The Limits of Our Knowledge: Tracking the Size and Scope of Police Involvement with Persons with Mental Illness

Significant public discourse has focused recently on police–civilian interactions involving with persons with mental illness (PMI). Despite increasing public attention, and growing demands for policy change, little is actually known about the myriad of ways in which Canadian police encounter PMI in the context of routine police work. To assist policymakers in developing evidence-informed policy, this paper attempts to shed light on present difficulties associated with addressing fundamental questions, such as the prevalence of mental health related issues in police calls for service. To do this, we attempt to map the size and scope of police calls for service involving PMI, drawing on both the available scientific data and the limited knowledge to be gleaned from available police reports. Our focus is on two broad categories of police interactions with citizens: public safety concerns (wellness checks, suicide threats, missing persons, mental health apprehensions) and crime prevention and response (encountering PMI as victims–complainants and (or) as potential suspects). We also explore the challenges policy-makers face in relying on police data and the importance of overcoming weaknesses in data collection and sharing in relation to the policing of uniquely vulnerable groups. This paper concludes with some key recommendations for addressing gaps highlighted.


Executive Summary
Following protests of police involvement in several high-profile deaths of individuals experiencing a mental health crisis, a number of Canadian individuals and groups have advocated in response for various policy positions-ranging from enhanced police training and new policing programs to reallocating police funding to healthcare and social work groups and, in some instances, outright abolishment of public police. Consistently lacking from much of the public discourse is a thorough understanding of the myriad of ways in which policing intersects with mental health issues, the prevalence of mental health issues within police calls for service, the nature of interactions between persons with mental illness (PMI) and police, and the unique challenges in this area facing Indigenous communities in Canada.
The purpose of the present working paper is to begin to flesh out those areas in which we have some knowledge of this topic, as well as identify key areas in which little is known. The structure of this working paper is as follows. First, we speak to the challenges researchers face in trying to use existing data sources to develop estimates of the size and scope of mental health-related demands upon police. Then, we begin to explore the various intersections of policing and mental health. Where possible, we use police figures to highlight the volume of service calls for different types of interactions. Third, we look at previous attempts by researchers at estimating the overall prevalence of mental health-related calls in policing, noting both the strengths and limitations of previous work. We also consider what recent police statistics may add to the discussion, before presenting some recent data on police calls for service during the COVID-19 pandemic. Next, our focus then shifts to what the research can tell us about police interactions with PMI, before examining the important topic of Indigenous communities and the challenges faced by Indigenous police services, particularly those in rural and remote areas. Finally, based on an analysis of the data and research gathered, we present a series of recommendations for policy-makers and practitioners.
We are still in the early stages of understanding how resulting economic, social, community and familial disruptions and fractures have and will continue to influence the use of policing as a response to individuals in crisis. Given the enormous strains the COVID-19 pandemic is placing on different segments of the population-particularly those who are among our most vulnerable, including the economically disadvantaged, Indigenous communities, and members of other racialized groups-combined with the fact of the already over-burdened healthcare systems in Canada, this is clearly an area that requires surveillance. Unfortunately, as we have documented throughout this report, Canada is currently ill-prepared to take on this vital work because of a lack of standardized approaches to data collection and reporting systems in police services. To improve our knowledge base-and thus support the goal of evidence-based policies in such a critical area of public concern-Canadians will have to consider supporting strategies for improving the collection and use of policing data at provincial and national levels.

Recommendation #1-Provinicial data capturing standards
Developing provincial frameworks for police services to improve data accuracy and more consistently capture the nature of policework with PMI. A primary consideration will be to ensure that the data is captured consistently across police services. For example, all police services should be using MHAs as a conservative proxy for measuring the number of PMI police calls for service. In this case, the definition of a mental health event is provided by the provincial mental health legislation. Capturing patient or subject level information could also be extracted from police data sources though these data will not likely be as consistent as aforementioned MHA-derived data. To improve data quality and consistency, police services should consider establishing various standards for how to document PMI in their data. Hartford et al., (2005) provides a framework for identifying PMI cases as definite, probable, and possible. A similar approach could be used for range of estimation values in a Province.

Recommendation #2-Improve data sharing
Efforts should be made to enhance information sharing on PMI between both local and regional police services. The transient nature of some PMI is likely to result in neighbouring police services interacting with the same subset of individuals. Sharing information across services may enhance understanding and help officers in their interactions. Additionally, improving data sharing between police and health services is recommended to conduct further research on policing PMI, bring about effective ways to collaborate on this matter, and connect both social services together to engage in evidence-informed prevention and reduction strategies. To connect this back to the common theme throughout this report that we simply do not know much about PMI and policing, improving data sharing between police services and between police and health services serves as a means to build capacity to conduct research and enhance the knowledge base on this matter.

Recommendation #3-PMI data analysis
Police administrators should require civilian analysts to study the nature of PMI within police databases. Crime analysts are likely to be knowledgeable on how to query, analyse and report on patterns in their jurisdictions for scheduled reports as well as ad-hoc reports at the request of their supervisors. The analytic skills that many civilian crime analysts possess may be extended to PMI interactions with the police at both the subject level (e.g., EDP) and at the event level (e.g., MHAs). Aggregate information produced should be publicly available in annual reports. That being said, without improving data capturing and data sharing, PMI data analysis is not likely to be helpful as it would be upon the foundation of poor or inadequate data. Therefore, fulfilling this recommendation hinges upon achieving better data capturing practices and improved data sharing amongst local and regional police services, as well as health services.

Recommendation #4-Research collaboration
To address emerging issues or those that require additional expertise, research collaborations between police and universities/colleges/other expert groups is likely to assist in the understanding of the factors correlated with police interactions with PMI. Areas of research may include rigorous and externally evaluated studies of policing programs that are intended to respond to PMI in the community. Alternatively, scholars could explore the intersection between different community groups, police response, and mental health, with a particular emphasis on those groups experiencing various forms of marginalization. Such knowledge recognizes that that PMI-police interactions are multifaceted and should aid in the development of culturally appropriate and inclusive policing approaches. The federal government should provide long-term funding to aid in the development of such work.
Recommendation #5-Task force on Indigenous communities, Indigenous policing, and mental health.
More knowledge is needed to more fully understand the issues related to police mental health responses within Indigenous communities. Working closely with Indigenous communities, the federal government should create a task force to study this area for the future development of culturally-and contextually appropriate responses.

Introduction
Following protests of police involvement in several high-profile deaths of individuals experiencing a mental health crisis, a number of Canadian individuals and groups have advocated in response for various policy positions-ranging from enhanced police training and new policing programs to reallocating police funding to healthcare and social work groups and, in some instances, outright abolishment of public police. Consistently lacking from much of the public discourse is a thorough understanding of the myriad of ways in which policing intersects with mental health issues, the prevalence of mental health issues within police calls for service, the nature of interactions between persons with mental illness (PMI) and police, and the unique challenges in this area facing Indigenous communities in Canada.
The purpose of the present working paper is to begin to flesh out those areas in which we have some knowledge of this topic, as well as identify key areas in which little is known. The structure of this working paper is as follows. First, we speak to the challenges researchers face in trying to use existing data sources to develop estimates of the size and scope of mental health-related demands upon police. Then, we begin to explore the various intersections of policing and mental health. Where possible, we use police figures to highlight the volume of service calls for different types of interactions. Third, we look at previous attempts by researchers at estimating the overall prevalence of mental health-related calls in policing, noting both the strengths and limitations of previous work. We also consider what recent police statistics may add to the discussion, before presenting some recent data on police calls for service during the COVID-19 pandemic. Next, our focus then shifts to what the research can tell us about police interactions with PMI, before examining the important topic of Indigenous communities and the challenges faced by Indigenous police services, particularly those in rural and remote areas. Finally, based on an analysis of the data and research gathered, we present a series of recommendations for policy-makers and practitioners.

A Warning: The Challenges of Working with Police Data
The present evidence review draws heavily upon police data analyzed by two separate sources: academic researchers and Canadian police services. Before we begin this review, we need to highlight some of the significant challenges researchers face when it comes to working with police data. To do that, it might be helpful to first explain a little bit about how police data are produced.
Police services use multiple data capture and analysis tools. Most relevant to our work in the Canadian context are Computer Aided Dispatch (CAD) programs and Record Management Systems (RMS). When a call for service is generated through emergency communications, call details are logged by a dispatcher and this basic information is relayed to responders. Details tracked include the time, date, location of the incident, incident type and any particulars provided by the caller. Depending on the nature of the event, responding officers can provide more details as to the call particulars and then, for calls dealt with informally, 'clear' the CAD call from the system, thereby closing it. Alternatively, if information needs to be tracked, further investigation or the deployment of other police resources is required, or a criminal case needs to be logged, the officer files a separate report in the agency's RMS.
In theory, RMS data contain more information than CAD data about the details of a call. For example, a RMS will typically require an officer to fill out detailed information on the incident particulars-including, although not always correctly, an individual's age, ethnicity, health factors, and so on-and this has the potential to be very useful for research. In reality, though, RMS data quality is often riddled with duplicate entries and missing information (Huey, Koziarski and Ferguson in review). Compounding these data quality issues is the lack of uniformity of definitions, terms, and reporting methods across and within police services (Sinha 2009;e.g., Ferguson and Picknell in review). RMS are also difficult to work with from a research standpoint because they are designed for police use and not for easy extraction and research use. Each report will have an UCR code 1 attached as the primary call type (such as 'assault' or 'theft under 5000'). To find out whether there was a mental health component to the call requires a police service to utilize a mental health 'flag' in their data. Until fairly recently, this was not routinely performed by many services and, as a result, researchers would have had to access case synopses to find that information (e.g., Schulenberg 2016). The inability to easily extract calls for service with a mental health component represents a significant limitation in terms of our ability to gain a more accurate understanding of the impacts of mental health issues on police calls for service. To illustrate, as researchers, we approached six police departments across Canada and found that only one routinely flags mental health calls in RMS data. Of those that do not, we can include the Royal Canadian Mounted Police (RCMP). In their most recent report it was observed that, when it comes to mental health occurrences, this data does not represent all RCMP interactions with individuals suffering from mental illness. Some occurrences may be filed as an assault, a weapons complaint, a suspicious occurrence, or noise complaint as a result of someone shouting, for example. The language used by complainants, witnesses, family members, or the subject of the complaint varies. For these reasons, not all incidences of mental health-related calls for service may be captured in police RMS as mental health-related (RCMP 2020).
As we noted, to fill in some of the gaps in our knowledge of police contacts with PMI, we also draw on statistics reported by police services across Canada in their annual reports. Using keyword searches 2 , we reviewed 80 police websites to locate 19 annual reports containing statistics related to mental health and, where relevant information was presented, included that information here 3 4 . However, as Boyce and colleagues (2015) observe that, as a result of the lack of standardized data collection and reporting mechanisms (beyond the UCR system), there is no standardization in terms of what gets reported in annual reports or how it is reported, rendering it difficult to not only understand what is being reported, but moreover to do effective comparisons across jurisdictions.
This lack of standardized terms to reference is a significant challenge in this area of policing research. The absence of consistent terminology/nomenclature leads to quantitatively and qualitatively different lenses being used to understand police interactions with PMI both across and within jurisdictions. For example, our search terms for the present study includes the keyword "crisis" and "crisis team" which is likely to encompass those incidences where provincial/territorial legislation has been used to detain a patient for medical evaluation at a designated facility. Of course, the same patient may have additional interactions police services (perhaps even on the same day) whereby they are identified as having a mental illness by police but that during their interaction, the patient was not in a state of crises. Our approach to use a broad array of search terms enables us to include relevant research to this area of policing but that within these studies, we recognize there will be differences in how the interaction between police and PMI will be documented. Another significant problem in relation to relying on criminal justice data to enhance our understanding of justice-involved PMI is its lack of diverse and inclusive information on a range of demographic and other variables, such as race and ethnicity, that would provide more indication of the extent and scope of this issue. For instance, writing some 20 years ago, Wortley (1999) observed the existence of what he termed an 'informal ban' among police services on the creation of race-based data. This gap in our knowledge, he contends, means that we know very little about different phenomena that affect diverse communities, and how best to develop and test solutions including to issues of over-representation (Wortley and Tanner 2003;Wortley and Owusu-Bempah 2012). Moving ahead to 2020, we note that most, if not all, police services collect race/ethnicity-based data in their RMS records. That data has not, however, been part of the UCR shared with Statistics Canada and released publicly. This fact is about to change, though, as a move to include race-and ethnicity-based statistics in the UCR was announced in 2020 (Tunney 2020).
All of the above being said, the absence of available national and regional information on policeinvolved PMI and police calls for service involving PMI necessitates that we can only use what is presently available in an attempt to provide an understanding of this matter. Put another way, the limitations of police data aside, there is still value in utilizing these data as a means to advance our knowledge of PMI involvement with police as long as we recognize its constraints in the information and conclusions it can offer. Throughout the following sections, police data is relied upon only as a means to contribute to this conversation. We recognize that the above limitations are present in these data, and as so are cognizant of the fact that we simply cannot offer definitive conclusions on this topic without further investigation and addressing an array of recommendations (of which are present at the end of this document). This being said, without using such data, we would have no starting points to bring about an initial, base of awareness on this pervasive matter.

Mental Health as 'Police Property'
Much has been written on the role of de-institutionalization, changes in mental health legislation, and the lack of coordinated community services and supports for those with serious mental health disorders, so these issues need not be rehashed here (e.g., Iacobucci 2014; Cotton and Coleman 2010). Our focus instead is on the narrower question of the ways in which the management of mental illness has become a criminal justice matter, one requiring police response. In this section we explore two aspects of the police role and mandate that bring police officers into contact with individuals dealing with mental health disorders. Although the governing legislation for police officers in each of the Canadian provinces and territories employs variations on how these roles are framed, we have identified them as public safety and crime prevention and response (e.g., Schulenberg 2013).

Public Safety
Under public safety, we identify three ways in which police interact with PMI: 1. Conducting apprehensions under provincial and territorial mental health legislation; 2. Investigating reports of PMI who have been reported missing from home, shelters, or from hospitals or in-patient facilities; and 3. Conducting wellness checks, including responding to reports of individuals who are at risk of suicide.

Mental Health Apprehensions (MHAs)
Provincial and territorial legislation sets out powers of apprehension that permit police officers to take someone into custody for the purposes of escorting them to the nearest designated psychiatric facility for assessment. Little is known about police decision-making with respect to decisions to effect a mental health apprehension. One of the few Canadian studies to explore this phenomenon found that two sets of concerns inform police decision-making: whether the individual meets the criteria for civil commitment and lack of alternate available resources for someone in crisis and potentially at risk (Schulenberg 2016). In relation to both, officers interviewed expressed frustration over what they saw as stringent commitment criteria that often left them with little alternative but to either arrest someone "for their own good" or release an "at risk" individual 5 (ibid.).
Provincial and territorial mental health legislation also creates a mandated role for police to apprehend PMI who are under a warrant or other authorization ordering them to be apprehended and transported for assessment, examination, and/or treatment. In British Columbia (B.C.), for instance, police may be called upon to bring someone to a facility for an assessment under a Form 4. In Ontario and Alberta, these are known as Form 1s. These are typically situations in which someone is already under some form of community treatment order, but who has not complied with treatment demands. When that happens, mental health professionals call upon police to respond. In one study, Honer et al. (2017: 489) observed that "approximately 20% of apprehensions involved patients with more than 1 apprehension per year" and concluded this fact "is an indication of poor effectiveness of clinical and service delivery." An earlier study observed that only 6.3% of interventions were initiated by hospital staff (Charette et al. 2011). A more recent evaluation of the Saskatoon Police Service's (SPS) Police Crisis Team found that 11% of calls received were for apprehensions under the provincial Mental Health Act initiated by healthcare providers (SPS 2019). Similarly, the Vancouver Police has reported that 26% of their mental healthrelated calls for service were from healthcare or social work staff requesting police assistance (Vancouver Police Department [VPD] 2020).
As these types of calls for service are typically grouped under 'mental health' occurrences, it is difficult to know the overall volume of such calls. That said, we have tried to gain some sense of the overall volume of MHAs conducted by police services. We looked to police sources, searching the websites of 81 different Canadian police services 6 for annual reports for the year 2019, or, if not available, for 2018. Some of what we were able to glean is presented in Table 1 below. Unfortunately, what is not clear is the extent to which these are MHAs based on an officer's fieldbased judgment 7 versus the exercise of a warrant issued by a judge or healthcare professional. Nonetheless, we can see from the limited available information on MHAs that their percentages in relation to all calls for service generally varies by service but often accounts for fewer than 2% of service calls at these agency's.

Wellness Checks, Suicides, and Possible Suicides
Wellness checks, also referred to as 'well-being checks' or 'compassionate to locate' calls, are calls for service in which police are asked to verify the safety and well-being of an individual for whom someone is concerned. These are one of the call types that can have a mental health component. Previous research has, for example, documented police attendance at calls with individuals threatening self-harm or experiencing possible thoughts of suicide (Wilson-Bates 2008; Schulenberg 2016; Shore and Lavoie 2020). In some instances, the latter is coded as 'suicide threat' or 'suicide' within police data, but we have also observed instances where the primary code was 'missing' (if the person had left the residence) or 'wellness check' (if they had not). Although a wellness check may occur for reasons related to a mental health issue, we must be aware of the limitations of relying on this data. Wellness checks can also occur if someone has physical health problems or may be the victim of an accident, among other reasons. With this significant limitation in mind, we present data below on the volume of wellness checks observed within the available annual police reports (see Table 2). Where data on reports of possible suicide or attempted suicide and/or suicidal ideation or behaviour existed in the annual reports examined, we also include those figures below (Table 3). With there being such few agency's represented in both Table 2 and Table 3, we cannot say much about the findings on wellness checks and suicide-related calls. Ultimately, this area represents another matter related to PMI and persons experiencing crisis involvement with police requiring further investigation. In order to begin to understand this matter and develop research from which sound policy and practice can be drawn, much more information is needed.

Missing (Absconding from Facilities)
Each year, hundreds of thousands of Canadians are reported missing to police (Statistics Canada 2018). This is hardly a homogenous group, and previous research has well documented the variety of factors that contribute to this phenomenon (Hirschel and Lab 1988;Taylor et al. 2019). In this section, we focus solely on individuals with serious psychiatric illnesses who have absconded from forensic and/or other healthcare facilities. When such individuals leave or otherwise fail to return to a healthcare setting (such as failing to return after a cigarette break outside the facility), it is a common practice in most, if not all, Canadian jurisdictions for a police report to be filed. In provinces such as B.C., legislation dictates that such a report will be treated as a high priority call ('person at risk' or 'high-risk'), necessitating urgent police response.
Information on the prevalence of absconding cases among missing persons reports is not readily available. Fortunately, we can refer to some recent Canadian research, which used closed police files to look at various aspects of missing persons cases. For example, in one recent study, approximately 11.5% of those reported missing to a municipal police service were found to have absconded from a hospital or mental health facility (Ferguson and Huey 2020). In this study, the strongest predictors of absconding were cognitive impairment, substance dependency, and mental health issues (ibid.). A follow-up study examined the demographic, health and other risk profiles of those who abscond from hospitals and mental health facilities, finding that individuals with schizophrenia and/or dementia were more likely to abscond from long-stay mental health facilities, whereas those diagnosed with depression or bipolar disorders were more frequently reported missing from emergency psychiatric units at hospitals (Ferguson in review). In terms demographic factors, White males were the most frequent absconders from both settings; however, Indigenous patients were over-represented in relation to absconding at 9.7% of the total sample (ibid.). Conversely, Black patients and individuals from other racial or ethnic groups were significantly less likely to abscond from these locations (ibid.) 9 .
In another study using the same dataset, researchers identified adults and youth with histories of repeatedly being reported missing (Huey and Ferguson 2020). Armed with this information, they then looked at the locations from which 'repeat missing people' are most likely to abscond. They found that for adults, the top five locations included three health facilities and two homeless shelters/mission centers, thus indicating that individuals under MHAs not only abscond, but some do so repeatedly (ibid). In terms of the rate at which such individuals are reported missing, we have little go on. An earlier study found that within a one-year span, 230 individuals under MHAs were reported missing to police from two Vancouver area hospitals (Thompson 2015).

As Victims
Given the fact that vulnerable groups are at increased risk of victimization, it is imperative that we note that many PMI come into contact with police as victims of a crime (Brink et al. 2011;Frederick et al. 2018). In contrast to several international studies which highlight that factors such as substance use, homelessness that are likely to increase the rate of victimization in PMI (Hiday et al. 1999), this area has not received a significant amount of attention in Canada. One Canadian study of 547 individuals with mental illness found that they had experienced 518 formal contacts with police, of which 213 were as a result of some form of victimization (Koyoumdijian et al. 2019). Similarly, a study of homeless citizens across five Canadian cities found that those with a diagnosis of post-traumatic stress disorder (PTSD) and/or substance dependence disorder were at an increased risk of victimization (Edalti et al. 2017). Such results align with findings from other countries that have identified mental illness as a significant risk factor for victimization over an individual's life course (Choe et al. 2008;Maniglio 2009). Summarizing nine epidemiological studies, Latalova andcolleagues (2014:1925) state, "patients with severe mental illness are more likely to be violently victimized than other community members." This finding is echoed the Vancouver Police Department's Mental Health Strategy (2020:9), in which they observe that PMI "are often targeted by offenders-resulting in them being 15 times more likely to be the victim of a crime, and 23 times more likely to be the victim of a violent crime." 10 Put another way, analysis by Burczyka (2018) from data from the 2014 General Social Survey (GSS) resulted in a finding of "one in ten (10%) people living with a mental health-related disability report [ing] having been the victim of violence in the preceding 12 months". This figure is contrasted to a rate of violent victimization of "1 in 33 people with no such condition" (ibid.).
One consideration with respect to victimization is the potential unwillingness of victims to come forward to report their victimization to the police from fear of not being believed. Research with victims who are homeless indicates this is a strong concern (Huey and Quirouette 2009). Drawing on the GSS, Burczyka (2018) noted that "individuals with mental health-related disabilities were less likely to report their victimization to the police (22% versus 31%). Among those who did, levels of satisfaction with police action were similar to those reported by victims with no such condition." This research suggests a need for sensitive approaches to understanding and addressing this phenomenon.

As Suspects-Crime and Disorder
One aspect of criminal offending that has drawn some research attention is in the area of violent behaviours. With mental illness not likely to lead to violence and crime (Stuart 2003), other risk factors (which also exist in non-PMI cohorts) should be considered. In their seminal work Monahan and colleagues (2001) highlight that through computationally combining a variety of clinical information, five risk classes explain the vast majority of violence of PMI who were recently discharged into the community. Pertinent factors for predicting the five risk classes include: the patients demographic and clinical information (e.g., substance use), prior arrests, child abuse, underlying psychiatric diagnosis, and information surrounding their father.
Though the merits of clinical work surrounding risk assessment are important in terms of understanding re-offending patterns and developing risk assessment tools, the police will also be called upon to respond to calls in which PMI are the subject of disorder complaints or suspects in crimes. While it is commonly recognized that PMI are more likely to have multiple police contacts, the research suggests that mental health status is not treated by arresting officers as a significant criminogenic factor (Vaughan et al. 2015; see also Schulenberg 2016). Indeed, three separate studies have indicated that when "PMI do offend, they are arrested for the same reasons as other offenders" (Vaughan et al. 2015: 122; see also Schulenberg 2016). The Schulenberg study (2016: 470), that examined police decision-making in cases involving PMI interactions, found that officers are "2.1 times more likely to criminally charge a citizen" in cases involving a more serious offense, suggesting that the perceived seriousness of the alleged conduct has the strongest influence on arrest and charge decisions (see also Charette et al. 2014). A recent study by Shore and Lavoie (2020) offers support for this contention: they found that only 5% of police interactions with PMI resulted in arrests (ibid.), suggesting that most offenses were minor in nature as police are less likely to exercise discretion over more serious allegations (see also Lemieux and Leclair 2020). That said, a recent report by the Vancouver Police (2020) observed that 12% of mental healthrelated calls for service involved a PMI with a weapon.
For those patients who come to police attention, many do so sometimes as a result of engaging in violent behaviour, from uttering threats to displaying a weapon to acts of physical aggression (Stuart and Arboleda-Florez 2001). In some instances, police are mobilized by healthcare professionals as a result of threats or an actual assault (i.e., physical). One of the first studies to explore violence against healthcare workers was a survey of 136 Canadian psychiatry residents which found that approximately 40% had been physically assaulted by a patient (Chaimowitz and Moscovitch 1991). Subsequent research in other countries has produced similar results (Schwartz and Park 1999;Basfra, Hamdan and al-Habib 2019; see also Baby, Glue, and Carlyle 2014). Among these, is a survey of 513 United States (U.S.) psychiatric residents in which 73% reported being threatened, and 36% physically assaulted (Schwartz and Parks 1999).
In relation to the more general calls for service, a study by Shore and Lavoie (2020) found that of 400 PMI police interactions, some form of violent behaviour towards others was observed in 11% of incidents and weapons were present during 16% encounters. Research on police responses to these acts varies. When such individuals do come to police attention, research suggests that "their charge profile [is] similar to that of offenders without mental illness" insomuch as they are "not more likely [than non-PMI] to be charged with a minor offense" (Stuart and Arboleda-Florez 2001: 658). Notably, Charette et al. (2014) found the reverse: that PMI with minor offenses were more likely to be arrested, but that this did not hold true for more serous offenses. The explanation they provide is individual demeanour: the more 'difficult', 'uncooperative' or 'aggressive' the individual appeared, the more likely they were to be arrested independently of their mental health status (ibid.).
Other research has looked at the degree to which violent flags in Canadian Police Information Centre (CPIC) files is predictive of PMI behaviour during police contact. In Schulenberg's (2016) sample, it was found that PMI had more than twice the number of flags for violent behaviour than non-PMI offenders. This would seem to indicate higher levels of violence in PMI who come into contact with police. However, a study by Hoch et al. (2009) revealed that a violence flag is often not a good predictor of whether an individual will be violent in a given situation. They found that some individuals without violent flags were violent, whereas others (approximately 59% of the sample) whose files contained such flags were not. In other cases, where police had had multiple contacts with someone who had engaged in violent behaviour (20.7% of the sample), previous violent behaviour was not flagged in the system (Hoch et al. 2009). In short, these flags can, and do, fail to provide reliable information.
It is commonly suggested that PMI are likely to be the subject of complaints or concerns surrounding disorder or involved in minor offences, such as vandalism or theft of property, and thus come to police attention this way and to be charged accordingly. However, the available research in this area is not at all clear. A recently released study by Lemieux and Leclair (2020) of justice-involved homeless citizens with mental illness observed that the most frequent offenses for which these individuals were charged included thefts (29%), administration of justice offenses (23%) (such as breaches of probation or parole), and assaults (14%). These results led the researchers to conclude that impoverished PMI are more likely to "engage in property and acquisitive offenses as a means of survival" (ibid.: 800). Similarly, Kouyoumdjian and colleagues (2019) observed that, among their sample of homeless PMI who had been charged with an offense, approximately one in five charges were for administration of justice offenses and one in eight were related to what they term "an act of living" (ibid.: 723). However, the Charette (2014: 513) study, which used both a sample of PMI and a control group, found that police interventions involving PMI "were less likely … to be categorized as offenses against property" and a study of charge rates for PMI by Hoch et al. (2009) noted that 21% of charges were for crimes against property, whereas 29% of charges were for crimes against the person. Ultimately, across the existing Canadian literature, understanding the quantity and quality of complaints, disorder, and minor offences that police respond to that involve PMI vary considerably. However, what we can conclude is that administrative and non-violent events represent a considerable number of points in time where police services will intersect with PMI as suspects in the community (Vaughan et al. 2016).

Previous Research Attempts at Estimating Mental Health Prevalence Rates in Police Calls
When it comes to estimating the extent to which Canadian police are dispatched, summoned, or proactively engage in service calls involving individuals with a mental illness, estimates vary. For ease of reading, we present these estimates in Table 4 below. Why is it so difficult to have a clear-cut answer as to how prevalent mental health issues are in relation to police calls for service? One obvious answer is difficulties with how police data is collected and used, as previously mentioned. Another factor to consider is the nature of the data researchers have used and the assumptions made with respect to that usage. For example, the Charette (2014) study drew on data from only three days of police reports in 2006. The overall sample size was, therefore, relatively small and subsequent research has shown that calls for service involving mental health issues can cluster both temporally and spatially (Vaughan et al. 2018;Koziarski forthcoming). This means that dates selected might not be representative-in essence, that cross-sectional research designs may not be suitable for understanding this phenomenon. In the section on strengths and limitations of police data, we provided a detailed discussion of the challenges with relying on police data to draw conclusions on the overall volume of calls that have a mental health component. Problems with how police have historically recorded data has clearly created methodological issues for previous researchers, which is likely reflected in the range of prevalence rates-from an estimated 1% to 30%. A third important factor is that mental health prevalence rates will vary to some extent by location, as a result of, among other things, population characteristics and the ability of PMI to access services.
Another area of possible contention is whether police contacts with individuals with mental illness are increasing (Coleman and Cotton 2014). Certainly, this has been a consistent claim of police/ police services. In its 2018 annual report, the Ontario Provincial Police (2018) stated they had an approximate annual increase of 10% in their volume of mental health occurrences, an increase they stated was consistent over the previous two years. In less precise terms, Laval  noted "in 2019, the volume of disturbed mental state events responded to by our patrollers reached 1,941, the highest number in the past five years." Other police services reporting similar increases include Timmins, Chatham-Kent, and Camrose. Some studies support these claims (see Heslop et al. 2011). For example, in a 2019 study, Vaughan and Andresen found that calls for service involving PMI were increasing each year by some 9.7% among police services in B.C. What cannot be said from these studies, though, are the suspected reasons behind the potential increase in police contacts with PMI. Indeed, definitional changes and awareness of the issue by police officers may or may not be at play here, but, alternatively, it could be due to social changes, population growth, and other macro and micro impacts unrelated to (yet impacting) policing. As such, and to connect to it being of contention, pinpointing whether police contacts with PMI are, in fact, increasing is not possible at this time based on the available information.

Police Attempts at Estimating Prevalence Rates
As we noted, one of the ways in which we tried to overcome these challenges was by reaching out to police services that routinely flag mental health-related calls regardless of the UCR or dispatch code assigned. We did this keeping in mind the limitations associated with relying on flag use as a proxy for 'mental health.' These mental health flags are reported in Table 5 below with a comparison to the total number of calls for service at each agency. As can be seen in Table 5, there is variation in the percentages across the police services that flagged mental health-related calls. The large dispersion across the percentage of mental health flags could represent the following: 1) what is captured across police services as a mental healthrelated flag greatly varies, and so the percentages are reflective of service-by-service data capturing procedures, 2) awareness levels of mental health-related calls for service varies, pointing out the need for greater understanding on this matter, and/or 3) mental health-related flags may fluctuate according to other factors not captured by this procedure. Ultimately, we do not know what is interplaying in this variation, highlighting another policing area warranting attention in future research. What can be said, though, is that there are no systematized way that police officers and/ or services define what a 'mental health flag' means operationally and when they should place the flag on the system.
Other than what is included within Table 5, the RCMP (2020) recently released their statistics on 'mental health occurrences' across each of the provinces and territories, including those for which the RCMP is not contracted as a provincial or a municipal service. They reported more than 123,000 'mental health occurrences' out of approximately 3 million calls for service (ibid.). As with many other police agencies, the RCMP does not flag mental health-related calls when the primary call purpose is something else (such as an assault or theft) (ibid.), thus this figure is not reported above.
In relation to prevalence rates, one area that remains underexamined is the extent to which some individuals at increased risk of experiencing crisis may generate multiple calls for service. We were able to locate one police service that reported this figure. In 2020, Timmins Police observed that of 656 'Mental Health Act' calls received, 348 incidents involved PMI who had generated 2 or more calls.

COVID-19 and Mental Health Calls to Police
We were fortunate to be able to access a recently released report from Statistics Canada (2020) on police calls for service during the early part of the COVID-19 pandemic (March 2020 to June 2020 in Canada). Calls for service data were collected from 17 police services across Canada, including data on 'mental health apprehensions', 'mental health-other', 'suicide/attempted suicide', and 'welfare check'. One significant drawback of using these figures is they are not parsed out based on the police services selected, so regional or other variations cannot be noted. Further, as we found ourselves when looking at mental health statistics across police service reports, variations exist in how different types of calls for service are categorized that render it difficult to provide accurate counts. Statistics Canada issues a stark warning: "interpret calls for service data with caution" (ibid.: 3). That said, we reproduce selected figures below in Table 6. Examining the above figures provides an interesting perspective in that we have a snapshot of possible changes in both mental health rates due to COVID-19, as well as in resulting demands for increased police service. To note, though, it only offers a potential glimpse of the COVID-19 pandemic impacts on mental health demands on police with respect to public safety, but not on crime. 12 A year-over-year comparison would seem to be useful as it is unlikely that changes to call coding among individual police services would have occurred, thereby skewing the results. Ultimately, a key takeaway from this snapshot from Statistics Canada (2020) is that COVID-19 may have impacted mental health calls for service. That being said, we also urge some caution here: most researchers would require several years of historical data to rule out the possibility of increases or decreases not being the result of normal fluctuations over time.

What is Known About PMI Who Come into Contact with Police
A consistent finding across several Canadian studies is that certain socio-demographic groups are over-represented in analyses of police contact with PMI. One of those groups is homeless citizens; a not entirely surprising finding for several reasons. One factor is that individuals without permanent shelter are more likely to be visible in public spaces, and thus are seen as a sign of social disorder and reported to police for 'status offenses' (Huey 2007). Relatedly, individuals without sustainable shelter and income are also more likely to engage in survival offenses, such as petty thefts (Schulenberg 2016). Another factor is that homeless populations often have higher rates of mental disorders and substance abuse. For example, in a Quebec study of individuals using homeless services, researchers found that 72% of their sample had experienced a significant mental illness in the preceding year (Bonin, Fournier and Blais 2007). Similarly, a study in Vancouver of residents of the city's impoverished Downtown Eastside found an association between poverty and emergency room visits related to mental illness and substance abuse (Honer et al. 2017). In Ontario, the combination of mental illness and substance abuse was seen in the majority of cases involving homeless citizens who not only came into contact with police but did so at a "higher rate of police contact than other PMI" (Schulenberg 2016: 476).
While we have focused primarily here on police contacts with PMI who are also homeless, it is important to keep in mind that research has consistently found that PMI-whether housed or not-are more likely to have police contact than those without a mental illness. To illustrate, the 2012 Canadian Community Health survey found that 34.4% of PMI or individuals with substance dependency reported at least one police contact within the previous year (Boyce et al. 2012). This figure is twice the rate of those without a mental illness or substance dependency (ibid.). The authors of the resulting report observed that, among the general public, "the presence of a mental or substance use disorder was associated with increased odds of coming into contact with police, even after controlling for related demographic and socioeconomic factors" (ibid.: 18).
Another matter is the use of force in police arrests and apprehensions. Although it has been reported that fewer than 1% of contacts with police result in the use of force (Adams 1999;Baldwin et al. 2019), PMI have reported experiencing some level of force, from being pushed or punched to having a baton used or conducted energy weapon deployed while being apprehended . One Ontario agency, the Durham Regional Police Service (2018) has begun collecting use of force statistics related to incidents involving individuals apprehended under the provincial Mental Health Act. The rationale for including this metric is that the organization committed to reducing the number of use of force incidents involving MHAs and have been documenting the results of this effort.
Who initiates police calls for service regarding PMI? To date, little Canadian research has explored this question or the dynamics that led to a request for police intervention. One of the few studies that has explored this phenomenon found that PMI were among those mostly likely to initiate an intervention (21%), followed by a relative (20%) or bystander (19.5%) (Charette et al. 2011). In the previously cited evaluation of its police crisis team, the Saskatoon Police collected the following information on where such calls originated. They found that of those calls that came through dispatch, 42% (n=191) were initiated by family members or friends, whereas approximately 13% (n=60) were self-initiated (SPS 2019). An Ontario-based study found "the most common initiator of police-PMI contact was a service provider (i.e., paramedic, community mental health worker, school or hospital staff, or the police officer themselves) (29%)" (Shore and Lavoie 2020: 162).
If Canadian research on PMI contacts with police is generally scarce, this is doubly the case in relation to police contacts with individuals with mental illness who are also members of minority and/or marginalized groups. We can assume that such contacts among, for example, racialized persons are increasing based on recent research that shows a "steady increase in forensic admissions" in Ontario from 1987 to 2012 of "individuals with of non-Caucasian ethno-racial background" (Penney et al. 2019: 635). One of the few studies to look specifically at the ethnic composition of a sample of PMI who have come into contact with police found that 14% were non-White (Shore and Lavoie 2020). As these scholars note, ethnic minorities comprise only 10% of the population in the area studied (ibid.). They also note a finding that we believe requires substantially more research to thoroughly understand and unpack the underlying dynamics: analysis of their data revealed that PMIs who were White were 6.6 times more likely to receive a community health referral from attending police officers than PMI from an ethnic minority group (ibid.). That said, even less is known about other groups, particularly in relation to police contacts with PMI from the LGBTQ2IS+ community. This lack of research is critical to highlight because it reduces our ability to craft thoughtful, culturally appropriate, and inclusive approaches to addressing any both current and potential issues.
When it comes to police-involved shootings, a recent report of the Office of the Independent Police Review Director (OIPRD) states "we cannot ignore the fact that, in many of these cases, the deceased was Black or a person of colour" (McNeilly 2017: 4). The report then notes, "it is unknown exactly how many civilians killed by the police in Ontario have been Black or of colour as no agency maintains race-based statistics on this issue" (ibid.). Attempts have been made to catalog (CBC News 2017) and describe unique predictors of police-involved fatalities of PMI (Ouellet 2019). Problems with case inclusion criteria have been noted in the literature (Prevett 2018). A more recent effort is currently underway by researchers at Carleton University. Unfortunately, preliminary results were not available at the time of writing. That said, we were able to locate Coroner's inquest records 13 for the provinces of Ontario and B.C. for police-involved shootings for the period of 2003 to 2019 14 . Data from this source-cross-checked against news reports-provides some basic information regarding race/ethnicity 15 , gender and age of PMI who died in a police-involved shooting 16 .
Of the 34 Coroner's Inquest cases selected for analysis 17 , 21 were from Ontario and 13 from B.C. 18 The gender composition of this sample comprised one female and 33 males, with an average age was 36 years. In terms of the racial or ethnic makeup of PMI in the cases analyzed, 23 were identified as White (67.6%), five as Black (14.7%), four as Indigenous (11.7%), one as Middle Eastern (2.94%) and one as Asian (2.94%). As with previous research (see McNeilly 2017), we observe that Black individuals (14.7%) were-relative to their representation in the larger populationdisproportionately represented in our findings, as were Indigenous persons. We also note-in line with Shore and Lavoie's (2020) findings-that 32% of this sample were persons of colour. While our sample is too small from which to draw significant conclusions and contains several limitations, we do see the utility in drawing attention to this phenomenon and the limits of current knowledge.

A Special Consideration: Indigenous Communities and Indigenous Policing
In this section we pay particular attention to the policing of Indigenous communities for two important reasons. The first is the unique history of Indigenous peoples in Canada and the resulting social, cultural, and economic devastation of colonialism and the role that both the criminal justice and healthcare systems have played in that devastation-along with continued discrimination and structural inequities, among other impacts-and in responses to its subsequent impacts on Indigenous peoples (Razack 2015;McCallum and Perry 2018). One of these impacts, for example, is the overrepresentation of Indigenous peoples in the criminal justice systems. While this is not a matter specific to the Canadian context 19 (Cunneen 2014), it is important when recognizing Canada's history related to Indigenous peoples. The second reason is centered on a distinctive feature of the Canadian policing landscape: The First Nations Policing Program, a federal program establishing and funding Indigenous policing services across Canada with comparable mandates and powers as municipal police agencies. These two factors, both individually and combined, create enormous challenges for Indigenous communities, as we discuss here.
The history of colonization and the resulting cultural fractures and individual and collective trauma forced on victims, families and communities through such colonization projects as the residential schools, reserve system, and the 1960s scoop are well-known. We see this trauma reflected in high rates of mental disorders, suicide, substance dependency, and homelessness, among other tragic effects (Wilk, Maltby, and Cooke 2017;Curran et al. 2018;Pearce et al. 2018). In one recent study, researchers found that "Indigenous ethnicity was independently associated with being homeless at a younger age, having a lifetime duration of homelessness longer than 3 years, PTSD, less severe mental disorder, alcohol dependence, more severe substance use in the past month and infectious disease" (Bingham et al. 2019:1). Indigenous persons, particularly Indigenous women, are also at increased risk of violent victimization relative to the general population, a fact which also contributes to ongoing trauma, emotional distress and/or worsening of pre-existing mental health issues (ibid.; RCMP 2015; Sochting et al. 2007). Sexual and other forms of violent victimization of women in other populations has shown to be tied to repeat cycles of homelessness, which is also an important consideration (Broll and Huey 2020).
Compounding the problems Indigenous PMI face are issues with the Canadian healthcare system. It has frequently been suggested in popular media that social workers and healthcare professionals should be utilized as frontline response to calls involving those in emotional or mental crisis. Such calls ignore the historic and contemporary reality of many Indigenous peoples, who have experienced racism and other forms of marginalization within the health care system (Allan and Smylie 2015;McCallum and Perry 2018;Browne et al. 2016), as well as in relation to dealing with the social workers through social service provisions and/or child welfare systems (McCauley and Mathison 2018). Other barriers to healthcare include limited resources in rural and remote communities, and structural problems for Metis and non-status Indigenous peoples with the provision of health care benefits (Lavoie 2013;Allan and Smylie 2015). In relation to the former, a Health Canada report (2015: 30-31) on the healthcare challenges experienced by Indigenous communities observed that "the [northern] territories rely heavily on provincial health systems for services they are unable to provide, particularly for treatment of acute mental illness and tertiary care. As a result of this, the continuum of care is disrupted, which creates significant challenges that exacerbate mental illnesses." Lack of access to mental health treatment leaves members of rural and remote Indigenous communities feeling "frustrated and helpless with the prevalence of depression and suicide and their limited means to address it" (Curran et al. 2018: 7).
When it comes to the policing of Indigenous communities, there are two systems: one for those persons 'off-reserve', who are policed by provincial, municipal or regional services along with the rest of the population, and those 'on-reserve'. The latter may be policed through arrangements with a provincial police agency 20 or through their own service established under the First Nations Policing Program (FNPP). The FNPP was instituted in 1992 to provide opportunities for Indigenous communities to set up their own services under local authority with funding support from the federal government. We could not determine the exact number of these services in operation today because several do not maintain a social media space or have disbanded (Kiedrowski, Jones, and Ruddell 2016). Still, according to a 2017 study of Indigenous policing, the researchers observed that 453 communities in Canada are policed by a service established through the FNPP (ibid.).
Far too little is known about Indigenous policing in Canada. In 2020, one of the authors observed that there were only eight published, peer-reviewed research studies on the Indigenous policing (Huey 2020). One of the few exceptions is work by Kiedrowski, Jones, and Ruddell (2017) who conducted a survey of Indigenous police services to explore their strengths and challenges.
What they documented was a system in which these services were chronically underfunded, struggled to meet local demands, faced difficulties in recruitment and retention, and several other challenges, leading the researchers to conclude that these agencies had been 'set up to fail'. Among the issues with which FNPP officers had to grapple with were significant rates of mental health disorders emerging in disproportionately high levels of suicide within their communities, substance dependency, and violence (ibid.; see also Curran et al. 2018).
We are fortunate in that, as part of their survey of FNPP officers, Kiedrowski et al. (2017) examined issues related to police mental health responses within Indigenous communities. They detailed that there were several major challenges, including high rates of mental health-related calls and lengthy distances to transport individuals detained under provincial mental health acts, which put additional pressures on under-resourced services (ibid.). Another challenge cited was inadequate training in crisis response and community mental health, a training deficit which forces police members to try to identify external resources.
In our search of police annual reports for statistics related to mental health calls for service, we located 16 Indigenous police services online. Of these, five provided some statistical information on mental health-related occurrences, of which is contextualized with information on their location and population size served (see Table 7).

Summary
We are still in the early stages of understanding how resulting economic, social, community and familial disruptions and fractures have and will continue to influence the use of policing as a response to individuals in crisis. Given the enormous strains the COVID-19 pandemic is placing on different segments of the population-particularly those who are among our most vulnerable, including the economically disadvantaged, Indigenous communities, and members of other racialized groups-combined with the fact of the already over-burdened healthcare systems in Canada, this is clearly an area that requires surveillance. Unfortunately, as we have documented throughout this report, Canada is currently ill-prepared to take on this vital work because of a lack of standardized approaches to data collection and reporting systems in police services. To improve our knowledge base-and thus support the goal of evidence-based policies in such a critical area of public concern-Canadians will have to consider supporting strategies for improving the collection and use of policing data at provincial and national levels.

Recommendations
The following recommendations are based on the preceding review of the research evidence, as well as on those critical gaps we have identified.

Recommendation #1-Provinicial data capturing standards
Developing provincial frameworks for police services to improve data accuracy and more consistently capture the nature of policework with PMI. A primary consideration will be to ensure that the data is captured consistently across police services. For example, all police services should be using MHAs as conservative a proxy for measuring the number of PMI police calls for service. In this case, the definition of a mental health event is provided by the provincial mental health legislation. Capturing patient or subject level information could also be extracted from police data sources though these data will not likely be as consistent as aforementioned MHA-derived data. To improve data quality and consistency, police services should consider establishing various standards for how to document PMI in their data. Hartford et al., (2005) provides a framework for identifying PMI cases as definite, probable, and possible. A similar approach could be used for range of estimation values in a Province.

Recommendation #2-Improve data sharing
Efforts should be made to enhance information sharing on PMI between both local and regional police services. The transient nature of some PMI is likely to result in neighbouring police services interacting with the same subset of individuals. Sharing information across services may enhance understanding and help officers in their interactions. Additionally, improving data sharing between police and health services is recommended to conduct further research on policing PMI, bring about effective ways to collaborate on this matter, and connect both social services together to engage in evidence-informed prevention and reduction strategies. To connect this back to the common theme throughout this report that we simply do not know much about PMI and policing, improving data sharing between police services and between police and health services serves as a means to build capacity to conduct research and enhance the knowledge base on this matter.

Recommendation #3-PMI data analysis
Police administrators should require civilian analysts to study the nature of PMI within police databases. Crime analysts are likely to be knowledgeable on how to query, analyse and report on patterns in their jurisdictions for scheduled reports as well as ad-hoc reports at the request of their supervisors. The analytic skills that many civilian crime analysts possess may be extended to PMI interactions with the police at both the subject level (e.g., EDP) and at the event level (e.g., MHAs). Aggregate information produced should be publicly available in annual reports. That being said, without improving data capturing and data sharing, PMI data analysis is not likely to be helpful as it would be upon the foundation of poor or inadequate data. Therefore, fulfilling this recommendation hinges upon achieving better data capturing practices and improved data sharing amongst local and regional police services, as well as health services.

Recommendation #4-Research collaboration
To address emerging issues or those that require additional expertise, research collaborations between police and universities/colleges/other expert groups is likely to assist in the understanding of the factors correlated with police interactions with PMI. Areas of research may include rigorous and externally evaluated studies of policing programs that are intended to respond to PMI in the community. Alternatively, scholars could explore the intersection between different community groups, police response, and mental health, with a particular emphasis on those groups experiencing various forms of marginalization. Such knowledge recognizes that that PMI-police interactions are multifaceted and should aid in the development of culturally appropriate and inclusive policing approaches. The federal government should provide long-term funding to aid in the development of such work.
Recommendation #5-Task force on Indigenous communities, Indigenous policing, and mental health.
More knowledge is needed to more fully understand the issues related to police mental health responses within Indigenous communities. Working closely with Indigenous communities, the federal government should create a task force to study this area for the future development of culturally-and contextually appropriate responses.