar to the bivariate analyses, in 14 out of 20 models, the odds of reporting a risk factor as very/extremely important was significantly lower among US officers than UK officers. Being an American officer reduced the odds of considering any risk factor to be very/extremely important by an average of 50%. [This was calculated using the formula [percent = (1-(SUM(Exp B)/1))* – 1], see Long (1997, p. 228).] In short, the importance of country of employment, relative to the other variables included in the model, was pronounced.

Table 4. Predicting the odds of perceiving a risk factor to be very/extremely important.

	USA	Male	Age	Minority	College degree	Tenure at current agency	Domestic abuse training	Domestic abuse specialist	Large domestic abuse caseload	Nagelkerke R^2
Risk factor
Escalation of abuse	0.21**	0.59	1.01	0.82	0.61	.996	1.38	0.28**	0.93	.09
Relationship breakdown/separation/divorce	1.02	0.65	1.04**	1.07	1.01	1.00	1.58	1.01	1.51*	.04
Using a weapon or threatening to use a weapon	0.05**	1.43	1.00	1.11	0.67	1.00	3.72	0.36	1.03	.10
Making threats to kill	0.43*	1.22	1.02	1.71	0.50**	0.998	0.34	1.36	0.67	.07
Strangulation	0.18*	0.83	1.03	1.69	1.14	0.99	1.61	0.73	0.99	.06
Jealousy/controlling behaviours	0.28**	0.91	1.04*	1.49	0.85	0.998	2.67*	0.77	1.17	.10
Unemployment	1.74*	0.57	1.05**	1.31	1.36	0.999	1.06	0.81	1.15	.04
Threats of suicide	0.87	1.38	0.98	0.99	0.80	1.00	1.33	0.98	0.81	.02
Blended family (step-children)	0.86	0.48*	1.02	0.96	0.89	0.999	3.19	0.95	1.31	.03
Spying/stalking/harassing/threatening messages	0.26**	1.03	1.01	0.89	0.98	0.998	2.85*	0.75	1.16	.08
Access to a firearm	0.06**	0.89	1.04*	1.63	0.81	0.99**	1.37	0.44	0.93	.21
Victim's level of fear	0.19**	0.81	1.03*	1.07	0.67*	0.996*	0.90	0.82	1.10	.15
Physical assault resulting in injury	0.19**	1.43	1.04	1.44	0.63	0.99*	0.61	1.02	0.69	.09
Conflict over child contact	0.59**	0.81	1.03*	1.45	0.81	1.00	0.43	0.86	0.81	.06
Victim pregnant or recently had a baby	0.24**	0.73	1.01	1.07	0.85	0.997	0.59	0.91	0.95	.14
Sexual assault	0.13**	2.11	0.96	0.68	0.80	1.00	0.85	0.87	1.40	.09
Mistreatment of an animal or family pet	1.18	0.50**	1.01	0.69	0.82	1.00	0.82	1.31	0.96	.03
Drug/alcohol abuse	1.03	0.87	1.01	0.83	0.76	0.998	1.74	0.65	0.93	.02
Bail/injunction/protection order	0.39**	1.09	0.99	1.17	0.70	1.00	1.53	1.30	0.75	.06
Criminal history for domestic or sexual violence	0.13**	1.46	1.01	1.04	0.82	0.997	1.91	0.56	0.82	.09

Notes: **p ≤ .01; *p ≤ .05.

Exp (B) reported. Models can be read horizontally.

There was less support for the effect of other experience measures (tenure, domestic abuse training, domestic abuse specialist, heavy domestic abuse caseload) for impacting the odds of evaluating a given risk factor as important. For 3 of the 20 risk factors, tenure at current agency was significantly related to the reduced odds of finding the risk factor to be very/extremely important. However, substantively, the effect of tenure was negligible: the odds of reporting that access to a firearm, a victim's level of fear and physical assault resulting in an injury were very/extremely important for evaluating future risk, were only between .4% and 1% lower for officers who were employed at their current agency for longer compared to those employed at the agency for a shorter period of time.

The odds of evaluating evidence of jealousy/controlling behaviours and spying/stalking/harassing/threatening messages as very/extremely important were significantly greater for officers who had received domestic abuse training. Specifically, the odds of considering these risk factors to be very/extremely important for evaluating risk were more than two times higher for officers who had received domestic abuse training relative to those who had not. The odds of reporting that the escalation of abuse is a very/extremely important risk factor for evaluating future risk was lower for officers who had ever held a specialist position at their current agency; however, this predictor was not associated with any of the other risk factors. Finally, officers who held a large domestic abuse caseload were one and a half times more likely than their counterparts with lower domestic abuse caseloads to view that a relationship breakdown was very/extremely important for evaluating future risk; however, caseload had no effect on any other outcome.

Qualitative data

In response to whether the officers had additional comments to make about the risk factors under study, 32 British officers and 24 American officers commented. Analysis of their comments revealed two major themes related to risk assessment: (1) the context of the situation shaped the meaning of the risk factor and (2) an awareness of victims’ perceptions was important when assessing risk.

The situational context matters

As instructed in the survey, officers first rated the importance of each risk factor individually; however, officers often took the opportunity to express the need for taking into consideration all of the risk factors present in an incident to determine future risk. Numerous British (10/32) and American (6/24) officers identified the importance of an incident's overall context and noted that every risk factor is – or can be – important depending on the situation, whereas others may be less important when considering context because of the relationship between the risk factors.

For example, one UK officer noted that when asked about the importance levels of risk factors while on the close-ended portion of the survey, the officer marked only a few risk factors as somewhat important

because on their own they wouldn't necessarily cause concern, (i.e., just because someone is unemployed or a step-parent wouldn't mean a risk of harm); however, if accompanied by other identified factors then it obviously would add concern to the risk assessment.

Another UK officer responded that, ‘Depending on individual circumstances of each case, all areas could easily fall into the extremely important category.’

Similarly, US officers’ comments indicate support for weighing the risk factors in relation to each other, in order to get a sense of the ‘whole picture’ when determining risk. For example, one US officer stated, ‘Some factors would need to be evaluated along with the circumstances for each incident, and certain factors alone may not be of any concern; however, that same factor may be of great concern in a different situation.’ Similarly, another officer noted that, ‘Some of the factors listed by themselves mean nothing to me (i.e. “level of fear,” or “access to a firearm”), but combined with other circumstances they could greatly affect that factor's relevance in the threat level.’ Comments such as these imply that individual risk factors can only be judged in combination and in context; however, another US officer indicated a slightly different point of view when he said that, ‘I believe that [considering the] totality of circumstances [is] very important; however, I believe that some [risk factors] are red flags in and of themselves (i.e. sexual assault, weapon offenses).’

This ‘totality of the circumstances’ approach is certainly not unique to risk assessment, as it also guides a range of other police actions (e.g. stop, arrest) and underscores that each situation is unique and must be considered on a case-by-case basis. Overall, these comments signify that ‘context matters’ to officers as they evaluate the importance of certain risk factors. Although the quantitative analyses revealed some level of ‘ranking’ of the risk factors in terms of their importance, analysis of officers’ own words suggest a more subtle interpretation of risk that seems to depend on the unique circumstances of each case.

The victim's perception matters

British officers (4/32) in the sample also mentioned the role of the victim in determining future risk. Specifically, some officers noted that victim's perceptions of the apparent risk factors and assessment of future risk at the hands of the offender are important to consider when responding to domestic abuse cases. For example, one UK officer noted that, ‘All of the above [risk factors] are important in evaluating the risk if the victim expresses concerns over them.’ Another UK officer stated that

All of the above can affect people in different ways and hold varying levels of importance to individuals. Therefore as a police officer attending a domestic related incident all of these risk factors may be relevant to assess/consider in the situation.

Another UK officer provided additional insight into why a victim's perceptions of his/her future risk is important to consider. According to this officer,

They are all very important and their importance will be different depending on the impact they have on the victim (e.g. threats to kill from someone who is incapable of carrying it out will be different to someone who can carry the threat out). [A] victim with family support may not feel as threatened by certain things as one who is isolated or not independent or [not a] British citizen.

In addition, officers in both countries (UK: 3/32; US: 5/24) noted that the victim's decision to prosecute or move forward with criminal justice system intervention was important to consider as it was seen to impact how officers can respond in a domestic abuse incident as well as the likelihood of future abuse. For example, one US officer stated,

Every situation is different. I try to rely heavily on what I can prove or what gives me the ability to act (i.e., arrest); other factors add to it but don't give me probable cause to arrest (i.e., victim's fear, or suspect's threat of suicide). Sure they can be indicators of bad things to happen, but unless the victim takes action to improve their circumstance, I can only do what the law allows/mandates me to do.

In a similar vein, officers in the UK also alluded to their risk assessments to some extent incorporating the victim's perspective. For example, one UK officer stated, ‘Victim's willingness to assist with prosecution – Massive risk factor for me as officers often lose ability to safeguard the victim and their family without a certain level of cooperation.’ The focus on the victim in this regard suggests that officers believe victim participation is essential to successful criminal justice intervention, and that without it, the odds are reduced that the criminal justice system can sanction the offender, and subsequently help reduce re-abuse.

Discussion

This study contributes to a small but growing area of research investigating police officers’ understanding of risk and risk assessment in domestic abuse cases. Our results clearly indicated that a majority of participating officers recognised the utility of a broad array of factors to the risk assessment process (e.g. over 75% of officers agreed that 12 of the 20 risk factors were very/extremely important). Despite this, officers also appeared to focus on a small constellation of risk factors that they considered to be both very important as well as essential when evaluating risk: using or threatening to use a weapon; strangulation; physical assault resulting in injury and escalation of abuse. These factors seem to signify to police officers – both British and American – what it means for a victim to be at high risk. This level of consensus should not be overlooked, particularly given the different legal and policy contexts surrounding the officers working in our research sites, which reflect broader differences in crime and policing evident in the US versus the UK. The results are also consistent with prior research which suggests that practitioners often rely on a small subset of risk factors, despite the presence of other available information which may be relevant for evaluating risk (Belfrage and Strand 2008, Trujillo and Ross 2008). Officers in our study appeared to have a common language of risk they draw upon when evaluating domestic abuse situations, which reveals how police perceptions have changed over time and are now influenced by ideas of risk (Sinden and Stephens 1999).

Our research also revealed how it is those factors explicitly related to physical violence that are in the forefront of officers’ minds. Many of these factors are consistent with prior research highlighting the salience of risk factors such as the escalation of violence and physical injuries for evaluating risk (Bennett Cattaneo 2007, Robinson and Howarth 2012). In another study stemming from the same sample of officers, respondents were randomly assignment to a hypothetical vignette, in which one version depicted a physically violent domestic abuse situation, whereas the other was identical but had the depiction of violence removed (Robinson, et al. forthcoming); officers in the sample were significantly more inclined to take formal action when responding to the violent vignette. Therefore, both studies suggest that violence-related risk factors are particularly influential in police interpretations of risk, providing an insight into the police mind-set or schema when dealing with domestic abuse. The research reflects long-standing community, cultural and criminal justice conceptualisations of domestic abuse which have been criticized for predominantly focussing on physical violence when a broad range of harmful behaviours, both violent and non-violent, are often involved and which require identification and intervention by practitioners (Dutton and Goodman 2005, Stark 2007, Belknap and Sharma 2014).

Despite the aforementioned consistencies across UK and US officers, a noteworthy pattern of between-country differences emerged in this study. Specifically, a significantly greater percentage of British than American officers perceived 14 of the 20 risk factors to be very or extremely important for determining risk, perhaps indicative of a heightened sensitivity towards risk relative to their American counterparts. This finding might be interpreted as a reflection of Kemshall's (1998) ‘precautionary principle,’ where practitioners choose to err on the side of caution rather than underestimate risk. In the UK, there has been widespread implementation of the DASH tool for several years, and periodic reminders via domestic homicide reviews that a failure to properly assess risk can have devastating consequences, a fact also highlighted in the recent national inspection of the police response to domestic abuse conducted by HMIC (2014). In this study, the UK agency had a more established track-record employing a ‘risk-led’ approach to domestic abuse, going back more than a decade, which is one possible explanation for their relatively higher importance ratings across most risk factors. Conversely, the US agency could be considered to be at the beginning stages of implementing a risk-led approach, with a risk checklist (the LAP) implemented approximately 5 years ago. Officers did not receive any specialised training on administering the LAP, nor does the agency have a police response that can be considered to be informed by risk (e.g. with certain police actions and multi-agency arrangements prioritised for high-risk cases, as has been policy for some time across UK police forces, including the UK agency in this study; see Robinson 2006, 2010). The possibility that these factors impact the role and meaning of risk assessment in cases of domestic abuse and might lead to officers choosing to err on the side of caution (rather than underestimate risk) would seem to be an avenue worthy of further exploration. Our research only contained two research sites, but a larger multi-site study might be able to reveal certain organisational characteristics that significantly affect officers’ risk assessments.

In addition, this study did reveal some discrepancies with previous literature regarding the risk assessment process for practitioners: perhaps most significantly was the effect of victim's fear on perceptions of risk. Although many officers were indeed aware of the role of victim's fear in predicting future abuse, the quantitative results suggest that officers did not rank the victim's level of fear as highly as some of the other risk factors. Recall, approximately 68% of the total sample (roughly 60% and 86% in the US and UK, respectively) viewed victim's level of fear as very or extremely important, and far fewer (approximately 13% overall; roughly 7% and 27% of US and UK officers, respectively) rated it as one of the top five most essential factors for predicting future risk relative to other factors. Although the quantitative results are inconsistent with previous studies that have found victim's fear to be influential in the risk judgements of both police and victim advocates (Belfrage and Strand 2008, Trujillo and Ross 2008, Robinson and Howarth 2012), the qualitative data offer some tentative support. Given that extant research strongly supports the utility and accuracy of victims’ risk assessments, this area seems worthy of additional attention within police training (see, e.g. Weisz et al. 2000, Bennett Cattaneo and Goodman 2003, Heckert and Gondolf 2004).

Finally, analyses indicated that prior training and experiences responding to domestic abuse were less salient than expected. However, two results related to the impact of domestic abuse training are worth discussion. Specifically, officers who had ever received training on domestic abuse were significantly more likely to view jealousy/controlling behaviours and spying/stalking/harassing/threatening messages as very or extremely important for evaluating victims’ future risk. These findings are important given that these two risk factors are reflective of coercive control (Stark 2007), yet have been shown to fly ‘under the radar’ of police due to the lack of violence directly associated with them (Robinson et al. forthcoming). Furthermore, the criminal justice system has been criticized for failing to acknowledge non-violent, yet controlling behaviours under the realm of domestic abuse (Tuerkheimer 2004, Burke 2007, Stark 2007). These findings suggest the importance of domestic abuse training to ensure that front-line officers are aware of the risk associated with ostensibly non-violent behaviours, such as coercive control and stalking. The College of Policing in the UK is currently revising its training materials in an attempt to address this issue (Myhill 2015); a focus on both a higher quantity and better quality of training could have important implications for how officers understand risk-led policing.

Limitations and suggestions for future research

Although this study makes a valuable contribution to the extant literature on risk assessment and domestic abuse by providing the first cross-national comparative research on this topic, there are methodological limitations worth noting. First, the survey we employed raises a few issues that future research may consider addressing. For example, because the list of 20 risk factors was created using a combination of items from the LAP and the DASH, officers were responding to items that are differently worded to those listed on their agency's risk assessment tool, not included on their agency's tool, or that combined multiple risk factors from their agency's tool into a single item. Although this facilitated our statistical analysis (i.e. by enabling us to compare like-with-like), future research should endeavour to provide officers with the opportunity to respond using their own agencies’ risk tool rather than using hypothetical items. This would complement our findings and progress the field further. Similarly, in this study, officers were asked about their perceptions more generally, rather than referring to a particular case. This area of research may benefit from asking officers to engage in a risk assessment exercise using a few specific case studies. This would help address the possibility that officers’ endorsement levels become inflated when they are asked to respond to general questions rather than specific examples (i.e. perhaps they respond that a risk factor is important because it is on their agency's tool and they have received training to reinforce this point, but it is not clear whether they would actually incorporate this risk information into their decision-making around a specific case). Furthermore, other studies might contribute to the limited evidence on police officers’ actual (rather than hypothetical) risk assessment practices and how the identification of certain risk factors informs their perceptions about which victims are at high or low risk and what actions they take in response (Belfrage et al. 2012). Future research might consider exploring how officers assess risk in situations which do not seem to have any of the risk factors from the ‘small constellation’; for example, it would be interesting to explore whether officers are likely to underestimate risk in cases that do not have evidence of physical violence, weapons and/or escalation. Likewise, examining the extent to which officers recognise the relationship between coercive controlling behaviours and future risk of domestic abuse should guide future research endeavours, especially given the introduction of this concept into national policy in the UK.

Second, this study draws upon information provided by officers from two police agencies and thus is limited in its generalisability. Given the pervasiveness of officer-led risk assessments used internationally, future research should consider expanding this line of research to additional agencies using a variety of risk assessment tools. Doing so would not only increase generalisability and provide additional information about risk-led policing approaches, but participation from more agencies would also help to boost sample sizes, which may be limited due to the well-known difficulty of achieving high response rates using online surveys (see e.g. Puleston 2011).

Third, although this study employed both quantitative and qualitative components, there may be some limitations with the analytic strategies utilised. As suggested by the model fit statistics, the logistic regression models had little utility in predicting the outcomes. It appears that many other relevant factors – either at the individual/offense level, officer level and perhaps even community level – are missing from our models. Additional factors could have implications both for understanding the occurrence of domestic abuse as well as officers’ perceptions of future risk (Bennett Cattaneo and Goodman 2005, Pinchevsky and Wright 2012).

Finally, the qualitative themes were developed off of an open-ended question asking officers if they had any other comments to make about the 20 risk factors; thus, officers were not prompted to respond about a specific type of risk factor. As a result, it is not possible to discern how the results provided here would have differed were officers asked more specific prompts about particular types of risk factors (e.g. violence-related, victim-centred). Given the paucity of qualitative research about practitioners’ perceptions of risk – including domestic abuse risk – future research would benefit from incorporating a qualitative methodological approach.

Despite these limitations, this study adds to the growing area of research focusing on criminal justice practitioners’ perceptions of risk assessment tools. Only with continued research will scholars be better able to understand how officers respond to domestic abuse incidents and view mandated risk assessments; such research has implications for policy, practice and training of frontline officers.

Disclosure statement

No potential conflict of interest was reported by the authors.

Appendix A. Comparison of risk factors.

Table

Current study (20 items)	LAP (11 items)	DASH (27 items)
1) Escalation of abuse		1) ‘Is the abuse happening more often?’ 2) ‘Is the abuse getting worse?’
2) Relationship breakdown/separation/ divorce	1) ‘Have you left him/her or separated after living together or being married?’	3) ‘Have you separated or tried to separate within the past year?’
3) Using a weapon or threatening to use a weapon	2) ‘Ever used a weapon against you or threatened you with a weapon?’	4) ‘Ever used weapons or objects to hurt you?’
4) Making threats to kill	3) ‘Ever threatened to kill you or your children?’ 4) ‘Do you think he/she might try to kill you?’	5) ‘Ever threatened to hurt or kill the children/dependants?’
		6) ‘Ever threatened to kill you or someone else and you believed them?’
5) Strangulation	5) ‘Ever tried to strangle you?’	7) ‘Ever attempted to strangle/choke/suffocate/drown you?’
6) Jealously/controlling behaviours	6) ‘Violently or constantly jealous or do they control most or all of your daily activities?’	8) ‘Try to control everything you do and/or are they excessively jealous?’
7) Unemployment	7) ‘Is he/she unemployed?’	9) ‘Are there any financial issues/are you dependent for money/have they recently lost their job/other financial issues?’
8) Threats of suicide	8) ‘Ever tried to kill himself/herself?’	10) ‘Are you feeling depressed or having suicidal thoughts?’ 11) ‘Has he/she ever threatened or attempted suicide?’
9) Blended family (step-children)	9) ‘Do you have a blended family (step children?’	12) ‘Are there any step-children in the household, or are there other dependants in the household (i.e. older relative)?’
10) Spying/stalking/harassing/threatening messages	10) ‘Follow or spy on you or leave threatening messages?’	13) ‘Constantly text, call, contact, follow, stalk or harass you?’
11) Access to a firearm	11) ‘Have a gun or can he/she get one easily?’
12) Victim's level of fear		14) ‘Are you very frightened?’ 15) ‘Is there any other person that has threatened you or that you are afraid of?’ 16) ‘What are you afraid of (specifically)?’
13) Physical assault resulting in injury		17) ‘Has the current incident resulted in injury?’
14) Conflict over child contact		18) ‘Is there conflict over child contact?’
15) Victim pregnant or recently had a baby		19) ‘Are you currently pregnant or have you recently had a baby in the past 18 months?’
16) Sexual assault		20) ‘Do or say things of a sexual nature that makes you feel bad or that physically hurt you or someone else?’
17) Mistreatment of an animal or family pet		21) ‘Ever mistreated an animal or the family pet?’
18) Drug/ alcohol abuse		22) ‘Problems in the past year with drugs (prescription or other), alcohol or mental health leading to problems in leading a normal life?’
19) Bail/injunction/protection order		23) ‘Ever breached bail/an injunction and/or any agreement for when they can see you and/or the children?’
20) Criminal history for domestic or sexual violence		24) ‘Has he/she ever been in trouble with the police or has a criminal history?’
		25) ‘Do you feel isolated from family/friends?’
		26) ‘Has he/she ever hurt the children/dependants?’
		27) ‘Do you know if he/she has hurt anyone else?’