October 2006, Vol 96, No. 10 | American Journal of Public Health Ajdacic-Gross et al. | Peer Reviewed | Research and Practice | 1  RESEARCH AND PRACTICE  Changing Times: A Longitudinal Analysis of International Firearm Suicide Data | Vladeta Ajdacic-Gross, PhD, Martin Killias, PhD, Urs Hepp, MD, Erika Gadola, MA, Matthias Bopp, MSci, Christoph Lauber, MD, Ulrich Schnyder, MD, Felix Gutzwiller, MD, DrPH, and Wulf Rössler, MD, MA We investigated changes in the proportion of firearm suicides in Western countries since the 1980s and the relation of these changes to the change in the proportion of households owning firearms. Sev- eral countries had an obvious de- cline in firearm suicides: Norway, United Kingdom, Canada, Aus- tralia, and New Zealand. Multilevel modeling of longitudinal data con- firmed the effect of the proportion of households owning firearms. Legislation and regulatory mea- sures reducing the availability of firearms in private households can distinctly strengthen the prevention of firearm suicides. (Am J Public Health. 2006;96:XXX–XXX. doi:10. 2105/AJPH.2005.075812) Our analysis focused on international data on firearm suicides over different periods, which enabled a longitudinal perspective. To our knowledge, this is the first analysis to have such a focus. Proportions of firearm suicides in Western countries since the 1980s were compared with the proportions of households owning firearms within those countries, a well-known predictor of firearm suicide 1 and, moreover, homicide, 2 particu- larly femicide. 3,4 Specifically, we analyzed whether changes in the proportions of house- holds owning firearms were associated with changes in legislation and regulatory mea- sures and whether changes in firearm legisla- tion contributed to the prevention of firearm suicides. METHODS The data in this analysis were obtained from the following sources: • Suicide data published in 1992 by Killias 5 from the World Health Organization (WHO) database, which used International Classifi- cation of Diseases, Eighth Revision, and Inter- national Classification of Diseases, Ninth Revision (ICD-8 and ICD-9) codes, 6,7 and data derived from the WHO database or, in some instances, data delivered on request from national statistical offices; • Suicide data published in 1998 by Krug et al. 8 (ICD-8 and ICD-9 codes, mostly from the year 1993 and 1994); • Data published by the WHO in the mortality database (http://www.who.int/research/en/); the data are limited to those countries using International Classification of Diseases, 10th Revision (ICD-10), 9 codes and include up to 3 years (1999–2001), if available; data com- pilation was performed by the authors; • Data from the 1989, 1992, 1996, and 2000 International Crime Victims Surveys (http:// www.unicri.it/wwd/analysis/icvs/data.php) on gun ownership (proportion of households owning firearms—legal or illegal) 3,10,11 ; • Gun ownership data, for the United States only, from the National Opinion Research Center General Social Survey data series, published in the Sourcebook of Criminal Justice Statistics (http://www.albany.edu/ sourcebook/pdf/t256.pdf and 258.pdf); Na- tional Opinion Research Center data were used because they are collected annually. Analysis of longitudinal data was done within the framework of random coefficient models (also called multilevel model for change, mixed models, or hierarchical linear models) 12 ; the proportion of firearm suicides as modeled as a function of time, suicide rates, and the proportion of households own- ing firearms. The proportion of firearm sui- cides and the proportion of households own- ing firearms were introduced as logits to mitigate floor and ceiling effects. The follow- ing countries were included in the analysis: United States, Switzerland, Finland, France, Canada, Sweden, Australia, the Netherlands, England and Wales, Scotland, Norway, Spain, and New Zealand. The 1992 values for the proportion of households owning firearms were not used in modeling except if necessary to replace missing 1989 values (applies to Sweden and New Zealand). This analysis was performed with the PROC MIXED procedure in SAS for Macintosh, Version 6.12 (SAS In- stitute Inc, Cary, NC). RESULTS The proportion of firearm suicides within overall suicides and the proportion of house- holds owning firearms varied considerably across countries, and the proportions also varied over time (Table 1). On the one hand, firearm suicides have become more common in countries such as Switzerland, Germany, and the Netherlands, while on the other, firearm suicides are clearly declining, some dramatically, in England and Wales, Scotland, Canada, Australia, New Zealand, and Norway. The statistical analysis of longitudinal data via random coefficient models (Table 2) con- firmed that the main relevant effect on the pro- portion of firearm suicides derived from the proportion of households owning firearms. Model C, which was accepted as the final model, showed that the latter variable ab- sorbed almost all variance from the 4 variance components. Additional effects (e.g., the inter- action term of time and proportion of house- holds owning firearms) did not contribute any noteworthy improvement to the model. DISCUSSION This longitudinal study with international data showed that the proportion of firearm suicides decreased simultaneously with the proportion of households owning firearms. This result is in line with the well-established association between availability of firearms at home and risk of firearm suicide. 1,3–5,13–16 Firearm suicides depend on the availability of the method more than other suicide meth- ods. Firearm suicides result more often from impulsive decisions than other suicide methods and tend to be associated more often with al- cohol abuse. 17 Victims of firearm suicides were shown to have distinctly fewer previous suicide attempts (22%) in their psychiatric history than were victims of other suicide methods http://www.ajph.org/cgi/doi/10.2105/AJPH.2005.075812 The latest version is at Published Ahead of Print on August 31, 2006, as 10.2105/AJPH.2005.075812