Alcohol use disorder and risk of specific methods of suicide death in a national cohort

Abstract

Introduction

Alcohol use disorder (AUD) is among the strongest correlates of suicide death, but it is unclear whether AUD status is differentially associated with risk of suicide by particular methods.

Methods

The authors used competing risks models to evaluate the association between AUD status and risk of suicide by poisoning, suffocation, drowning, firearm, instruments, jumping, or other means in a large Swedish cohort born 1932-1995 (total N=6,581,827; 48.8% female). Data were derived from Swedish national registers, including the Cause of Death Register and a range of medical registers.

Results

After adjusting for sociodemographic factors and familial liability to suicidal behavior, AUD was positively associated with risk of suicide for each method evaluated (cumulative incidence differences: 0.006 to 1.040 for females, 0.046 to 0.680 for males), except the association with firearm suicide in females. AUD was most strongly associated with risk of suicide by poisoning. Sex differences in the effects of AUD and family liability were observed for some, but not all, methods. Furthermore, high familial liability for suicidal behavior exacerbated AUD’s impact on risk for suicide by poisoning (both sexes) and suffocation and jumping (males only), while the inverse interaction was observed for firearm suicide (males only).

Conclusions

AUD increases risk of suicide by all methods examined and is particularly potent with respect to risk of suicide by poisoning. Differences in risk related to sex and familial liability to suicidal behavior underscore AUD’s nuanced role in suicide risk. Future research should investigate targeted means restriction effectiveness among persons with AUD.

Introduction

Suicide is a leading cause of death in many countries and accounted for more than 700,000 deaths worldwide in 2019 (1). Methods of suicide vary widely, in part as a function of accessibility (2): Self-poisoning with pesticides is especially common in low- and middle-income countries (3), while approximately half of US suicides involve firearms (3, 4). Although there are inconsistencies regarding the extent to which suicide occurs in the context of mental illness (5-7), psychiatric disorders are prominent risk factors for suicide death, and alcohol use disorder (AUD) is among the most consistently and strongly implicated (8-12).

Many nuances of the association between AUD and suicide have yet to be examined, including whether AUD is differentially associated with specific suicide methods, e.g., is the elevated risk of suicide death among those with AUD driven primarily by a propensity toward the use of poisonous substances? Given the substantial proportion of suicide deaths that occur among those with AUD, this group represents a potentially important clinical target for prevention and intervention. Understanding whether these individuals are especially susceptible to using particular methods in their suicide attempts could lead to improved means restriction efforts, which are effective in some contexts (13, 14).

Liability to suicidal behavior, including suicide death, is influenced in part by genetic factors, with heritability estimates ranging from 0.30-0.55 (15, 16), and suicidal behavior and AUD are genetically correlated (17, 18). Prior studies on the relationship between genetic factors and method of suicide attempt are, however, quite limited. A recent study examined the association between polygenic scores for a range of clinical antecedents (e.g., risk tolerance, anxiety, schizophrenia) and risk of suicide death by selected means, but did not find statistically significant differences (19). Another report identified differentially expressed genes within the dorsolateral prefrontal cortex of suicide decedents who had died by violent versus non-violent means (20). We are unaware of studies that have evaluated associations between aggregate risk for suicidal behavior itself and suicide method.

In the current study, we sought to address gaps in our understanding of how AUD relates to suicide death by using competing risk models to evaluate whether AUD is differentially associated with specific methods of suicide. This approach allows us to account for the fact that suicide methods are mutually exclusive. Given suicide’s low frequency, the use of Swedish national registry data to maximize sample size greatly enhances our ability to estimate these associations. We stratify models by sex due to prior evidence that the distribution of methods (21, 22) and the association between AUD and suicide (23) differs across females and males. Furthermore, we estimate the extent to which familial liability for suicidality, operationalized using family genetic risk scores for suicidal behavior (FGRS_SB), is associated with different methods of suicide death, and whether FGRS_SB interacts with AUD status to impact risk of specific methods. By clarifying how AUD relates to method of suicide, we hope to provide empirical insight that could be informative to means restriction efforts (13), particularly among the high-risk population of individuals with AUD.

Materials and Methods

Sample.

Information for this study was collected from nationwide Swedish registers (see Supplementary Material for details). Each person’s unique identification number, replaced with a serial number for confidentiality, was used for registry linkage. We included all individuals born in or immigrated to Sweden prior to 1975, with birth years 1932-1995. Individuals had to reside in Sweden in 1975 at age 12 or older, or were born into the study at a later time point. Follow-up time of suicide death was from January 1, 1975 or age 12, whichever occurred later, until December 31, 2017, death or emigration, whichever occurred first. The year 1975 was chosen to maximize the coverage and specificity of registry data related to AUD and method of suicide death (e.g., for some municipalities, data availability improves in 1975) to minimize the likelihood of bias.

Measures.

Method of suicide death was the outcome of interest. Information on cause of death was taken from the Swedish Cause of Death Register and divided into eight mutually exclusive categories: non-suicide, or suicide by a specific method (i.e. poisoning, suffocation, drowning, firearm, instrument [e.g., sharp or blunt object], jumping and other). International Classification of Diseases (ICD) codes for the seven inclusive groupings of methods are shown in Supplementary Table 1. The category of “other” included no ICD-8 codes, therefore follow-up time for that method started from January 1, 1987 or age 12, whichever occurred later.

In our primary analyses, deaths of undetermined intent (UDI; see Supplementary Table 1 for corresponding ICD codes) were considered to be non-suicide deaths due to prior evidence that UDI deaths are considerably more common among individuals with AUD (23): Their inclusion could lead to erroneous conclusions if a non-trivial number of UDI deaths among those with AUD were in fact unintentional. However, other studies include UDI deaths as suicides due to concerns that excluding them can lead to underestimates of the true prevalence and cost of suicide (24-29). Therefore, UDI deaths were recoded as suicide deaths under the appropriate method (e.g., poisoning, suffocation) in supplementary analyses.

The primary independent variable was AUD registration, which was identified using multiple medical and criminal registers as described in Supplementary Table 2. As problematic drinking may start some time before an actual registration of AUD, and to avoid overestimating the potential effect of AUD on suicidal death, we treated AUD as a time-fixed variable.

Family genetic risk score for suicidal behavior, FGRS_SB, (including both suicide deaths and attempts) was also included as an independent variable. This was derived according to methods described in Supplementary Table 3 and reported previously (30). Briefly, the FGRS_SB is calculated from rates of suicidal behavior in first- to fifth-degree relatives, controlling for cohabitation effects, rather than from molecular variation in DNA. Prior work demonstrates that FGRS is likely a good indicator of genetic liability (31), and a recent study found that phenotype- and genotype-based genetic instruments explain comparable proportions of disease liability to psychiatric disorders (32).

As a proxy for socioeconomic status, we used parental educational level (pre-secondary, secondary or post-secondary) retrieved from the LISA and the Population and Housing Censuses, by linking individuals to their biological parents via the Multi-Generation Register, described in the Supplementary Material.

Statistical analyses.

When analyzing time to method-specific suicide, all other deaths (non-suicidal and suicide by other methods) were treated as competing events, rather than as non-informative censoring, i.e., we do not make the simplified assumption that a person dying by suicide by poisoning would have the same risk as any other person to die by suicide by suffocation, had he not been using poisoning: likely the person is at increased risk of all different types of suicide. Taking into account the competing events, we produce descriptive uncorrected cumulative incidence functions (CIFs) for the different suicide methods, as a function of AUD status, using the Aalen-Johansen estimation procedure (33). Corrected effects of AUD on method-specific CIFs at age 70 are produced for males and females separately, and also, for supplementary analyses, including UDI deaths as suicide deaths. We present point estimates for the additive effect of AUD on the CIFs at age 70, including conservative 95% confidence intervals (CI), produced by applying non-parametric bootstrap methodology (see Supplementary Material). To investigate the interaction between AUD and FGRS_SB, we use an approach proposed by Scheike and Zhang (34) and described in the Supplementary Material.

We ran a series of models with the effects of AUD on the CIFs (using age as the timeline) of primary interest in the main text. These are modeled, stratified by sex, both uncorrected (Model 1), corrected for birth year and average parental educational level (Model 2), and further corrected for FGRS_SB (Model 3). The interaction term described above (AUD × FGRS_SB) was included in a subsequent model (Model 4) only if main effects of both AUD and FGRS_SB were observed. In supplementary analyses, the same modeling approach was applied, replacing AUD with FGRS_SB in Models 1 and 2. Finally, to evaluate whether effects of AUD, FGRS_SB, and AUD × FGRS_SB differed by sex, we also tested a AUD × FGRS_SB × sex term.

We further performed supplementary multinomial regression models, estimating relative risk ratios for specific suicide methods associated with AUD. Each method acted as reference level, one at a time, to simplify comparisons. Covariates were the same as in the full competing risk model (without the interaction term), the exception being inclusion of calendar year of suicide, as a multinomial regression model is not based on any timeline.

Holm-Bonferroni correction procedures for multiple testing, using a significance level of 0.05, were applied to all tables including p-values or CIs. Data analysis was conducted from October 13, 2021, to October 20, 2023. Statistical analyses were performed using R Statistical Software (35) and SAS software v9.4 (36).

Ethical approval was secured from the regional ethical review board in Lund, Sweden.

Results

Descriptive statistics.

The selected cohort resulted in a total size of N=6,581,827 (3,208,730 females [48.8%] and 3,373,097 males [51.2%]). Registrations for AUD were identified for 103,936 (3.2%) of females and 295,511 (8.8%) of males. There were 10,245 (0.3%) and 25,816 (0.8%) suicide deaths among females and males, respectively, of which 1669 (16.3%) and 6380 (24.7%) were registered among AUD-registered individuals. Total follow-up time was 106,872,996 years for females and 110,042,491 years for males. Table 1 provides additional details, including the distribution of suicide methods for the full sample and within those with AUD. Age at suicide varied by method. As shown in Figure 1 and Figure 2, cumulative incidences of suicide were higher among those with AUD for each method.

Table 1.

Descriptive statistics for females (top panel) and males (bottom panel).

Method of suicide	N (% of total suicide deaths)	Mean (SD) birth year	Mean (SD) age at end of follow- up/suicide death	AUD N (% of total suicide deaths among those with AUD)
Females
Population sample size	3,208,730	1963.0 (18.3)	52.9 (18.0)	103,936
Poisoning	4694 (45.8)	1952.2 (13.2)	43.7 (14.0)	1015 (60.8)
Suffocation	2775 (27.1)	1958.2 (16.6)	40.5 (14.9)	337 (20.2)
Drowning	961 (9.4)	1947.3 (12.2)	49.4 (14.3)	75 (4.5)
Firearm	124 (1.2)	1956.2 (14.0)	37.9 (14.4)	10 (0.6)
Instrument	171 (1.7)	1950.6 (11.3)	46.6 (13.5)	24 (1.4)
Jumping	1329 (13.0)	1957.2 (16.1)	39.7 (14.8)	176 (10.5)
Other	191 (1.9)	1955.7 (15.3)	45.6 (14.7)	32 (1.9)
Total	10,245	1954.1 (14.9)	42.9 (14.6)	1669
Males
Population sample size	3,373,097	1963.2 (18.2)	52.1 (17.6)	295,511
Poisoning	7053 (27.3)	1952.1 (12.6)	41.4 (12.9)	2288 (35.9)
Suffocation	9944 (38.5)	1956.0 (15.6)	43.0 (15.0)	2253 (35.3)
Drowning	902 (3.5)	1950.0 (12.9)	46.2 (15.0)	191 (3.0)
Firearm	4084 (15.8)	1951.5 (13.4)	45.0 (15.6)	692 (10.8)
Instrument	742 (2.9)	1952.2 (12.7)	46.5 (13.8)	238 (3.7)
Jumping	2532 (9.8)	1959.0 (16.1)	39.1 (15.0)	593 (9.3)
Other	559 (2.2)	1958.9 (15.4)	45.3 (14.8)	125 (2.0)
Total	25,816	1954.2 (14.6)	42.8 (14.6)	6380

SD = standard deviation; AUD = alcohol use disorder

Figure 1.

For females, descriptive uncorrected cumulative incidence functions for different suicide methods as a function of alcohol use disorder (AUD) status, using the Aalen-Johansen estimation procedure. The pink line represents the incidence among those with AUD and the blue represents the incidence among those without AUD. Note that the y-axis scale differs across methods; Supplementary Figure 5 provides these results with the y-axis held constant across methods.

Figure 2.

For males, descriptive uncorrected cumulative incidence functions for different suicide methods as a function of alcohol use disorder (AUD) status, using the Aalen-Johansen estimation procedure. The pink line represents the incidence among those with AUD and the blue represents the incidence among those without AUD. Note that the y-axis scale differs across methods; Supplementary Figure 6 provides these results with the y-axis held constant across methods.

Competing risk models.

We next computed competing risk models, with each method category “competing” with the others. We first evaluated a crude model (Table 2, Model 1). For both females and males, AUD was associated with a significantly increased CIF for all methods. The most pronounced increase was observed for poisoning: the cumulative incidence (i.e., the probability of suicide by poisoning between ages 12 and 70 years) was 1.1 percentage points higher among females with AUD compared to those without, and 0.8 percentage points higher among males with AUD compared to those without (Table 2, Model 1). In Model 2, which was adjusted for sociodemographic variables, and Model 3, which was further adjusted for familial liability to suicidal behavior (FGRS_SB), the CIFs based on AUD status were only modestly attenuated and remained significant after correcting for multiple tests with only one exception: suicide by firearm among females, for which only 10 registrations were recorded, yielding very low power for statistical tests (Table 2).

Table 2.

Difference in cumulative incidence (probability in % of having experienced event between ages 12-70) of suicide method, for persons with an alcohol use disorder registration compared to those without. Estimates are presented for age 70, along with 95% confidence intervals. Results for females are in the top panel; results for males are in the bottom panel.

	Model 1	Model 2	Model 3
Females
Poisoning	1.106 (1.017-1.183)	1.079 (0.991-1.154)	1.040 (0.950-1.119)
Suffocation	0.332 (0.283-0.379)	0.326 (0.276-0.373)	0.316 (0.270-0.364)
Drowning	0.060 (0.039-0.085)	0.055 (0.033-0.079)	0.048 (0.026-0.073)
Firearm	0.009 (0.001-0.019)	0.008 (0.000-0.018)	0.006 (−0.001-0.015)
Instrument	0.024 (0.012-0.037)	0.023 (0.011-0.036)	0.017 (0.007-0.030)
Jumping	0.170 (0.136-0.205)	0.166 (0.132-0.201)	0.158 (0.125-0.193)
Other	0.053 (0.032-0.075)	0.052 (0.031-0.074)	0.051 (0.030-0.074)
Males
Poisoning	0.813 (0.757-0.848)	0.746 (0.691-0.781)	0.680 (0.629-0.717)
Suffocation	0.708 (0.654-0.746)	0.652 (0.599-0.691)	0.620 (0.569-0.662)
Drowning	0.060 (0.047-0.074)	0.050 (0.037-0.064)	0.046 (0.033-0.059)
Firearm	0.171 (0.144-0.194)	0.128 (0.101-0.151)	0.116 (0.089-0.140)
Instrument	0.089 (0.076-0.104)	0.083 (0.069-0.097)	0.077 (0.063-0.091)
Jumping	0.181 (0.158-0.205)	0.176 (0.153-0.199)	0.159 (0.137-0.181)
Other	0.052 (0.038-0.066)	0.048 (0.035-0.063)	0.046 (0.031-0.059)

Model 1 represents the crude effect of alcohol use disorder registration.

Model 2 is adjusted for birth year and parental education level.

Model 3 is adjusted for birth year, parental education level, and FGRS_SB.

Because we observed main effects of AUD and FGRS_SB (Table 2, Model 3 and Supplementary Table 4), we next tested Model 4, which included an additive interaction term between these predictors. Results are presented in Figure 3 and Supplementary Table 5. In the final competing risks model, the marginal effect of AUD remained significant in all but one case (suicide by firearm among females). Likewise, the marginal effect of FGRS_SB was significant, but approximately an order of magnitude smaller than for AUD. Among females, the interaction between these predictors (AUD × FGRS_SB) was positive, and significant after Holm-Bonferroni correction for multiple tests, only for suicide by poisoning. For males, we observed positive interaction terms for suicide by poisoning, suffocation, and jumping; i.e., males with AUD were disproportionately likely to die by one of those methods if they had high familial liability for suicidal behavior. We also observed one significant negative interaction among males, for suicide by firearm: Males with AUD were less likely to die by firearm if they had a high familial liability to suicidality.

Figure 3.

Difference in cumulative incidence (probability in % of having experienced event between ages 12-70) of suicide method, for three predictors included in Model 4: An alcohol use disorder (AUD) registration in the left panel, family genetic risk score for suicidal behavior (FGRS_SB) in the center panel, and the interaction between these predictors in the right panel. The model is corrected for birth year and parental education and is stratified by sex: Blue points represent estimates for females, and yellow points represent estimates for males. Estimates are presented for age 70, along with 95% confidence intervals. The horizontal dashed line represents the null. Note that the y-axis scale varies across panels.

We conducted formal tests of sex differences in the final competing risks model described above. As shown in Figure 3 and Supplementary Table 5, the additional effect of AUD on CIF differed by sex for four suicide methods. For poisoning, the effect was higher among females, while it was higher among males for suffocation, firearm, and instrument suicides. The additional effect of FGRS_SB on CIF differed by sex in three cases – poisoning, suffocation, and firearm – wherein the effect was higher for males. There were no significant sex differences in the interaction effects.

Multinomial regression models.

To consider how AUD status relates to risk of suicide by different methods using a complementary approach, we ran multinomial regression models. Relative risk ratios (RRRs) for specific suicide methods are presented in Supplementary Tables 6-7 for females and males, respectively. For both sexes, AUD was associated with an increased risk of suicide by poisoning compared to every other method, though after correcting for multiple tests, comparisons to the “other” category in females and the “instrument” category in males were no longer significant. In contrast, AUD was inversely associated with suicide by firearm compared to nearly every other method (the exception being relative to drowning, among females only).

Sensitivity analyses.

We repeated the competing risks models, including undetermined intent deaths as suicides. Descriptive statistics are provided in Supplementary Table 8 and uncorrected CIFs are presented in Supplementary Figures 1 and 2. Results from competing risk models are presented in Supplementary Tables 9-11 and in Supplementary Figure 3. As shown in Supplementary Figure 4, the additional effects on the CIFs attributable to AUD, and to a lesser extent to FGRS_SB, are most pronounced for poisoning deaths.

Discussion

In the current study, we sought to clarify the association between AUD and risk of different methods of suicide death using competing risk models and complementary analyses. Results indicated that a history of AUD was associated with increased risk of every suicide method examined, and most prominently with suicide by poisoning. Familial liability to suicidality, though of considerably smaller effect than AUD, was also positively associated with elevated risk of each method, and among those with AUD, risk of suicide by certain methods was exacerbated by one’s genetic liability to suicidal behavior. Furthermore, we observed sex differences in the impact of AUD on risk of suicide by poisoning, suffocation, instrument, and firearms. These findings underscore the importance of suicide prevention among individuals with AUD and raise the possibility that means restriction efforts within this group could be particularly effective if focused on potentially poisonous substances (including alcohol, narcotics, analgesics, and other chemicals).

The CIFs observed for each suicide method associated with AUD in the current study, while small in terms of absolute increases in risk, are non-trivial in the context of low-frequency events of high public health importance, such as suicide death. For example, while only 3.2% of females in the cohort have an AUD registration, these females account for 16.3% of suicide deaths. Furthermore, while 0.12% of females without AUD died by suicide by poisoning, the corresponding figure is 0.98% among females with AUD, a >8-fold increase in risk, which is reflected in the Model 3 adjusted CIF of 1.04. Considering the disproportionate number of suicide deaths that occur among individuals with AUD (8, 9, 11, 12, 23), our findings speak to the potential benefits of targeted intervention and prevention in this group.

While AUD was associated with increased risk of each method of suicide, except by firearm among females in fully adjusted models, the strongest association was with suicide by poisoning. As is true in many countries (3), poisoning was the most common method employed in this cohort, and it was disproportionately used among those with an AUD registration. This could be due simply to availability: Prior studies indicate that suicide methods vary geographically in part as a function of accessibility (37-40). Individuals whose AUD was sufficiently severe to be recognized in the medical and/or criminal registers are more likely to have comorbid psychiatric and/or substance use problems (41), which would afford greater access to psychotropic medications or drugs of abuse. The use of poisonous substances among those with AUD is conceptually consistent with suicide risk models that include “acquired capability” as a key component, such as the Interpersonal Psychological Theory (42, 43) and the Three-Step Theory (44). In this case, individuals with AUD could more readily meet the threshold for acquired capability through their consistent ingestion of a substance toxic in high doses (ethanol and potentially other drugs). Indeed, Baer, Tull (45) found that the frequency of substance use was positively associated with suicidal ideation only indirectly, via the path from substance use to acquired capability, though findings are inconsistent (46). Thus, while the overall increased risk of all suicide methods among those with AUD could reflect other aspects of theories of suicide risk – for example, the perceived burdensomeness and/or social isolation experienced by those with substance use problems (47-50) – the observation that risk is most strongly increased for suicide by poisoning could be of particular relevance to future research on acquired capability, and to prevention efforts.

In the current study, AUD was more strongly associated with risk of suicide by poisoning among females than males; AUD CIFs were higher for males than females for suffocation, firearms, and instruments. These findings are consistent with prior evidence that females employ less violent means in suicide attempts (22, 51, 52), though findings vary across countries (53). The current results underscore the nuanced nature of sex differences in the relationship between AUD and suicide.

We also found sex differences in the effects of FGRS_SB. Genetic risk was more strongly associated with suicide by poisoning among females, and with suicide by suffocation and firearm by males. We have previously reported that the total heritability of suicide death is quite similar across sexes (16), and are unaware of prior studies that have examined potential sex differences in suicide method as a function of genetic liability. Furthermore, we observed several significant interactions between AUD and FGRS_SB. Higher FGRS_SB exacerbated the effects of AUD on risk of suicide by poisoning (both sexes), suffocation, and jumping (males only). Perhaps due to Swedish firearm regulation (see below), we observed an inverse association between AUD and FGRS_SB for male firearm suicides. These findings provided tentative support for considering family history of suicidal behavior when evaluating risk of suicide among those with AUD, perhaps in particular with respect to means restriction. However, due to the paucity of prior studies, replication is warranted.

Sensitivity analyses including UDI deaths were concordant with the primary analyses’ findings. However, the magnitude of AUD’s effect was higher when UDI deaths were included, particularly for poisoning. Poisoning accounts for a very large proportion of total UDI deaths: 74.2% and 66.2% of UDI deaths are due to poisoning among females and males, respectively, and among those with AUD, these figures are even higher (84.9% and 77.2%). This underscores the difficulty in making determinations about intent in substance-induced deaths among those with AUD, and should be taken into consideration in the ongoing debate regarding how UDI deaths are treated in suicide research (27, 28). Given that some proportion of UDI deaths are in fact suicides, the “true” effects of AUD are likely intermediate to those in our primary and sensitivity analyses.

We note that the current findings warrant replication in other countries and cultural contexts, as important differences could lead to disparate conclusions. For example, in the United States, 54.6% of suicide deaths in 2021 were firearm-related (54), whereas 11.7% of suicides were due to firearms across the observation period in the current study. In contrast to permissive firearm access in the US, in Sweden, a license is required for firearm ownership or possession and ammunition purchase, subsequent to a check in the police registry (55, 56); background checks may include an assessment of mental health, which could preclude ownership among individuals with AUD (56); and licenses require training in firearm safety (55, 57). Thus, firearm-related means restriction could prevent far more suicides in the US than in Sweden.

Means restriction in general has been identified as an effective suicide deterrent. As noted by Yip, Caine (40) the unavailability of a particular method can either defer a suicide attempt altogether or lead to the selection of less-lethal means. Restriction at the social level, e.g., through erecting physical barriers to prevent jumping (58) or reducing access to specific pesticides (59, 60), has the capacity to prevent suicide even among those who are otherwise considered at low risk, but efforts targeted at individual risk groups are also feasible. Perhaps particularly relevant to those with AUD who might be especially inclined to attempt suicide by poisoning, a review by Lim, Buckley (14) found that reducing the number of doses available per pack for certain medications, and requiring prescriptions for medications previously available over-the-counter, corresponded to marked reductions in suicide deaths by those methods.

Few prior studies have directly examined the relationship between AUD and method of suicide death or attempt, complicating efforts to contextualize the current findings. One study found that, among 143 individuals with AUD who survived an attempted suicide, poisoning was the most common method used (61). However, that study did not include non-AUD individuals for comparison, and the distribution of methods across non-fatal attempts is not expected to parallel that for fatal attempts. A one-year Finnish study, which relied on informant reports to determine whether suicide decedents had misused alcohol, reported differences in methods used across alcohol status, wherein overdose was more common among those with an alcohol problem, particularly females (62). These studies provide some insight, but their different analytic approaches, ascertainment strategies, and small sample sizes preclude clear comparison with the current report. The sparseness of prior studies underscores the need for replication.

In addition to the need for replication, we note a number of limitations to the current study. First, registry data captures severe AUD but likely results in false negatives for more moderate cases that do not come to the attention of medical personnel or the police. Whether identification of those cases would impact the distribution of suicide methods, and by extension our CIF estimates, is unknown. Second, our FGRS_SB are derived from family-based data rather than molecular genetic data. However, family history remains a clinically useful indicator of risk for many psychiatric outcomes (63, 64). Third, we collapsed methods of suicide into cohesive groups due to small numbers of deaths in more precise categories; a more granular approach could impact our findings, but is unlikely to be possible without a much larger dataset. Fourth, we did not account for psychiatric comorbidity, which could attenuate associations between AUD and methods of suicide death (23); such an approach was outside the scope of the current research question, but could be addressed in future studies. Finally, the methods applied in these analyses do not enable causal inference. While we have previously reported that the association between AUD and suicide death overall is likely due in part to a causal pathway (23), further research is necessary to determine whether there is support for a causal association between AUD and specific methods of suicide death.

In conclusion, our analyses of the association between AUD and method of suicide death in a large Swedish cohort indicate that those with AUD are at higher risk for suicide by any method, but particularly for suicide by poisoning. Familial liability to suicidal behavior plays a statistically significant, but less prominent, role in risk. In the context of limited resources for clinical intervention and prevention, these findings suggest that conversations about limiting access to toxic substances might be an effective approach to reducing suicide deaths among those with AUD. Further research efforts are necessary to confirm the current findings, particularly in other cultural contexts.

Significant Outcomes

Risk of suicide death by any method is elevated among individuals with alcohol use disorder.

The magnitude of increased risk among those with alcohol use disorder varies as a function of sex.

High familial liability to suicidal behavior is associated with increased risk of suicide by all means; it exacerbates the effects of alcohol use disorder for specific methods, including suicide by poisoning.

Limitations

Findings from this Swedish cohort might not generalize to other countries or cultures, particularly considering differences in access to particular methods (e.g., firearms).

Mild or moderate cases of alcohol use disorder might not be captured by national registers; inclusion of such cases could impact the strength of associations between alcohol use disorder and suicide by various means.

Data availability statement:

The data that support the findings of this study are available from Statistics Sweden. Restrictions apply to the availability of these data, which were used under license for this study. Data are available only from Statistics Sweden.