Higher rates of firearm ownership are strongly associated with higher rates of overall suicide and firearm suicide, but not with nonfirearm suicide (Table 1). Suicide attempt rates are not significantly associated with suicide mortality rates in unadjusted models (correlation coefficient = −0.08, P = 0.60) (data not shown) or in models that control for firearm ownership (partial correlation coefficient = 0.02, P = 0.89) (Table 1). Suicide attempt rates are also not significantly associated with rates of household gun ownership (data not shown). The prevalence of household firearm ownership, which ranges from 10% to 66% across the 50 states, explains 67% of the variation in firearm suicide, 42% of the variation in overall suicide, and less than 2% of the variation in nonfirearm suicide. By contrast, suicide attempt rates, which range from 0.1% to 1.5%, explain less than 1% of the variation in rates of overall suicide, firearm suicide, and nonfirearm suicide. Indeed, suicide attempt rates are not significantly related to suicide mortality rates overall or by method, even in crude comparisons. Adjustment for suicide attempt data in regressions, therefore, has little influence on the magnitude of the associations between rates of firearm ownership and suicide mortality. For example, the partial correlation coefficient relating rates of household firearm ownership and suicide mortality in our primary analysis is 0.6 whether or not suicide attempts are included in the regression (data not shown). Likewise, regardless of whether suicide attempt rates are included in regressions, the partial correlation coefficient relating household firearm ownership and firearm suicide is 0.8 (data not shown).
Regression analyses further quantify these relationships. Suicide rates are, on average, 0.22 deaths (per 100,000 population) higher in states where firearm ownership rates are 1 percentage point higher. The relationship between firearm ownership and suicide rates is entirely accounted for by the relationship between firearm ownership and firearm suicides, as reflected in a β coefficient associated with firearm suicide of 0.22 (95% confidence interval: 0.18, 0.27), which is virtually identical to that for overall suicide (β = 0.21, 95% confidence interval: 0.14, 0.28); the β coefficient relating firearm ownership and nonfirearm suicide is essentially null (β = –0.02, 95% confidence interval: –0.05, 0.02). By contrast, suicide rates are only slightly higher (0.30 deaths per 100,000 population) in states where rates of suicide attempts were 1 percentage point higher (Table 1). Because suicide attempt rates vary from 0.1 to 1.5 per 100,000 population, the maximum influence of suicide attempt rates on the suicide mortality rates observed across the 50 states is small (0.30 × 1.4 = 0.42 deaths per 100,000 population on a suicide rate scale that ranges from 8.6 to 28.9 deaths per 100,000 population). By contrast, because firearm ownership prevalence ranges from 10% to 66%, the corresponding (maximum) influence of household firearm ownership on suicide rates is 0.22 × 56 = 12.3 deaths per 100,000 population. Weighted and unweighted analyses produce virtually identical coefficients for household firearm ownership in relationship to suicide mortality (Table 1). Likewise, the relationship between variation in suicide attempt rates and suicide mortality in both weighted and unweighted analyses is similar, nonsignificant, and materially trivial (Table 1). Secondary stratified analyses by sex and by age groupings for which suicide attempt data were available produce patterns similar to those in primary analyses (Table 2).
Although the aggregate number of people residing in the 16 high–gun ownership states and the 6 low–gun ownership states is approximately equal, and the suicide attempt rates are similar, almost twice as many adults completed suicide in the high–gun ownership states compared with the low–gun ownership states (Table 3). This difference in total suicides over a 2-year period is almost entirely attributable to differences in firearm suicides (7,275 vs. 1,697), with virtually no difference in the number of nonfirearm suicides (4,153 vs. 4,341).
Figure 1 illustrates the strong association between rates of household firearm ownership and mortality from overall suicide (Figure 1A) and suicide involving firearms (Figure 1B) and the weak association between rates of firearm ownership and suicide involving methods other than firearms (Figure 1C). In addition, Figure 1 illustrates that suicide attempt rates have little influence on the relationship between firearm ownership rates and mortality from suicide overall and from suicide by firearms and correlate weakly with suicide by methods other than firearms. For example, adjustment for suicide attempt rates hardly moves the cross-hatches off the regression line linking firearm ownership and firearm suicide rates (or off the regression line linking firearm ownership to overall suicide rates), providing a visual representation of how little the observed association between suicide mortality and firearm ownership depends on confounding by suicide attempt rates. These visual renderings directly mirror results from linear regression analyses.