Study type

A scoping review following the Joana Briggs Institute (JBI) recommendations and the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) checklist was conducted [19–22]. A protocol for this review was registered on the Open Science Framework [23].

Inclusion and exclusion criteria

This review focused on studies examining suicide and self-harm surveillance systems. The inclusion criteria encompassed original studies investigating surveillance systems for suicide and self-harm, using qualitative, quantitative, or mixed methods approaches. No filters restricting publication date or language were used at this point. However, the search for grey literature was only conducted using strategies in English, Portuguese, and Spanish. When deemed relevant to the research question, studies not available online were retrieved through either a) university-affiliated Virtual Private Networks (VPNs) for access to subscribed scientific journals or b) direct contact with the studies’ authors. These measures were employed to mitigate limitations arising from a potentially limited pool of available studies.

Studies were excluded from the analysis process when: a) they discussed or focused on euthanasia and assisted suicide; b) they did not explicitly describe suicide and self-harm, limiting the understanding of the field of study and the methodological design; c) they did not delineate the characteristics of surveillance systems (e.g., studies disseminating epidemiological data collected via surveillance systems or providing general recommendations for systems); and d) they explored the reporting of suicide and self-harm in the media or to families and friends rather than the registration of cases within surveillance systems.

Search strategy

To identify relevant studies, a comprehensive three-step approach was used (initial search of terms in databases and Mesh terms, retrieval of results from databases, and the search for results in grey literature) [19–21]. Initially, a preliminary search, in collaboration with a librarian, was conducted on PubMed (NCBI). This initial search aimed to identify relevant keywords in study titles and abstracts, enabling the formulation of the final search strategy. The identified keywords were strategically combined using Boolean operators (AND/OR/NOT) in a systematic search aligned with the research question and a mnemonic for population, concept, and context—PCC framework developed specifically for this study (S1 Table).

The final systematic search was conducted on 20^th June 2023, in the following databases: Cinahl, Embase, Lilacs (Latin American and Caribbean Health Sciences Literature), PubMed (US National Library of Medicine), Scopus, and WebScience. Additionally, a search was done on Google Scholar to identify grey literature. For Google Scholar and Lilacs, the authors also used terms in Portuguese and Spanish alongside English, considering the particularities of both databases. Due to its coverage, only the first 200 results were extracted from Google Scholar [24]. Terms in Portuguese and Spanish were added to the Lilacs and Google Scholar searches. References were identified and screened for eligibility (n = 2,801). To ensure a comprehensive search, two researchers independently screened the references, facilitating the identification and inclusion of additional relevant studies.

Source of evidence screening and selection

Identified references were imported into the EndNote reference manager, and de-duplicated, before being transferred to Ryyan software for screening. To ensure a rigorous and impartial selection process, two independent researchers (AV and AS) assessed titles and abstracts against a predefined inclusion criteria [25]. Discrepancies or disagreements during the screening process were resolved through discussions between the two researchers without requiring the involvement of a third party.

Data extraction and analysis

Two researchers (AV and AS) independently extracted data using a charting instrument based on the JBI Template Source of Evidence Details, Characteristics, and Results Extraction Instrument [19, 26]. The extraction chart included a) study details, such as characteristics, year of publication, country, context, and other pertinent information, and b) an adapted structure for collecting data on specific components. This study used components related to system characteristics (e.g., data custodians, data items, data format, data security, privacy, and confidentiality). Any discrepancies that emerged during the process were resolved upon completion.

The data were analyzed using Content Analysis [27], involving pre-analysis steps, material exploration, categorization, treatment, and interpretation of results. The data were presented based on evidence details, characterization components, and descriptions of surveillance systems. The results are presented descriptively following PRISMA-ScR guidelines, using charts, figures, and tables as necessary.

Surveillance systems for general behaviours (including suicide and self-harm)

Fifteen surveillance systems covering general behaviours, including suicide and self-harm, were identified. Most of these systems addressed health-related behaviours related to injuries broadly [31, 43, 44, 48, 49, 52], injuries due to violence [36], health-related incidents [15, 36, 39], poisoning, external causes, suicide [34], mental health [30, 33], mortality [39, 50], mortality due to violence [45], or non-natural causes [51]. These systems predominantly operated at a national level for data collection [15, 31, 33, 34, 36, 39, 43, 44, 49, 51], primarily in a healthcare context [30, 31, 33, 39, 41, 43, 44, 48, 49, 50, 52], and specifically within emergency departments [30, 44, 48, 49, 52].

References that focused on surveillance systems in emergency departments highlighted challenges in defining precise variables for recording, data incompleteness [31, 44, 49, 52], or incorrect data entry [49]. Additional challenges included: poor hospital engagement in the surveillance system’s performance improvement [31], a lack of training and capacity building for human resources [49], the absence of records for injured individuals referred to other services [44], and limited dissemination of data [31, 44, 49] (S2 Table). In this context, the importance of coding for case recording, the use of protocols and reference materials for consultation and training, and team awareness regarding the importance and necessity of data recording were emphasised [48].

The systems focused on non-fatal outcomes [15, 31, 33, 36, 39, 48, 49, 52] or both fatal and non-fatal outcomes [30, 36, 40, 43, 44] and provided different definitions of the behaviour based on the intentionality of the acts (S1 Table). In most cases, socio-demographic, psychosocial, and psychiatric information was collected, often through clinical assessments conducted by medical professionals (Table 2). Most systems used the International Classification of Diseases—ICD [15, 34, 36, 39, 45, 50, 51] (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Main characteristics of surveillance systems for general health behaviours (including suicide and self-harm).

https://doi.org/10.1371/journal.pgph.0003292.t002

Three mortality surveillance systems were identified, including one initiative in the United States that collected data about violent deaths from primary systems [38, 45, 47]. A Brazilian initiative collected data from death certificates for mortality surveillance in the country [39, 50]. Both systems used the ICD as the classification criterion [38, 39, 50] and publicly disseminated data through websites and reports (DataSUS) [39, 50], annual webinar (CDC) and written reports [39]. Additionally, a South African initiative collected data from medicolegal investigations of non-natural deaths [51]. These highlighted the challenges in using different case definitions [39], coding and data entry [39, 49, 51], inconsistency in information sources [39], organisational acceptance difficulties, funding, and system continuity [51]. The papers on emergency department and mortality systems pointed out the challenge of assessing data sensitivity [49, 51].

Furthermore, a Brazilian surveillance system for violence and accidents [36] was identified, gathering suicide attempt data nationally with two components: continuous surveillance based on information systems [15, 36, 39] and sentinel surveillance with epidemiological surveys. Continuous surveillance was conducted in an intersectoral context involving all three levels of the healthcare system (primary, secondary, and tertiary), social assistance, education, child protection councils (permanent and autonomous bodies responsible for ensuring children’s and adolescents’ rights compliance), rights councils, protection councils, justice councils, as well as governmental, non-governmental, and private sector organisations. Also, an international system in Central America and the Dominican Republic collected data about mental in primary care for planning interventions at different levels (national level in Central American and regional level in Dominican Republic) [30] (S2 Table).

Specific suicide and self-harm surveillance systems

Fifteen references described systems specifically for suicide and/or self-harm surveillance. These systems expanded the identification of cases with non-fatal outcomes (ideation, threat, nonsuicidal self-injury—NSSI, attempt) and had varying levels of data collection (state-based, district, regional, and local). Most of these systems were situated in a healthcare context [35, 37, 41, 53], especially in emergency departments [40, 41, 53], and one system for war veterans, a perceived high-risk group in the United States [37]. Additionally, there were mortality systems based on coroner’s investigations [26, 29]. Most systems developed their own classification criteria [26, 28, 32, 37, 41, 42, 46], with a notable emphasis on using Rosemberg et al.’s (1988) criteria (operational criteria to assist coroners and medical examiners in determining suicide) and consensus procedures. In cases of suspected suicide, most systems used coronial investigations (Table 3 and S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Main characteristics of specific suicide and self-harm surveillance systems.

https://doi.org/10.1371/journal.pgph.0003292.t003

Systems using coronial investigations [26, 29] collected cases labelled as "suspected suicide", "open verdict" or "identity not yet confirmed," primarily from coroners who conducted investigations based on various sources of information (medical records, police reports, witness statements, family reports). Cases were reviewed and classified after the coronial investigation or inquiry [26]. The coronial systems require nine months to two years to complete an investigation process. In this sense, the real-time suicide surveillance described in the study contributes to the sharing and analysis of data (with different timelines among the systems) to aid in the planning and implementation of immediate suicide prevention actions.

This data was used for surveillance and local responses to suspected suicides (CDS), early responses to groups (clusters or contagion) (CDS, TV-RT-SSS), investigation of specific clusters or subgroups (iQSR), and facilitating early responses to suicides, associated contagions, and grieving communities (SSHO). Subsequently, the data is reviewed based on coroners’ verdicts, but the emphasis of the real-time surveillance system is on the potential for early prevention strategies. The systems collectively reported high sensitivity and satisfaction, serving as a model suitable for replication in other locations [26]. Studies also highlighted the use of police-led systems for data collection, as law enforcement is primarily called in cases of suspected suicides [29]. However, concerns were raised regarding the need for education and destigmatisation and concerns about police involvement in mental health actions [26].

A national integrated health system, including data from hospitals, fire departments, social services, and education systems, was described in Tawan [46]. This system aimed to identify suicide attempts and refer individuals to mental health or social services, support families (high-risk groups), and provide data for tailoring prevention strategies. However, reluctance or lack of motivation to record cases, the need for feedback and organisational incentives, alternative recording channels, and administrative support were identified as challenges [46].

In community initiatives, particularly those areas with diverse socio-cultural backgrounds (indigenous or remote communities), integration of different social resources (medical, school, and social service personnel, first responders, religious leaders, family members, and peers) was emphasised [28]. Data collection in community systems also extended beyond socio-demographic, psychosocial, and psychiatric characteristics to include cultural aspects (e.g., access to housing and traditional ceremonies) [28, 42].

For example, Celebrating Life, a community-based surveillance in the Fort Apache Indian Reservation region (on the border of New Mexico and Arizona, United States), collecteds data from individuals, schools, and other social sectors through admission forms. These forms were submitted to the Celebrating Life team, which conducted follow-up visits, risk identification, and monitoring. In cases of suicide, the suicide death form was completed by the system’s responsible team. Event classification was based on declared intent, congruence of method with intent, method lethality, reported behaviour function, and simultaneous substance use. Behaviours could also be confirmed by accessing other information sources (police reports, IHS medical records, local providers, and first responders) and through consensus procedures within the team [42].

An implementation study of a suicide surveillance system in 124 villages in Mehsana, North Gujarat, India, utilized data triangulation from key informants in the community, health records, and police records. The findings from triangulating data collected through community sources, health services, and police records revealed a noteworthy number of cases identified by the community that had not been documented in health and/or police systems [32].

Studies on community-based surveillance systems highlighted collaborative development involving identifying and engaging partners, available active resources, and recognizing cultural assets such as traditional individuals, natural helpers, ceremonies, and elders familiar with traditions and language [28]. Sustainability factors included community involvement, funding, organisational capacity, recognition, communication, partnerships with clear expectations, assigned responsibilities, actions aligned with partner capacity and policies, and evaluation [28].