Access to non-police crisis response for people who use drugs: lessons from community safety response programs in two U.S. cities

Abstract

Introduction

Community Safety Response (CSR) programs increasingly provide non-police alternatives to behavioral health crises, but their role in addressing substance use calls remains poorly understood. This qualitative study examined how municipal stakeholders in Denver, Colorado and San Francisco, California perceive and operationalize CSR protocols for substance use-related emergencies.

Methods

Between 2021 and 2023, we conducted 82 semi-structured interviews and 32 field observations with 911 dispatchers, law enforcement, city officials, and alternative responders from Denver and San Francisco’s respective CSR programs. Transcripts and field notes were analyzed using thematic analysis to identify attitudes toward substance use, decision-making processes for call deployment, and barriers and facilitators to routing of substance use related calls to alternative response.

Results

Although written protocols permitted non-police response to substance use calls absent violence or weapons, findings revealed a gap between policy and practice. Dispatcher and responder discretion played a central role in determining CSR deployment: while some stakeholders regularly sent alternative responders on substance use calls, others described any mention of drugs —particularly methamphetamines—as automatically triggering police involvement. Barriers to effective CSR response included limited referral options for people who use drugs, fragmented service systems, community tensions around visible drug use, and perceived safety concerns. Facilitators included CSR teams’ unique engagement capacity and potential for peer specialist involvement.

Conclusion

To promote effectiveness of CSR as a public health intervention to reduce morbidity and mortality, programs must address discretionary decision-making, enhance training on stigma and harm reduction principles, and coordinate with broader service systems to ensure equitable access to non-police responses for people who use drugs.

Introduction

Substance use has become an increasingly visible and urgent public health crisis in communities across the United States. The opioid epidemic led to an unprecedented number of overdose deaths over the past two decades, and recent epidemiologic trends in stimulant and polysubstance use have further amplified this burden [1]. Despite recent and encouraging reductions in mortality, overdoses continue to account for tens of thousands of preventable deaths annually, concentrated among racial/ethnic minorities and individuals experiencing poverty, housing insecurity, or mental illness [2, 3]. Moreover, people who use drugs (PWUD) continue to experience elevated levels of drug morbidity (e.g., wounds, hepatitis, HIV, and skin and soft tissue infections), and are at increased risk for unemployment, unstable housing, food insecurity, and exposure to the criminal legal system [4–6]. The intersections between substance use and these social and structural determinants of health have compounded poor health outcomes in vulnerable populations [5].

Traditionally, emergency responses to behavioral health crises, including those involving substance use, have relied heavily on law enforcement. Police often are dispatched to respond to 911 calls about public drug use, suspected intoxication, and overdose emergencies. Yet, officers may lack the training and tools to respond effectively in these situations, particularly when a health- and trauma-informed approach is required [7, 8]. Research has shown that aggressive policing of PWUD contributes to elevated overdose risk by deterring individuals from calling 911 or accessing harm-reduction services due to fear of arrest or other legal consequences [9, 10]. Moreover, criminalization can force drug use into isolated or unsafe environments, increasing the likelihood of fatal outcomes [10]. These risks are compounded in non-white communities, which experience disproportionate rates of drug law enforcement and related harms: nationally, Black people are 3.6 times more likely to be arrested for drug possession than white people, despite similar prevalence of use [11, 12]. Racial disparities in drug enforcement, alongside broader social and structural determinants of health such as housing and healthcare access, are reflected in overdose mortality rates: among Black Americans, overdose deaths increased by 44% between 2019 and 2020, compared to 22% among white Americans [13].

Consequently, there has been a growing call to shift responsibility for behavioral health crises away from law enforcement and toward behavioral health professionals, including peer specialists and community-based care teams [14]. Surveys indicate considerable public appetite shifts away from traditional policing toward diversifying first response and including health-promoting alternatives [15, 16]. Approaches to redirecting individuals away from the criminal legal system toward health and social services vary substantially by degree of law enforcement involvement. Police-led models include deflection, which occurs pre-arrest, and diversion, occurring after arrest or system involvement, which redirect individuals to services before advancing legal action [17]. Co-response models include both police and behavioral health professionals in the response itself to provide a hybrid response that integrates behavioral health expertise while retaining law enforcement presence [14]. In contrast, alternative non-police responses, referred to as Community Safety Response (CSR) programs, aim to eliminate police involvement entirely from eligible crisis calls where possible. CSR first emerged in 1989 with the launch of CAHOOTS in Eugene, Oregon, and have since expanded rapidly; as of 2024, more than 100 cities have been implemented or piloted in the U.S [18]. These programs vary in structure and scope but commonly include clinicians, social workers, emergency medical technicians (EMT), and/or peer support specialists who respond to 911 calls without law enforcement. However, their sustainability remains uncertain: in 2025, the elimination of CAHOOTS was announced due to funding challenges [19]. Moreover, while evidence on outcomes for police-involved models is accumulating [20], research on the impacts of non-police crisis response on community health outcomes for PWUD remains nascent [21].

An estimated half of individuals with mental illness also engage in illicit substance use, and co-morbidities between serious mental illness and substance use disorders are substantial and well-documented [22, 23]. Given this, programs responding to mental health crises will likely often encounter substance use, but CSR responses to drug-related calls are not well characterized in the literature. Integrating drug-related calls into CSR remains a complex and often contested domain [8, 24], as they can present unique challenges that influence whether calls are diverted entirely from law enforcement’s purview and what kinds of teams, skills, and resources are needed (or perceived to be needed) on the scene. Perceived risks of health complications (e.g., overdose) or of intoxicated individuals as potentially threatening may complicate the likelihood of call diversion, responder safety assessments, and care planning. Additionally, given the possible presence of illicit drugs or activity at the scene, jurisdictions may be reluctant not to have law enforcement respond in some capacity. The extent to which substance use-related calls are referred to CSR programs, and what parameters guide these decisions, remains poorly understood. Many studies focus broadly on mental health crisis response, but to understand the full potential for CSR to reduce morbidity and mortality among those with behavioral health conditions there is a need for research to explore CSR eligibility, triage, and response processes among PWUD [25].

This study addresses this gap by examining how municipal actors perceive and operationalize protocols and policies related substance use calls within their CSR programs in Denver, CO and San Francisco, CA. Overdose fatalities peaked in both cities in 2023 before exhibiting slight declines in 2024, consistent with national trends [26]. In Denver, the estimated number of people experiencing homelessness increased a staggering 87% rise since 2019 [27]. San Francisco has similarly seen increases in overall and especially youth homelessness in recent years, with recent data showing an 11% increase in chronic homelessness since 2022 [28]. Leadership in both jurisdictions have committed to addressing the intersecting crises of mental health, homelessness and substance use, and the degree to which police should be involved in this effort remains politically contentious [29].

In 2020, both Denver and San Francisco launched CSR programs that vary in infrastructure and scope but share the objective of reducing law enforcement’s footprint in health-related crisis response. We utilize qualitative data collected from 2021 to 2023 as part of the Reimagining Public Safety Initiative of the Policing Project at New York University School of Law, which partnered with several U.S. cities to examine the design and early implementation of these emerging community safety response models.

Materials and methods

Study design and setting

The current analysis draws on in-depth interviews and field observations collected in Denver, Colorado and San Francisco, California between December 1, 2021 and September 1, 2023. While cities in both programs share the goal of reducing law enforcement’s role in crisis response and use 911 dispatch diversion as the primary routing mechanism: important ways, the program in Denver operates as a single generalist team responding to a range of nonviolent crises, whereas San Francisco developed distinct program units by crisis type, including a dedicated overdose response team. Programs also differ in staffing composition, institutional governance, and operational hours. These sites were not selected to enable systematic cross-site comparison; rather, they are treated as two contemporaneous examples of an emerging model, examined together to identify common implementation patterns and challenges related to substance use calls. Both cities have a high burden of fatal and non-fatal overdose and drug markets characterized by synthetic opioids (e.g., fentanyl) and methamphetamine, with increasing rates of polysubstance use [30–32].

Denver launched the Support Team Assisted Response (STAR) program in 2020, which sends a clinician–paramedic team to nonviolent crisis calls instead of police, including some involving substance-related issues. Calls are diverted from police to STAR through Denver’s emergency communications center. STAR responders include a licensed mental health clinician from the Mental Health Center of Denver and a paramedic from Denver Health. STAR was fully implemented in six neighborhoods from 10am-6pm Monday through Friday and expanded operational hours to 6am-10pm during the period of data collection. Between 2020 and 2023, dispatchers flagged 38,375 calls as eligible for STAR [33]. STAR responded to roughly a quarter of eligible calls during this time, though capacity was noted to be increasing throughout the study period [33].

San Francisco has different forms of CSR programs organized by crisis type. The Street Crisis Response Team (SCRT), launched in 2020, responds to behavioral health crises through the 911 system using Emergency Medical Dispatch protocols, which direct eligible calls to crisis teams instead of police. SCRT units are composed of a community paramedic (i.e., paramedics with primary care certification enabling them to provide a broader scope of services, including mental health and harm reduction), an EMT, and either a peer counselor or a member of the Homeless Outreach Team. From 2020 to 2022, SCRT responded to roughly 43% of the 15,065 calls flagged as eligible [34]. The Street Overdose Response Team (SORT) likewise is dispatched through 911 and responds specifically to overdose-related calls with a team that includes a community paramedic, EMT, and harm-reduction or peer support staff. SORT also operates as a non-police alternative, providing on-scene medical and harm reduction support and conducting follow-up, with police engaged only when there is a clear safety threat.

Recruitment and study sample

We conducted 36 semi-structured interviews and 9 field observations in Denver, and 46 interviews and 23 field observations in San Francisco (Table 1). Stakeholders involved in crisis call triage, dispatch, and response purposively sampled to capture diverse institutional perspectives across local public safety systems were eligible for participation. They included 911/dispatch staff (call-takers, dispatchers, supervisors, and managers), law enforcement (line officers, supervisors, and command staff), and city officials (elected officials, appointed administrative staff, one staff member at a non-profit that funded CSR programs). CSR staff included both frontline responders who attended calls and supervisors or program managers responsible for coordination, training, and operations. Individual demographic data were not collected, as our primary unit of analysis was job role rather than individual-level characteristics. Participants were identified through direct outreach to local government agencies and city-affiliated organizations, referrals from program administrators, and snowball sampling initiated through established program contacts. We recruited a mix of employees across different shifts and varying levels of tenure to capture a range of perspectives within each agency, and all participants whom we approached in this manner accepted to participate. In San Francisco, participation was facilitated by contacts within the Department of Emergency Management, public health services, and program-specific teams such as SCRT. In Denver, stakeholders were drawn from STAR program leadership, police department command staff, and behavioral health partners.

Table 1.

Sample sizes for field observations and interviews conducted with public safety stakeholders in Denver, CO and San Francisco, CA

	Stakeholder	Interviews	Field observation sessions
Denver (N=45)
	911 dispatch	13	4
	Law enforcement	9	2
	Community safety response	7	3
	City and program officials	7	-
San Francisco (N=69)
	911 dispatch	16	6
	Law enforcement	18	4
	Community safety response	3	13
	City and program officials	9	-

Eligible participants were those over the age of 18 with operational, supervisory, or policy-level knowledge of their city’s CSR design and implementation who could complete interviews in English. The study was approved by the Institutional Review Boards (IRB) at NYU (FY2022-6203) and Suffolk University (1820359-1).

Data collection

Interview guides were developed collaboratively by a multidisciplinary team of public health, legal, and policy scholars at NYU’s Policing Project and Suffolk University. Guides were tailored by city and stakeholder type to elicit perspectives on the design and implementation of CSRs, including processes of 911 triage and dispatch, interagency coordination, and stakeholder experiences with behavioral health and substance use crises. Data collection included semi-structured interviews with municipal actors and structured field observations of CSR operations in both cities. Interviews were semi-structured and organized around three core thematic domains: familiarity with and goals of alternative response policy changes; triage and response practices, including call eligibility, dispatch decision-making, interagency coordination, and experiences with specific call types; and perceptions of public safety. Protocols were tailored by stakeholder group (alternative responders, dispatch staff, law enforcement, and government officials) with approximately 8–12 core questions per protocol plus follow-up probes. Interviews occurred in person or on Zoom, lasted approximately 60 min and were audio-recorded with participant consent.

Field observations in Denver and San Francisco consisted of ride-alongs and dispatch sessions during two week-long site visits embedded with the respective CSR programs and 911 operations in each city. Observations lasted between 2 and 4 h and were conducted by between 2 and 3 researchers who took handwritten notes in real time. Following each observation team members recorded individual voice memos elaborating on their observations and field impressions. Ride-alongs with on-duty personnel in police vehicles or alternative response vans took place in precincts identified in advance by our site liaisons in San Francisco and Denver. Downtown precincts, including the Tenderloin in San Francisco and central downtown areas of Denver, were prioritized, as were areas of each city with larger unhoused populations. Dispatch sessions were conducted either inside 911 call centers, where the research team used headsets connected to call-takers’ and dispatchers’ workstations to listen to live 911 calls in real time, or by attending CSR response units in the field. Team members engaged in informal conversation with personnel about the nature of their work, CSR implementation, and the barriers they encountered in practice.

Written IRB-approved consent forms were used for interviews; verbal IRB-approved consent was obtained for field observations, where the informal and fast-moving nature of operational settings made written consent infeasible. Participants did not receive reimbursement, as the vast majority were municipal employees not permitted to accept compensation. All interviews and field notes were transcribed using Rev.com and coded together under a single codebook, reflecting the consistency of interview protocols across both sites. All data were stored within a secure cloud environment per IRB protocols and tagged with site identifiers. Coding was conducted in Dedoose v.9, which permitted filtering and sorting by site during systematic qualitative analysis. Given the integrated coding approach across stakeholders and sites, the term “participants” refers collectively to interview respondents and individuals observed throughout; where appropriate, findings are attributed to specific roles (e.g., dispatchers, patrol officers, CSR responders, city officials) to distinguish perspectives.

Data analysis

Data were analyzed using a thematic analysis approach [35]. A codebook was using inductively from the data and deductively [36] based on study objectives and prior literature on crisis response alternatives and harm reduction framework. Transcripts and field notes were coded in Dedoose by a team of researchers including members with backgrounds in public health, sociology, criminal justice and community-based crisis response. Principal investigators of the study, along with two trained research assistants, applied a preliminary codebook to a subset of transcripts. The team met weekly to compare coding decisions, resolve discrepancies through discussion, and reach consensus before proceeding. Research assistants then individually coded the remaining transcripts while continuing to meet the team met weekly to review coding decisions, discuss emergent themes, resolve any inconsistencies across coders, and ensure analytic consistency. Comparative analysis was conducted across cities and stakeholder groups to explore variation in experiences, perceptions, and system-level implementation dynamics. For this analysis, the coded data were specifically analyzed to examine attitudes towards substance use among stakeholders, perceptions of CSR’s role in dealing with substance use, mechanisms for including or excluding substance use calls from CSR response, and facilitators and barriers to implementation of CSR to address substance use crises.

Results

Attitudes towards the substance use crisis

Across both Denver and San Francisco, frontline public safety personnel expressed frustration regarding the overwhelming scale of public substance use, the perceived inadequacy of existing systems and resources, and the belief that police are ill-equipped to manage the medical, behavioral, and social complexities of substance use. Participants in both cities also spoke about substance use as deeply intertwined with the housing crisis, perceiving homelessness and public drug use as a dual crisis to which they did not have the tools or resources to effectively respond. Perspectives in this section reflect primarily law enforcement, dispatch, and municipal stakeholders, as CSR responders’ interview guides were oriented toward operational and implementation questions.

Police officers in Denver expressed frustration with existing methods of drug enforcement, which they perceived to be ineffective and burdensome to officers. Some described exercising discretion not to issue citations, citing futility of traditional enforcement tactics. However, they also expressed skepticism about alternative models such as decriminalization and reduced enforcement, linking them to increased public drug use. One patrol officer noted, “Possession is just a misdemeanor now… that’s why you see people on the street corner smoking drugs right in front of cops.” Across various roles, participants commonly highlighted that officers are ill-equipped for the complex social, behavioral, and medical aspects of substance use crises. One patrol officer lamented, “We’re probably not very effective at treating these individuals,” while a police leader said, “I can’t imagine we get many… women and men in our police academy that say, ‘Yeah, I’d like to arrest the same person for shooting up heroin in the park every single day.” Some city stakeholders framed substance use within a broader context of systemic failures, pointing to chronic underinvestment in mental health and substance use services. As one city official put it, “We, as a society, have not done a good job with mental health and substance misuse… and so I think the public safety part of this is saying like, okay, what’s our role?” Together, data from interviews and field notes in Denver illustrated frustration with traditional enforcement related to the substance use crisis and a willingness to explore alternative policies.

Data from San Francisco revealed a more contradictory set of attitudes among frontline public safety personnel regarding substance use and their role in addressing it. Many responders acknowledged the increasing normalization of public drug use, particularly among unhoused populations, while patrol officers cited a lack of deterrence due to shifts in city policy. One patrol officer described, “They’re not even discreet about it anymore… I think there’s just a lack of belief in police, law enforcement.” Frontline responders and officers described the cyclical nature of encounters with PWUD, many of whom they perceived to be resistant to services. While some support was expressed by those involved in direct medical or crisis response, other stakeholders including 911 dispatchers, city officials and police were skeptical or outright critical, with one 911 official saying: “We’re tellin’ people, ‘We don’t want you to do drugs,’ but yet we’re gonna provide you with the needles to do it and the Narcan to bring you back if you do overdose. We’re enabling them to do that.” Notes from field observations corroborated this sentiment among dispatchers and police: “Most of the folks, both in dispatch and police that we’ve talked to mentioned this idea of city-funded facilities to do drugs as being, sort of, ridiculous and absurd in their eyes.”

Factors determining inclusion or exclusion of substance use calls from CSR

Across Denver and San Francisco, decisions related to dispatching CSR programs to substance use calls were related more to stakeholders’ perceptions of risk and resource availability than to strict adherence to written protocols. Both cities maintained explicit eligibility criteria that permitted non-police response for substance use calls absent the threat of violence or acute medical emergencies. Official protocols for Denver’s STAR program included calls involving mental health, substance use, or homelessness (e.g., citizen assists, intoxicated persons, suicidal persons, welfare checks, indecent exposure, trespassing, and syringe disposal), provided there was no indication of violence, weapons, or acute medical emergencies requiring hospital transport. Similarly, San Francisco’s protocols explicitly permitted or recommended SORT and SCRT dispatch for substance-related calls when no immediate violence or weapons were reported. Exclusions of PWUD from CSR response operated at multiple levels: at the point of dispatch, where calls were redirected to police based on dispatcher interpretation of substance use as a disqualifying factor regardless of written protocol; at the point of response, where CSR responders made independent safety assessments that resulted in police involvement based on perceived substance-specific risk; and downstream, at the level of treatment facilities, whose admission criteria further limited referral options for PWUD even when CSR response had been successfully initiated.

Dispatcher discretion

In practice, a mix of perceived safety threats and logistical realities shaped whether community safety responders were dispatched instead of police. Denver’s single-team model relied heavily on dispatcher and responder discretion to navigate ambiguous situations, which one 911 Operator described as “reading the call between the lines.” San Francisco’s multi-team structure, with SORT handling acute drug issues and SCRT addressing behavioral crises, provided more formalized pathways for inclusion of PWUD but still depended on discretionary decisions at the point of dispatch. The following sections detail how discretion created opportunities for subjective concerns and assumptions about drug use and PWUD to influence deployment decisions, sometimes resulting in exclusions that contradicted official protocols.

Substance use as a basis for service exclusion

Despite formal policies permitting response to substance use calls, field observations revealed that any indication of substance use—even cannabis—was sometimes sufficient to trigger police involvement. In Denver, there was variability among both 911 dispatch and responders (STAR dispatch) in facilitating CSR dispatch to substance-use related calls. While some responders clearly stated that responding to ‘intoxicated’ individuals was a routine affair, others expressed reticence and tended to send police as cover even without a direct impression of violent threat. One community safety responder expressed confidence in responding to individuals managing substance use issues, describing various avenues for alternative response to triage and support their needs:

STAR responds to intoxicated individuals all the time… it’s the same as every other scenario. They need to assess whether this individual is at a level of intoxication, then they need to be transported to detox, or if it’s to a medical level that they need to get to a hospital. And at that point, then the clinician and paramedic can absolutely get on the radio and request for the appropriate party to respond. If the individual is just somewhat intoxicated, but is still able, you know, to manage their ADLs [activities of daily living] and take care of themselves, then STAR can find a solution there. If it’s more intense, then they might assess and then get it to the appropriate party.

One 911 operator said that the STAR program could “100% deal with that [substance use],” and in alignment with protocol added that “if someone is on substances and is now punching people in the face, that would maybe be more of a co-responder police response, rather than a STAR response.” However, dispatchers and responders also described situations where any mention of substance use was an automatic reason to either include police cover (i.e., enact a co-response model) or simply direct the call to only the police. Examples included a welfare check where cannabis use led to deployment of police to cover for STAR despite evident mental health needs, as described by one 911 operator: “Because the man had been using marijuana… the dispatcher sent the police to cover STAR… even though on this call, there was clearly, there were some mental health issues going on.” One STAR dispatcher explicitly articulated an exclusion related to substance use: “There are guidelines… she doesn’t send STAR to calls when there are weapons involved, when there is drug use involved, or when there is intoxication involved.” This contradicted STAR’s formal policy, which does not prohibit all substance use-related calls absent violence or weapons. Similarly, field notes documented instances where calls were held from STAR simply due to suspected drug involvement: “Not sending STAR van because poss[ible] meth involved.” When questioned about this decision, responders cited potential safety concerns and the unpredictability associated with methamphetamine use, even in the absence of overt threats of violence.

In contrast, sending CSR-eligible calls to police was observed less frequently in San Francisco, whose SORT team specialized in dealing with intoxication and in harm reduction practices. For example, a SORT responder described their role as: “Going out to people who are using, making sure they’re alive, offering Narcan and linking them to services.” Accounts of 911 dispatchers and field notes illustrated that SORT responders frequently assigned themselves to drug related calls, and that drug use alone without accompanying aggression often did result in an alternative non-police response in this setting. A peer responder stated: “We’ll still go if they’re high, as long as no one’s being hurt – the point is to keep them alive and connected.” Moreover, multiple participants stated that police themselves often called SORT or SCRT to address cases with suspected intoxication or overdose. In one instance, a police officer recounted a case where an individual reported taking ‘some pills’ and was ‘getting too wild,’ resulting in a decision to involve SCRT to deescalate the situation and ensure everybody’s safety.

Perceived safety concerns

Perceptions of certain substances resulted in heightened fears about safety even when overt threats of violence were absent. In both cities, methamphetamine use was consistently perceived as distinct from other substances and was implicated in decisions to deploy police rather than a CSR program. This extended beyond situations involving actual aggression, illustrating perceptions and assumptions about potential unpredictability of those who use methamphetamines.

In Denver, a STAR clinician’s recommendation during roll call exemplified these attitudes: “We could send STAR, but he likes meth too much,” recommending that the individual be picked up on an outstanding warrant to “sober up in jail” rather than receiving alternative services. Similarly, responders made distinctions based on perceived threat levels associated with different substances, with one explaining that suspected methamphetamine use could indicate “unpredictability or potential drug dealing,” which were considered safety concerns warranting police involvement. Similar concerns were heard in San Francisco with an SCRT responder stating: “With meth, you can’t predict,” though practices generally aligned more closely with formal protocols; rather than categorically refusing these calls, this individual noted, “I’ll ask for police cover if things feel unsafe.”

Barriers/facilitators to integration of substance use calls into CSR response

Across both cities, stakeholders described how broader systemic and operational constraints affected all crisis calls, and by extension, substance use-related incidents. CSR programs in both Denver and San Francisco highlighted broader structural issues, including shortage of capacity at treatment and psychiatric facilities, coordination gaps between agencies and staffing limitations, as limiting their capacity to effectively respond to substance use calls. However, participants also identified opportunities to improve substance use call responses through better service integration and inclusion of peer voices in program design.

Facility exclusions and service fragmentation

In Denver, participants identified a critical paradox: treatment facility exclusions related to drug use limited referral options for PWUD. Mental health treatment facilities and residential programs often require clients to test negative for illicit and even licit substances (cannabis, medication for opioid use disorder, alcohol) as a condition of admission, and even some substance use treatment settings may suspend intake for individuals presenting above a certain intoxication threshold, creating barriers to care linkage during substance use calls. In one case, a STAR responder learned that a client had used methamphetamine the night before, prompting the realization that the facility they planned to link them to wouldn’t “accept anyone who might test positive,” and forcing a shift to a day shelter instead. As one co-responder explained: “We can’t arrest our way out of homelessness and drug use, but we also can’t refer our way out of it.” Responders suggested that, in such cases, they may be able to stabilize an immediate crisis but lack viable referral pathways for sustained care. Fragmented operations between alternative response teams and substance use treatment providers added to this problem, leading to inconsistent handoffs during substance use calls. Denver participants described disconnected systems between STAR, detox centers, and treatment providers that complicated follow-up care for people encountered during substance use calls. A city official noted: “We have these different responses sometimes falling under different organizations, which… further complicates it particularly for our community.” Similar coordination and communication barriers were mentioned in San Francisco, highlighting a broader systemic fragmentation. In San Francisco, siloed service systems and lack of interagency communication hindered effective response to substance use calls. One SCRT clinician likened the system to “a Frankenstein composition” in which “different body parts [are] not talking to each other;” this is particularly problematic when coordinating care for people with substance use needs and other co-morbid conditions.

Lastly, participants also described scenarios where individuals with suspected substance use did not get STAR response simply due to the lack of STAR’s availability, which resulted in police stepping in. Though this did not constitute a substance-use specific exclusion, it represents an additional avenue through which PWUD may continue to encounter police instead of alternative forms of response.

Community expectations

Stakeholders in San Francisco described community tensions specifically related to visible substance use that could complicate response decisions. One 911 operator recounted one caller asking for help during substance use emergencies but pleading, “Don’t shoot him… he’s not hurting anybody,” while another described being criticized for insufficient action regarding drug use in public spaces. In one scenario, an operator discussed a conflicted caller who expressed guilt about placing a call related to substance use, while also wanting to make the problem less disruptive to his immediate environment. The operator recalled the man, who was crying, as saying “I’m just trying to get to work…And I’m not a bad person. Like I understand like these people are hurting and everything…But I am just at my wit’s end and I don’t know what to do.” An alternative responder observed: “They [callers] want to be seen as compassionate, but they also want people out of their yards.” This was viewed as especially challenging in light of perceived resistance on the part of PWUD to engage to referral and treatment, for example, refuse medications for opioid use disorder due to concerns about side effects and withdrawal-like symptoms. One responder expressed frustration with refusal of services: “It’s such a bitch when they don’t want help.”

Facilitators and recommendations for effective CSR response to substance use calls

Despite these challenges, participants in both cities recognized CSR programs’ distinct capacity to engage people in substance use calls. A Denver patrol officer emphasized STAR’s unique ability to build connections with PWUD, recalling a case where STAR “developed an immediate rapport” with someone high on methamphetamine—an interaction he believed could not have navigated as effectively through traditional police response. Denver stakeholders saw particular promise in incorporating lived experience of substance use into CSR programs; one described efforts to explore incorporating peer specialists into STAR teams specifically for substance use calls, recognizing that people with lived experience of drug use could provide unique insights and engagement capabilities during these encounters. These findings suggest that while CSR programs face barriers in responding to substance use calls—particularly around treatment access and system coordination—their unique engagement capacity and potential for peer involvement represent important facilitators for improving non-police responses to PWUD.

Discussion

This study examined how CSR staff and municipal stakeholders perceive and operationalize community safety response programs to substance use calls in Denver and San Francisco. Our findings reveal a disconnect between the potential of CSRs and their implementation for people who use drugs, with implications for equity in crisis response and the broader goals of police reform.

Although participants acknowledged that written protocols explicitly permit alternative responders to be sent in lieu of police for substance use calls in the absence of violence or weapons, findings illustrate that dispatch of CSR relies heavily on dispatcher and responder discretion. Research has detailed the considerable role of discretion on the part of criminal legal actors—including police officers, prosecutors, and judges—in determining legal outcomes for PWUD [37–40]. In these studies, actors exercise discretion in determining whether to confiscate tools for risk reduction (e.g., sterile syringes), to arrest an individual, what charges to bring, and length of sentences. Scholars have argued that this discretion is to some extent inherent and necessary for individuals in these roles, but that it poses a very real risk of creating and reproducing inequities along a variety of racial/ethnic and socioeconomic lines [37]. Here, we illustrate that discretion is also a factor in access to non-police crisis response, an intervention which has a stated aim of reducing the negative health and social consequences of police contact. Specifically, data demonstrate variation in access to CSR among PWUD, with some hesitancy among both 911 operators and community safety responders to dispatch CSR to substance-related calls. This may lead to increased likelihood of encountering police among PWUD in crisis, even without overt threats of violence, creating a risk that CSR programs perpetuate some of the health inequities they aim to address.

The finding that for some dispatchers, any indication of substance use—including cannabis—could trigger police involvement highlights how the enduring criminalization and stigmatization of drug use may override and undermine crisis response protocols and harm reduction principles. It is important to note that the present study data was collected early on in the implementation of these CSR programs (within the first year), so some of these findings may be due to the need for the gradual adaption to change among those implementing novel and alternative models. However, research documents pervasive stigma toward people who use drugs among criminal justice professionals, including police officers, prosecutors, judges, and correctional staff who view drug-involved offenders as deserving of punishment rather than treatment [41, 42]. There is also evidence of stigma towards PWUD in the healthcare sector, and studies show how this may compromise their access to quality interventions, delay care, and result in worse health outcomes [43, 44]. Among first responders, research demonstrates varied levels of support for harm reduction, with some holding negative attitudes toward naloxone use and medication for opioid use disorder which are evidence-based interventions to reduce mortality and promote recovery [45, 46]. Qualitative research with emergency response personnel (medical providers, firefighters and police officers) suggests that the likelihood of patients’ acceptance of medical care, which can determine their survival, may depend in part on the level of compassion of responders [46]. Our research adds to this literature by demonstrating varied levels of support for harm reduction among stakeholders designing and implementing alternative models of emergency response.

Although some participants utilized stigmatizing language about drugs or described a blanket practice of denying non-police response to PWUD, others referred specifically to concerns about responder safety in the presence of drugs, describing a more nuanced case-by-case practice of alerting police if perceived a threat. The extent to which responder safety concerns and experiences differ across alternatives to traditional policing, such as co-response versus fully non-police CSR models, represents an important avenue for future research. Pronounced concerns related to methamphetamine use were observed across both cities. While there is some evidence that agitation and aggression can be increased among those with chronic methamphetamine use (or withdrawal), this also applies to individuals presenting with other factors, such as psychosis or alcohol use [47] which did not arise in the data as descriptions that would trigger automatic exclusion from services. However, it remains unclear whether these heightened concerns around methamphetamine reflect stigma toward people who use stimulants or stem from responders’ prior experiences of aggression or unpredictability in such encounters. Nonetheless these findings add to a growing body of international literature highlighting concerns related to responding to methamphetamine-related events among first responders [48–50]; specifically, studies highlight fear and compassion fatigue expressed by these actors due to the lack of standardized algorithms and protocols for response, and insufficient resources and support during and following these interactions. This is occurring within a context of increasing levels of stimulant-involved overdoses, particularly those involving methamphetamine: in the US, nearly 60,000 overdose deaths involved cocaine and/or psychostimulants (mostly methamphetamine) in 2023, representing about 57% of all overdose deaths that year [6], Relative to opioid use, stimulant use disorder and overdose present complex clinical challenges: there are no overdose reversal medications (equivalent to naloxone) or efficacious medications for methamphetamine use disorder (equivalent to methadone or buprenorphine), and research on evidence-based interventions remains limited. The need to establish guidelines for both clinical and emergency response personnel in the face of this growing issue is critical for promoting responder safety and achieving just and equitable outcomes for PWUD in crisis.

Stakeholders in both cities also emphasized the inextricable link between the substance use and housing crises, describing encounters with the same individuals repeatedly in public spaces. The community frustrations observed in San Francisco—wanting to appear compassionate while also wanting people “out of their yards”—reflect the broader tension between addressing these issues through criminalization and forced removal instead of comprehensive social policy (e.g., meaningful investment in affordable housing). Responders operating within this context face the impossible task of addressing individuals in crisis, often repeatedly, while broader structural barriers to their circumstances remain unchanged. The ‘not in my back yard’ (NIMBY) sentiment is an established barrier to the implementation of treatment and harm reduction programs in the United States and elsewhere [51, 52]. More recently, evidence of community opposition and NIMBYism related to visible drug use and homelessness has emerged as a challenge to sustainability of criminal justice reforms such as decriminalization of low-level drug possession [53, 54]. Data presented here highlight a possible risk that community frustration with the toll of homelessness and an insufficient social service landscape will similarly be attributed to CSR and non-police approaches, introducing a possible threat to their acceptability and sustainability. In contrast, an interesting observation in San Francisco was that data from field notes and interviews alike illustrated situations in which the police themselves called upon or diverted substance use cases to alternative responders. There is literature highlighting frustrations among police officers themselves at the burden of drug and mental health calls that their departments currently face, and a desire to share that workload with stakeholders who have more professionalized training in these areas [54, 55]. Research further indicates that leadership and supervisor support are key facilitators of officers’ willingness to use non-arrest alternatives and engage in harm reduction practices with PWUD, and that police culture, in the absence of such championship, remains the primary barrier [56, 57]. Considering that police departments and unions often wield disproportionate power in shaping and sustaining criminal justice policy reforms, understanding the impacts of a robust non-police response network on police officers’ workload and perceived role in public health and safety response represents an important avenue for future research.

Lastly, the underfunded and fragmented health and social service ecosystem was seen as constraining what CSR can achieve. Treatment facility exclusions especially may threaten CSR effectiveness for PWUD should responders be unable to refer to treatment due to strict abstinence requirements. CSR programs cannot fulfill the expectations placed upon them to address public behavioral health crises and address the shortcomings of police response if they are perpetually constrained by systemic problems rooted in underfunded social services, housing inequality, and drug criminalization. Homelessness, for example, is both a driver of the public crises CSR programs are called to address and a product of longstanding policy failures like chronic underinvestment in affordable housing and the use of encampment sweeps that displace rather than stabilize PWUD [58]. The high prevalence of homelessness encountered by CSR teams also has direct implications for program operations: without access to stable housing as a referral destination, responders are left to manage acute crises without the ability to address their underlying drivers, likely contributing to repeat encounters with the same individuals and limiting the sustained impact of any single intervention. Therefore, though CRS represent an important step toward more humane crisis response, they are not sufficient substitutes for comprehensive policy changes addressing the root social and structural determinants of mental health and substance use.

Our findings have implications for improving CSR programs’ response to substance use calls. There is a need for comprehensive training for dispatchers and responders on harm reduction principles, stigma reduction, and evidence-based approaches to responding to substance use calls. Dispatch protocols in both cities stated that substance use calls were eligible for CSR, but findings suggest that that guidance was not being consistently followed. As with other criminal legal actors (police, prosecutors, judges), dispatcher discretion is to some extent unavoidable, but findings suggest that measures to reduce the possible influence of biases about drug use on referrals may be needed. However, there is a tension in how dispatcher discretion should be addressed: while emerging national guidance has emphasized the need for standardized call-taker training on behavioral health and substance use [59] research from this project has also shown that rigid, formulaic protocols not attuned to the complexity of real-world crisis calls may themselves compromise access to and quality of CSR [34]. A combination of protocols which include input from frontline workers and communities with additional forms of tailored training and support for dispatchers addressing how to root discretionary decisions in best practices could strengthen CSR services for PWUD. Recent initiatives focusing on integrating harm reduction, trauma-informed care, and public health in police response and co-response models have yielded training materials that may be adapted for individuals involved in dispatch and CSR specifically [14, 60]. The potential for peer involvement identified by Denver stakeholders represents a critical opportunity to address stigma and improve engagement with PWUD. Peers with lived experience of substance use can provide unique insights into effective engagement strategies and serve as powerful advocates within alternative response systems, though such integration must include adequate compensation, training, and institutional support [61]. The finding that CSR-eligible calls were less frequently routed to police in San Franciso may indicate that having a separate, specialized substance use and overdose response unit is a strength. However, the level of comorbidity between serious mental illness and substance use disorder precludes separate units from operating in isolation and ultimately, guidance on dealing with substance use and mental health crises should be provided to both groups. Finally, addressing treatment facility exclusions at the local level is an actionable policy recommendation that may help facilitate warm hand-offs between CSR and evidence-based care.

This study had several limitations. First, the original data collection instrument was not designed with substance use call routing as a primary focus, possibly limiting the depth of detail available on this topic. Further research should be explicitly designed to query the question of PWUD inclusion in CSR models. As with any purposive sample, participants may not fully represent the range of views held within their agencies: those who agreed to participate may differ from non-participants in their openness to or support for CSR models. Additionally, we did not collect demographic data or information on participants’ tenure in their roles, limiting our ability to assess how these factors may have shaped perspectives or to compare participants to those who were not recruited. Further, data were collected during an early phase of CSR implementation and findings may not fully reflect how these programs evolve over time as policies, staffing, and funding change. We therefore cannot know from this study whether barriers such as training gaps, protocol inconsistencies, and interagency coordination, reflect transitional issues or more structural and persistent features of crisis response systems (which stigma toward PWUD, treatment facility exclusion criteria, and housing-related barriers likely are). Distinguishing between these will require longitudinal research as programs mature. The study was conducted in two large urban contexts. Generalizability to smaller, more rural jurisdictions, or conversely larger or more populous settings, is limited, given differences behavioral health resources, public safety infrastructure, population density and crime. Lastly, research has documented that demographic characteristics of people in crisis may influence whether 911 is called, that dispatcher framing of calls can shape the type of response deployed and responder expectations upon arrival, and that neighborhood-level indicators that proxy for race and class may influence dispatch decision-making, suggesting that inequities may be embedded throughout the crisis response process in ways our data were not designed to [62–65]. Future research should incorporate the perspectives of PWUD and community members, systematically assess racial and structural inequities in CSR design and implementation and how they may interact with substance use call response, and evaluate outcomes across a wider range of geographic and resource contexts to strengthen the evidence base for non-police crisis response.

Conclusions

Findings reported here are novel and timely. Overlapping crises of mental illness, substance use disorders, and housing instability result in the occurrence of highly visible behavioral health emergencies that warrant responses that balance safety with compassionate and evidence-based care. CSR programs are rapidly proliferating rapidly nationwide as a promising strategy to achieve this balance. However, this study demonstrates that during early implementation CSR programs’ potential to serve PWUD remains unrealized. Key barriers include dispatcher discretion, exclusion criteria operating at the level of dispatch, response, and treatment facilities, and structural challenges, including housing insecurity and drug criminalization, that no crisis response program can address alone. Should some of the issues highlighted here go unchecked, they risk mirroring or reinforcing patterns of discrimination documented in healthcare, housing, and social services. For example, if treatment and housing programs reject people who test positive for substances, and community safety responders avoid calls involving drug use due to perceived safety concerns, this compounds an ongoing system that maintains criminalization as the primary response for substance use among those most vulnerable. For alternative community response programs to fulfill their promise of reducing police involvement in behavioral health crises, they must actively address bias in decision-making processes, provide community safety responders the necessary training and continued support to provide informed and effective responses, invest in peer and community involvement, and coordinate with broader service systems to ensure inclusive access.

Abbreviations

CSR

Community Safety Response Programs

PWUD

People who use drugs

STAR

Support Team Assisted Response (Denver)

SCRT

Street Crisis Response Team (San Francisco)

SORT

Street Overdose Response Team (San Francisco)

Authors’ contributions

BF and JG obtained funding for, conceptualized and executed the parent study, including data collection and primary data analysis. SR and TT obtained funding for and conceptualized and led the current analysis, for which KN and MW conducted data management, coding and analysis. SR and KN drafted the manuscript. BF, JG and TT provided critical feedback on study design and manuscript revisions. All authors reviewed and approved the final manuscript.

Funding

This work was supported by the Microsoft Justice Initiative and an NYU School of Global Public Health’s Cross-Center Collaboration Pilot Grant. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to the terms of the informed consent agreements and the Institutional Review Board approvals, which restrict public data sharing to protect participant anonymity. However, de-identified summaries or specific coding frameworks are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Institutional Review Boards (IRB) at NYU (FY2022-6203) and Suffolk University (1820359-1). All participants provided informed consent prior to participation in interviews and field observations, and study procedures adhered to the Declaration of Helsinki.

Consent for publication

All participants were informed during the consenting process that the data collected would be used for academic publication. Any specific quotes or case examples used in this manuscript have been anonymized to ensure that individual stakeholders cannot be identified by their specific responses. All authors and relevant institutional authorities have reviewed the manuscript and consent to its publication in its current form.

Competing interests

The authors declare no competing interests.