Age, period, and cohort effects on suicide death in the United States from 1999 to 2018: moderation by sex, race, and firearm involvement

Abstract

The role of sex, race, and suicide method on recent increases in suicide mortality in the United States remains unclear. Estimating the age, period, and cohort effects underlying suicide mortality trends can provide important insights for the causal hypothesis generating process. We generated updated age-period-cohort effect estimates of recent suicide mortality rates in the US, examining the putative roles of sex, race, and method for suicide, using data from all death certificates in the US between 1999 and 2018. After designating deaths as attributable to suicide according to ICD-10 underlying cause of death codes X60-X84, Y87.0, and U03, we (i) used hexagonal grids to describe rates of suicide by age, period, and cohort visually and (ii) modeled sex-, race-, and suicide method-specific age, period, and cohort effects. We found that, while suicide mortality increased in the US between 1999 and 2018 across age, sex, race, and suicide method, there was substantial heterogeneity in age and cohort effects by method, sex, and race, with a first peak of suicide risk in youth, a second peak in older ages-specific to male firearm suicide, and increased rates among younger cohorts of non-White individuals. Our findings should prompt discussion regarding age-specific clinical firearm safety interventions, drivers of minoritized populations' adverse early-life experiences, and racial differences in access to and quality of mental healthcare.

Fig. 1. Suicide rates in the US from 1999 to 2018 for men, stratified by firearm involvement, across age, period, and cohort.

A = Age, P = Period, C = Cohort. All isolines are spaced at 5 years. Suicide rates are stratified by firearm involvement. Suicide rates for each birth cohort can be visualized along its specific diagonal C isoline, with corresponding ages and periods along A and P isolines, respectively. For instance, diagonal isoline C:1964 represents the suicide rate specific to the cohort of men born in 1964. The hexagon where C:1964 intersects with vertical isoline P:2010 and diagonal isoline A:46 represents their suicide rate in year 2010, when they were 46 years of age. Likewise, suicide rates for each period can be visualized along its specific column or vertical P isoline; and suicide rates for each age group along its specific diagonal A isoline.

Fig. 2. Suicide rates in the US from 1999 to 2018 for women, stratified by firearm involvement, across age, period, and cohort.

A = Age, P = Period, C = Cohort. All isolines are spaced at 5 years. Suicide rates are stratified by firearm involvement. Suicide rates for each birth cohort can be visualized along its specific diagonal C isoline, with corresponding ages and periods along A and P isolines, respectively. For instance, diagonal isoline C:1964 represents the suicide rate specific to the cohort of women born in 1964. The hexagon where C:1964 intersects with vertical isoline P:2010 and diagonal isoline A:46 represents their suicide rate in year 2010, when they were 46 years of age. Likewise, suicide rates for each period can be visualized along its specific column or vertical P isoline; and suicide rates for each age group along its specific diagonal A isoline.

Fig. 3. Age, period, and cohort effects on suicide in the US from 1999 to 2018 by firearm involvement and race, among men.

The left axis coincides with the age effect and indicates suicide rates per 100,000 men and is anchored to the reference cohort (1960). The right axis corresponds to the risk ratio comparing each cohort and period to the reference cohort (1960) and period (2010).

Fig. 4. Age, period, and cohort effects on suicide in the US from 1999 to 2018 by firearm involvement and race, among women.

The left axis coincides with the age effect and indicates suicide rates per 100,000 women and is anchored to the reference cohort (1960). The right axis corresponds to the risk ratio comparing each cohort and period to the reference cohort (1960) and period (2010).