Comparing Substance-involved Suicide and Unintentional Deaths from 2011 to 2021 in Missouri, United States

Abstract

Background:

Substance-involved suicide and unintentional overdose deaths share risk factors, yet our understanding of how these deaths overlap and differ remains limited. Direct comparisons of substance-involved suicide and unintentional deaths are lacking.

Objective:

To guide effective prevention and intervention efforts regarding substance-involved suicide, we examined risk factors, demographic and substance-related, of substance-involved suicide and unintentional deaths.

Methods:

Using logistic and multinomial regression, we used medical examiner records obtained annually (N=6,467, 72% male) to examine characteristics associated with suicide and unintentional substance-involved deaths in St. Louis, Missouri between 2011 and 2021.

Results:

Between 2011 and 2021, age-standardized suicide rates slightly decreased (3.55 to 3.33), while unintentional deaths nearly tripled (23.2 to 68.2). Many deaths involved both alcohol and traumatic injury, and almost a fifth of suicides involved prescription opioids. In the logistic model (combining data across 11 years), the following factors were associated with increased odds of suicide, relative to unintentional deaths: White race (OR=5.42, 95%CI[3.95,7.56]), greater age (OR=1.01, 95%CI[1.00,1.02]), traumatic injury (OR=4.40, 95%CI[3.17,6.13]), and presence of not otherwise classified substances (including prescription medications; OR=4.36, 95%CI[3.11,6.10]). The following were associated with decreased odds of suicide: presence of medical condition (OR=0.32, 95%CI[0.23,0.45]), fentanyl (OR=0.04, 95%CI[0.02,0.06]), ethanol (OR=0.64, 95%CI[0.46,0.88]), cocaine (OR=0.40, 95%CI[0.24,0.63]), heroin (OR=0.05, 95%CI[0.03,0.09]), and other stimulants (OR=0.32, 95%CI[0.20,0.50]).

Conclusions:

Results suggest a divergence over time in the rates of classified substance-involved suicide and unintentional deaths. A distinctive set of demographic and substance use characteristics differentiated the two manners of death, highlighting potential risk factors to inform further research and targeted interventions.

1. Introduction

Substance use and substance use disorder (SUD) greatly increase suicide risk (1–4), and intentional substance overdose is one of the most common methods of suicide (5,6). However, substance use and SUD are also associated with risk of unintentional overdose, and distinguishing between suicide and unintentional overdose can be difficult (5). This is especially the case because our understanding of how the characteristics of these deaths overlap and differ is limited, and few studies have used mortality data to directly compare substance-involved suicides to unintentional substance-involved deaths. This is an important gap because, though both groups of people are likely at an increased risk of death, prevention efforts may differ for people using substances who have intentions to die versus those who do not. For example, harm reduction practices to help people use more safely may be insufficient by themselves for people at high risk of suicide, who would likely also benefit from enhanced support and suicide-focused intervention. There is, thus, a need to be able to identify people who are using substances and are at risk of suicide. Bohnert et al. (2010) conducted a key review of suicide and unintentional overdose among people who use substances and found that people who have experienced unintentional overdose are more likely to have attempted suicide and vice versa (5). Though valuable in establishing a link between suicide and unintentional overdose, the review was published at the start of the second wave and prior to the most recent wave of the opioid overdose epidemic, and focused on heroin and cocaine use. Most of the reviewed studies exclusively examined non-fatal overdoses and suicide attempts, with only two examining fatal overdoses and suicide. Finally, most studies did not specifically examine substance-involved suicides, which represent a potentially distinct group of suicide deaths with their own risk factors.

More recent work focused on opioids and suicide has found that, in addition to being associated with unintentional overdose, opioids are implicated in more than a fifth of suicide deaths in the United States (1, 3–4) and experiencing a nonfatal opioid overdose is associated with greater risk of later dying by suicide (3). Sometimes collectively called “deaths of despair,” opioid overdose deaths and suicide share risk factors, including social and environmental disadvantages, physical and psychological pain, and negative affective states (4).

The links between suicide and both alcohol use disorder (AUD) and acute alcohol use are also well established (7–10). A high proportion of US suicide deaths involve acute alcohol use, particularly those using more lethal methods, such as firearms (4,11). Moreover, higher blood alcohol concentrations have been linked to higher injury severity in suicidal and self-injurious behaviors (12). Alcohol intoxication can increase the risk of suicidal behavior through various mechanisms, such as limiting the ability to use alternative coping strategies and increasing impulsivity, disinhibition, and dysphoria (4,9). Alcohol is also prevalent in unintentional overdose, including approximately one-fifth of opioid overdose deaths (e.g., 13).

Although research has linked substances such as alcohol and opioids to both unintentional overdoses and substance-involved suicide, fewer studies have directly compared these two manners of death. Studies find that suicides, compared to unintentional overdoses, more often involve over-the-counter medications, opioid and non-opioid analgesics, and prescription medications (e.g., antidepressants and antipsychotics; 14–18). Conversely, unintentional substance-involved deaths are more likely to involve fentanyl, heroin, cocaine, and other illicit substances (15,17). However, Lundgren et al. (2022) found that, although recency and frequency of non-medical opioid use were associated only with unintentional deaths, other opioid use-related variables (e.g., early onset of drug use, drug injection) were also associated with suicide (27).

Regarding demographics, when compared to substance-involved unintentional decedents, substance-involved suicide decedents are typically older and more often White (18,19), with drug poisoning rates highest among people aged 35–49, while suicide and alcohol-induced death rates peaked among people aged 50–64 (20). Athey et al. (2020) found, in 63 people who used opiates, that factors including depression and a previous suicide attempt were associated with suicide death, while a chronic pattern of severe drug use was associated with unintentional overdose (39).

There is also important temporal variation in whether substance-involved deaths have been categorized as suicide or unintentional (e.g., 20,21), creating a need to examine classification of deaths over time. For example, despite significant increases in opioid-related suicide rates among many demographic groups in the US between 2000–2017, there was an observed proportional decline in the percentage of opioid-related deaths that were suicides over the same period, indicating that the increase in suicide rate did not increase as quickly as the rate of unintentional overdoses (19). Researchers have argued that many deaths classified as unintentional or undetermined may, in fact, represent suicides (19,22), with 20–30% of opioid overdoses estimated to be suicides (40). These data raise concerns about the effectiveness of methods to accurately identify suicide deaths.

Additional direct comparisons are needed to understand what predictors distinguish substance-involved suicides and unintentional overdose deaths and to guide effective prevention efforts. Importantly, many existing studies were conducted outside of the US (e.g., 16,18,27,30,37). Both unintentional substance overdoses and suicide deaths have been rising in the US, posing an urgent public health concern (23–24,52–53). Given that patterns of drug use vary significantly across and within countries, it is important to further examine these patterns within the US. This is particularly true regarding parts of the country that face significant challenges related to substance use but may receive relatively less attention, including Midwestern cities. The limited US-based literature (e.g., 14,17 in North Carolina; 39 in Ohio) highlights regional variations, including significant differences in the associations of race with manner of death. Given inconsistent definitions of substance-involved suicides and unintentional overdoses (e.g.,5), there is also a need for research on suicides and fatal overdoses using more standardized definitions of deaths. Moreover, concerns about potential biases regarding classification of death (e.g.,22) emphasize the need to examine trends over time regarding the prevalence of substance-involved suicide classifications compared to overdoses. Changes over time may reflect true shifts in the prevalence of substance-involved suicides, but may also signal the possibility of misclassification, particularly if suicide and overdose trends diverge significantly (26).

The current study used medical examiner data to compare demographic and substance use characteristics associated with suicide and unintentional substance-involved deaths in St. Louis, Missouri between 2011–2021. Although past work has identified factors potentially associated with substance-involved suicide and unintentional overdose, given that limited work has directly compared these types of deaths and associations may vary geographically, we did not have directional hypotheses for the current study. Based on existing work, we examined multiple theoretically-driven demographic (e.g., age, race, gender) and substance-related variables (e.g., types of substances present at death), expecting to see patterns of differences between substance-involved suicide and unintentional deaths.

2. Methods

2.1. Data Source

This preregistered project (osf.io/9n3ef) used data provided annually by medical examiners’ offices in St. Louis City and St. Louis County from 2011–2021. As this project relied exclusively on de-identified data regarding deceased individuals, human subjects review was not required. The dataset comprises all individuals whose cause of death was determined to be drug-involved (N=6,777). Data included age at death, race, biological sex, manner of death, death date, location of death (St. Louis City or County), substances determined by the medical examiner to be involved in the cause of death, and other factors involved in the cause of death (i.e., traumatic injury, co-occurring medical condition). Consistent with CDC guidelines (54,55), deaths were classified as drug-involved if the medical examiner listed one or more drugs as involved in the cause of death on the death certificate, medical examiner report, or postmortem toxicology report. When the cause of death indicated involvement of multiple drugs but did not indicate specific drugs, all drugs detected by postmortem toxicology at a level that could contribute to death were classified as involved in the cause of death. For example, if the cause of death was “multidrug overdose” and toxicology results were positive for five drugs at levels that could contribute to death, all five were classified as involved in the cause of death.

2.2. Data Preparation

Manner of death categories included: accident (i.e., unintentional), motor vehicle accident, stillborn, homicide, suicide, natural, and undetermined (i.e., the classification given when no single manner of death is more compelling than another given the information available). For analyses, the current study focused on suicide deaths (n=589) and unintentional deaths (n=5,878), with a total final sample of N=6,467 (95% of the original sample). Unintentional deaths included unintentional overdose deaths (n=5,576) and motor vehicle accidents (n=302). Stillborn (n=15), homicide (n=86), natural (n=45), undetermined deaths (n=164), and those with missing manner of death (n=8), were excluded, although undetermined deaths were included in the descriptive assessment of trends over time to provide context.

Substances involved in deaths were coded into the following categories: ethanol, benzodiazepine, cocaine, methamphetamine, amphetamine, prescription stimulant, fentanyl (and analogs), heroin, methadone, morphine, buprenorphine, prescription opioids, non-specified opioids, non-specified stimulants, other prescription drugs, other non-prescription drugs, club drugs, MDMA, kratom, xylazine, anticonvulsants, muscle relaxants, antipsychotics, gabapentin, sedatives, cannabis, inhalants, quaalude, PCP, unspecified drugs, and over-the-counter drugs. Categories were made accounting for all generic and brand names and misspellings by medical examiners. Ethanol, benzodiazepines, cocaine, fentanyl, and heroin were each examined as categories. As pre-registered, we collapsed the remaining categories as follows. Methamphetamine, amphetamine, prescription stimulants, and non-specified stimulants were combined into an other stimulants category. Methadone, morphine, buprenorphine, prescription opioids, and non-specified opioids were combined into a prescription opioid category. All other drugs were combined into a not otherwise classified substances category.

2.3. Analytic approach

Analyses were pre-registered, unless noted as exploratory, and completed in R 4.1. Descriptive analyses examined prevalence and demographic characteristics of suicide, unintentional, and undetermined deaths overall and over time. Direct standardization (58) was used to calculate age-adjusted mortality rates over time (using thirteen age groups). Estimates of the St. Louis City and County populations for each year between 2011–2021, as reported by the US Census, were used to calculate age-specific rates of deaths per 100,000. Further, the standard 2000 US population weights (e.g., 58) were used to standardize the age-adjusted mortality rates for each year.

In exploratory analyses, we examined the specific means of suicide (e.g., gunshot wound). We coded qualitative responses recorded in the cause of death field, with the following categories: 1) gunshot wound; 2) substance or medication intoxication/toxicity; 3) blunt trauma; 4) hanging/asphyxiation/suffocation; 4) carbon monoxide intoxication; and 5) other means. A logistic regression model assessed associations with manner of death (suicide vs. unintentional death). Predictors were: ethanol, cocaine, other stimulants, heroin, fentanyl, prescription opioids, benzodiazepines, other drugs, presence of a co-occurring medical condition in the cause of death (yes/no), presence of a traumatic injury involved in the cause of death (yes/no), sex (male/female), race (Black/White/Asian/Hispanic), age in years (linear effect), and year of death (linear effect). Age and year of deaths were mean-centered. In secondary analyses, we also planned to explore the characteristics of fentanyl-involved suicides. However, because only 22 suicides (3.7%) involved fentanyl, we did not examine them further. We also conducted planned secondary analyses to assess 1) whether the number of substances present at death was associated with the manner of death and 2) whether season of death was associated with manner of death. Predictors were not significant and are not reported further.

After observing that ethanol was present in a large number of suicides but, in the logistic regression model, was associated with decreased odds of suicide, we conducted exploratory follow-up analyses to examine whether ethanol involvement differed across traumatic and non-traumatic unintentional and suicide deaths (classified based on the presence of traumatic injury involved in the cause of death). Specifically, we conducted a multinomial regression with four outcome categories (traumatic unintentional deaths, non-traumatic unintentional deaths, traumatic suicides, non-traumatic suicides).

We additionally calculated relative risk, which can diverge from odds ratios depending, in part, on the prevalence of the outcome (59). Detailed methods and results are outlined in the supplement. Risk ratios were largely similar to the odds ratios from the logistic models (Tables S1–S2), and overall interpretation of the findings remained unchanged.

3. Results

3.1. Descriptive results

Suicides comprised 8.7%, unintentional 86.6%, and undetermined deaths 2.4% of all substance-involved deaths (including all manners of death) reported in St. Louis City and County between 2011–2021 (N=6,785). Examination of trends over time showed that age-adjusted rates of suicide (3.55 to 3.33) and undetermined deaths (2.71 to 1.62) decreased slightly from 2011–2021, while unintentional deaths nearly tripled from 2011 to 2021 (23.2 to 68.2; Figure 1). Proportionally, the number of suicide deaths decreased from 12.9% (n=48) of 2011 deaths to 4.7% (n=44) of 2021 deaths. Table 1 summarizes the demographic and cause of death characteristics by manner of death. Across the sample, average age at death was 41.9 (SD=13.5), and most decedents were White (57.3%) or Black (41.2%). Co-occurring medical conditions complicating the cause of death were present in 18.2% of deaths, while traumatic injuries were present in 12.1% of deaths. Fentanyl, heroin, and ethanol were the most frequently present substances, although fentanyl and heroin were much more common in unintentional deaths and ethanol was more common in suicides. The average number of substances present at death was 1.65 (SD=0.82, range 1–7), with one substance present in 52.6% and two substances present in 33.9% of deaths.

Figure 1.

Age-adjusted rates of accidental, suicide, and undetermined deaths from 2011–2021.

Table 1.

Demographic and death characteristics of the sample.

	Suicide N=589	Unintentional N=5,878	Total Deaths N=6,467
Age M (SD)	46.0 (16.4)	41.5 (13.1)	41.9 (13.5)
Sex N (%)
Male	392 (66.6)	4,290 (73.0)	4,682 (72.4)
Female	197 (33.4)	1,588 (27.0)	1,785 (27.6)
Race N (%)
Black	60 (10.2)	2,605 (44.3)	2,665 (41.2)
White	514 (87.3)	3,193 (54.3)	3,707 (57.3)
Asian	12 (2.0)	16 (0.3)	28 (0.4)
Hispanic	3 (0.5)	46 (0.8)	49 (0.8)
Conditions present at death N (%)
Medical condition	63 (10.7)	1,113 (18.9)	1,176 (18.2)
Traumatic injury	305 (51.8)	479 (8.1)	784 (12.1)
Substances present at death N (%)
Fentanyl	22 (3.7)	3,239 (55.1)	3,261 (50.4)
Ethanol	307 (52.1)	1,413 (24.0)	1,720 (26.6)
Benzodiazepines	35 (5.9)	215 (3.7)	250 (3.9)
Cocaine	24 (4.1)	1,100 (18.7)	1,124 (17.4)
Heroin	14 (2.4)	1,830 (31.1)	1,844 (28.5)
Other stimulants	34 (5.8)	715 (12.2)	749 (11.6)
Prescription opioids	111 (18.8)	833 (14.2)	944 (14.6)
Other drugs	143 (24.3)	277 (4.7)	420 (6.5)
Number of drugs present at death N (%)
One drug	411 (75.6)	2,945 (50.1)	3,356 (52.6)
Two drugs	97 (17.8)	2,065 (35.4)	2,162 (33.9)
Three drugs	30 (5.5)	663 (11.4)	693 (10.9)
Four or more drugs	6 (1.1)	167 (2.9)	173 (2.7)

3.2. Means of suicide

Figure 2 presents suicides by means of death. As expected, among this sample of substance-involved suicides, substance intoxication was listed as an additional cause of death for all suicides (e.g., hanging and acute opioid intoxication), except for 9 carbon monoxide deaths. Close to half of suicide deaths (n=251, 42.6%) solely involved substance intoxication/toxicity, 201 suicides (34.1%) involved a gunshot wound, 85 suicides (14.4%) involved hanging/asphyxiation, 16 suicides (2.7%) involved carbon monoxide intoxication, 10 (1.7%) involved blunt trauma, and 17 (2.9%) involved other means (e.g., drowning, stabbing).

Figure 2.

The frequencies of various means of suicide among drug-involved suicide deaths in St. Louis, Missouri.

Note: Substance intoxication = substance intoxication/toxicity listed as the only cause of death. Some type of substance intoxication was also involved in all other deaths, with the exception of 9 carbon monoxide intoxication deaths.

3.3. Primary analyses

3.3.1. Associations between demographic/cause of death characteristics and manner of death

A logistic regression examined associations for demographics and cause of death characteristics with classification of the death as a suicide (vs. unintentional). White race (OR=5.42, 95%CI [3.95,7.56], p<.001), Asian race (OR=24.35, 95%CI [7.94,78.22], p<.001), greater age (OR=1.01, 95%CI [1.00,1.02], p=.002), a traumatic injury (OR=4.40, 95%CI [3.17,6.13], p<.001), and the presence of not otherwise classified drugs (OR=4.36, 95%CI [3.11,6.10], p<.001) were significantly associated with increased odds of dying by suicide (Table 2). Predictors with low prevalence (e.g., Asian race) were included in the models to adjust for their effect and should only be interpreted with caution.

Table 2.

Multivariate logistic regression examining associations for demographic factors, cause of death characteristics, and substances present at death with classification of the death as a suicide (vs. unintentional).

	OR a	95% CI	p
White race	5.42	3.95, 7.56	<0.001
Asian race	24.35 b	7.94, 78.22	<0.001
Hispanic race	2.29	0.48, 8.23	0.238
Male sex	0.80	0.62, 1.01	0.064
Age	1.01	1.00, 1.02	0.002
Year of death	1.00	0.96, 1.04	0.878
Medical condition	0.32	0.23, 0.45	<0.001
Trauma	4.40	3.17, 6.13	<0.001
Fentanyl	0.04	0.02, 0.06	<0.001
Ethanol	0.64	0.46, 0.88	0.007
Benzodiazepines	0.99	0.61, 1.53	0.926
Cocaine	0.40	0.24, 0.63	<0.001
Heroin	0.05	0.03, 0.09	<0.001
Other Stimulants	0.32	0.20, 0.50	<0.001
Other Drugs	4.36	3.11, 6.10	<0.001
Prescription Opioids	0.81	0.58, 1.12	0.205

Note:

^aOR = Odds Ratios. OR > 1 indicates increased likelihood of a suicide classification (unintentional death is the reference group).

^bAs there was a very small number of Asian decedents, this finding should be interpreted with caution. Bolding represents significant values at p < .05.

Presence of a medical condition (OR=0.32, 95%CI [0.23,0.45], p<.001), fentanyl (OR=0.04, 95%CI [0.02,0.06], p<.001), ethanol (OR=0.64, 95%CI [0.46,0.88], p=.007), cocaine (OR=0.40, 95%CI [0.24,0.63], p<.001), heroin (OR=0.05, 95%CI [0.03,0.09], p<.001), and other stimulants (OR=0.32, 95%CI [0.20,0.50], p<.001) were associated with decreased odds of dying by suicide (Table 2). Gender, Hispanic race, year of death, benzodiazepines, and prescription opioids were not associated with manner of death.

3.3.2. Associations for traumatic versus non-traumatic deaths

An exploratory multinomial logistic regression examined associations between demographic and cause of death characteristics and the following outcomes: traumatic suicides (n=305), non-traumatic suicides (n=284), traumatic unintentional deaths (n=479), and non-traumatic unintentional deaths (n=5,399). Non-traumatic unintentional deaths were the reference category. Compared to non-traumatic accidents, ethanol was associated with a greater likelihood of both traumatic unintentional deaths (OR=4.06, 95%CI [3.09,5.35], p<.001) and traumatic suicides (OR=4.73, 95%CI [3.11,7.19], p<.001), but a lower likelihood of non-traumatic suicides (OR=0.52, 95%CI [0.35,0.77], p<.001; Table 3). In sensitivity analyses (three separate binomial regressions with relative risk), ethanol was not significant for non-traumatic suicides (RR=0.85, 95%CI [0.65,1.11], p=0.23), but remained significant for traumatic unintentional deaths (RR=2.69, 95%CI [2.19,3.31], p<.001) and traumatic suicides (RR=2.80, 95%CI [2.12,3.69], p<.001; Table S2). Thus, ethanol was associated with traumatic death, regardless of whether it was categorized as a suicide or unintentional. As seen in Table 3, all other drug categories were either associated with a lower likelihood of traumatic unintentional death and traumatic suicide or not associated with these outcomes.

Table 3.

Multinomial logistic regression examining associations between demographic/cause of death characteristics (e.g., substances present at death) and traumatic versus non-traumatic deaths, with drug overdoses as the reference group

Predictor	Outcome
	Traumatic unintentional deaths			Non-traumatic suicides			Traumatic suicides
	OR	95% CI	p-value	OR	95% CI	p-value	OR	95% CI	p-value
White race	0.88	0.68, 1.14	0.3	5.62	3.47, 9.08	<0.001	3.89	2.55, 5.93	<0.001
Other race	0.94	0.35, 2.52	0.9	12.2	3.89, 38.0	<0.001	4.38	1.45, 13.2	0.009
Male sex	1.05	0.81, 1.37	0.7	0.59	0.43, 0.80	<0.001	1.69	1.15, 2.47	0.007
Age	1.00	0.99, 1.01	0.9	1.02	1.01, 1.04	<0.001	1.00	0.99, 1.01	0.5
Year of death	1.04	1.00, 1.09	0.054	0.99	0.94, 1.04	0.7	1.06	1.01, 1.12	0.040
Medical condition	0.24	0.17, 0.34	<0.001	0.35	0.24, 0.51	<0.001	0.04	0.02, 0.09	<0.001
Ethanol	4.06	3.09, 5.35	<0.001	0.52	0.35, 0.77	0.001	4.73	3.11, 7.19	<0.001
Fentanyl	0.03	0.02, 0.05	<0.001	0.01	0.01, 0.03	<0.001	0.01	0.01, 0.02	<0.001
Heroin	0.05	0.03, 0.07	<0.001	0.03	0.01, 0.06	<0.001	0.01	0.01, 0.02	<0.001
Benzodiazepines	0.29	0.13, 0.63	0.002	1.01	0.61, 1.67	>0.9	0.13	0.04, 0.43	<0.001
Cocaine	0.53	0.36, 0.77	<0.001	0.16	0.07, 0.35	<0.001	0.58	0.31, 1.08	0.086
Other Stimulants	1.07	0.73, 1.57	0.7	0.19	0.09, 0.39	<0.001	0.53	0.31, 0.98	0.042
Other Drugs	0.60	0.37, 0.97	0.038	4.99	3.36, 7.40	<0.001	0.52	0.27, 1.04	0.066
Prescription Opioids	0.12	0.07, 0.19	<0.001	0.60	0.41, 0.89	0.010	0.07	0.04, 0.14	<0.001

Note: OR = Odds Ratios. Bolding represents significant values at p < .05.

4. Discussion

The current study explored the characteristics and trends of substance-involved suicide and unintentional deaths in St. Louis, Missouri between 2011 and 2021. Findings add to existing research on the distinctions between intentional and unintentional substance-involved deaths, with findings largely replicating past work. Nearly 70% of suicide decedents consume a substance prior to death (6), and almost a quarter are intoxicated at the time of death (41), meaning that effective suicide prevention requires understanding and addressing the role of substance use in suicide. The differences and similarities between suicides and unintentional deaths suggest that many of the demographic and substance use characteristics we examined may represent important risk factors, both separate and shared, that can inform further research and aid in targeted prevention and intervention efforts.

Suicides comprised 8.7% of all substance-involved deaths, with unintentional and undetermined deaths accounting for 86.6% and 2.4%, respectively. These estimates follow US-wide patterns (24,52). The majority of both substance-involved suicides and unintentional deaths were men. However, compared to unintentional decedents, suicide decedents were older and more likely to be White, consistent with other studies (e.g., 18,20,25). Notably, White individuals comprised close to 90% of all suicides. It is unclear whether this reflects actual racial disparities in suicide (e.g., due to cultural factors) or systematic racial bias in suicide classification (26). Considering specific substances, and consistent with previous work, findings indicated associations of illicit opioids and stimulants with unintentional deaths (18,27–29) and not otherwise classified substances, such as prescription and over-the-counter non-opioid medication, with suicides (15–16,29–30,42).

4.1. Accuracy of classification of substance-involved suicides

The current study cannot directly examine the accuracy of classifications of deaths as suicide or unintentional. However, while speculative, patterns in the data, combined with existing research, suggest that substance-involved suicides may be underestimated. Overall, fewer than 4% of suicides involved fentanyl and heroin, while a much larger percentage involved prescription medications, including prescription opioids, which are culturally more strongly associated with suicide. Further, between 2011 and 2021, the annual rates of substance-involved suicides remained largely the same, declining slightly, which aligns with another report of declines in intentional overdose deaths between 2012–2019 in the US (57). This is despite the fact that, over the same time period, unintentional drug-involved deaths nearly tripled between 2011–2021 in St. Louis, primarily reflecting increases in overdose deaths involving heroin and fentanyl (19) and increases during COVID-19-related lockdowns (31). In contrast to the slight decline of substance-involved suicides, there was an overall increase in the total number of suicides in the US between 2011–2021 (53). It may be that these trends reflect true patterns, in which suicides and substance-involved overdoses have increased, while substance-involved suicides have not. Similarly, it may be the case that people rarely attempt suicide via fentanyl or heroin overdose. However, existing work suggests a historical tendency of misclassification of suicide generally and specifically for substance-involved deaths to be classified as unintentional, in part because classifying deaths as suicides typically requires affirmative evidence (19,22,26,42–44). Further, as noted by Oquendo & Volkow (2018), opioid use disorder is strongly associated with risk for attempting suicide. Based on this and emergency department visit data, Oquendo & Volkow (2018) estimated that 20–30% of opioid overdoses may be suicides, whereas only 2.4% of opioid-involved deaths, including those involving prescription opioids, were classified in the current data as suicides (40).

As already mentioned, this study relied upon medical examiner data and cannot directly evaluate whether unintentional and suicide deaths were accurately classified. Indeed, intentionality is often extremely difficult, if not impossible, to determine and is not all-or-nothing, as people may engage in dangerous behavior while holding ambivalent feelings about whether they wish to live or die. Nevertheless, there is a significant need for accurate classification of substance-involved deaths. Underestimating the rate of substance-involved suicides may result in fewer resources put toward suicide prevention and intervention amongst populations using substances, and may also lead to inaccurate conclusions about risk factors for substance-involved suicides. Although efforts to prevent substance-involved suicide and unintentional overdose overlap, as both involve dangerous substance use, there are also important differences. For example, harm reduction approaches are effective in helping people use substances in a way that decreases risk for overdose (32). However, someone who is contemplating an intentional overdose likely would benefit from additional intervention, such as developing a safety plan. Thus, there is a significant need for suicide risk assessment to be integrated into routine care of people who use substances. Importantly, as the current evidence base for effective suicide prevention for people who use drugs is limited (56), more efforts are needed to improve and develop effective interventions.

4.2. Ethanol and traumatic death

Alcohol was strongly associated with traumatic death, both in terms of unintentional and suicide deaths, with approximately a quarter of unintentional and half of suicide deaths involving alcohol. The rate of alcohol involvement in suicide was also slightly above national reports (29). Strikingly, nearly all firearm suicides involved acute alcohol intoxication. Given the link between substance use and firearm availability, gun-related behaviors, and firearm suicide (45–49), suicide prevention strategies should emphasize firearm safety, as well as increased access to AUD and SUD treatment.

The links between suicide and both AUD and acute alcohol use are well established (7–10). Further, alcohol use proximal to an attempt increases the risk of a more lethal method, which is likely correlated with greater risk of trauma, and young adults and middle-aged men are more likely to attempt suicide with a firearm while intoxicated (9,33). These findings underscore the immense impact of alcohol on both intentional and unintentional traumatic death (34–35,50–51). Further research is needed to examine suicide deaths that involve both alcohol and traumatic injury, as a better understanding of how alcohol use impacts the intent, method, and lethality of suicide attempts is a crucial step toward using targeted and personalized prevention and treatment approaches (37).

4.3. Limitations and future directions

This study has important limitations, including its use of medical examiner data, which relies on examiners’ judgements for classifying substances and the cause and manner of death. Inconsistent reporting standards and inadequate training for medical examiners can make it more challenging to accurately identify suicides and suicide deaths may be underreported (19,44). We do not know, for example, to what extent significant associations (e.g., for race) indicate actual differences between unintentional overdoses and substance-involved suicides versus potential systematic biases in death categorization (e.g., 38). This is an important topic for future work to explore. These data also include subjective and potentially inaccurate postmortem reports of race/ethnicity and consider biological sex but not gender, limiting interpretation of the reported demographic differences. To improve the comparison and synthesis of findings across studies, future research should address the inconsistencies in the conceptualization and reporting of substance-involved deaths (e.g., terms such as “substance-related,” “overdose,” “poisoning”).

5. Conclusions

Suicide and substance overdose are a major public health concern in the US, with mortality rates continuing to rise. This study explored the different predictors of suicide and unintentional substance-involved deaths between 2011–2021 in Missouri. Largely consistent with other literature, unintentional deaths involved more illicit substances, such as opioids and stimulants, while suicide deaths involved other substances, such as prescription medications. Alcohol was particularly involved in traumatic deaths, and almost a fifth of suicides involved prescription opioids. The rise of unintentional opioid-involved deaths, in particular, underscores the need to increase harm reduction efforts, including statewide distribution of naloxone, a medication for reversing opioid overdoses. Together with other research, our findings point to the possibility that suicides may be under-reported and classified as unintentional, as disentangling substance use behaviors and intentional self-harm can be challenging. Given that efforts to prevent suicide and unintentional overdose meaningfully differ, despite overlap, there is a critical need for additional research to better understand to what extent the present results represent actual differences between suicide and unintentional deaths versus cultural biases in the classification of substance-involved deaths. Taken together, research suggests that substance use treatment and mental health/suicide care would benefit from enhanced integration (e.g., incorporating suicide risk screening in SUD treatment and vice versa), as the field hopefully moves closer towards a cohesive healthcare system to prevent these deaths of despair.