INTRODUCTION

The use of digital technology in healthcare settings has increased over recent years and presents new opportunities for the delivery of physical and mental health services in the United Kingdom (Department of Health and Social Care 2019). In 2005, CCTV was first introduced as a technological surveillance tool for maintaining staff and patient safety in a range of healthcare settings (Desai 2009). Continued technological advances in this field have led to the development of wearable camera technologies referred to here as Body Worn Cameras (BWCs). These small transportable devices are usually worn on the outside of clothing, glasses, or headwear to produce video and audio recordings (The Metropolitan Police, 2020).

This technology has been in use by police in the UK since 2005 (The Home Office 2007) and it is estimated that over 70% of police forces have started to adopt the use of BWCs (Lum et al. 2020). In a police setting, research suggests BWCs may enhance transparency, supplement documentation, and deter illegal and inappropriate behaviours from both officers and citizens (Bureau of Justice Assistance 2015). However, there is a lack of evidence supporting their efficacy, cost effectiveness, and wider social impact. The only comprehensive systematic review of literature to date found that the use of BWCs by police officers had no significant impact on police (mis)use of force or assaults against officers (Lum et al. 2020).

Despite a lack of evidence to support the use of BWCs in public sector services, The National Health Service (NHS) has pledged to invest (UK) £8 million in pilot testing BWCs as a way to enhance staff safety and assist in prosecuting violence against staff (Department of Health and Social Care 2019). Since this announcement, we have seen the roll out of BWCs in ambulance services across the country (London Ambulance Service 2021) and initial pilot testing in inpatient mental health services (Ellis et al. 2019). Safety is a considerable issue within mental health services, where staff are 7.5 times more likely to report they have been attacked than staff in other NHS services (Royal College of Nursing 2018b). In 2020, 14.9% of staff in mental health trusts claimed that they experienced physical violence from service users or other members of the public (NHS Staff Survey 2020: National Results Briefing 2021). Despite the higher rates of physical assault against staff, only 4% of alleged physical assaults by psychiatric inpatients were reported to the police in 2018 and even fewer resulted in action taken (Doedens et al. 2020; Young & Ready 2016). In a mental health setting, BWCs allow staff to record situations where conflict may occur, and containment measures may be used. However, the use of BWCs in mental health settings is in its infancy and is a strongly contested intervention (Royal College of Nursing 2018a).

The high rate of assault against mental health nurses contributes to the high rate of staff turnover in mental health services (NHS Improvement 2019; Royal College of Nursing 2018b). Despite identifying the staffing crisis as a priority in the Five Year Forward View 2017–2019 report (Department of Health and Social Care 2017) and the NHS Long Term Plan (Department of Health and Social Care 2019), the number of mental health nurses increased by <0.5% over 2019 (Buchan et al. 2019) and recruitment and retention remain huge challenges (Launder 2020). Understaffing plays a key role in the deterioration of patient care (Baker et al. 2019), and BWCs may be implemented in an attempt to improve staff safety and ultimately improve patient care.

However, conflict and violence in mental health wards is a complex issue that affects both staff and patients (Bowers 2014; Fletcher et al. 2021; Kumar et al. 2001). Research and audits conducted in inpatient services in England revealed patients often experience verbal abuse, fighting, bullying, theft, racism, and sexual assault (Care Quality Commission 2018; Jones et al. 2010). Mental health staff have a responsibility to protect their patients from physical and psychological harm during their stay; yet, frequently staff interactions with patients can fuel conflict (Papadopoulos et al. 2012). Mental health nurses are the staff group most likely to be involved in face-to-face interactions with service users who may be highly distressed and/or frustrated by the restrictions typically imposed within inpatient settings to maintain safety, such as removing personal items, restricting smoking, or limiting movements (Bowers et al. 2015). There are currently evidence-based interventions, such as the Safewards model, which have been found to reduce incidents of conflict and use of containment measures (e.g., seclusion, restraint) on mental health wards (Bowers 2014). However, chronic understaffing can make it difficult to implement these effective interventions more widely (McAllister et al. 2019).

Ethical concerns remain central to wider debates surrounding the implementation of BWCs in mental healthcare settings (Royal College of Nursing 2018a). Those who require care in an inpatient mental health ward are often admitted against their will and are at the most vulnerable point in their patient journey (Care Quality Commission 2019). Further, Black and South Asian minority groups are disproportionately detained under mental health legislation in the UK, and Black patients are more likely to be subject to measures to maintain safety, such as use of physical restraint (Barnett et al. 2021; Payne-Gill et al. 2021; Rodrigues et al. 2020). Implementation of BWCs would likely have the greatest impact on this already vulnerable group.

While policy makers hope this new technology will bring improvements to the delivery of mental health services, it is essential that patients receive care based upon the best current evidence in conjunction with clinical expertise and patient values (Reid et al. 2017). A small number of BWC evaluations in mental health wards in England have been undertaken (Ellis et al. 2019; Hardy et al. 2017), but given their relatively small scale and localised focus, a wider review is required. The lack of research on BWCs in a mental health setting means it is essential to draw upon the wider literature in the public sector to explore its effects and consequences. A systematic review provides the opportunity to appraise and synthesise existing evidence across the public sector and make recommendations regarding the potential future use of BWCs in a mental health environment (Pati & Lorusso 2018).

AIMS

This review seeks to answer the research question: Are BWCs likely to enhance safety in mental health inpatient wards based on the literature regarding BWC use across a range of public sector services? The primary objective is to identify and systematically review literature relating to the use of BWCs in the public sector, assessing where/how BWCs are being deployed, the methods used to conduct such research, and the effect of BWCs in these settings. The findings and implications associated with BWCs in public sector services will be discussed in relation to their suitability for mental health services.

METHODS Design

We conducted a systematic review using narrative synthesis in accordance with the Guidance on the Conduct of Narrative Synthesis (Popay et al. 2006). This approach facilitates synthesis of a range of methodologies and study designs and allows a focus on the wide range of BWC applications in public services. A meta-analysis of this literature was not a suitable way to address the research questions proposed in this review (Borenstein 2009). For the purposes of this review, public sector services are defined as central government, local government, and public corporations delivering services to citizens, including healthcare, law enforcement, and public transportation (Office for National Statistics 2019).

Protocol and registration

The review protocol was registered with PROSPERO before commencement (CRD42020164878).

Eligibility criteria

Studies were only included where they met the following criteria: BWC was a standalone, visible device able to provide continuous video playback; BWC was attached to a member of staff working in a public sector service; BWC was used to record face-to-face interactions with the public or service users; and the study reported the methodology for evaluation. Studies were excluded if the BWC was used as a data collection tool; BWC was used outside the public sector; or BWC use was reported in an internal evaluation without reporting the methodology.

Database coverage

A search of all relevant literature was undertaken with librarian assistance using the following databases: Medline (via Ovid, 1966 to 25/05/2021); Embase (via Dialog 1974 to 25/05/2021); PsycINFO (via Datastar 1806 to 25/05/2021); Global health (via Ovid, 1966 to 25/05/2021); HMIC (via Ovid, 1979 to 25/05/2021); Web of Science (via Clarivate 1975- 25/05/2021); Sage Journals (via Datastar 1994- 25/05/2021); OpenGrey (1972 to 25/05/2021) and Google Scholar (2004- 25/05/2021). No limits or filters were imposed. Searches were supplemented by reference list screening and BWC websites were reviewed for publications (Reveal, Axon, Google Glass, Calla, VIEVU, Panasonic and Puma). Email alerts from all journals and RSS feeds from camera websites were utilised to ensure identification of new articles.

Search strategy

A keyword search was applied to all databases using the search terms; Body adj3 camera* OR Wearable video OR Wearable camera* OR Body worn video. An initial scoping search revealed most journal articles had yet to be indexed under a MeSH term, therefore, MeSH terms were omitted from the facet analysis.

Study selection

In May 2019, the search was conducted by JE. In May 2021, the search was repeated by KW to capture new research that was published in the time elapsed since the initial search. PRISMA reporting guidelines were followed, and a full PRISMA chart can be found in Figure 1. Referencing software Zotero was used to de-duplicate, and articles were screened in Microsoft Excel. To ensure unbiased selection, title and abstract screening was undertaken by two independent researchers (First search: JE, UF; Second search: KW, UF). Full text screening utilising the same criteria was undertaken. Full text exclusions were documented along with reasons. Disagreements at any stage were resolved by a third reviewer (AS; n = 6).

PRISMA flow diagram (Moher et al. 2009). Data collection

Data was extracted using a pre-designed data collection tool included in the PROSPERO registration. The tool includes country, study design, sample size, setting, main method, camera user, recording subject, camera model, comparison arm, study length, outcome measures, and funding source. The form was piloted on 10 studies during the initial search to ensure suitability. No amendments were made. No limit on summary measures were implemented and all relevant results were collected. Efforts were made to contact authors for missing information wherever possible.

Quality appraisal

All studies were reviewed for selection, performance, attrition, detection, and publication bias. The Joanna Briggs Institute (JBI) Critical Appraisal tool was utilised (Aromataris & Munn 2020). Critical appraisal of all included studies was undertaken individually by JE and UF. No discrepancies were apparent. Individual scores attained from the tool were then collated to review bias across studies.

Analysis

Results were divided into healthcare, law enforcement, and public transport sectors and analysed using a narrative synthesis approach according to an established framework (Popay et al. 2006). With this approach, main findings from the quantitative studies in each sector were summarised and further supplemented with examples from the qualitative studies. KW, JE, and ME identified the effects, outcomes, and experiences of BWCs from the qualitative data.

RESULTS Study characteristics

An overview of the 52 studies included in this review can be found in Appendix A. Most were conducted in law enforcement settings (including police officers, prison guards, and traffic wardens; N = 43), followed by healthcare (including physicians, frontline clinical staff, paramedics, and family carers; N = 8) and transportation (railway ticket inspectors; N = 1). There was a wide range of study designs, such as randomised control trials, pilot evaluations, and qualitative interviews; however, not all studies explicitly reported their chosen design.

Participants

BWC research is comprised of various units of analysis reported alongside participant population data. A total of 10 articles reported the number of cameras used in the study (7 in law enforcement, 2 in healthcare, 1 in public transportation). Participant figures are more commonly reported than number of BWCs due to frequent sharing of a limited number of cameras across entire departments. Across all studies, 46 reported a specific count of participants (6 in healthcare, 39 in law enforcement, 1 in public transportation). Other units of measurement reported in lieu of participants or number of cameras include shift patterns (Ariel et al. 2015), contacts or interactions with police (Young & Ready, 2016), geographic regions (Grossmith et al., 2015; Mitchell et al. 2018), caregiving pairs (Matthews et al. 2015), and physician consultations (Gupta et al. 2016).

BWC intervention

The model of camera used also varied widely across the studies. Police studies primarily used TASER cameras; however, 23 of these studies neglected to declare a model. One study (Ho et al. 2017) used TASER cameras in a healthcare setting with paramedics. However, Google Glass (4) and Calla (2) were most common in healthcare settings. There was also variability in protocol reporting. Only 17 out of 52 (32%) studies reported recording protocols, and these were primarily law enforcement. The practice of camera wearers verbally announcing recording varied and a range of visual warnings to alert others varied from audible sounds, flashing lights and activation of screens. Camera activation was either at staff discretion or mandatory for the duration of a shift.

Research quality

All studies were reviewed for selection, performance, attrition, detection, and publication bias. The JBI Critical Appraisal tool was utilised. Critical appraisal of the included studies was undertaken by three researchers (JE, UF, KW). No discrepancies were apparent. Individual scores attained from the tool were then collated to review bias across studies and assess quality. Overall, 30 out of the 52 studies included in this review were of poor quality, with a high risk of bias. Only five studies displayed low enough risk to be classified as good quality research. Quality assessment can be found in Appendix A.

Funding and costs

A total of 19 studies reported funding sources. A further six studies did not receive funding, and the remaining 30 studies neglected to report on funding sources. Of the 19 reports of funding, 8 were funded by a federal government body (e.g., Bureau of Justice, Home Office, US Department of Justice), and 4 were funded by other government bodies such as local task force and police budgets. Five studies (Ariel et al. 2015; Ariel 2016b; Ellis et al. 2019; Hardy et al. 2017; Mitchell et al. 2018) reported receiving camera equipment free of charge from the company. Overall, law enforcement studies account for 80% of studies with undeclared funding.

Three studies (Braga et al. 2018; Hardy et al. 2017; ODS Consulting 2011) report on the total costs of BWC implementation, including hardware, software, and training. On average, the cost per BWC was £1,750. However, none of the studies in this review included an economic analysis in which wider consideration of resource use, training costs, health benefits, and opportunity costs can be calculated and analysed.

Outcomes in transportation sector

Only one study examined the use of BWCs by staff at railway stations across England (Ariel et al. 2019). The results from this randomised controlled trial indicated a 47% reduction in odds of assaults against staff when wearing BWCs. This article positions BWCs as beneficial to employee health and safety by reducing rates of violence against staff.

Outcomes in law enforcement sector

Most studies included in this review were from law enforcement settings (43 out of 52). Twenty of those studies examined the impact of BWCs on one or more of the following outcomes (Table 1): police behaviour (use of force, arrest rates), citizen behaviour (complaints, assault against officers). Combinations of these four variables were the most frequently examined outcomes. Additionally, 12 studies examined both civilian and police opinions on BWCs. Other less frequently examined outcomes were court processes (Morrow et al. 2016; Owens et al. 2014), traffic stops (Peterson et al. 2018), stop and frisks (Young & Ready 2016), response time/time spent on scene (Wallace et al. 2018), camera activation (Roy 2014; Young & Ready 2016), and public reporting of crime (Ariel 2016a).

TABLE 1. Impact of BWC on police use of force, arrest rates, citizen complaints, and officer assaults Paper Use of force Arrests Complaints Assaults Ariel et al. (2015) −61.4%*** – −91.3%** – Ariel (2016b) No significant change – –14%* – Ariel et al. (2017) – – −93%*** – Ariel et al. (2018) – – −61%* Braga et al. (2018) −40.7%* + 5.2%** −30.2%* – Braga et al. (2019) –63.6%* – –50.5%* – Ellis et al. (2015) – – −15%† – Gaub et al. (2020) No significant change – – – Groff (2020) −38.3%*** – −39.2%*** – Grossmith et al. (2015) – No significant change No significant change – Headley et al. (2017) – −16.1%* – No significant change Hedberg et al. (2017) – No significant change −96%*** – Huff et al. (2020) +0.10%** + 34.8%** No significant change – Jennings et al. (2017) −8.4%† – −65.4%† – Rankin (2013) –75%† – –40%† – Mitchell et al. (2018) – – −86%*** – Morrow et al. (2016) – + 6.6%* – – ODS Consulting (2011) – – – Decrease† Ready and Young (2015) – −6.9%* – – Peterson et al. (2018) No significant change No significant change No significant change No significant change Pope et al. (2020) No significant change No significant change – – Wallace et al. (2018) – No significant change – – White et al. (2018) No significant change – No significant change – Yokum et al. (2017) No significant change – No significant change – *p < 0.05; **p < 0.01; ***p < 0.001. – Not measured. † Significance level not reported. Officer Behaviour

Two studies reported an increase in officer-initiated contact (Wallace et al. 2018; Young & Ready 2016), and a third reported a decrease (Huff et al. 2020). Although (Young & Ready 2016) reported an increase in officer-initiated contact, they also reported that officers were less likely to perform stop and frisks while wearing BWCs. However, BWCs were not found to impact response time or time spent on scene once contact was initiated (Wallace et al. 2018). Research also shows a mix of statistically significant increases (Braga et al. 2018; Huff et al. 2020, Morrow et al. 2016) and decreases (Headley et al. 2017; Young & Ready 2016) in arrest rates, and one study reported no change at all (Wallace et al. 2018). No conclusions can be drawn about the impact of BWCs on officer-initiated contact or arrest rates based on this literature.

Four studies reported significant decreases in police use of force resulting from BWC use (Ariel et al. 2015; Braga et al. 2018, 2020; Groff 2020). Two further studies also reported a slight decrease in use of force but neglected to report the statistical significance of this change (Jennings et al. 2017; Rankin 2013). Additionally, one study reported a decrease in use of force rates only when controlling for compliant handcuffing (Henstock & Ariel 2017). However, (Huff et al. 2020) reported a small but significant increase in use of force following BWC activation. There appears to be a trend toward decreased use of force after implementing BWCs, but inconsistencies in methods and measures across the literature make it difficult to draw solid conclusions.

Officer opinions

When asked about BWCs, the ten studies examining police perspectives report mixed opinions. Some studies have found police officers believe the cameras are helpful and should be implemented (George & Meadows 2016; Pelfrey & Keener 2016; Ready & Young 2015), while others report neutral and negative leaning opinions (Hyatt et al. 2017). For example, officers in one study believed that BWCs enhance the quality of evidence, particularly in prosecuting domestic violence cases (Gaub et al. 2016). However, another study reported contradictory findings in which officer perspectives on BWC helpfulness for interpersonal violence prosecution was less favourable after BWC implementation (Morrow et al. 2016).

Qualitative studies examining police officer perspectives highlighted some of the perceived benefits of BWCs. One common belief was that BWCs can protect officers against citizen complaints:

You give the command, the dog pops off and comes back to you, and that’s captured on camera. I mean that’s gold to us. Later down the road if there is any civil litigation, it’s there. It’s captured for the argument that the dog stayed on too long. Canine Officer (Gaub et al. 2020)

…don’t think they will create any problems for the officers although they may hold some officers to a higher standard of professionalism, will assist in stopping complaints about officers if they occur. Police officer (Makin 2016)

Officer beliefs about the impact of BWCs on civilian behaviour were mixed. One study found that while some officers believed the cameras would benefit the officers, they did not believe the cameras would enhance officer safety or change civilian behaviour (Pelfrey & Keener 2016). This contradicts the results of another police survey in which officers reported beliefs that BWCs do change civilian behaviour (George & Meadows 2016). Similarly, a BWC pilot evaluation in the UK found that prison staff showed increased perceptions of safety after BWC implementation (Pope et al. 2020).

There have also been contradictory findings on whether police officers feel the implementation of BWCs has impacted their own behaviour. One qualitative study demonstrates possible positive behaviour changes resulting from BWC use:

This was a pretty straightforward situation. But when you know, you are having a bad day, or you are in an intense situation you know? I would normally maybe let a curse word fly. And we are not allowed to curse at citizens, but you know, we all do that from time to time. Now, I tend to watch myself a little more. Patrol officer (Koen et al. 2019)

However, George and Meadows (2016) found that officers believed the cameras would improve the behaviour of their colleagues, but not their own.

These studies also highlighted perceived challenges and limitations to BWC use. Volume of calls, technical limitations of recording and battery capacity, and documentation procedures were all identified as problems (Gaub et al. 2020). Another concern raised by officers was that BWCs would be used as a tool to control and discipline the officers:

Regardless of statements used by management, the primary use of these videos will be to police the officers’ actions and to be used for disciplinary actions.Police officer (Makin 2016)

In addition to concerns about BWCs being used to discipline officers, another officer raised a concern about officers manipulating footage:

Cameras have limited abilities and perceptions and could be used negatively by outside influences to create problems. How many times have you seen video where the clips were cut to create the perception someone wanted to portray, rather than the entire circumstance? Police officer (Makin 2016)

Ultimately, officer opinions appear to be the most robust and widely researched aspect of BWC implementation in law enforcement. The opinions of officers vary widely, and qualitative research has captured this in detail.

Citizen behaviour

Rates of citizen complaints were frequently reported alongside police use of force and arrest rates. Despite being a measure of citizen activity, complaints are most frequently operationalised as a measure of officer behaviour. Ten studies reported decreases in citizen complaints; however, three of those studies neglected to report statistical significance of that decrease (Ellis et al. 2015; Jennings et al. 2017; Rankin 2013).

Methods for reporting changes in officer assaults varied across the five studies which examined this outcome. A pilot study of BWC use by a police force in Scotland reported a decrease in assaults with no statistics to support this statement (ODS Consulting 2011). Another study also reported a significant difference (Ariel et al. 2018), but this is contradicted by an earlier study which reported an increase in assaults against officers (Ariel et al. 2016). A further two studies reported no significant changes (Headley et al. 2017; Peterson et al. 2018). Additionally, one study reported assaults by comparing the frequency of change in assault rates per prison, rather than reporting specific changes in actual assault rates (Pope et al. 2020); therefore, this study cannot be directly compared to studies that reported figures for assaults. This study reports that prison sites with BWCs experienced fewer officer assaults after implementation; however, this was statistically insignificant. It is unclear from the existing