Crisis Response Model Preferences of Mental Health Care Clients With Prior Misdemeanor Arrests and of Their Family and Friends

Reliance on police as default responders to people experiencing a mental health crisis has led to frequent entanglements of people with serious mental illnesses within the criminal legal system (1, 2). People with serious mental illnesses (such as psychotic or mood disorders) are overrepresented in jails and prisons, and an estimated 25% of people with mental disorders have been arrested (3–5). People with mental illnesses are also overrepresented among those killed by police (6). Meanwhile, the public and policy makers have increasingly recognized that many encounters with people experiencing a behavioral health crisis could be addressed by mental health professionals working in collaboration with or independently of police officers. Social justice movements have created momentum for alternative crisis response models, and recent implementation of an easily accessible national crisis hotline, 988, has invigorated discussions about how to strengthen a long-neglected crisis care system. A range of crisis response models now exists, along with guidance for practitioners and policy makers (7–9).

Among the best-known models of crisis response are the crisis intervention team (CIT) model and the co-responder model. The CIT model involves a specialized, police-based program with strong community partnerships that includes 40 hours of officer training on identifying people with mental health issues, deescalating encounters, and linking people to care. Trained officers then serve as specialized responders to mental health–related calls (10). Co-responder models use mental health clinicians and police officers, who collaboratively respond to crisis calls to deescalate crises and link individuals to services (11). A third type of response involves providing phone-based support to police officers in the field by linking them to a mental health professional who can discuss presenting behaviors and options for intervention (12, 13).

Outside of police-based responses, recommendations have been made for wider implementation of mobile crisis teams (14, 15). Mobile crisis teams are dispatched from the mental health system and provide onsite assessment, intervention, consultation, and referral (16). Recently, a broader range of nonpolice crisis response models have emerged. These models are housed in different agencies (e.g., mental health agencies, nonprofits, and fire and emergency medical service departments) and are staffed in multiple ways (e.g., by clinicians, medical practitioners, crisis workers, and peer workers). Such models are proliferating, but their outcomes have not been rigorously evaluated; existing studies are primarily observational (17).

Individuals with serious mental illnesses and their family members are the end users of crisis response models and have the most to gain or lose in a crisis. To date, their preferences related to response models have received little attention. Qualitative studies outside the United States suggest that negative experiences and low expectations related to crisis response among clients and family members make them hesitant to access crisis services until the crisis is severe (18–20). These studies have found that service users report a variety of experiences with both police and mental health crisis teams. Clients and families have reported that police involvement can be stigmatizing and can exacerbate crises (19, 20). Experiences with nonpolice crisis teams are not uniformly positive either, particularly because of wait times and limited availability.

This study aimed to elicit the perspectives of mental health clients with a history of arrest and of their family members or close friends regarding four established crisis response models: the CIT model, co-responder model, telephone linkage model, and nonpolice model (i.e., mobile crisis or community responder model). We designed a qualitative supplement to an ongoing randomized controlled trial of a new police–mental health linkage system in Georgia to understand the experiences of people with serious mental illnesses interacting with police and to learn about their preferences for the crisis response models listed above. We aimed to identify clients’ needs when they are in crisis, the models they most and least prefer, and key factors shaping their preferences.

Methods

We conducted semistructured interviews with clients enrolled in an ongoing trial of a new police–mental health linkage system. The linkage system consisted of three steps. First, individuals with serious mental illnesses and a history of arrest gave consent to be enrolled in a statewide database that allowed for brief disclosure of their mental health status to responding police officers and for the responding officer to talk with a mental health professional. Second, if a police officer had an encounter with an enrolled client and ran a routine background check, he or she received an electronic message that the individual had mental health considerations, and the officer was given a phone number to call. Third, the linkage specialist provided brief assistance to the officer by telephone (12).

Recruitment and Data Collection

Linkage system participants.

Participants were eligible for this study if they were enrolled in the parent study and had ever been arrested for at least one of the following five misdemeanor charges: criminal trespass; disorderly conduct; willful obstruction of a law enforcement officer; minor theft, petit larceny, shoplifting, or theft by taking; or misdemeanor assault. Eligibility for the parent study included the ability to speak and read English; age ≥18 years; diagnosis of a psychotic or mood disorder (with or without comorbid conditions), based on the referring clinician’s report of the most recent diagnosis in the patient’s electronic medical record; at least one previous arrest within the past 5 years, as reported by the client; having no known or suspected intellectual disability or dementia, according to clinician or client report; and ability to give informed consent for participation in the police–mental health linkage system. The misdemeanors of interest were chosen on the basis of work by the research team about the most common or overrepresented misdemeanor charges among people with serious mental illnesses (21–23).

Clients in the linkage system were recruited for interviews from March to June 2021. We aimed to recruit 50 clients for interviews to balance the goal of information saturation with the complexity of analysis (24). Participants who enrolled in the linkage system during this recruitment period were asked whether they were also interested in participating in an interview and, if so, were assessed for eligibility. Additional participants were recruited by calling people already enrolled in the linkage system in reverse chronological order, starting with the individual most recently enrolled. Although 1,081 clients had already been enrolled in the parent study of the linkage system when this qualitative study began, the rationale for this approach was that most recently enrolled individuals would be easier to contact or would be more likely to respond, compared with clients who had enrolled up to 2 years earlier. Research assistants approached or called 92 clients and could reach 66; of these clients, 60 met eligibility criteria and 50 provided informed consent and completed interviews. At that point, having achieved informational redundancy, we planned no further interviews (25).

Two research assistants conducted 60- to 90-minute, audio-recorded interviews via HIPAA-compliant WebEx technology. Interviews were conducted one on one, and participants were compensated with a $75 gift card. The interview guide was divided into three parts. First, clients were asked about their experiences being arrested for one or more of the five misdemeanor charges of interest. Second, clients were asked about their perceptions of the criminal legal system and interactions with various criminal legal professionals. Third, clients were asked about their experiences of mental health crisis, including about their needs during crisis, the crisis response models they experienced, and their perceptions of the advantages and disadvantages of each model.

This article focuses on the part of the interview about mental health crisis and response. Table 1 provides information on how each model was described to participants. Participants were shown a visual depiction of the models via WebEx, and the research assistant described the models by using a prepared script. To avoid using jargon, research assistants referred to the CIT model as a “specialized police response” and the co-responder model as the “ride-along model.” (For clarity and consistency with the literature, we use “CIT” and “co-response” in this article.) After a description of each model, participants were invited to ask clarifying questions. The research assistant asked about the following topics: the models that had been dispatched to participants, relevant experiences among participants, advantages and disadvantages of each model on the basis of direct experience or the description provided, and the crisis response model participants would most and least prefer if they were experiencing a crisis and needed immediate help from a trained professional.

Family members and friends of enrolled participants.

The research team also recruited close family members or friends of the 50 study participants. These individuals were a first-degree relative, a spouse or partner, or in close contact with the individual (in contact for at least 1 hour per week). Of the 50 clients, 30 referred a family member or friend. Research assistants could contact 25, and 18 provided informed consent and completed interviews. Interviews of family and friends were conducted via WebEx, designed to last 60 minutes, and modeled on client interviews. Participating family members and friends were compensated with a $50 gift card.

Data Analysis

This study was approved by the New York State Psychiatric Institute Institutional Review Board. All interviews were recorded and transcribed verbatim. A thematic approach was used for analysis (26, 27). The research team used inductive and deductive approaches for analyzing data. The team read a sample of interview transcripts and used open coding, creating an initial list of distinct concepts for categorization. Additional rounds of transcript review and discussion allowed the team to further refine these concepts until a final codebook was produced. The final codebook also included codes that were deductive, given the team’s interest in articulating advantages and disadvantages of specific response models. The lead author (L.G.P.) and one research assistant (A. Warnock) coded the interviews in Dedoose, and a 10% sample of text was double coded for reliability; coding discrepancies were resolved through consensus discussions.

The research team then reviewed code reports and code co-occurrence tables, focusing the analysis on the broader level of themes and drafting summaries related to the findings reported below. These themes were reviewed alongside a count of the crisis response models that individuals named as their most and least preferred. Team discussions were held to review the results and consider alternative ways of organizing the data to ensure meaningful triangulation across investigators.

Results

Demographic characteristics of clients and their family and friends are presented in Tables 2 and 3. Most interviewed clients identified as men (62%). Slightly more than half were Black (54%), and almost all identified as non-Hispanic (98%). Their mean±SD age was 34.7±7.8 years. Overall, 40% (N=20) reported living with a family member, 20% (N=10) reported living alone, and only one person (2%) reported being homeless. Clients reported a history of arrest for an average of 2.2 of the five charges of interest. (We did not ask about arrests on other charges.) Very few clients (N=8, 16%) endorsed having knowingly had experiences with any of the described models: CIT model (N=4, 8%), co-responder model (N=1, 2%), telephone linkage (N=1, 2%), or nonpolice response (N=5, 10%).

Most family and friends were women (72%), Black (56%), and non-Hispanic (100%). Seven (39%) were first-degree relatives, six (33%) were spouses or partners, and five (28%) were friends or roommates.

Client Needs During Mental Health Crisis

Clients and their family and friends were aligned in describing clients’ needs during a crisis. Participants discussed concrete needs, including crisis line access, medications, safety plans, sobriety, and prayer or religion. However, the most common needs cited related to the environment and interpersonal connections. Participants highlighted the importance of being calm, relaxed, and in a safe environment. One client stated, “No boisterous, loud, chaotic situations. I’m talking about calm, collected approaches” (client 1). They also discussed the value of having somebody to talk to and of feeling supported. Notably, clients described how they often reach out to informal supports first when in crisis, rather than involving a professional. As one client stated, “I need to get out of that [situation] I’m in. . . . That’s usually what triggers my emotions the most. . . . So I would have to figure out where my [support] person is . . . because he knows how to calm me down the most and get me out of the situation” (client 2).

Participants generally agreed about the need for empathy and understanding. One family member stated, “She needs to know that she’s not alone. She needs to know that everybody else is going through some similar problems” (family 1). Similarly, a client described the need for an empathetic approach: “All they need is compassion . . . they need to feel loved . . . they need to feel that ‘Hey, I’m—I’m not here to harm you, I’m not here to do anything to you. I’m here to help’” (client 3).

Most and Least Preferred Models

Clients and their family and friends were mostly consistent in selecting their most and least preferred crisis response models. As shown in Figure 1, the most preferred model was a nonpolice response (clients: N=24, 48%; family and friends: N=11, 61%), followed by the co-responder model (clients: N=13, 26%; family and friends: N=3, 17%). CIT was the least preferred model (clients: N=28, 56%; family and friends: N=12, 67%). Although the nonpolice response option was preferred by the most participants, almost one-quarter of clients (N=11, 22%) chose the nonpolice response as their least preferred option.

Importance of Trained Responders

The most commonly cited advantage of the nonpolice response model was that it involves a range of professionals trained to respond to mental health crises (clients: N=26, 52%; family and friends: N=10, 56%). “They actually know how to talk to you, know how to handle you, deal with you at that time in that situation. . . . I feel like they’re trained for that,” remarked one client (client 4). A family member said, “The advantage is they’re all trained. They all know what they’re dealing with. And so, you know, if there’s a decision that this person needs to be committed, the decision is made onsite by who has the training for it. And it’s not just somebody guessing” (family 2).

The presence of a responder with mental health training was also cited as an advantage for co-response, phone linkage, and CIT, compared with traditional police responses. With co-response, participants noted how a mental health professional “would be able to explain what’s going on to the [police] officer better” (client 5). Similarly, with the phone linkage model, participants described how a “mental health worker could give [the officer] insight as to how to go about it and approach the situation” (client 6). With respect to CIT, participants described how a police officer with mental health training would be better equipped than untrained officers, because they “would know how to interact with whichever person they need to, without getting overwhelmed and frightened” (client 2).

Participants also differentiated the models by the amount of training responders received; the most preferred models were those in which the professionals on the scene had the most training and experience with mental health crises. Thus, although participants appreciated that police receive such training, they also highlighted the pitfalls of not having enough mental health expertise present during crisis situations. One of the primary disadvantages of the phone linkage model was that the mental health professional was not physically present and that the outcome of the situation would depend on the police officer accurately relaying information to the clinician on the phone. The primary disadvantage cited for CIT was that 40 hours of mental health training was not sufficient (clients: N=19, 38%; family and friends: N=10, 56%). As one participant reflected, “I don’t think that’s anywhere near sufficient enough training to be able to handle . . . a mental health crisis” (client 7).

Negative Associations With Police

Participants reflected on how the involvement of police in crisis response could be detrimental. Clients spoke from experience, given that all of them had been arrested and that all but two reported having spent time in jail. First, 14 clients (28%) were not convinced that a CIT response was much different from a traditional police response and worried that CIT would still lead to negative consequences, such as arrest or hospitalization. “Even though they’re trained, it’s still just the police,” noted one client (client 8). Second, almost one-third of clients (N=15, 30%) and half of all family and friends (N=9) described how police could be “intimidating” or “threatening” to people in crisis. As one family member remarked, for “people who are going through mental health crisis, when they see the blue lights and see the police, sometimes it makes it worse. The stress is even more” (family 3). A friend described how police uniforms can produce a triggering response: “A uniformed officer creates a different response from people automatically, just because of the uniform. [The uniform] changes the person’s perspective altogether, rather than just plainclothe[s] people approaching the situation” (family 4). Clients agreed that police could be “triggering” because they either are afraid of police or fear the consequences of a police response. One client noted that a police response would make them “feel more anxiety about being in serious trouble, rather than actually getting help” (client 9).

Safety Concerns

Although clients and their family and friends reported negative experiences with police and cited how police could escalate crises, they also acknowledged that certain situations might require police presence: those involving the safety of the person in crisis, of the people around them, or of the professional responders. The primary disadvantage raised about the nonpolice response model concerned the lack of police presence in “unsafe” situations. This factor was cited by 12 clients (24%) and seven family members (39%), who described a range of scenarios involving safety concerns. Some participants thought police might be needed “if the person was really out of control” or “had a breakdown” or was “acting out.” Others thought that police might be required in more serious situations such as those involving serious crimes, violence, weapons, and harm to self. “If that person is at the point where they have a weapon . . . sometimes the police are needed” (family 1). Three participants used the word “authority” when referencing the type of skills and presence police bring to such situations.

The issue of ensuring the safety of mental health professionals was also raised with respect to the co-response model. Nine clients (18%) and two family members or friends (11%) expressed concerns regarding the physical danger that mental health professionals could face when responding to a crisis, even with police. One client noted, “The dangers of the social worker getting hurt, in my opinion, outweigh the benefits of getting contact [from a social worker]” (client 10). Some participants suggested that the very value of a co-response was that police would be on hand if needed. As one friend of a client noted, mental health professionals “would also be protected if something, you know, went haywire and the person overreacted” (family 5). These participants valued having two types of professionals respond so that various types of expertise could be deployed as needed.

Concerns With Nonpolice Response

One final theme emerged with respect to the nonpolice response model. Although the nonpolice response model was the most preferred overall, 11 clients (22%) and one family member (6%) indicated that a nonpolice response was their least preferred option. Most of these participants (N=8) felt that the model as described involved too many professionals at the scene. One participant noted it would feel “like an ambush” (client 6), whereas another reflected, “It’s a lot of attention . . . that would be overwhelming for me” (client 11). Furthermore, two participants expressed concern that a nonpolice response could result in hospitalization or the threat of hospitalization, which would further trigger the client. A friend of a client stated, “Some people don’t like talking to doctors. . . . They may not be as open and honest with what’s going on with them because those are professionals that sometimes we look at in a negative light as well. ‘Oh, they’re just here to take me away’” (family 6).

Discussion

This study builds understanding about clients’ and family members’ preferences for crisis response models at a time when novel models are proliferating as an alternative to police response. At least four findings from this study are noteworthy. First, similar to the findings by Boscarato et al. (18), participants desired a calm environment and empathy during crises and described seeking informal supports before contacting formal systems. The rollout of the 988 Suicide & Crisis Lifeline, along with data indicating that 80% of crisis calls can be resolved by phone (7), suggests that raising awareness about low-touch crisis support services could be valuable. Further research is needed to assess how successfully various response options—ranging from informal supports to call, text, and chat support and field-based responses—resolve crisis situations.

Second, the fact that participants described reaching out to informal supports suggests that there is more to learn about the value of peer versus professional support. The quality of interactions with professionals and perceptions about coercion have emerged as important factors influencing how people with mental illnesses experience contact with police or the mental health service system (28–30). Client perspectives should continue to inform how to redesign service systems.

Third, more than half of the participants wanted a nonpolice response when they (or their family member or friend) are experiencing a crisis, and almost one-quarter desired a co-response model. The presence of professionals with mental health expertise in the field was identified as a clear advantage over more distant forms of support from mental health professionals (i.e., via telephone) or from police officers with mental health training. This finding aligns with what advocates have long been saying: “Mental health crises deserve a mental health response” (31). But it also invites further consideration about how communities will develop mental health–led responses. Additionally, systemic inequities are built into the mental health system (32), and it is no coincidence that those with the least access to mental health care—Black and Latinx populations—are among the most adversely affected by disparities in the criminal legal system (33, 34). The implementation of new models will need to be followed closely to understand whether these models reduce criminal legal system involvement and for whom. We will address issues of interpersonal and systemic racism in another analysis that will focus on participants’ reports of their experiences with the criminal legal system.

Fourth, concerns about safety were articulated by participants in weighing the advantages and disadvantages of crisis response models, although these concerns were often defined vaguely. Participants preferred nonpolice responses but also described situations where police might be necessary. Safety is a legitimate concern in some crisis situations, but varied perceptions about safety risk can unduly complicate crisis response situations in which the risk is minimal. Research is needed to define and operationalize what safety means in mental health crisis response so that appropriate responses can be dispatched.

This study had several limitations. First, it did not formally assess clients’ previous experiences with mental health crises, which is an important area for future work. However, it is fair to assume, by virtue of the parent project’s inclusion criteria, that all participants had previous experiences in this area. Second, participants had limited direct experience with crisis response models, and stated model preferences were based on having participants think hypothetically. Engaging in hypothetical thinking or forming expectations about possible outcomes can be difficult (35, 36). Future research should include participants who have more experience with specific models. Third, to facilitate participants’ understanding, we simplified definitions of the four response models. Models are designed and operate in various ways in the field, and models may also operate together in many communities, rather than functioning as discrete entities. Therefore, our findings relate to these models’ prototypes. Fourth, all clients had been arrested on misdemeanor charges, which has the potential to bias views toward police-involved crisis response. Fifth, although we acknowledge that individuals with intellectual or developmental disabilities (37, 38) and dementia (39) are at risk for criminal legal involvement, those groups were excluded from this study, given the nature of the parent project (i.e., the need for a high threshold of capacity to give informed consent for participation in the randomized controlled trial). Sixth, because of COVID-19 pandemic restrictions, all interviews were conducted via videoconferencing, and some participants may have been excluded from participating because they didn’t have active phone service at the time of recruitment. Finally, the sample was relatively small. Although this sample size is appropriate for a qualitative study of this nature, further research with larger samples is needed to enhance the transferability of the findings.

Conclusions

The expansion of crisis response alternatives raises important questions about how to best create a comprehensive and equitable crisis care continuum. The findings reported here highlight the perspectives of people with serious mental illnesses and histories of arrest and of their family and friends. The findings of this study indicate a clear preference for health-centered responses and reveal the practical benefits and concerns about various models, which are relevant for both practitioners and policy makers. Continued research that explores stakeholder perspectives and documents lived experiences will be critical for tracking progress in the field.