Growing Alcohol Use Preceding Death by Suicide Among Women Compared with Men: Age-specific Temporal Trends, 2003-18

Abstract

Background and aims

In the United States (US), up until 2018, both the prevalence of heavy alcohol use and the suicide mortality rate increased among men and women; however, women had experienced a notably higher increase in both. Given that heavy alcohol use is a risk factor for suicide, we postulate that heavy alcohol use may have contributed to the observed sex disparity in the suicide mortality rate increase, which may be reflected in the proportion of suicides that were alcohol-involved over time. Therefore, the aim of the current study was to investigate the temporal trend of the sex- and age group-specific proportion of suicides that were alcohol-involved in the US.

Design

Using restricted-access data from the National Violent Death Reporting System, we performed joinpoint regression analyses to investigate temporal trends in the sex- and age-group (young adults: 18-34 years; middle-aged adults: 35-64 years; and older adults: 65+ years) specific proportion of suicides that were alcohol-involved.

Setting

United States.

Participants

A total of 115,193 suicide decedents 18+ years of age from 2003 to 2018.

Measurements

The sex- and age-group specific proportion of suicides that were alcohol-involved, among all suicide decedent, for which the decedent had a blood alcohol concentration (BAC) a) ≥0.04 g/dl and b) ≥0.08 g/dl.

Findings

For 2003-18, the proportion of suicides that were alcohol-involved wherein the decedent had a BAC ≥0.08 g/dl significantly increased on average annually for women of all age groups (young women: 2.80% [95% CI: 1.86%, 3.75%]; middle-aged women: 2.20% [95% CI: 1.20%, 3.21%]; older women: 10.48% [95% CI: 1.17%, 20.65%], while only middle-aged men experienced a significant average annual percent increase (0.81% [95% CI: 0.003%, 1.62%]).

Conclusion

An increase in alcohol use preceding death by suicide among women compared to men over time in the US was found. These findings of the present study support the need for additional research into the link between heavy alcohol use and the accelerated increase in the suicide mortality rate among women in the US.

INTRODUCTION

Until 2018, the suicide mortality rate in the United States (US) had been increasing for about two decades for both men and women (1, 2). However, the rate of increase had been disproportionate, with the rate among women increasing year over year at an accelerated pace compared to that of men (1). This accelerated increase in the suicide mortality rate among women mirrored that of heavy alcohol use and alcohol use disorders. From 2001/02 to 2012/13, the prevalence of heavy alcohol use and alcohol use disorders among women in the US increased by 58% and 84%, respectively, while the prevalence among men increased by only 15% and 35%, respectively (3). More recently it has been shown that, among adults, the well-documented gender gap is narrowing primarily due to women drinking more (4, 5).

Alcohol use has been identified as an important risk factor for suicidal behavior, having both a precipitating and predisposing effect (6, 7). The risk relationship between heavy alcohol use and death by suicide has been shown to differ for men and women (8–10). For instance, women were found to have a 10-fold higher risk of death by suicide while intoxicated, while men were found to have a six-fold higher risk (11).

The finding that the association between alcohol use and death by suicide is more pronounced in women than in men may be due to the fact that, compared to men, women have a heightened susceptibility to the effects of alcohol (manifested by a more rapid progression of alcohol use disorders (12) and by an earlier onset of alcohol-related health and psychosocial complications (13)), and heavy drinking women are more likely than men to experience social stigmatization (14). Additionally, women with alcohol use problems are about 60% less likely to access treatment services compared to men (15), which may also contribute to their heightened suicide risk.

Taken together, the evidence suggests that alcohol use may be contributing to the observed sex disparity in the rate of increase of suicide in the US. As such, we hypothesized that we would see a higher increase in the proportion of suicides that were alcohol-involved among women compared to men over time. Accordingly, the purpose of the current study was to investigate the temporal trend of the sex- and age group-specific proportion of suicides that were alcohol-involved, where the decedent had a blood alcohol concentration (BAC) a) ≥0.04 g/dl and b) ≥0.08 g/dl in the US from 2003-18.

METHODS

Data sources

Restricted access data from the National Violent Death Reporting System (NVDRS) – a state-based active surveillance system in the US that provides a detailed account of violent deaths that occur in participating states–for 2003-2018 were used. The NVDRS gathers and links detailed investigative information from numerous sources including coroners and medical examiners reports, toxicology reports, death certificates, law enforcement records, and supplementary homicide reports. Although a relatively small number of states participated in the NVDRS in 2003, by 2018 data were available for a total of 38 states (Appendix S1).

The NVDRS system is coordinated and funded at the federal level by the Centers for Disease Control and Prevention (CDC), but depends on separate data collection efforts in each state managed by the state health departments. The participating states receive funding from the CDC, and a requirement of that funding is that they must submit data. CDC has developed case definitions specifically for NVDRS, and provides training to all state-level encoders to ensure protocols for extracting data from its multiple, complementary data sources are applied uniformly (for additional details on the NVDRS see (16)). There are a few states who only provided data for a subset of their violent deaths, in accordance with requirements under which the respective state was funded; such data was excluded from the current analyses (Appendix S1).

Measures

Suicide decedents were identified as those with death certificates that listed International Classification of Diseases, 10th Revision codes X60-X84 or Y87.0 as the cause of death. All present analyses were restricted to single suicides only (i.e., those suicides in which there was a single victim). Alcohol-involved suicides were defined as those suicides where the decedent was tested for alcohol and had a BAC a) ≥0.04 g/dl and b) ≥0.08 g/dl. These two cut-points were selected to 1) capture moderate and heavy alcohol use (17); and 2) to ensure that we capture the precipitating effects of alcohol on suicide, which likely differ for men and women, as women are more sensitive to the impairing effects of alcohol than men, as per a variety of subjective and performance measures (18, 19). As such, it can be surmised that the precipitating effects of alcohol on suicide occur among women at a lower BAC threshold than for men. In the NVDRS, blood alcohol levels are coded as a continuous measure in terms of weight by volume. As such, the continuous measure of BAC was used to create two categorical variables (one for each BAC cut-point). Based on the recommendation of the NVDRS, cases where the BAC was >0.60 g/dl were excluded as they are suspected to be in error, given that a BAC above this threshold is highly unlikely (20). The current analyses were restricted to those 18+ years of age, and the following age groups were used: young adults (18-34 years), middle-aged adults (35-64 years), and older adults (65+ years). Suicide decedents with incomplete data were excluded from the current study.

Statistical analysis

First, the sex- and age group-specific proportion of suicides that were alcohol-involved, where the decedent had a BAC a) ≥0.04 g/dl and b) ≥0.08 g/dl, among all suicide decedents 18+ years of age tested for alcohol was calculated by year. Second, in order to investigate the temporal trends joinpoint regression analyses were performed (21). A joinpoint regression analysis is a data-driven statistical technique that identifies inflection points in the data and 95% confidence intervals (CIs), based on a pre-specified number of joinpoints (21). For the present analyses a maximum of two joinpoints was specified, as is standard for joinpoint analyses of 12-16 data points (21). Based on the maximum number of joinpoints, linear segments were fitted to the data. Using a Monte Carlo Permutation method, the fewest number of linear segments such that an additional joinpoint does not add a statistically significant linear trend is selected (21). The slope coefficient for each regression line was then transformed to an annual percent change. Using a weighted average of the slope coefficients, with the weights equal to the length of each segment over the interval, the average annual percent change over the total study period (2003-18) was calculated. The parametric method was used to estimate the 95% CI for both the annual percent change and average annual percent change (AAPC).

Calculations of prevalence were performed using Stata 16 (22) and the joinpoint regression analyses were conducted using the Joinpoint Regression Program, version 4.8.0.1 (23). Statistical significance was determined using an α of 0.05.

Sensitivity analysis

In order to determine whether the addition of states over time had an impact on the temporal trend in the proportion of suicides that were alcohol-involved, joinpoint regression analyses were conducted on the sex-specific age-adjusted proportion of suicides that were alcohol involved, at a BAC threshold of ≥0.08 g/dl, for 2005-2018 for a) all participating states, and b) 16 states that participated across years: Alaska, Colorado, Georgia, Kentucky, Maryland, Massachusetts, New Jersey, New Mexico, North Carolina, Oklahoma, Oregon, Rhode Island, South Carolina, Utah, Virginia, and Wisconsin (Appendix S1). The age-adjusted proportion was estimated by adjusting the NVDRS sample of suicides to have the same age structure as that of all suicides in the US in 2018 (24).

RESULTS

Data for total of 115,202 suicide decedents (87,771 men and 27,431 women) 18+ years of age with data on the presence of alcohol (i.e., a BAC) were available in the NVDRS from 2003-18 (Figure 1).

Figure 1.

Study flow chart

*Excluded data for states where only a subset of violent deaths was reported to the NVDRS: California (2017, 2018), Illinois (2016, 2017, 2018), Pennsylvania (2016, 2017, 2018), and Washington (2016, 2017).

Women of all age groups experienced a significant average annual percent increase in the proportion of suicides that were alcohol-involved at both BAC thresholds from 2003-18. In contrast, only middle-aged men experienced a significant average annual percent increase in the proportion of suicides that were alcohol-involved where the decedent had a BAC ≥0.08 g/dl (0.81%; 95% CI: 0.003%, 1.62%); however, it was by less than half the average annual percent increase among suicide decedents with a BAC ≥0.08 g/dl who were middle-aged women (2.03%; 95% CI: 1.12%, 3.21%; see Table 1). Even so, it should be noted that the AAPC CIs for men and women overlapped for all age groups except for young adults–meaning the AAPC in the proportion of alcohol-involved suicides in the US from 2003-2018 among young women was significantly higher than among young adult men (for both BAC thresholds).

Table 1.

Joinpoint analysis of sex- and age group-specific proportion of suicides that were alcohol-involved among all suicide decedents (18+ years of age) in the NVDRS from 2003-2018

Sex	Age group	Proportion (%)		Total study perioda		Period 1			Period 2			Period 3
		2003	2018	AAPC (%)	95% CI	Years	APC (%)	95% CI	Years	APC (%)	95% CI	Years	APC (%)	95% CI
Alcohol-involved suicides, BAC ≥0.04 g/dl
Women	Young adults	18.95	31.64	2.80	1.86, 3.75	2003-18	2.80	1.86, 3.75	-	-	-	-	-	-
	Middle-aged adults	21.58	29.88	2.20	1.20, 3.21	2003-18	2.20	1.20, 3.21
	Older adults	9.26	12.74	4.11	1.93, 6.33	2003-09	11.21	6.09, 16.58	2009-18	−0.37	−2.74, 2.05	-	-	-
Men	Young adults	30.88	35.80	0.62	−0.44, 1.69	2003-14	2.02	1.24, 2.80	2014-18	−3.14	−6.88, 0.76	-	-	-
	Middle-aged adults	33.77	35.85	0.73	−0.31, 1.78	2003-15	1.77	1.29, 2.25	2015-18	−3.33	−8.50, 2.14	-	-	-
	Older adults	13.13	15.61	1.67	−1.26, 4.69	2003-10	2.03	−0.34, 4.47	2010-14	11.12	0.07, 23.40	2014-18	−7.55	−13.21, −1.51
Alcohol-involved suicides, BAC ≥0.08 g/dl
Women	Young adults	12.63	26.56	3.09	1.70, 4.49	2003-18	3.09	1.70, 4.49	-	-	-	-	-	-
	Middle-aged adults	18.71	25.16	2.03	1.12, 2.95	2003-18	2.03	1.12, 2.95	-	-	-	-	-	-
	Older adults	1.85	8.11	10.48	1.17, 20.65	2003-05	87.40	−9.97, 290.07	2005-18	1.86	0.10, 3.64	-	-	-
Men	Young adults	25.35	29.62	0.48	−0.67, 1.64	2003-14	2.16	1.30, 3.03	2014-18	−4.00	−8.01, −0.18	-	-	-
	Middle-aged adults	27.46	29.95	0.81	0.003, 1.62	2003-14	1.97	1.42, 2.52	2014-18	−2.33	−5.24, 0.68	-	-	-
	Older adults	12.12	11.84	−0.01	−5.18, 5.44	2003-05	−14.38	−43.57, 29.91	2005-14	6.47	3.68, 9.33	2014-18	−6.18	−14.18, 2.56
Sensitivity analysisb
Women	All states	15.75	21.27	2.16	0.78, 3.57	2005-18	2.16	0.78, 3.57	-	-	-	-	-	-
	16 states	12.30	15.06	0.93	−0.20, 2.08	2005-18	0.93	−0.20, 2.08	-	-	-	-	-	-
Men	All states	24.65	25.81	0.63	−0.26, 1.52	2005-14	2.35	1.59, 3.11	2014-18	−3.13	−5.86, −0.32	-	-	-
	16 states	21.33	21.00	−0.27	−1.32, 0.78	2005-12	1.86	0.32, 3.42	2012-18	−2.70	−4.59, −0.78	-	-	-

AAPC: Average annual percent change; APC: Annual percent change; CI: Confidence interval

^aYears 2003 to 2018

^bJoinpoint regression analyses on the sex-specific age-adjusted proportion of suicides that were alcohol involved, at a BAC threshold of ≥0.08 g/dl, for 2005-2018 for a) all participating states, and b) 16 states that participated across years.

Note. Boldface indicates statistical significance (p<0.05).

The temporal trend in the proportion of suicides that were alcohol-involved at both BAC thresholds was best described by a single linear upward trend for both young and middle-aged women, while the proportion among older women was best described by two linear segments (Figure 2 and 3). In contrast, the temporal trend in the proportion of suicides that were alcohol-involved at both BAC thresholds was best described by two linear segments for both young and middle-aged men, while the prevalence among older men was best described by three linear segments. Of those temporal trends with more than one linear segment, the most recent period had an annual percent decrease for all age-groups of men for both BAC thresholds, albeit not all statistically significant, while older women experienced a slight non-significant annual percent decrease in the most recent period for a BAC ≥0.04 g/dl and a significant annual percent increase in the most recent period for a BAC ≥0.08 g/dl.

Figure 2.

Observed sex- and age group-specific proportion of suicides that were alcohol-involved for which the decedent had a blood alcohol content ≥0.04 g/dl among all suicide decedents (18+ years of age) in the NVDRS from 2003-2018 and the joinpoint trend

Note. P-values are presented for the annual percent change of each linear segment found to be statistically significant.

Figure 3.

Observed sex- and age group-specific proportion of suicides that were alcohol-involved for which the decedent had a blood alcohol content ≥0.08 g/dl among all suicide decedents (18+ years of age) in the NVDRS from 2003-2018 and the joinpoint trend

Note. P-values are presented for the annual percent change of each linear segment found to be statistically significant.

The sensitivity analysis found that the overall trends for the sex-specific age-adjusted proportion of suicides that were alcohol-involved, where the suicide decedent had a BAC ≥0.08 g/dl, for 2005-2018 for a) all participating states, and b) 16 states that participated across years were largely the same. There was an overall upward trend across all years for women, while men experienced an overall upward trend in earlier years and a downward trend in the most recent period (Table 1 and Figure 4).

Figure 4. Sensitivity analysis:

Observed sex-specific age-adjusted proportion of suicides that were alcohol-involved for which the decedent had a blood alcohol content ≥0.08 g/dl among all suicide decedents (18+ years of age) in the NVDRS for 2005-2018 for a) all participating states, and b) 16 states that participated across years and the joinpoint trend

DISCUSSION

Overall, the proportion of suicides that were alcohol-involved had significantly increased during the study period among all age-groups of women, while among men, the temporal trends were less consistent. The results of the present study are important for both clinical and preventative efforts. By providing evidence that alcohol use may have been a core driver in the accelerated increase in the rate of suicide among women will, hopefully, promote additional research in this area. If additional research substantiates this conjecture, we would expect to see a push for routine screening for suicidal ideation and behavior of all heavy drinking women who present to a healthcare professional (primary care, emergency department, etc.).

It is worth noting that since 2018, the suicide mortality rate in the US has decreased, likely in part due to the novel coronavirus disease 2019 pandemic (for reasons unknown). However, the suicide mortality rate trends may be more nuanced (for example, see (25)).

To our knowledge, this is the first study to use joinpoint regression analyses to investigate the temporal trend of the sex- and age group-specific proportion of suicides that were alcohol-involved in the US. Among the strengths of the present study is the relatively long time period covered (2003-2018), as well as the data source itself. NVDRS is the only surveillance system for violent deaths in the US, collecting information from multiple sources including toxicology reports and thus, providing BAC for suicide decedents. Further, the analytical method of choice allowed us to test whether a multi-segmented line best fit the data, as compared to a straight line, which provides a much more detailed overview of what has been happening with respect to the proportion of suicides that were alcohol-involved over time than a single summary trend statistic, for example. There are a few limitations that should be acknowledged however. First, the NVDRS collected data on violent deaths from only seven states in 2003, with progressive additions to the system over the years. Thus, the number of states with available data varied across years. However, we did not find evidence that the addition of states over time had an impact on the temporal trend in the proportion of suicides that were alcohol-involved involved. Also, a limitation of the NVDRS that is important to acknowledge is that given that toxicological testing depends greatly on local resources, there is variation across states in the percent of suicide decedents who were tested for blood alcohol levels (26). This limitation is particularly relevant to the current study as only suicide decedents with complete data (i.e., they had to of had a BAC ascertained) were included in the current analyses. Very clearly, the distribution of alcohol use among those not tested is unknown. Lastly, joinpoint regression analysis fits linear segments to data, which may or may not be linear, and as such, may represent a simplification of the observed temporal trend. It should also be noted that the present analysis was not pre-registered and therefore the results should be considered exploratory.

Despite the need for additional research into the link between alcohol use and the accelerated increase in the suicide mortality rate among women in the US, heavy alcohol use is an established risk factor for death by suicide (6, 7). Thus, it is clear that education and awareness of this relationship must be increased exponentially to help prevent future suicides, and better detection and more accurate assessments must be achieved, especially among women, as shown here.

WHAT THIS RESEARCH ADDS.

• Data from the National Violent Death Reporting System show a growing trend of alcohol-involved suicides among women in the US.

• The average annual percent increase in the proportion of alcohol-involved suicides in the US from 2003-2018 was significantly higher among young adult (18-34 years) women compared to their male counterparts.

• The average annual percent increase in the proportion of suicides that were alcohol-involved, where the decedent had a blood alcohol concentration of 0.08 g/dl or greater was two-times higher among middle-aged women than their male counterparts.