Mobile crisis effectiveness: a systematic review and associated functions and forms framework

Abstract

Background

Increased demand for mental health crisis services and falling supply of inpatient psychiatric beds in the United States (U.S.) has led to heavy investment in mobile crisis units (MCUs), teams of mental health professionals that de-escalate crises in the community and aim to prevent emergency department visits, arrests, self-harm, etc. Despite nearly two thousand such services being active in the U.S., the last review of MCUs, conducted in 1995, found no evidence of their effectiveness. It is not known if newer research validates the effectiveness of this widely used intervention, nor how aligned the goals and functions of the many MCUs are. This article systematically reviews the literature since 2000 to determine the effectiveness of MCUs in the U.S. and create a theoretical framework of their functions and forms to aid future research efforts.

Methods

A systematic review was conducted by using a keyword search in PubMed, PsycINFO, and Scopus. Articles were included through June 2025 and excluded if they were published before 2000 or were not from the U.S. Quality of included articles was assessed using the Mixed Method Appraisal Tool. The Functions and Forms Framework for describing complex healthcare interventions was used to identify the core elements of MCUs and design a theoretical framework for their functions.

Results

Of the 134 studies included in title/abstract screening, nine met inclusion criteria and were found to be medium or high quality. One quasi-experimental study found evidence against MCU effectiveness, one randomized controlled trial found little evidence of MCU effectiveness, and four quasi-experimental studies, one pretest-posttest, one qualitative, and one retrospective cohort study found evidence supporting MCU effectiveness. Primary functions of MCUs included hospital diversion, justice-system diversion, and providing accessible patient-centered care in the community.

Conclusions

The limited evidence uncovered in this review offers tentative support for MCU effectiveness. However, the current literature is too sparse to draw firm or generalized conclusions. An initial theoretical framework was created to help practitioners and policymakers document the functions and forms of MCUs to increase the comparability of these services, and research recommendations were generated to fill gaps in the literature.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-025-13806-2.

Background

The United States (U.S.) faces a dual crisis of rising demand for mental health (MH) crisis services and falling supply [1–4]. Suicide rates have significantly risen over the past two decades [5], MH-related ED visits have risen in adolescents and adults [1–3], and return emergency department (ED) visits have risen in adolescents [6, 7]. Increasing demand for MH crisis services has co-occurred with a reduction in inpatient psychiatric beds from 337 to 12 per 100,000 people since the 1950s [4]. Patients seeking MH care in EDs may face wait times lasting days or weeks, adding stress to individuals experiencing psychiatric crises [8]. The declining number of inpatient psychiatric beds is correlated with rising suicides [9]. Moreover, law enforcement (LE) is frequently called upon when individuals are experiencing a MH crisis in the U.S. However, LE officers often have limited training in MH crisis management and feel underprepared for this role [10]. Individuals with mental illness are disproportionately incarcerated [11], and make up about one in five LE-related deaths [12]. Black individuals are overrepresented among the jail MH population [13], and die during LE encounters at 2.8 times the rate of White individuals [12], while they are only half as likely as White individuals to receive any MH services [14]. The pattern of medicalization of White individuals and criminalization of Black individuals leads to excess Black mortality, highlighting the need for equitable systems where race does not influence the choice of crisis response.

Multiple professional leadership organizations [15, 16], as well as the Substance Abuse and Mental Health Services Administration (SAMHSA) [17, 18], have encouraged states to implement Mobile Crisis Units (MCUs), also referred to as Mobile Crisis Teams, to decrease the number of psychiatric ED visits and decrease LE involvement in crisis response [17, 18]. MCUs are an integral part of SAMHSA’s National Guidelines for Behavioral Health Crisis Care (National Guidelines), which also includes the implementation of MH crisis hotlines and crisis stabilization units [17]. According to SAMHSA’s National Guidelines, MCUs should be able to rapidly respond to individuals in crisis, wherever they are in the community [17]. Teams ideally operate continuously year-round [17]. The primary directive of MCUs is to provide patients, especially those who are hard to reach, with mental healthcare in the least restrictive setting [17], and de-escalate MH crises when possible to avoid unnecessary ED utilization and inpatient hospitalization [16]. A stated advantage of engaging with individuals in their own environments is that MCU providers can help patients utilize the coping skills and supports they already possess, which may include their family members [19]. In this way, community-based crisis de-escalation is meant to further the ethical mandate of minimally interfering with patient autonomy and preventing the unnecessary utilization of resources that could be used to help others [20]. Ideally, this has the added benefit of destigmatizing psychiatric treatment, since remaining at home is thought to reduce the self-stigma and external stigma associated with inpatient psychiatric hospitalization [21]. Finally, MCUs differ from LE co-responder models, as they are staffed exclusively by MH professionals and do not include LE [17]. For this reason, MCUs have been put forward as a potential solution to the disproportionate risk of police violence faced by Black Americans in crisis [22].

Per SAMHSA’s National Guidelines [17], MCUs are being implemented across the U.S. The Consolidated Appropriations Act (2023) expanded Medicare coverage for MCUs and authorized new grant funds for MCU pilots and evaluation [23]. States have also increased spending on MCUs. A survey conducted in 2022 found that 33 of 45 responding states covered mobile crisis services through Medicaid [24]. The Certified Community Behavioral Health Clinic (CCBHC) Medicaid demonstration program requires participating clinics to provide 24-hour mobile crisis services and has recently expanded to include sites in 18 states [25, 26]. Furthermore, in the summer of 2022, the implementation of the “988” phone number for the National Suicide Prevention Lifeline, an easy-to-remember, standardized way to connect to MH crisis care, drastically increased the number of crisis calls from communities [27]. Advocacy organizations like the National Alliance on Mental Illness have called upon policymakers to use this as an opportunity to enact a comprehensive crisis response system [28]. Others in the field of Community Mental Health have also urged lawmakers to implement MCUs so these crisis calls do not leave patients in crisis stranded [29, 30].

Purpose

This review study aims to: (1) Evaluate MCUs in the U.S. across multiple dimensions of effectiveness and (2) synthesize functions and forms from the included studies to clarify core functions of MCUs and support research and implementation.

Aim 1: systematic review

In 1995, Geller, Fisher, and McDermeit published a national survey of MCUs across all U.S. states and territories, finding that MCUs were widely used but “based on little or no empirical evidence” [31]. While the widespread use of MCUs has persisted, with at least 1,820 MCUs operating across 50 states as of 2023 [32], data on their effectiveness is limited. The only systematic review of MCUs aligned with SAMHSA’s National Guidelines since 2000 was a 2015 Cochrane review of Randomized-Controlled Trials (RCTs) by Murphy et al., which found no U.S.-based RCTs measuring MCU effectiveness [33]. The lack of up-to-date evidence calls into question the effectiveness of MCUs in the U.S. Moreover, given the changing landscape of supply and demand for MH services, an updated systematic review is needed to evaluate MCU effectiveness.

Aim 2: functions and forms

For MCUs to become evidence-based, there must be a theoretical framework (model) clarifying core functions and any adaptable components [34]. We seek to generate a model from currently available literature using the Functions and Forms Framework, used to define core functions of an intervention and variable forms that may accomplish those functions [35–37]. Defining the functions and forms of MCU care is a first step in creating an evidence-based theoretical framework upon which future investigators can measure the effectiveness of MCUs or propose testable innovations [35, 37]. Based on our review of the evidence, we will discuss knowledge gaps and make methodological recommendations to spur needed research on MCU effectiveness.

Methods

We conducted a systematic review of the literature following standard practices and the reporting guidelines of the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) [38]. We did not register/publish our protocol [38]. Covidence (Veritas Health Innovation, Melbourne, Australia) was used to organize and conduct our screening and reporting procedure (Fig. 1).

Fig. 1.

PRISMA flowchart of systematic review procedure

Search strategy

Search terms were created by authors JF and KHL, with the assistance of a health-sciences librarian. Searches were initially conducted on 8/21/2023 and updated on 7/1/2025. The following PsycINFO and Scopus search terms were used: {(“mobile crisis”) AND (effectiv* OR effica* OR evaluat*)}. In PubMed, the following search terms were used: {(“mobile crisis“[TIAB]) AND (effectiv* OR effica* OR evaluat*)}.

Inclusion and exclusion criteria

Studies met inclusion criteria if they were: (1) published in/after the year 2000, (2) conducted in the U.S. (3) focused on an MCU intervention without LE involvement consistent with some or all of SAMHSA’s National Guidelines (teams of MH professionals offering rapid response, assessment, treatment, and linkage to further care) [17], and (4) used a comparator to empirically measure patient-level effectiveness outcomes of the MCU. Studies were excluded if they were purely descriptive, or if they measured only implementation outcomes (e.g. those at the service-level) [39]. Studies were also excluded if the crisis response was LE only or LE co-response because these do not align with SAMHSA’s goal of avoiding unnecessary LE involvement. Requesting occasional backup from LE when deemed appropriate would not preclude inclusion, rather we sought to exclude studies with routine LE engagement.

Screening and data extraction

Lead authors JF and RP independently screened the titles and abstracts (N = 134), with moderate agreement (Cohen’s κ = 0.68) [40], before meeting to resolve conflicts. They then independently reviewed the remaining full-texts (N = 48, κ = 0.75) and met to resolve conflicts. Finally, both authors independently performed extraction (N = 9), before reconciling conflicts and creating the extraction table which included a description of the intervention, study design, comparison group (if applicable), and measured outcomes. Authors independently analyzed the completed table before reconvening to synthesize their independent narrative interpretations of effectiveness results. Study results were compared to assess the certainty of evidence related to each MCU effectiveness outcome.

Quality assessment

The quality of all studies was assessed using the Mixed Method Appraisal Tool (MMAT) [41], which was deemed to be appropriate for this study because both quantitative and qualitative studies were included. RP and JF independently completed the MMAT screening for each included study and assigned a score of high ([6]– [7]), medium [3–5], or low quality (< 3) in line with previous work [42, 43]. The two screeners met to resolve disagreements in MMAT scores by consensus, and independently reviewed each study holistically to ensure that the assigned quality scores matched the validity and interpretability of reviewed studies. Where this was agreed not to be the case, study limitations were described qualitatively.

Functions and forms

Insights about MCUs across studies were synthesized using the Functions and Forms Framework, used to describe the core components of complex health interventions [35–37]. Functions define the main purposes of the intervention, and the forms are the specific ways in which these functions are carried out [37]. Multiple functions may address a central motivating problem within the system [37]. The effectiveness of an intervention can be described in terms of the extent to which the stated functions are performed and the impact they have on associated motivating problems, while the forms can vary based on context [35]. For example, firefighters address the motivating problem of a fire in the community through functions that include rapid response and fire suppression. Forms that allow for rapid response include maintaining 24/7 availability through adequate staffing, and training firefighters to leave the station quickly.

Included studies were inductively coded for motivating problems, functions, and forms of MCUs endorsed by authors of reviewed studies. Content receiving a given code was synthesized into a Functions and Forms Table, from which key highlights were further consolidated.

Research recommendations

From the included studies, the authors generated a list of research recommendations based on the level of evidence supporting key outcomes identified and the methodological limitations of the existing literature.

Results

Description of studies

Our literature search identified 222 studies (131 after removing duplicates) for abstract screening. An additional 3 articles were identified through citation mining of reviewed full-text articles. Of these, 48 full texts were screened for eligibility, and nine studies were included for extraction (Supplemental Table 1). Common reasons for exclusion were lack of outcome measures [44], and for studying a police co-responder model [45]. The mean study publication year was 2014 (sd = 9.0). Only two included studies were published before 2010 [46, 47]. Fig. 2 presents the geographic distribution of study locations and summarizes outcomes.

Fig. 2.

Geographic distribution of mobile crisis units from included studies. MCU = Mobile Crisis Unit; ED = Emergency Department

The majority of studies focused on MCUs that directly responded to patients in MH crises and attempted to link them to care after the encounter, but two studies described an MCU that could provide ongoing care for up to 45 days [48, 49]. Two studies included only children/adolescents [48, 49], six included only adults [46, 47, 50–53], and one included both groups [54], but mainly served adults. Data for studies came from multiple sources, including trial-collected data [50], Medicaid claims [48, 51], administrative records [46, 47, 49, 51, 52, 54], Electronic Health Record data [54], and interviews with MCU service users [53]. Fig. 3 provides a visual summary of the nine included studies categorized by population type, study design, MCU functions, and key outcomes.

Fig. 3.

Study categorization by population, study design, mobile crisis unit functions, and outcomes. CRT is a law enforcement co-responder team; MCU = Mobile Crisis Unit; ED = Emergency Department.

Study designs and detailed outcome descriptions are reported in Supplemental Table 1. The most common study designs were quasi-experiments [46–48, 51, 54] (N = 5) and other variations of quantitative non-randomized studies (pretest-posttest design and a retrospective observational cohort study, N = 2) [49, 52]. One qualitative study and one RCT were also identified [50, 53]. Measures of effectiveness varied widely, although the most common were ED utilization frequency [48, 50, 54] and linkage to care [46, 50, 52]. Other effectiveness outcomes included the risk of inpatient psychiatric hospitalization [47], rate of arrest [51], psychiatric symptom severity [49], and satisfaction with care [53]. Only one study described workforce and infrastructure-related resource requirements for the MCU [49], but no study reported the costs associated with ongoing MCU operation.

Table 1 is a summarized version of the extraction table. Two pairs of studies in Table 1, Guo et al./Dyches et al. and Fendrich et al./Vanderploeg et al., investigated the same MCU [46–49]. MCU effectiveness outcomes from the nine included studies were grouped into six categories described below.

Table 1.

Summarized mobile crisis unit effectiveness outcomes

Effectiveness outcomes	MCU inferior to comparator	MCU no different than comparator	MCU superior to comparator
Symptom-related outcomes		1 (Currier et al.)	1 (Vanderploeg)
Linkage-related outcomes		1 (Currier et al.*)	3 (Currier et al*., Dyches et al., Holland & Tran)
ED diversion	1 (Vakkalanka et al.)	1 (Currier et al.)	1 (Fendrich et al.)
Inpatient diversion			1 (Guo et al.)
Arrest diversion			1 (Swanson et al.)
Satisfaction with care			1 (McDaniels et al.)

MCU = Mobile Crisis Unit

Note: The study by Currier et al. tracked linkage-related outcomes at multiple time points; the MCU was more effective at increasing initial linkage, while there was no significant effect when measured at further time points

Types of effectiveness outcomes

Symptom-related outcomes

Currier et al., in an RCT of ED-stabilized individuals assigned to either receive MCU services within 48 hours of ED discharge or standard outpatient follow-up within five business days, found no symptom or functional differences between the two groups after 2 weeks and 3 months [50]. These null findings are notable given successful randomization across demographic, clinical, and social variables. Interestingly, individuals who received at least one follow-up—regardless of group—showed similar symptom improvements to those who received none, raising questions about the utility of post-crisis follow-up itself. By 3 months, substantial symptom burden persisted across all subgroups [50]. Vanderploeg et al. evaluated symptoms/function using the Ohio Scales at intake and discharge from MCU services and found significant decreases in severity at the end of care when the patient is stabilized or escalated (up to 45 days) [55]. While compelling, the interpretation of this finding is limited by the lack of a control group.

Linkage-related outcomes

Currier et al. found that those assigned to receive MCU services were significantly more likely to attend their first follow-up appointment than those who were assigned to receive standard outpatient care, but participants in both groups were equally likely to attend subsequent appointments [50]. Dyches et al. studied the impact of shifting support from a specialized psychiatric emergency room in a hospital-based setting to a mobile crisis model and found that propensity score-matched MCU users were more likely than psychiatric emergency room users to receive community-based MH services in the 90 days after the initial episode of care [46].

Holland and Tran found that individuals issued an involuntary transport and ED-assessment order by an MCU had significantly higher odds of linkage to care compared to a co-responder unit housed within the MCU [52]. However, interpretation of this study requires caution due to the use of a composite outcome variable. The authors defined linkage to care as either: (1) admission of the patient to the hospital by the ED physician, or (2) discharge from the ED with a plan for follow-up. Individuals were considered unlinked to care if they were discharged with no plan. The authors note that individuals working in the MCU and co-responder unit can collaborate with ED staff to create discharge plans [52]. Therefore, it is not clear if the MCU primarily exerted its effect by: (1) initiating involuntary treatment proceedings for individuals who were more likely to require hospitalization (i.e. having a lower rate of “false positives” than the comparator), (2) collaborating during discharge planning more often than the comparator, or (3) a combination of both and/or another unmeasured confounder [52].

ED diversion

Vakkalanka et al., Currier et al., and Fendrich et al. measured ED utilization [48, 50, 54]. Vakkalanka et al. found that those receiving care by an MCU within 30 days of an ED visit for MH had a higher risk for ED re-admission in the following year [54]. Notably, there were significant and uncontrolled socioeconomic differences between the MCU and control groups which may have confounded results. Those in the MCU group were far more likely to be unhoused, on Medicaid, and lacking social support, among other differences [54]. Currier et al. found no difference in the likelihood of returning to the ED for a psychiatric cause within six months between those in the MCU follow-up and outpatient follow-up groups [50]. Fendrich et al., in contrast, found a significantly reduced odds ratio for ED return within a year-and-a-half of an index crisis for Medicaid-eligible youth whose crisis care began with an MCU versus an ED visit [48]. Selection bias was addressed in this study by stratifying individuals into propensity-score quintiles based on their likelihood of receiving mobile crisis services, using variables such as region, demographics, prior ED use, and diagnosis category (from ICD-9 codes) [48]. Regression models run within and across quintiles ultimately allowed for comparisons that adjusted for baseline likelihood of group assignment [48].

Inpatient diversion

Similar to Dyches et al., Guo et al. compared propensity score matched-individuals in crisis who were treated in a specialized psychiatric emergency room to those treated by an MCU, but instead measured 30-day inpatient hospitalization risks and time-to-hospitalization among those receiving MCU services compared to psychiatric emergency room controls [47]. Those with MCU care had a lower risk of inpatient hospitalization [47]. Particularly, the two days immediately following the initial crisis contained significantly more hospitalizations for the psychiatric emergency room group compared to the MCU group, before dropping to similar levels for the rest of the follow-up period [47].

Arrest diversion

Swanson et al. completed a study designed to measure the impact of various psychiatric crisis interventions on future arrests, in comparison to psychiatric crises managed by LE [51]. By using LE data and data from MH crisis providers in 5 counties in Michigan, mobile crisis proved to be the only intervention that led to significantly lower incidence of arrest in the year following the initial crisis [51]. Co-response and office-based crisis response had no significant impact on arrest rate [51]. Inverse probability of treatment weighting was used to match the control group to the experimental group on demographic variables, arrest history, and jurisdiction. Arrests that occurred during the initial crises were excluded from the analysis [51].

Satisfaction with care

In a qualitative study by McDaniel et al. that was designed to determine how well a San Francisco MCU supported the goals of unhoused individuals in MH crisis, participants discussed how MCU staff made them feel heard, and provided for short-term needs like food, water, and blankets [53]. They contrasted this with police and emergency medical service providers, who were described as escalating crisis scenarios [53]. However, the police were described as helpful in ensuring interviewees and their possessions could remain safe [53].

Adverse events

Currier et al. and Vakkalanka et al. both measured the prevalence of adverse events (e.g. suicide, homicide) between groups during the study period, and neither recorded any incidents [50,54]. Notably, relatively small cohort sizes (n = 120 and n = 302) may have limited the capacity to detect such rare events.

Quality assessment

Results of the MMAT quality assessment are presented in Supplementary Table 2. All studies met the two screening questions used by the MMAT (“Are there clear research questions?” and “Do the collected data allow to address the research questions?”) [41]. Overall, six included studies were ranked high quality (score of 6–7), three were ranked medium quality (3–5), and none were ranked low-quality (< 3). Each study received one point for every “yes” response to the seven quality assessment questions, including the two initial screening questions. Common reasons for which studies lost points included lack of clarity about whether the exposure or outcome were correctly measured (N = 4) and inability to account for confounders (N = 3). RP and JF had 79.4% agreement in their coding before reaching consensus (κ = 0.26, slight agreement). Cohen’s kappa was decreased by low agreement on the specific questions to which the small number (N = 11) of “Can’t tell” or “no” scores were assigned, although RP and JF’s total scores for each study never differed by more than one point. All disagreements were discussed and reconciled, with author KHL serving as a tiebreaker when needed. Notably, two of the studies ranked medium quality had substantial limitations. Vakkalanka et al. (MMAT score = 5) reported major differences between the MCU and control groups, most notably a fifteen-fold difference in the likelihood of being unhoused, that was not accounted for in propensity score matching [54]. This likely introduced confounding, which may threaten the conclusions drawn by the study. The study by Holland and Tran (MMAT score = 3) also scored at medium quality. However, as discussed above in the section Linkage-related Outcomes, the results of this study are difficult to interpret due to the choice of dependent variable.

Functions and forms table

As described in reviewed studies, MCUs addressed five motivating problems arising from the standard response to MH crises: delayed provision of care, the high cost of hospital care, the potential for harm in the community, unnecessary diversion to the justice system, and poor linkage to care. Multiple functions addressed each motivating problem. For example, the functions “Provide effective care anywhere in the community,” “Respond rapidly,” and “Assure 24/7 coverage” addressed the motivating problem, “Delays in crisis care can lead to poor outcomes.” Forms represented ways of operationalizing the functions [35]. “Assure smooth shift transitions” and “Operate 24/7…” were forms of the function “Assure 24/7 coverage.” Table 2 presents motivating problems, functions, and forms of MCUs as identified in reviewed studies. Additionally, the table includes Outcomes assessing functionality—performance metrics that evaluate whether the forms fulfill their functions. These outcomes include those that have been measured and those that were recommended for future study but have not yet been investigated.

Table 2.

Functions and forms of mobile crisis units based on included studies

Motivating problems	Functions: “How are these problems solved?”	Forms: “What might this solution look like?”	Outcomes assessing functionality*
1) Delays in crisis care can lead to worse outcomes	Provide effective care anywhere in the community	• Institute an interprofessional staffing model capable of providing crisis de-escalation, triaging, short-term care, and linkage to next steps in care • Establish clear catchment areas with sufficient MCU coverage	• Field contact rate (share of dispatched units who are successfully reach patients in the community)[49] • Symptom reduction[50] • Mobility rate (share of crisis calls resulting in in-person dispatch))[49]
	Respond rapidly	• Operate call center to quicky respond to patient • Provide crisis de-escalation or therapy by phone • Coordinate and dispatch MCU to patient when appropriate	• Response time from initial phone call to in-person visit) [49]
	Assure 24/7 coverage	• Operate 24/7, with clear plans and communication when closed or unavailable • Assure smooth shift transitions	• Audit of hours among MCU teams/facilities?
2) Hospital-based crisis care is costly and often not patient-centered	Operate MCUs more affordably than hospital-based care	• Provide care is community-based, lower-cost settings • Hire care providers who work at the top of their credential • Reduce overhead (e.g., physical office space, cleaning) through mobile care	• Total cost associated with a mental health crisis episode? • Cost-effectiveness (relative to hospital or facility-based care)?
	Divert from EDs/hospitals; provide care in therapeutic settings	• MCU teams attempt to stabilize patients in community whenever possible	• Rate of mental-health related inpatient hospitalization )[47] • Rate of ED utilization)[48, 54] • Rate of crisis service re-utilization [50]
	Maximize voluntary treatment	• MCUs de-escalate patients and share the importance of and discuss diverse available mental health services with patients – increasing the likelihood they will voluntarily engage	• Involuntary commitment rates of MCU patients?
	Assure outpatient follow-up	• Institute community-based assessment as an alternative to outpatient follow-up • Follow-up by telehealth if appropriate	• Completion of first follow-up visit after index crisis event [50] • Outpatient contacts in the 6 months post-crisis [50]
	Reduce wait time in ED settings (“boarding”)	• Triage effectively to reduce psychiatric patient volume in ED to better align acute crisis care (e.g., inpatient bed) supply and demand	• Boarding duration?
	Intervene early to reduce peak crisis severity	• Respond rapidly to minimize the time that individuals are in crisis without care • Start treatment by phone if appropriate to de-escalate individuals as quickly as possible	• Symptom severity [50] • Mortality [50, 54] • Setting of first contact? • Age of diagnosis? • Time in crisis before receiving care?
3) Unaddressed mental health crises may lead to violence in the community	Improve community safety	• De-escalate mental health crises, which may prevent violent incidents in the community and victimization of those with mental health conditions • Post-crisis linkage to care will reduce subsequent crises (which carry a risk of violence)	• Rate of crisis service re-utilization[50, 54] • Homicide and other violent crime rates? • Rate of violence against those with mental health conditions? • Mortality [50, 54]
4) People with mental health disorders are disproportionately diverted into the justice system	Divert individuals from the justice system	• Train MCU providers to de-escalate crises to reduce criminal offenses • Assure MCU providers are able to help LE appreciate the medical nature of mental health crises and value of justice diversion whenever appropriate • Coordination between LE and MCU dispatch to encourage appropriate referral (e.g., 911 dispatches MCU over LE whenever appropriate)	• Odds of arrests [51] • Frequency of incidents involving use of force on individuals with mental illness?
5) Many people face barriers to navigating, accessing, and receiving community-based mental health care	Provide care navigation and referral to appropriate resources	• Familiarize MCU providers with community mental health resources and build connections to establish referral pathways • Include case-management and referral under MCU providers job description	• Community-based mental health service utilization within 90 days of a crisis [47] • Rate of completed community referrals? • Number of referrals offered? • Staffing composition?
	Coordinate community resources	• Leverage relationships with varied partners to identify system gaps and suggest ways to fill them • Learn from individuals why they are struggling to access regular care and relay these concerns to the appropriate providers or state/local authorities	• Receipt of supportive services post crisis (e.g., supportive housing, supported employment, etc.)? • Screening for health-related social needs?
	Be accessible for anyone needing a first step to care	• Provide services without insurance requirements or out-of-pocket costs through state funding or other arrangements • Be available 24/7 through a well-publicized access line such as 988	• Crisis service utilization [49] • Public awareness of community MCUs?
	Reduce stigma around seeking mental health services	• Treat individuals in community, reducing the sense that mental health crises must be treated at stigmatized institutions • Staff with peer-support specialists and use plain clothes and unmarked vehicles to minimize the otherness of the intervention	• Patient-reported willingness to seek care following crisis intervention [53] Reported stigma related to mental health diagnosis?

MCU = Mobile Crisis Unit; LE = Law enforcement; * citations indicate that the function was evaluated in the study specified; ? = function of MCUs mentioned in included studies but not evaluated

Highlights

Six key elements of MCUs were synthesized from the functions and forms table. These were selected because of their importance to multiple functions and emphasis placed on them by authors of reviewed studies. These highlights are not meant to imply that an MCU will necessarily be effective if accomplished. Rather they are set out to define, based on reviewed studies, the core elements of MCUs as currently described in the literature.

1. Team-based staffing: Several studies expressed that MCUs should be equipped with at least two MH professionals (dyads), who may have a variety of qualifications, but are both trained in crisis stabilization and have knowledge of community services to provide effective linkage [46–49, 52]. Moreover, several MCUs included a psychiatrist, either in-person or on-call, for decision-making and medical treatment planning [46–49, 52].

2. Rapid availability: Studies endorsed that MCUs should operate the maximum feasible number of hours, with many operating 24/7 [48, 49, 51, 53, 54]. Studies indicated that MCUs should respond rapidly – one MCU responded regularly within 45 min [49].

3. Crisis response and triage: Apart from the study by Currier et al., [50] Every study endorsed the primary function of MCUs to serve as first-line MH crises responders [46–49, 51–54]. Studies affirmed that MCUs could respond to a variety of locations in the community [46–49, 51, 53, 54]. Additionally, every MCU studied conducted triage to determine whether the individual needed escalated care or could be treated in the community [46–54].

4. Continuing care: Studies indicated that MCUs should link service users to long-term supports [48–54]. To achieve this goal, MCUs should provide follow-up, referral, and/or case management services to help individuals understand the complex mental healthcare system and connect individuals with appropriate supports [48–51, 53, 54].

5. Operational crisis line: Studies indicated that MCUs should be easily reachable through a well-known access line [48, 49, 54]. Those staffing the line should be MH professionals able to discern caller needs, e.g., telephone-based stabilization, linkage to community services, or an MCU dispatch [49, 54].

6. Community integration: Several studies highlighted that MCUs should function as bridge-builders between community resources, individuals receiving care, service providers, schools, LE, etc [49, 53]. As such, MCUs may become well-positioned to build communication channels across organizations across the MH continuum of care and use their frontline knowledge to point out and address emerging challenges [49].

Discussion

To our knowledge, this is the first comprehensive review of MCUs in the U.S. since 1995 [31]. Though the number of operational MCUs has grown significantly [32], there is still limited evidence of MCU effectiveness. Nine studies met our inclusion criteria and were high or medium quality, although some studies had substantial limitations that compromised the interpretability of their findings. Most included studies had quasi-experimental designs, and studies measured a variety of outcomes related to repeat crises, symptom severity, linkage to care, and satisfaction with care. The strongest evidence for MCU effectiveness came from three quasi-experimental studies—Swanson et al. (2025) [51], Guo et al. (2001) [47], and Fendrich et al. (2019)— [48] which found that MCU treatment reduces individuals’ likelihood of arrest, inpatient admission, and ED visits, respectively, in the period after MCU treatment compared to controls similarly at risk for these outcomes.

Evidence against MCU effectiveness came from two studies. Currier et al. (2010) found no difference in the likelihood of outpatient MH contact or ED revisit after individuals were randomly assigned to an MCU or outpatient visit [50], and Vakkalanka et al. (2021) found that individuals treated by MCUs were more likely to have an ED revisit than controls [54]. Notably, both studies are challenging to integrate into the interpretation of the results of this review. Currier et al. studied the impact of assigning individuals to MCU follow-up care after an ED visit, rather than studying the effect of the MCU on crisis first-response and triage [50]. The study by Vakkalanka et al. suffered because sociodemographic variables such as housing status were not included as covariates in propensity score matching. This likely led to substantial confounding. Even ignoring these null findings, the other seven studies that supported MCU effectiveness investigated six distinct outcomes, with one of the two studies on linkage being difficult to interpret due to a composite dependent variable. Results for each outcome must be replicated to draw stronger conclusions. Even more, the MCU models in these seven studies were not standardized, underscoring the potential lack of generalizability of the findings from each study to MCUs in other contexts. Overall, our findings tentatively suggest that MCUs may improve symptoms and linkage to care, and reduce ED revisits, inpatient admissions, and arrests. However, more research is required before these conclusions can be drawn with confidence.

To illustrate the sparsity of evidence, it is useful to compare MCUs, which first appeared in the literature in 1971 [31], to Assertive Community Treatment Teams (ACTTs), first described in the literature in 1980 [56]. There are an estimated 1600–1900 active MCUs and a similar number of ACTTs in the U.S [32, 57]. While ACTTs likely have higher hourly labor costs associated with employing psychiatrists and psychiatric assistants, many MCUs can be quite large and bear the cost of staffing for a 24/7 in-person response, so total costs are plausibly within a similar range. Despite this, their respective research bases are remarkably different. By 2001, in the 21 years since the creation of ACTTs, 22 RCTs had measured the effectiveness of ACTTs in the U.S. for treating severe mental illness and preventing inpatient hospitalization [58]. In comparison, a previous review found no RCTs of MCUs published before 1995 [31], and we found only one RCT [50], with dubious comparability to the standard MCU model, published in the 25 years from 2000 to 2025. Additional research is needed to establish an evidence base for MCUs similar to ACTTs.

Comparing our findings to the existing literature from outside the U.S. is desirable but challenging. For example, Crisis Resolution Teams are a relatively well-studied intervention in the United Kingdom that perform similar functions to U.S. MCUs [59]. However, these teams also provide intensive home treatment after the crisis [59], involving up to six weeks of frequent community-based visits, up to twice daily, prior to discharge from the team [59]. Other well studied analogous interventions abroad, such as Crisis Assessment and Treatment Teams in Australia [60], have similar differences in service definition that make comparisons to U.S. MCUs challenging [61].

Based on the nine studies found in our literature search, we synthesized functions and forms of reviewed MCUs. We believe this can provide the theoretical foundation for research, implementation, and adaptation of this intervention. What follows is a discussion of MCU functions and how they compare to SAMHSA’s existing guidelines, followed by a set of research recommendations.

MCU functions

Functions and forms were compiled from all reviewed studies, which included a diverse range of reviewed MCUs. Still, reviewed studies assessed relatively well-resourced MCUs, and we recognize that there is more diversity among the estimated 1820 MCUs in the U.S [32]. Some MCUs serve children/adolescents and primarily coordinate with school systems; others serve exclusively adults. Some are part of statewide networks; others are non-profit organizations. Some are urban and some are rural. Some serve areas with many outside resources to link individuals to, and others do not. Because of this, MCUs will and likely should look different depending on the context. Staffing capacity and composition, service availability, and target response times are adaptable forms that may not be standardizable because communities have different resource and operational constraints and goals [62].

At the same time, a minimum set of core functions should be established by which all MCUs can be evaluated [62]. This would make studies more comparable and prevent the issue identified in this review where one study evaluated an MCU limited to post-crisis follow-up care while every other study evaluated MCUs as a first-line crisis response. The Functions and Forms Framework encourages a focus on functions to create a strong theoretical foundation for MCUs. For example, do all MCUs need to respond to crises and conduct de-escalation, or can the service begin with follow-up after an ED visit? Do all MCUs need to link individuals to community support (e.g., outpatient care)? These were functions endorsed by some MCUs but not all. The functions and forms table developed begins to differentiate core MCU functions and the forms used to achieve them. Future work should edit this list and build upon it, including the list of measurable outcomes to evaluate MCU success according to specific functions.

In comparison to the most relevant existing guidelines for MCU functions—SAMHSA’s minimum expectations and best practices for MCUs—the functions identified here appear generally consistent, with a few key differences [17]. SAMHSA’s four minimum expectations (that MCU response is on-demand and rapid, mobile, in-person, and inclusive of a licensed or credentialed provider) all conform with functions and forms identified in this study [17]. Two of the six best practices (responding without LE if possible and responding in community settings) also aligned with MCU practices in included studies [17]. However, four of SAMHSA’s best practices (initiating an MCU encounter with an intentional triage/screening process, incorporating peer support specialists, configuring teams to be attentive to safety of individuals in crisis and team members, and including families in MCU responses to youth), were not well described across studies [17], indicating a need for future research that determines whether these practices are core functions or adaptable forms of MCU operations, and if/how these practices moderate effectiveness.

Research recommendations

Studying MCUs is challenging. Different MCUs currently accomplish different functions, from crisis response to linkage and even community-based treatment, as shown in Fig. 3. Study design must be appropriate to measure the MCU functions/forms being evaluated and consider confounding and bias. Those who receive MCU care may at baseline differ significantly from those who receive ED care regarding insurance status/access to care, frequency of prior emergencies, and sociodemographic variables [48, 54]. Even more, the urgency of MH crises requires providers to prioritize immediate care over obtaining consent and randomizing treatment groups, creating selection bias. Recognizing this, we offer the following recommendations, in no particular order, to enhance understanding of MCU effectiveness.

Research recommendation 1: Robustly investigate safety

Two reviewed studies measured adverse events (e.g., harm to self or others) [50, 54], and although neither found evidence that MCUs were associated with them, they were likely underpowered to do so. It remains possible that inappropriate diversion from EDs by MCUs could lead to undertreatment of crises. This fear is not unfounded: at least one study of a MH crisis diversion model in the UK was associated with increased adverse events compared to control groups [63]. This study investigated a British Crisis Resolution Treatment Team [63], which has a different service definition than an MCU, as has been described, but serves as a reminder that the potential for adverse events, though rare, is real. We believe that a sufficiently powered longitudinal study of adverse events comparing MCU-diverted individuals to those receiving ED care should be performed in the U.S.

Research recommendation 2: Use representative samples and informative experimental groups

Although all reviewed studies focused on users of MH crisis services, the specific samples of individuals studied varied significantly. Not all studies had a sample that contained individuals typically served by an MCU. For example, the RCT by Currier et al. excluded individuals who already had outpatient providers, which likely decreased the representation of individuals with chronic conditions [50]. Experimental groups also varied in ways that made drawing overarching conclusions across studies challenging. In the case of ED diversion, for example, Vakkalanka et al.’s experimental group included both individuals referred to the MCU by the ED for step-down care, as well as individuals who stepped up to the ED from the MCU (i.e. an “MCU-exposed” group) [54]. Currier et al.’s experimental group only contained individuals who stepped down to MCU care [50], and Fendrich et al.’s experimental group contained youth who had an index MCU-managed crisis at the beginning of the 18-month time horizon (and may or may not have stepped up to the ED) [48]. Of all these groupings, Fendrich et al.’s experimental group appears to be the most well-equipped to study causal inferences regarding MCU use and subsequent ED utilization; we recommend that similarly defined experimental groups be used in future studies of MCU effectiveness.

Research recommendation 3: When possible, utilize Zelen-blinded RCTs

RCTs may at first appear to be ethically impermissible for studying a MH crisis service, as individuals must possess the capacity to consent to enroll in a research study. However, many individuals in psychiatric crises do retain decisional capacity [64]. Even when this is not the case, a Double-Consent, Zelen-Randomized Controlled Trial (DCZ-RCT) is available, elements of which were performed by Currier et al. [50] DCZ-RCTs randomize eligible participants before seeking their consent, and only those who are deemed to have the capacity to consent are given the option to enroll [65]. After enrollment, individuals receive their group assignments, and individuals assigned to the experimental group are offered the opportunity to switch to receive usual care [65]. The ability to switch assignments means that the findings of these trials must be interpreted from an implementation perspective (e.g. “Does offering MCU services to a community reduce the rate of ED and inpatient utilization for MH crises over time?”) rather than from a pure effectiveness perspective (e.g. “Do MCUs decrease ED and inpatient utilization over time?”). These results may be invaluable for healthcare system administrators [66], and this design has been used to study psychiatric crisis services in the past [66, 67].

Research recommendation 4: Measure program cost and cost-effectiveness

Cost, cost-effectiveness, and economic evaluations are vital outcomes for persuading or dissuading decision-makers to invest in new healthcare interventions, and high-quality studies abiding by the CHEERS Standards for Health Economic Evaluations should be performed for MCUs [68]. However, no reviewed studies measured cost/cost-effectiveness, although several reviewed studies cite lower costs as a key benefit of MCUs [48, 49, 52]. Often-cited studies investigating MCU cost-effectiveness were not included either because they investigated LE co-responder models that do not fit the current understanding of an MCU [45], or because they were published before 2000 [69, 70]. The lack of health economic evaluations makes it challenging for public and private funders to determine how much to invest in MCUs.

Research recommendation 5: Study sub-populations in a variety of settings

Three of the nine included studies contained children/adolescents in their samples [48, 49, 54]. Due to the paucity of literature, we combined studies including children and adults in our review. However, these are distinct populations, and an effective psychiatric crisis intervention in one population may not generalize to the other. Future research could also focus on priority populations including individuals with co-occurring substance use disorders, developmental disabilities, dementia, or traumatic brain injuries, as well as sexual and gender minorities and individuals racially concordant vs. discordant from MCU providers.

Effective adaptation of MCUs requires an understanding of how they operate in varying geographic settings and service system contexts [71]. Our search revealed limited geographic diversity of MCUs studied, with no studies from the Southeastern or Southwestern U.S. All included studies assessed MCUs in urban areas. The Connecticut MCUs treating youth [48, 49], the focus of two of nine included studies, stood out for above-average staffing and financial capacity that may be unique to their circumstances and funding model. It will not be sufficient to conduct studies in well-resourced environments that do not match the capacity of most MCUs. MCUs in low-income and rural areas, especially those in the U.S. South, should be prioritized for future investigations.

Research recommendation 6: Conduct qualitative research

Based on the differences in MCU context, qualitative research is needed to explore different needs and challenges that MCUs face, which are nuanced and may not be drawn from quantitative studies. Research is needed to answer the question, “What are the necessary conditions within a community context for MCUs to function effectively?” It may be the case, for example, that certain services or structures must be in place for MCUs to be effective.

Additionally, it is notable that two of the three studies of ED diversion included individuals stepping down to MCU care from the ED, which diverges from SAMHSA’s vision for MCU care [17]. This level of heterogeneity is not uncommon among new interventions striving to address multiple challenges in varied contexts. Based on this review, MCU functions are currently not uniform. As a result, qualitative research is needed so that each MCU can be clearly understood in terms of its functions and forms. The motivation and approach to adapting the MCU intervention needs to be clearly explained until functions are standardized. Qualitative research informed by best practices from evaluation [72], implementation science [62], and systems science can help [73].

Research recommendation 7: Study the role of peer recovery specialists

One of SAMHSA’s best practices for MCUs is incorporating individuals with lived experience of behavioral health conditions (i.e. peer recovery specialists) within MCU dyads [17], which has been supported in a variety MH contexts [74]. However, of the reviewed studies, only McDaniel et al. mentioned that the MCU investigated included peer recovery specialists in the intervention, and their role was not discussed in qualitative findings [53]. Further research should investigate the impact of peer recovery specialist staffing in MCUs, appropriate roles and responsibilities, and barriers and facilitators to their participation in MCUs. While there is evidence surrounding the role of peer recovery specialists in behavioral healthcare, there is very little on their role in crisis settings [75].

Limitations

Our study had several limitations. First, it is possible we missed studies that investigated MCU effectiveness. We attempted to address this challenge through robust search terms and searching multiple databases. Second, we limited our search to the U.S. for clarity since our recommendations are aimed at researchers and practitioners within the U.S. However, we could have excluded MCU studies from outside the U.S. that functioned similarly enough to U.S.-based MCUs to warrant inclusion, although literature searches of crisis diversion models did outside of the U.S. did not reveal any that were directly comparable. In addition, for feasibility and to identify strongest evidence, we limited our review to peer-reviewed literature. Given the large number of MCUs in practice, we encourage evaluation teams to publish results to close this evidence gap. A limitation of our binary scoring system (1 point for “yes”, 0 points for “no” or “can’t tell”) for each question in the MMAT quality assessment tool was that the severity of limitations was not accounted for in the scoring system. To mitigate this, we holistically reviewed studies and described cases where both reviewers agreed that the limitations were not fully captured by the quality assessment score.

Our focus on documenting MCU effectiveness may mean that the sample of studies identified was not representative of all MCUs. As such, our synthesis of motivating problems, functions, and forms should be updated as evidence accrues. That said, we believe the identified studies have sufficient detail and variety to provide a basis for a useful framework. The alignment of core functions with SAMHSA’s MCU recommendations reinforces the framework’s breadth. Expanding research to exhaustively document MCU functions and more/less effective, context-responsive forms is an important next step. Future research might develop a repository of known MCUs, unique functions and forms, and MCU performance and cost metrics as the evidence base grows. Such a tool would allow researchers to refine and develop the MCU functions and forms framework and evidence of effectiveness and resource efficiency (e.g., cost-effectiveness, business case).

Conclusion

Mobile crisis units have been posited as a solution to unnecessary treatment of individuals in EDs, long ED boarding times, and avoidable incarceration of individuals in acute mental health crises [14, 15]. This systematic review offers tentative support for the effectiveness of mobile crisis units in reducing ED visits, inpatient hospitalization, arrests, and improving linkage to post-crisis care. However, our understanding of mobile crisis unit effectiveness is still limited, and more research is necessary to understand when and how mobile crisis units may be effective. To assist future evaluation efforts, we created an early version of a mobile crisis theoretical framework by synthesizing functions and forms emerging from the available effectiveness research. Our recommendations for future investigation chart a path to strengthening this model by establishing contextually appropriate evidence-based practices. Given the recent proliferation of mobile crisis units, it is urgent that this research be performed and the functions and forms of MCUs be rigorously tested and refined.

Supplementary information

Below is the link to the electronic supplementary material.

Abbreviations

ACT

Assertive Community Treatment

CHEERS

Consolidated Health Economic Evaluation Reporting Standards

ED

Emergency Department

DCZ-RCT

Double-Consent, Zelen-Randomized Controlled Trial

LE

Law Enforcement

MCU

Mobile Crisis Unit

MH

Mental Health

RCT

Randomized Controlled Trial

SAMHSA

Substance Abuse and Mental Health Services Administration

UK

United Kingdom

US

United States

Author contributions

RP and JF conducted the initial screening and data extraction for all included studies and drafted the manuscript. HN made substantial contributions to the development of Table. 2 and participated in drafting and revising the text. GD and AW provided critical input during the drafting and revision process. KHL conceptualized the review design, guided methodological decisions, and contributed to drafting and revising the manuscript. All authors read and approved of the final manuscript.

Funding

None.

Data availability

All data generated or analysed during this study are included in this published article [and its supplementary information files].

Declarations

Ethical approval

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.