The occurrence of suicide clusters has been reported for more than a century ago (Popov, 1911). The occurrence of this phenomenon is very rare (Niedzwiedz et al., 2014; Too, 2017) with 1–5 % of suicides typically occurring in a cluster (Gould et al., 1990; Gould and Lake, 2013). Although no standard definition exists, suicide clusters are commonly described as a higher number of suicides than expected by chance, occurring at a relatively small temporal and/or spatial scale (CDC, 1988; Joiner, 1999; Too, 2017; Hawton et al., 2020). In particular, the so-called point clusters, or space-time clusters, correspond to a high number of suicides occurring in a specific area over a short time window (Joiner, 1999; Haw et al., 2013; Olson, 2013). In this definition, the spatio-temporal co-occurrence of suicides is the primary criterion for identifying clusters, without reference to the mechanism(s) explaining the phenomenon. This limitation arises from the challenges of linking explanatory psychological, social, or biological mechanisms to statistical cluster detection.

Exposure to suicide and/or shared socio-demographic characteristics can be behind the occurrence of clusters. In this context, the most common explanation for the occurrence of suicide clusters is the idea of contagion or transmission (Gould et al., 1989; Pirkis et al., 2024a). This idea comes from infectious disease epidemiology (Haw et al., 2013). From this perspective, exposure to suicide can trigger new suicides (Gould et al., 1989; Pirkis et al., 2024a). This exposure can be direct, through personal or close contact, or indirect, for example, through the media (Haw et al. 2013 and references therein). Cheng et al. (2014) proposed the concept of contagion-as-imitation as the one with the best heuristic utility for studying suicide clusters at both individual and population levels. However, beyond the utility of the contagion and imitation concepts, it is important not to oversimplify the phenomenon. Suicide and suicide clusters depend on multiple factors, so imitation cannot completely explain the co-occurrence of suicides, and the social determinants of suicide have sometimes been insufficiently studied (Gould et al., 1989; Pirkis et al., 2024b). For example, cluster members can share similar socio-demographic and/or clinical characteristics (Haw et al., 2013; Pirkis et al., 2024b), such as personal risk level, precipitating life events, the tendency to form assortative relationships, and the environmental and social contexts (Hazell, 1993; Joiner, 1999; Exeter and Boyle, 2007; Blasco-Fontecilla, 2013; Haw et al., 2013; Johnson et al., 2017; Yamaoka et al., 2020).

This last perspective has been summarized using the community trauma assessment model called “circles of vulnerability” (Lahad and Cohen, 1998). This approach can be used to determine the degree of emotional impact that a negative event has on a community (Lahad and Cohen, 1998; Zenere, 2008, 2009). This method has also been used to explain the occurrence of suicide clusters and their impact on the community (Zenere, 2008, 2009). This method identifies individuals with a higher propensity to be affected based on their proximity to an event. Proximity is defined across three dimensions: geographical, psychological, and social. Geographical proximity refers to the physical distance between an individual and the primary suicide. For example, eyewitnesses or repetitive media coverage could increase the likelihood of additional suicides (Zenere, 2008). Psychological proximity refers to the degree of identification with the victim. In this case, the victim is perceived as being similar in terms of life circumstances or as a role model. For example, victims of bullying, teammates, classmates, and others (Zenere, 2008). Finally, social proximity is defined as the degree of relationship between an individual and the victim. Examples include family members, friends, neighbors, and others (Zenere, 2008). Thus, for all types of proximity, the closer individuals in the community are to the primary suicide, the higher their risk of additional suicides.

Unfortunately, geographical proximity is usually the only measure available for analysis, although it can at least partially connect to social proximity when the spatial scale is small (e.g., small towns or the surrounding neighborhood of the first suicide). Considering the absence of an explicit spatial or temporal scale to define a point cluster, they have been reported to occur in areas ranging from hundreds of meters to several kilometers (e.g., Perez-Costillas et al., 2015; Lu et al., 2023) and from a few days to several months or years (Olson, 2013). Thus, at small spatial scales, individuals living in the surrounding neighborhood would be at higher risk (Kirmayer et al., 2007; Olson, 2013), and the negative effects of a single suicide can resonate in the community for several months, affecting susceptible individuals (Masecar, 2009; Olson, 2013).

The occurrence of suicide clusters profoundly impacts affected communities (Robinson et al., 2016; Hawton et al., 2020). Fast-response actions to manage and contain these negative effects should be incorporated into public health policies (Robinson et al., 2016; Ivey-Stephenson et al., 2024; Pirkis et al., 2024a). These plans should include mourning support, assistance for at-risk individuals in the neighborhood, the dissemination of preventive messages (through media or direct interpersonal communication), among others (Hawton et al., 2020; Ivey-Stephenson et al., 2024; Pirkis et al., 2024a). However, effective prevention of suicide clusters and postvention actions requires a robust estimation of the spatial and temporal scales of their occurrence to precisely identify the affected community and efficiently allocate resources (Amin et al. 2022). This involves analyzing high-resolution data (fine grain scale) using sophisticated geostatistical methods (Gibbons et al., 1990; Benson et al., 2022; Trinh et al., 2024).

In Chile, the public health system is organized into several decentralized units distributed throughout the country (Goic and Armas, 2003). This structure enables health interventions at the local geographic scale to be potentially more effective and easier to design and implement. In southern Chile, the city of Valdivia accounts for 45 % of the regional population and 40 % of the region's suicides. In this study, we conducted a spatio-temporal analysis of suicide occurrences over a 26-year period to identify clusters at the city-block or neighborhood scale in the city oof Valdivia. The insights gained from this analysis can serve as a critical foundation for designing postvention strategies tailored to affected communities.