Most PRMHT interventions involved police officers working in partnership with MHPs, although the role, responsibilities and location of MHPs varied. There is very little UK-based evidence on the alternative model of providing specialist mental health training to police officers. Interventions were generally valued by staff and showed some positive effects on procedures (e.g. rates of detention) and resources, although these results were not entirely consistent and not all important outcomes were measured. In particular, mental health service-related and individual service user outcomes were largely absent.
As PRMHT interventions sit at the intersection of criminal justice and mental health services, their successful implementation may depend on strategic integration of these services at the relevant local or regional level. Effective sharing of information and integration of knowledge among police and MHPs appears to be crucial. There is some evidence on how partnerships, protocols and technology can influence integration and implementation.
Most of the evidence was at risk of multiple biases due to design flaws and/or a lack of reporting of methods, which might affect the results. All the included primary research was conducted in England and health equity data were largely absent.
What is the evidence base for models of police-related mental health triage interventions?
Which models have been described in the literature?
The schemes evaluated in UK studies were typically described as street triage; however, these incorporated aspects from a range of different models described by Parker et al.14 These included co-response (e.g. police officers and MHP jointly attending incidents), information-sharing agreements (e.g. information-sharing protocols and joint needs assessments), co-location (e.g. MHPs in police control rooms) and consultation (e.g. telephone advice and assistance from MHPs). All UK PRMHT schemes incorporated aspects of co-response and/or consultation, with a key difference being the role and location of MHPs. Roles of MHPs included joint in-person response to incidents alongside a police officer, providing telephone support with or without the ability to join police officers at an incident, or providing telephone and/or on-scene response without being accompanied by police. Locations of MHPs included a dedicated police car, police control room and an office in a local hospital, mental health trust or crisis assessment service.
Parker et al.14 identified ‘pre-arrest diversion models’ as those involving police officers with special mental health training serving as the first-line police response to mental health crises in the community and acting as liaisons to the formal mental health system. In these models, such as the US-based CIT model, MHPs may provide the initial specialist training but are not routinely involved in attending incidents or directly informing assessments. There is currently an absence of UK-based qualitative data on this particular model, and there are limited quantitative data from only one UK study.47
This report describes and classifies the different models of PRMHT on the basis of descriptions provided by study authors. However, none of the models appeared to have an explicit theoretical basis or followed a particular logic model. This has implications for evaluating the implementation of such models (see Limitations of the evidence and synthesis).
What evidence is there on the effectiveness of these models?
Whereas the scoping review by Parker et al.14 focused on the description and classification of PRMHT models, this current rapid evidence synthesis attempts also to draw together evidence on the effects and implementation of such models.
There is little robust evidence on the effectiveness of the PRMHT models. The limited quantitative evidence available suggests reductions in formal detentions, higher hospital admission rates, increased likelihood of follow-up by secondary mental health services if patients are not admitted, and an increase in the use of HBPOS. However, the results were not entirely consistent, with some regions seeing increases in S136 rates after introducing PRMHT interventions. A single study noted that far fewer S136 detentions were made during PRMHT operating hours than outside those hours.41 However, the absence of concurrent control groups in most evaluations means that it is difficult to separate any true effects of PRMHT interventions from statistical phenomena, such as confounding and regression to the mean. In addition, most included studies implied that a reduction in S136 rates was a desirable outcome, although this might not necessarily be true for every setting and context.
There is minimally reported, heterogeneous and conflicting evidence on the effects of PRMHT interventions on outcomes, such as on the quality/timeliness of assessment, referral and treatment, access to services, demand for police resources and number of repeated contacts with individuals.
There is limited qualitative evidence that PRMHT interventions may help identify people with mental health needs that have not previously been in contact with mental health services. However, there is a near-total absence of reliable quantitative evidence on measures such as individual mental health outcomes, changes in demand for mental health services and changes to case-finding or level of access to health services.
Although PRMHT might be conceived as a means to reduce longer-term police, health and social care costs at the expense of possible increased short-term costs, no true cost-effectiveness analyses of PRMHT schemes were found in this rapid evidence synthesis. Two studies reported police force cost savings43,46 (one46 noting that savings came from lower custody rates and reduced officer time attending S136 incidents), but these studies had conflicting findings with regard to NHS costs.
What evidence is there on the acceptability and feasibility of these models?
We identified several UK studies reporting qualitative evidence on the implementation of PRMHT models, but it should be borne in mind that this mostly consists of views from a relatively small number of police and mental health staff directly involved in delivering pilot interventions.
Acceptability
In general, police staff appeared to value PRMHTs, both officers with an interest in mental health and those who felt that mental health should not be a police responsibility. Both police and health staff noted an improvement in quality of care.
Although some service users reported improved experiences with PRMHT teams compared with previous arrangements, service user feedback was rare. Some qualitative evidence suggested that service users preferred to interact with MHPs rather than police officers. This was attributed to MHP communication skills and the association of police uniforms with authority and criminalisation. Any future provision of protective clothing (e.g. stab vests) for MHPs responding to incidents may need to consider the impact on user perceptions of MHPs as well as MHPs’ safety.
Feasibility
Most evaluated PRMHT models only addressed immediate concerns around the use of S136 by police called to mental health-related incidents. However, there are wider consequences of introducing these schemes for both police and health services, particularly risks of displacement or duplication of existing services. Strategic response to mental health-related incidents may need to consider, at the highest level, which pathways prove to be most effective for service users and make the most appropriate and efficient use of both police and NHS resources. Some barriers to successful outcomes lay outside the control of police or even PRMHT staff (e.g. lack of co-ordination between neighbouring NHS trusts). Similarly, the availability and resources of local services need to be taken into account. For example, PRMHT’s attempts to reduce the number of repeated crisis calls from an individual may prove unsuccessful if they are referred to local mental health services with long waiting times.
The reported advantages and disadvantages of retaining consistent staff on PRMHT duties may be of interest to service managers. Advantages may include enhanced relationships and understanding, greater efficiency and less frequent issues around police vetting procedures. However, disadvantages could include mental health knowledge being restricted to fewer police staff and poorer integration with the wider force. The service impact of reallocating police and MHP staff to PRMHT from other active roles should also be carefully considered.
What evidence is there on the barriers to, and facilitators of, the implementation of these models?
As for acceptability and feasibility issues, evidence on barriers and facilitators was based mostly on the views of a relatively small number of police and mental health staff directly involved in delivering pilot interventions.
Barriers
Some comments indicated uncertainty about how and when it is best to deploy MHPs to the scene of an incident. There may be a trade-off between MHPs having better access to records in a hospital/control room and using their hands-on skills to aid in incident resolution. In conjunction with other information-sharing measures, this kind of barrier might potentially be overcome by providing MHPs with improved mobile information technology. PRMHT staff from our advisory group have noted the value of being able to obtain information while en route to a service user.
Reveruzzi and Pilling’s9 evaluation appropriately recommended that the role of street triage schemes should be reviewed in relation to referrals from, and contacts in, private settings. However, since the recommendation was originally made, the Policing and Crime Act 20174 has resulted in changes to S135 and S136.2 In theory, these changes should mitigate some of the previously identified problems around police officers being unable to enter private premises to make an assessment. However, the effect of these changes has yet to be properly established. In addition, it was clear from the evidence that not all staff were aware of the nature of police powers in public places and private premises. All MHPs and control room staff involved in PRMHT interventions being knowledgeable about the constraints on police powers may prevent misunderstandings or inappropriate recommendations for action.
Data collection is often incomplete and restricted in scope, which limits the opportunity to continuously evaluate and improve services. Methods for comprehensive, accurate and efficient data collection (that do not place undue additional demands on front-line police or health staff) may need to be developed. As PRMHT interventions are multiagency, data collection and evaluation may benefit from being correspondingly integrated and strategic. Data collection (like information-sharing) may also benefit from being governed by relevant protocols and facilitated by appropriate technology. As well as routine data collection of outcomes, such as S136 rates, future evaluations would benefit from data on quality and timeliness of assessment, referral and treatment, mental health outcomes, demand on police resources and police officer time, demand for community mental health services, rates of hospitalisation via A&E or acute mental health services, level of service engagement, experience of services for service users, and costs and savings to health and police services, including estimates of skill mix and other workforce resources. One barrier to effective interagency data collection may be obtaining the necessary staffing and resources across all of the relevant services. Other barriers include a lack of clarity from commissioners about why certain data should be recorded, obtaining consent to share individuals’ data between services and difficulties where boundaries are not coterminous (e.g. different NHS trusts overlapping one police force area or people being referred to out-of-area hospital beds).
Some qualitative studies noted the challenge of disproportionately high demand created by repeat service users. Some comments suggested that efforts could be made to work towards a distinctive and separate partnership between police and health-care services for repeat callers,50 or towards reduced police input for callers for whom there are no immediate safety concerns.53
Such efforts may be worth further consideration and evaluation in any future implementation of PRMHT interventions.
Facilitators
The qualitative evidence as a whole emphasised Reveruzzi and Pilling’s9 conclusions around the importance of strong partnerships between police and health services, co-location of services and the value of shared information. The latter point in particular crossed several themes; shared information underpins the achievement of organisational objectives, such as reducing S136 arrests, speeding up user needs assessment and enhancing collaborative working. Future PRMHT interventions would probably benefit from immediate access to shared information across the police/health interface, facilitated by agreed protocols and underpinned by appropriate technology that permits compatibility of data across police and health systems.
In all cases, lines of accountability and responsibility need to be clear among all PRMHT staff. This is because differences in attitudes to risk between police officers and MHPs have been observed (e.g. around threats of self-harm and suicide), and because the presence of factors such as alcohol, drugs and/or the risk of violent behaviour can increase the complexity and difficulty of making judgments about the best course of action. Similarly, roles, responsibilities and reciprocal arrangements need to be clearly defined between PRMHT services, crisis teams and other related health services.
Immediate and consistent availability of MHP support was very important to police officers who were responding to mental health-related incidents, with immediacy sometimes seen as a key difference between PRMHT and crisis teams. As stated by Reveruzzi and Pilling,9 24-hour availability appears to be crucial to providing officers with a resource that is considered reliable. Appropriate communication technology may improve accessibility, such as mobile phones dedicated to the task of consulting MHPs that can have some advantages over police radio sets, although difficulties may be encountered in rural locations with poor signal coverage.
Many resource savings attributed to PRMHT interventions stem from their value in accelerating the assessment of user needs. Whereas most UK interventions have focused on retaining MHP partnerships to facilitate assessments, the US literature is more focused on interventions where police officers receive mental health training without the same level of ongoing specialist support from MHPs. No evidence was found comparing these two models, despite their potential to have quite different costs and benefits.
Limitations of the evidence and synthesis
The DHSC-funded street triage pilots were managed by local police forces in partnership with Clinical Commissioning Groups (CCGs), NHS England and Police and Crime Commissioners. Given the unique governance arrangements for delivering a mental health triage service in the UK, as well as important differences in social context and the delivery of health and criminal justice services between countries, qualitative data from non-UK settings were not incorporated. However, all identified studies evaluated PRMHT schemes in England alone, so care should be taken when extrapolating this evidence to other UK regions.
Discussion of health equity issues (e.g. for BAME communities, people for whom English is not their first language, people with neurodevelopmental disabilities) was largely absent from existing evidence. Similarly, when service user groups were involved in research, they were generally not well described.
We excluded traditional L&D services from our rapid evidence synthesis, but it is clear that many PRMHT schemes are closely integrated with L&D and are often conceptualised as a way of bringing some L&D functions to an earlier point in the pathway (e.g. assessing and referring individuals to an appropriate non-CJS treatment or support service). As such, the boundaries between PRMHT and L&D may be somewhat artificial.
Only a few systematic reviews met the inclusion criteria and there were variations between them in terms of the type of studies included and the nature of the interventions and outcomes, making it difficult to synthesise the evidence. The review authors similarly commented on the lack of robust primary evidence, which made it difficult to draw conclusions. Across the reviews, there were calls for more relevant and robust research, and a number of recommendations for research and practice were provided.
The included primary studies were limited, with very little concurrent comparative data and a wide variation in interventions generically labelled ‘street triage’ by study authors. This made it difficult to draw conclusions about effectiveness in our evidence synthesis. However, conclusions and recommendations by study authors for research and practice echoed those of the systematic reviews.
The methodological quality of the systematic reviews was formally assessed, with some having limitations that made it difficult to judge the reliability of the evidence presented. Although there were limitations in the reported review process, the authors reasonably stated that there were no robust studies to rigorously test the evidence.
The primary studies were also methodologically limited. One RCT was assessed using a recognised assessment tool, which indicated that it was largely at a low risk of bias. The remaining studies were largely descriptive with few concurrent comparisons, which made it difficult to evaluate the effectiveness across the interventions or of any particular model. Most studies included relatively small numbers of service users, precluding the generation of robust quantitative estimates of effects.
The volume of qualitative evidence presented in PRMHT studies in general is relatively limited. Even within the DHSC-funded evaluation of pilots, some of the subthemes are based on statements from just one or two individuals. Studies did not provide interview transcripts or other means of interrogating the primary study data, and it is unclear whether or not important perspectives have been overlooked. Given the lack of richness of data in individual studies, it was not possible to construct definitive context–mechanism–outcome configurations explaining why the observed outcomes developed as they did and how interventions reacted to underlying mechanisms and in what contexts. However, it was possible to identify recurrent themes across studies and further refine themes identified in previous research.
The methodological quality of included qualitative studies was evaluated using the CASP checklist. This used studies as the unit of analysis rather than outcomes, as is done in the more comprehensive GRADE-Confidence in the Evidence from Reviews of Qualitative Research (CERQual) approach.55 However, given the relatively small volume of evidence and the rapid nature of the current synthesis, CASP is likely to have captured the main relevant quality issues.
As well as the methodological limitations described above, the nature of the different sources of evidence (e.g. systematic reviews, primary quantitative studies and primary qualitative studies) precluded us from being able to consistently structure our synthesis across sections; in some cases, evidence could feasibly be presented only by intervention, and in others only by outcome.
Owing to time and resource limitations, our project advisory group included police and research professionals only. The lack of advisory group input from MHPs and service users should be considered a limitation.
Implications for future research
Although there is published evidence that aims to describe and evaluate various models of PRMHT interventions, most evaluations are limited in scope and methodologically weak.
As we have seen, most evaluations of PRMHT interventions have focused on the use of S136 and HBPOS as primary outcomes. At the same time, separate legislation4 and capital funding schemes7,8 have been directed specifically at ensuring that there are improvements on these outcomes. However, it is clear that PRMHT interventions have wider objectives and potential benefits.
All the systematic reviews in the synthesis, published between 2016 and 2018, recommended higher-quality evaluations of PRMHTs that measure outcomes beyond simple procedural measures, such as S136 rates.11,12,14,15,39,40
With the exception of one RCT of training,47 we have not identified any strong additional evidence in the most recent evaluative primary studies. These recent studies have echoed the recommendations of the systematic reviews, calling for prospective, comprehensive and streamlined collection of a wider variety of data to evaluate the impact of PRMHT interventions. In particular, there is a lack of information on the impact of these interventions on individuals in terms of their mental health and quality-of-life outcomes.
One frequently raised issue was that of people who often and repeatedly come into contact with services via the police. This is one important objective on which PRMHT interventions (and evaluations) could potentially focus. Some existing schemes already aim to work proactively with various partners (e.g. health, social care, local policing teams, ambulance service) on this issue.
Barriers to collecting such data may include the time and resources needed to obtain necessary ethics approval and the nature of routine police data collection. These and other methodological and practical challenges have been discussed by the authors of the only RCT in this area.48 They recommend that police IT systems should be designed with operational and research purposes in mind, as data (which may not be fully accurate or complete) is currently stored on a number of different and poorly integrated systems. Problems around the collection of police-related data fall into three categories: (1) the nature of the data being collected (e.g. routine procedural data vs. person-centred outcomes), (2) data collection procedures (e.g. how data collection has an impact on the workload of front-line police officers) and (3) data analytic considerations (e.g. the accessibility and usefulness of data for research). In order for researchers to understand the context in which interventions are being implemented, Scantlebury et al.48 appropriately recommend collaboration between the police and academia, with police officers embedded within trial management groups.
The available qualitative data around the implementation of PRMHTs are also relatively sparse, but provided clearer insights into possible future directions for research. Although police and health staff directly involved with PRMHTs seem to consider these interventions to be of value, the objective impact of PRMHTs (especially on wider services) is less clear. Although some, very limited, evidence suggests that PRMHT schemes may result in an overall saving of police resources, this same evidence paints a more mixed picture in terms of NHS resources, with the potential for increased resource use as well as savings. This, perhaps, should not be surprising given that co-response/consultation PRMHT models require the allocation of MHPs from elsewhere in mental health services. However, whether the impact of PRMHTs on case detection and referral pathways increases or decreases, the demand on health services remains unclear.
On the basis of the evidence included in this rapid evidence synthesis, future evaluations would be more informative if they addressed the following:
Clearly articulate the objectives of the PRMHT intervention.
Articulating the logic model underpinning the intervention may help illustrate the relationship between objectives and outcomes.
Involve all stakeholders (including people with mental health issues) in the design and evaluation of interventions, including the identification of these objectives.
Collect and analyse outcomes that relate directly to the stated objectives. Quantitative data should extend beyond S136 rates, places of safety and process data, to measuring the outcomes that are most important to the police, mental health and social-care services, and individual service users. These might provide greater insights into:
quality and timeliness of assessment, referral and treatment
mental health outcomes for service users
experience of services for service users
level of service engagement after encounters with PRMHT
characteristics and needs of people who frequently and repeatedly come into contact with services via the police
changes in case-finding and access to health services (e.g. mental health, substance misuse, sexual health and contraception)
demands on police resources and police officer time
demands for community mental health services
rates of hospitalisation via A&E or acute mental health services
costs and savings to health and police services.
Similarly, evaluations should take into consideration the shorter-, medium- and longer-term effects of PRMHT interventions, for example by evaluating the consequences of PRMHT referrals on individuals beyond the initial number and type of referral.
Evaluations of services to understand and address the needs of frequent/repeat service users who create a disproportionately large demand on resources may be of particular value.
Researchers need to make realistic allowance for data collection in budget allocations for new studies.
As stated in previous publications, it is likely that better data collection processes will be needed. However, these processes should not be overly burdensome to front-line police or health staff.
When possible, study designs should have an appropriate concurrent comparator. Although street triage interventions have been implemented in a number of ways (e.g. incorporating co-response, consultation and information-sharing components to different degrees), there may be an interest to compare the pragmatic implementation of such an approach with ‘pre-arrest diversion’ models that emphasise specialist training of police officers over ongoing collaboration with MHPs.
The collection of qualitative data may help better understand which approaches work best and why, although such research should capture dissenting views as well as those of advocates.
Given their potential to both incur costs and accrue benefits across multiple services, any future cost-effectiveness analysis of PRMHT should take a multiagency perspective to understand the relative impact of introducing a particular model on the resource use across police, health and social services.
