Objectives Patients in inpatient mental health settings face similar risks [eg, medication errors] to those in other areas of healthcare. In addition, some unsafe behaviours associated with serious mental health problems [eg, self-harm], and the measures taken to address these [eg, restraint], may result in further risks to patient safety. The objective of this review is to identify and synthesise the literature on patient safety within inpatient mental health settings using robust systematic methodology.
Design Systematic review and meta-synthesis. Embase, Cumulative Index to Nursing and Allied Health Literature, Health Management Information Consortium, MEDLINE, PsycINFO and Web of Science were systematically searched from 1999 to 2019. Search terms were related to ‘mental health’, ‘patient safety’, ‘inpatient setting’ and ‘research’. Study quality was assessed using the Hawker checklist. Data were extracted and grouped based on study focus and outcome. Safety incidents were meta-analysed where possible using a random-effects model.
Results Of the 57 637 article titles and abstracts, 364 met inclusion criteria. Included publications came from 31 countries and included data from over 150 000 participants. Study quality varied and statistical heterogeneity was high. Ten research categories were identified: interpersonal violence, coercive interventions, safety culture, harm to self, safety of the physical environment, medication safety, unauthorised leave, clinical decision making, falls and infection prevention and control.
Conclusions Patient safety in inpatient mental health settings is under-researched in comparison to other non-mental health inpatient settings. Findings demonstrate that inpatient mental health settings pose unique challenges for patient safety, which require investment in research, policy development, and translation into clinical practice.
PROSPERO registration number CRD42016034057.
- patient safety
- mental health
- inpatient settings
- systematic review
This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported [CC BY 4.0] license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: //creativecommons.org/licenses/by/4.0/.
- patient safety
- mental health
- inpatient settings
- systematic review
Strengths and limitations of this study
This is the first review to examine patient safety within inpatient mental health settings that uses robust systematic methodology.
The use of a robust patient safety taxonomy provides a comprehensive list of all incident types and resulted in a wide coverage of publications in terms of setting, country and population.
This review only included peer-reviewed studies with primary data.
The last systematic literature search was conducted on 27 June 2019, meaning that literature published since this date will not have been included.
Introduction
Patient safety has been defined as the ‘avoidance, prevention and amelioration of adverse outcomes or injuries stemming from the process of healthcare’.1 Those receiving care in inpatient mental health settings face similar risks [eg, medication errors] to patients in other areas of healthcare. In addition, some of the unsafe behaviours associated with serious mental health problems [eg, self-harm], and the measures taken to address these [eg, restraint], may result in further risks to patient safety.2–6 There may also be a tension between maximising patient safety and maintaining patient autonomy. Inpatient services will often include patients who are experiencing high levels of mental distress and are therefore at greatest risk.
While mental health research has focused on components of quality of care, published research lacks focus on the science of patient safety7–9; the stigma and discrimination associated with mental health problems may contribute to this relative neglect.7 Only two reviews have examined patient safety in a mental health context and described factors that influence patient safety.7 10 These reviews highlighted the complexity of patient safety in mental health, including the importance of wider organisational safety culture. While these reviews offer important insights into this complex topic, only a small number of specific patient safety incidents and concepts were examined. As such, the current breadth and depth of patient safety research in inpatient mental health settings is unknown.
The review presented here is exploratory in nature; building on previous reviews, we aimed to report an overview of the existing research base on patient safety in inpatient mental health settings. We also aimed to critically reflect on quality and methods used in included studies in the field.11 In addition to our original protocol,11 we aimed to collate, describe and construct the main research categories, allowing for an easily accessible reference index.
Search strategy and selection criteria
A systematic search was developed in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] guidelines.12 The protocol for this systematic review has been published elsewhere.11
Six databases were searched: Embase, Cumulative Index to Nursing and Allied Health Literature [CINAHL], Health Management Information Consortium [HMIC], MEDLINE, PsycINFO and Web of Science. The search was originally conducted on 5 April 2016 and then updated on 27 June 2019 using a comprehensive list of search terms [n=343] related to ‘mental health’ [n=73], ‘patient safety’ [n=206], ‘inpatient setting’ [n=13] and ‘research’ [n=51]; see online supplementary files 1 and 2 for full search criteria and terms. The search terms included in the ‘patient safety’ facet were based on the National Reporting and Learning System [NRLS] taxonomy for England and Wales13 to ensure all incident types were identified in the search. A Google Scholar search using the main search terms was also conducted; it was originally anticipated that the first 20 pages of Google scholar would need to be screened against criteria,11 but screening stopped at five pages as no new publications were retrieved. Similarly, we had anticipated hand-searching references of all included papers within the review. However, due to the large number of papers included in the review, only the reference lists of the two existing systematic reviews were searched for additional references.
Supplemental material
[bmjopen-2019-030230supp001.pdf]
Supplemental material
[bmjopen-2019-030230supp002.pdf]
Five reviewers [BT, CR, LD, DD and SAr] screened all titles against the inclusion and exclusion criteria, with 10% independently screened by a second reviewer [split equally between BT, CR, LD, DD and SAr]. Full definitions and descriptions of these criteria can be found in online supplementary file 1 and the protocol published elsewhere.11 Inclusion and exclusion criteria were developed over several iterative rounds among the research team to ensure consistency between reviewers [online supplementary file 1]. Any disagreements between reviewers were resolved through discussion and an overall consensus was obtained. Agreement between reviewers was calculated using Cohen’s kappa,14 which is a widely accepted measure of inter-rater reliability.15 16 Full-text papers were assessed for inclusion by two reviewers from the research team [BT and one other from CR, LD and SAr]; a third reviewer [DD] was consulted if necessary.
Inclusion criteria:
Population: mental health inpatients;
Intervention/outcomes: patient safety outcomes;
Setting: inpatient setting;
Comparators: no restriction;
General inclusion criteria: empirical peer-reviewed studies with a clear aim or research question, that used primary data and written up in the English language between 1 January 1999 and 27 June 2019 [in line with the publication of the Institute of Medicine’s report ‘To Err is Human: Building a Safer Health System’].17
Exclusion criteria:
Population: centres on physical healthcare patients;
Intervention/outcomes: patient safety was not the central aim, research question or outcome
Setting: amalgamation of data from inpatient and outpatient settings [where inpatient sample cannot be separated out]; primary care, outpatient mental health services, community or social care settings and risk assessment tool reliability/validity checks;
Comparators: no restrictions;
General exclusion criteria: secondary data, not in English language, protocols, editorials, commentaries/clinical case reviews/‘snapshot’ studies of a patient group, book chapters, conference abstracts, audits, dissertations, epidemiological studies and reviews.
Quality assessment
Quality assessment was performed to give an overview of the methodological rigour of included studies and to support readers’ interpretation of the literature. Publications were not excluded based on poor quality because the review was purposively exploratory and all-encompassing. Quality was assessed by four reviewers [BT, CR, LD and SAr] using the tool derived by Hawker et al,18 to allow appropriate assessment of the wide variety of studies included in this review. The checklist by Hawker et al evaluates nine domains: 1] abstract/title; 2] introduction and aims; 3] method and data; 4] sampling; 5] data analysis; 6] ethics and bias; 7] results; 8] transferability and generalisability and 9] implications and usefulness. For each study, the nine domains were assessed using one of four quality categories: very poor [10 points], poor [20 points], fair [30 points] and good [40 points]. The scores for each study were then summed and divided by nine to get an average score.
Data extraction
Data were extracted by five reviewers [BT, CR, LD, DD and SAr] using a standardised form that included study design information, participant characteristics, intervention description and patient safety outcomes. Extractions were compared within the research team to ensure reliability. Only published data were extracted; study authors were contacted only for confirmation or information clarity. If the contact attempt was unsuccessful, the article was assessed in its current form.
Data synthesis
Studies were grouped into research categories through consensus. First, four research team members [BT, CR, LD and SAr] individually re-read the included full-text publications and assigned each one based on the main topic area [eg, aggression]. Second, each assigned topic area was checked by another team member to ensure reliability. Third, topic areas were grouped into broader research categories [eg, interpersonal violence] that best described the patient safety focus for easier navigation of the literature. Finally, these categories and the related subcategories [initially called topic areas] from the previous stage were finalised after group discussion and consensus was reached. This was to ensure mutual exclusivity and appropriate definition [table 1 and online supplementary file 3]. Where data allowed, meta-analysis was performed applying a random-effects model, specifically calculating pooled prevalence considering both between-study and within-study variances that contributed to study weighting. Pooled values and 95% CIs were computed and represented on forest plots. Statistical heterogeneity was determined by the I2 statistic, where 60% is high. Analyses were performed using Stata V.15 [StataCorp, College Station, Texas, USA].
Supplemental material
[bmjopen-2019-030230supp003.pdf]
View this table:
- View inline
- View popup
Table 1
Overview of study characteristic identified within each category
Patient and public involvement
Patients and the public were not involved in this study.
Results
The search resulted in 79 672 records [figure 1] and reduced to 57 637 after de-duplication. Titles and abstracts were screened and excluded if they did not satisfy inclusion criteria [BT, CR, LD, DD and SAr]. Ten per cent were then screened [n=5763] by a second independent reviewer [split equally between BT, CR, LD, DD and SAr], in line with guidance on improving decision making by including more than one person in this process19; good agreement was found between pairs of reviewers [κ=0.72]. A total of 4758 publications were subjected to full-text review [BT, CR, LD and SAr]. Two reviewers independently screened the full-text articles against inclusion criteria [BT, CR, LD and SAr]. The third reviewer [DD] was consulted 59 times. Substantial agreement was reached [κ=0.64]. From the full-text review, 4394 publications were excluded. Three hundred and sixty-four publications met the inclusion criteria and data were extracted [online supplementary file 4].
Supplemental material
[bmjopen-2019-030230supp004.pdf]
- Download figure
- Open in new tab
- Download powerpoint
Figure 1
Flow chart of studies.
Study characteristics
Table 1 provides an overview of the study characteristics . The publications spanned 5 continents and 31 countries. The three countries contributing the greatest number of studies were the UK [n=102], the USA [n=55] and Australia [n=32]. The included studies collected data from over 150 000 participants. Studies included staff [n=165; 45%], patients [n=120; 33%] and a mixture of staff, patients and/or carers [n=77; 21%]. Only one study focused solely on patient family members [