Spatial clustering of suicide mortality and associated community characteristics in Kanagawa prefecture, Japan, 2011-2017

Space-time self-harm and suicide clusters in two cities in Taiwan

AIMS

Suicidal acts may cluster in time and space and lead to community concerns about further imitative suicidal episodes. Although suicide clusters have been researched in previous studies, less is known about the clustering of non-fatal suicidal behaviour (self-harm). Furthermore, most previous studies used crude temporal and spatial information, e.g., numbers aggregated by month and residence area, for cluster detection analysis. This study aimed to (i) identify space-time clusters of self-harm and suicide using daily incidence data and exact address and (ii) investigate the characteristics of cluster-related suicidal acts.

METHODS

Data on emergency department presentations for self-harm and suicide deaths in Taipei City and New Taipei City, Taiwan, were used in this study. In all-age and age-specific analyses, self-harm and suicide clusters were identified using space-time permutation scan statistics. A cut-off of 0.10 for the p value was used to identify possible clusters. Logistic regression was used to investigate the characteristics associated with cluster-related episodes.

RESULTS

A total of 5,291 self-harm episodes and 1,406 suicides in Taipei City (2004-2006) and 20,531 self-harm episodes and 2,329 suicides in New Taipei City (2012-2016) were included in the analysis. In the two cities, two self-harm clusters (n [number of self-harm episodes or suicide deaths in the cluster] = 4 and 8 in Taipei City), four suicide clusters (n = 3 in Taipei City and n = 4, 11 and 4 in New Taipei City) and two self-harm and suicide combined clusters (n = 4 in Taipei City and n = 8 in New Taipei City) were identified. Space-time clusters of self-harm, suicide, and self-harm and suicide combined accounted for 0.05%, 0.59%, and 0.08% of the respective groups of suicidal acts. Cluster-related episodes of self-harm and suicide were more likely to be male (adjusted odds ratio [aOR] = 2.22, 95% confidence interval [CI] 1.26, 3.89) and young people aged 10-29 years (aOR = 2.72, 95% CI 1.43, 5.21) than their cluster-unrelated counterparts.

CONCLUSIONS

Space-time clusters of self-harm, suicide, and self-harm and suicide combined accounted for a relatively small proportion of suicidal acts and were associated with some sex/age characteristics. Focusing on suicide deaths alone may underestimate the size of some clusters and/or lead to some clusters being overlooked. Future research could consider combining self-harm and suicide data and use social connection information to investigate possible clusters of suicidal acts.

The application of spatial analysis to understanding the association between area-level socio-economic factors and suicide: a systematic review

BACKGROUND

Little is known about what impact the use of different spatial methodological approaches may have on understanding the relationship between area-level socio-economic factors and suicide.

METHODS

In this systematic review, we searched PubMed, Embase, CINAHL and PsycInfo for original empirical studies examining the relationship between socio-economic factors and suicide with a spatial lens, published up to January 22, 2022. Data on applied spatial methods, indicators of socio-economic factors, and risk of suicide related to socio-economic factors were extracted. The protocol for this systematic review was registered with PROSPERO (CRD42021251387).

RESULTS

A systematic search yielded 6290 potentially relevant results; 58 studies met the inclusion criteria for review. Of the 58 included studies, more than half of the studies (n = 34; 58.6%) used methods that accounted for spatial effects in analyses of the association between socio-economic factors and suicide or examined spatial autocorrelation, while 24 (41.4%) studies applied univariate and multivariate models without considering spatial effects. Bayesian hierarchical models and spatial regression models were commonly used approaches to correct for spatial effects. The risk of suicide relating to socio-economic factors varied greatly by local areas and between studies using various socio-economic indicators. Areas with higher deprivation, higher unemployment, lower income, and lower education level were more likely to have higher suicide risk. There was no significant difference in results between studies using conventional versus spatial statistic methods.

CONCLUSION

An increasing number of studies have applied spatial methods, including Bayesian spatial models and spatial regression models, to explore the relationship between area-level socio-economic factors and suicide. This review of spatial studies provided further evidence that area-level socio-economic factors are generally inversely associated with suicide risk, with or without accounting for spatial autocorrelation.

A simulation study for geographic cluster detection analysis on population-based health survey data using spatial scan statistics

Background

In public health and epidemiology, spatial scan statistics can be used to identify spatial cluster patterns of health-related outcomes from population-based health survey data. Although it is appropriate to consider the complex sample design and sampling weight when analyzing complex sample survey data, the observed survey responses without these considerations are often used in many studies related to spatial cluster detection.

Methods

We conducted a simulation study to investigate which data type from complex survey data is more suitable for use by comparing the spatial cluster detection results of three approaches: (1) individual-level data, (2) weighted individual-level data, and (3) aggregated data.

Results

The results of the spatial cluster detection varied depending on the data type. To compare the performance of spatial cluster detection, sensitivity and positive predictive value (PPV) were evaluated over 100 iterations. The average sensitivity was high for all three approaches, but the average PPV was higher when using aggregated data than when using individual-level data with or without sampling weights.

Conclusions

Through the simulation study, we found that use of aggregate-level data is more appropriate than other types of data, when searching for spatial clusters using spatial scan statistics on population-based health survey data.

Exploring the epidemiology of suicide attempts: Risk modeling in Kermanshah-Iran

Background

Suicide attempt is a serious global public health issue. The patterns of suicide may vary depending on the individual characteristics, methods, causes, and the geographical area-also socio-cultural context that determine it. This study identifies the spatial patterns of suicide attempts in Kermanshah province, Iran.

Method

The sample size of this cross-sectional study is 18,331 people (7234 males and 11097 females) who attempted suicide in Kermanshah province between 2006 and 2014. Data was collected from the records of patients referred to the emergency department of hospitals in Kermanshah and analyzed using tests of Mean Center, Standard Distance, and Average Nearest Neighbor.

Results

The results of the mean center and standard distance tests show that drug overdose, poisoning with toxins and pesticides, and chemicals mostly were used in the central areas of Kermanshah province. The mean center of suicide attempts by self-immolation, hanging and firearms was in the western parts of the province, while the suicide attempts with narcotic drug were concentrated in the eastern regions of the province. Out of the 18,331 cases, 74% attempted suicide with drug overdose, 13% with toxins and pesticides, 0.59% with chemicals, 4% with fire, 1% by self-immolation, 1% by hanging, 0.16% with firearms and 0.7% with cold weapons. The spatial pattern of all suicide attempts in Kermanshah was clustered (Z-score < -2.58).

Conclusion

The results of this study show that the methods of suicide attempt vary with geographical areas in the province. Therefore, it is suggested that planning tailored to the geographical location can reduce suicide attempts in Kermanshah.

Spatial and space-time clusters of suicides in the contiguous USA (2000-2019)

The present study investigates the spatiotemporal variations in suicide mortality and tests associations between several covariates and suicides for the years 2000-2019 in the contiguous USA. The epidemiological disease surveillance software (SaTScan^TM) was used to identify spatiotemporal variations in suicide mortality rates and to test for significant spatial and space-time clusters with elevated relative suicide risk. The analysis was done with age-adjusted suicide mortality counts data from the Centers for Disease Control (CDC) with (International Classification of Diseases) ICD-10 codes. Specifically, data with codes ICD-10 codes X60-X84.9 and Y87.0, plus ICD-10 113 codes from the CDC, was used. Fourteen significant spatial clusters and five significant space-time clusters of suicide in the contiguous USA were found, including nine significant bivariate spatial clusters of suicide deaths and opioid deaths. Based on these data, there exist significant and non-random suicide mortality clusters after adjusting for multiple covariates or risk factors. The covariates studied provide evidence to develop a better understanding of possible associations in geographical areas where the suicide mortality rates are higher than expected. In addition, there is a significant association between several of the studied risk factors and suicide mortality. While most suicide clusters are also opioid clusters, there exist some clusters with high opioid deaths that are not suicide clusters. These results have the potential to provide a scientific framework that is based on surveillance, allowing health agencies to intervene and reduce elevated rates of suicides in selected counties in the U.S. The study is limited due to the resolution of the data at the county level, and some covariate data was unavailable for the entire period of the study.

Quantitative Methods to Detect Suicide and Self-Harm Clusters: A Systematic Review

Suicide and self-harm clusters exist in various forms, including point, mass, and echo clusters. The early identification of clusters is important to mitigate contagion and allocate timely interventions. A systematic review was conducted to synthesize existing evidence of quantitative analyses of suicide and self-harm clusters. Electronic databases including Medline, Embase, Web of Science, and Scopus were searched from date of inception to December 2020 for studies that statistically analyzed the presence of suicide or self-harm clusters. Extracted data were narratively synthesized due to heterogeneity among the statistical methods applied. Of 7268 identified studies, 79 were eligible for narrative synthesis. Most studies quantitatively verified the presence of suicide and self-harm clusters based on the scale of the data and type of cluster. A Poisson-based scan statistical model was found to be effective in accurately detecting point and echo clusters. Mass clusters are typically detected by a time-series regression model, although limitations exist. Recently, the statistical analysis of suicide and self-harm clusters has progressed due to advances in quantitative methods and geospatial analytical techniques, most notably spatial scanning software. The application of such techniques to real-time surveillance data could effectively detect emerging clusters and provide timely intervention.

Spatial statistical analysis of regional disparities in suicide among policy units in Japan: Using the Bayesian hierarchical model

Suicide prevention is a crucial policy issue in Japan to be addressed nationally. Nevertheless, if there are regional differences in suicide, even in adjacent sub-regions, measures may need to be taken at the sub-regional level. Previous studies have not compared regional differences in suicide based on the size of policy units, such as prefectures, secondary medical areas, and municipalities. This study used the number of suicides from open data for 10 years from 2009 to 2018 to obtain shrinkage estimates of the standardized mortality ratio (SMR) using the Bayesian hierarchical model. We visualized and compared the regional disparities in suicide for each policy unit. For each gender and policy unit, adjacent regions had similar clusters of SMRs and positive spatial autocorrelation of global Moran's I (p < 0.001 for each). Comparisons between each policy unit showed that even if the SMR was low for the prefectural units, there were regions with high SMRs in municipalities and secondary medical areas, and vice versa. It was found that assessing suicide solely on a prefecture-by-prefecture basis may overlook regional disparities in suicide. This research emphasizes the need to establish suicide indicators at the secondary medical or municipal level and execute individual suicide prevention interventions in neighboring communities. Prefectures can also play a role in developing collaborative cooperation between neighboring regions by acting as actors.

Geographic disparities, determinants, and temporal changes in the prevalence of pre-diabetes in Florida

Background

Left unchecked, pre-diabetes progresses to diabetes and its complications that are important health burdens in the United States. There is evidence of geographic disparities in the condition with some areas having a significantly high risks of the condition and its risk factors. Identifying these disparities, their determinants, and changes in burden are useful for guiding control programs and stopping the progression of pre-diabetes to diabetes. Therefore, the objectives of this study were to investigate geographic disparities of pre-diabetes prevalence in Florida, identify predictors of the observed spatial patterns, as well as changes in disease burden between 2013 and 2016.

Methods

The 2013 and 2016 Behavioral Risk Factor Surveillance System data were obtained from the Florida Department of Health. Counties with significant changes in the prevalence of the condition between 2013 and 2016 were identified using tests for equality of proportions adjusted for multiple comparisons using the Simes method. Flexible scan statistics were used to identify significant high prevalence geographic clusters. Multivariable regression models were used to identify determinants of county-level pre-diabetes prevalence.

Results

The state-wide age-adjusted prevalence of pre-diabetes increased significantly (p ≤ 0.05) from 8.0% in 2013 to 10.5% in 2016 with 72% (48/67) of the counties reporting statistically significant increases. Significant local geographic hotspots were identified. High prevalence of pre-diabetes tended to occur in counties with high proportions of non-Hispanic black population, low median household income, and low proportion of the population without health insurance coverage.

Conclusions

Geographic disparities of pre-diabetes continues to exist in Florida with most counties reporting significant increases in prevalence between 2013 and 2016. These findings are critical for guiding health planning, resource allocation and intervention programs.