Children and Youth Services Review 100 (2019) 461-467
Children and Youth Services Review 100 (2019) 461-467
Children and Youth Services Review 100 (2019) 461-467
A R T I C LE I N FO A B S T R A C T
Keywords: Purpose: Given increasing youth suicide rates and liberalization of marijuana-use laws, this study examined post-
Youth suicide mortem marijuana toxicology test results among suicide decedents aged 15–19 (youth) compared to those aged
Marijuana 20–29 (young adults).
Marijuana liberalization Methods: Data were from the 2005–2015 National Violent Death Reporting System (N = 6002 decedents aged
Alcohol
15–19 and N = 20,925 decedents aged 20–29). Following descriptive analysis, logistic regression models were
Toxicology
used to test associations of marijuana-positive toxicologies with incident year, state in which injury/death in-
Suicide means
curred, interactions between incident year and state, and suicide means.
Results: Marijuana was the most commonly substantiated substance among youth and the second most com-
monly substantiated substance, after alcohol, among young adults. The odds of a marijuana-positive toxicology
among youth were 2.21 (95% CI = 1.78–2.73) higher in 2012–2015 than in 2005–2011 and were 1.46 (95%
CI = 1.18–1.79) higher in states with marijuana legalization or decriminalization laws; however, the incident
year × state interaction term was not significant, and only the main effect of incident year remained significant
(AOR = 1.86, 95% CI = 1.34–2.59) in the interaction model. Among young adults, the main effects of incident
year, state, and interaction effects (AROR = 2.02, 95% CI = 1.61–2.53) were all significant. For both age groups,
the odds of a marijuana-positive toxicology were lower among decedents who used poisoning than firearms.
Conclusions: Liberalized marijuana policies do not appear to have influenced increases in marijuana-positive
toxicologies among youth, but marijuana-positive rates were higher among those who died in recent years and
by more violent/lethal means. Youth suicide prevention strategies should include monitoring marijuana use,
education on marijuana's harms, and substance use treatment.
⁎
Corresponding author at: 1925 San Jacinto Blvd, D3500, Austin, TX 78712, United States of America.
E-mail address: nchoi@austin.utexas.edu (N.G. Choi).
https://doi.org/10.1016/j.childyouth.2019.03.035
Received 11 December 2018; Received in revised form 18 March 2019; Accepted 18 March 2019
Available online 19 March 2019
0190-7409/ © 2019 Elsevier Ltd. All rights reserved.
N.G. Choi, et al. Children and Youth Services Review 100 (2019) 461–467
other studies have examined post-mortem toxicologies among youth legalization, youth are also more likely to be exposed to marijuana
suicide decedents. While much attention has been paid to youth mar- through their parents, older sibling, and peers, leading to higher use
ijuana use in the context of increasing legalization of marijuana use in (Bailey et al., 2016; Roditis, Delucchi, Chang, & Halpern-Felsher, 2016;
recent years, little is known about potential changes in marijuana-po- Shah & Stahre, 2016).
sitive toxicologies among youth suicide decedents. Given increasing youth suicide rates and increased access to mar-
Marijuana use among youth has been associated with development ijuana due to laws legalizing marijuana for medical and recreational
of psychosis and psychotic disorders; increases in psychosis-related use, research is needed to determine if marijuana-positive toxicology
outcomes; physical and other mental health problems; impairment in rates among youth suicide decedents have changed. Marijuana's many
the cognitive domains of learning, memory, and attention; poorer adverse effects during adolescence and young adulthood (psychosis,
academic performance; lower likelihood of advancement to post-sec- impaired cognitive and social functioning, impulsivity) may further
ondary education; impaired social functioning; delinquency; and in- contribute to suicidal behaviors, especially among those with other
juries (Hall & Degenhardt, 2009; Homel, Thompson, & Leadbeater, mental health and substance use problems. A review of psychological
2014; Meier et al., 2012; National Academies of Sciences, Engineering, autopsy studies of youth and young adult suicide decedents showed
& Medicine, 2017). Among young adults (mean age 23.7 [SD = 4.6] that nearly 90% met criteria for at least one mental disorder, most
years) with any substance dependence problems, marijuana use on a frequently, mood disorders, followed by substance use disorders and
given day (compared to days when marijuana was not used) has been disruptive behavior disorders (Fleischmann, Bertolote, Belfer, &
associated with increased impulsivity, hostile behaviors, and percep- Beautrais, 2005). Simon et al. (2001)also found that almost a quarter of
tions of hostility in others, independent of frequency of marijuana or suicide attempts among those aged 15–34 years were impulsive
alcohol use (Ansell, Laws, Roche, & Sinha, 2015). Similarly, 9th and (i.e., < 5 min passed between the decision to die by suicide and the
10th graders who increased their marijuana use over a six-year period suicide attempt). A study of Swedish young adults (18–25 years) seen at
had higher cigarette smoking, novelty-seeking, and aggressive and anti- psychiatric emergency services after a suicide attempt found that 44%
social behavior than non-escalating users (Passarotti, Crane, Hedeker, & had made an impulsive attempt (i.e., with no preparation and no pre-
Mermelstein, 2015). meditation), and it was often associated with a substance use disorder
The National Survey on Drug Use and Health, conducted annually, (Beckman et al., 2019).
shows that marijuana use among both young and older adults has in- Using 2005–2015 NVDRS data, we examined post-mortem mar-
creased steadily over the past decade (Substance Abuse and Mental ijuana toxicologies among youth (aged 15–19) suicide decedents. For
Health Services Administration [SAMHSA], 2018), and medical mar- comparison purpose, we also examined the same among young adult
ijuana laws (MML) in increasing numbers of states appear to have (aged 20–29) suicide decedents. Based on conflicting findings about
contributed to this trend (Hasin, 2018; Martins et al., 2016; Wen, youth marijuana use, we posed four research questions: (1) How do
Hockenberry, & Cummings, 2015). However, on a national level, youth marijuana-positive rates compare to other substance-positive rates?; (2)
marijuana use rates have not changed significantly and have decreased Did marijuana-positive toxicology rates change during the study period,
in recent years (Johnson et al., 2015; Salas-Wright, Vaughn, Todic, especially after 2012 when Colorado and Washington became the first
Córdova, & Perron, 2015; SAMHSA, 2018; Sarvet, Wall, Keyes, et al., states to pass recreational marijuana laws?; (3) Are marijuana-positive
2018). Attempts to determine whether marijuana legalization has had rates higher in states with medical or recreational use or decriminali-
any effect on youth marijuana use have produced inconsistent results. zation laws than states where marijuana use was illegal?; and (4) Do
Some studies report either no significant relationship between MML marijuana-positive rates differ by suicide means? Controlling for de-
and youth marijuana use or decreased marijuana and other substances mographic and suicide precipitating/risk factors, we tested the fol-
use among youth in MML states (Cerdá et al., 2018, Choo et al., 2014; lowing study hypotheses among both youth and young adult decedents.
Johnson, Hodgkin, & Harris, 2017; Sarvet, Wall, Fink, et al., 2018). H1: Marijuana-positive rates will be higher than other substance posi-
Other studies note significant associations between marijuana legali- tive rates. H2: Marijuana positive rates will be higher in recent years
zation and marijuana risk perception and use among youth. For ex- (2012–2015) than earlier years (2005–2011). H3a: Marijuana-positive
ample, Miech et al. (2015) found that since California decriminalized rates will be higher in marijuana-legal or -decriminalized states than in
marijuana use in 2010, the state's 12th graders had significantly higher marijuana-illegal states, and H3b: especially in recent years in mar-
rates of current marijuana use and future use expectations and lower ijuana-legal or -decriminalized states (interaction effects). H4:
risk perception and disapproval of marijuana use than their peers in Marijuana positive rates will be higher among decedents who used
other states. In Washington state, where recreational marijuana use more violent suicide means. The findings may enhance knowledge
became legal in 2012, perceived harmfulness declined and marijuana about youth suicide decedents in an environment where marijuana
use increased from 2.0% in 2010–2012 to 4.1% in 2013–2015 among availability and accessibility is increasing, and comparisons between
8th and 10th graders, whereas in states without recreational marijuana youth and young-adult decedents may increase insights about both age
laws, marijuana use among students in the same grades declined from groups.
1.3% to 0.9% during the same period (Cerdá et al., 2017). No sig-
nificant differences were found in perceived harmfulness or use of 2. Material and methods
marijuana among Washington 12th graders or among 8th, 10th, or 12th
graders in Colorado, where recreational marijuana use was also lega- 2.1. Data source
lized in 2012 (Cerdá et al., 2017). However, marijuana-related visits for
emergency and urgent care among those aged 13–20 in Colorado in- The NVDRS is an incident-based violent death reporting system that
creased from 1.8 per 1000 visits in 2009 to 4.9 in 2015 (Wang, Davies, provides detailed data on all individual victims and/or suspects of
Halmo, Sass, & Mistry, 2018). suicides, homicides, deaths from legal intervention (i.e., victim killed
In Oregon, legalization of recreational marijuana use in 2015 was by law enforcement acting in the line of duty), deaths of undetermined
associated with increased use among youth who were already using but intent, and unintentional firearm deaths in participating states since
not with uptake among nonusers (Rusby, Westling, Crowley, & Light, 2003 (Blair, Fowler, Jack, & Crosby, 2016; Centers for Disease Control
2018). However, Paschall, Grube, and Biglan (2017) found positive [CDC], 2017). In 2003 and 2004, seven and 13 states, respectively,
relationships between the numbers of registered marijuana patients and participated in the NVDRS. In 2005 through 2014, 16 states (Alaska,
growers per 1000 population in Oregon counties and marijuana use Colorado, Georgia, Kentucky, Maryland, Massachusetts, New Jersey,
prevalence among youth from 2006 to 2015, partially attributable to New Mexico, North Carolina, Oklahoma, Oregon, Rhode Island, South
perceived norms favorable towards marijuana use. With marijuana Carolina, Utah, Virginia, and Wisconsin) provided data; Ohio joined in
462
N.G. Choi, et al. Children and Youth Services Review 100 (2019) 461–467
2011; Michigan in 2014; and Arizona, Connecticut, Hawaii, Kansas, 2.2.2. Control variables
Maine, Minnesota, New Hampshire, New York, and Vermont in 2015 Demographic variables were gender and race/ethnicity.
(CDC, 2017). Studies of 17 NVDRS-participating states in 2011–2013 Precipitating/risk factors were: (1) report of current diagnosed mental
showed that these states closely resembled the nation in age and sex disorders/syndromes (as in the DSM-5 [American Psychiatric
composition, manner of death distribution, and crude and age-adjusted Association, 2015]); (2) recent (i.e., at the time of injury) depressed
all-cause, suicide, and undetermined-intent mortality rates, supporting mood (without need for a clinical diagnosis [CDC, 2017]); (3) a history
the generalizability of NVDRS-based findings (Rockett, Caine, Connery, of suicide attempt or explicit/indirect disclosure of suicide intent within
et al., 2018, Rockett, Caine, Stack, et al., 2018). a month (or recently) before suicide; (4) alcohol dependence or alcohol
NVDRS links data from death certificates and coroner/medical ex- problems (“perceived by self or others to have a problem with, or to be
aminer and law enforcement (CME/LE) reports that are based on the addicted to, alcohol”); (5) other substance use problems; (6) relation-
injury/death scene, ongoing investigations, or family/friend accounts ship problems (intimate partner, family, and other); (7) school pro-
and, when available, crime lab and toxicology reports. Decisions to blems; (8) job/finance/housing problems; (9) criminal/other legal
conduct alcohol and/or other toxicology testing are made by CMEs. In problems; (10) recent suicides or other deaths of family/friends; (11)
addition, the NVDRS includes variables on the circumstances of death physical health problems (e.g., terminal disease, debilitating condition,
that were calculated/coded (“Yes/Present” or “No/Not Present/ chronic pain) noted as having contributed to the death; (12) whether
Unknown”) by CDC NVDRS staff based on multiple original data the decedent was a perpetrator or victim of interpersonal violence; and
sources (e.g., reports from family/friends, suicide notes, CME/LE re- (13) whether the decedent had any recent crisis. Response categories
ports) (CDC, 2017). Some of these variables (e.g., relationship, legal, for each were “Yes/Present” or “No/Not Present/Unknown.” (CDC,
and physical health problems) are coded “Yes” only when they were 2017). Current mental health/substance abuse treatment including
believed or appeared to have caused/led to death (i.e., precipitants), pharmacotherapy and counseling and presence of a suicide note were
while others (e.g., depressed mood and other mental health, alcohol/ examined for descriptive purposes only.
substance, and job/financial/housing problems) are coded “Yes”
without the need for any indication that they directly contributed to the 2.3. Analysis
death (i.e., were precipitants/risk factors) (CDC, 2017).
The CDC granted the authors access to the NVDRS based on the data All statistical analyses were performed with Stata/MP 15. First, we
sharing agreement (DSA) for restricted access data release. We adhered used χ2 to describe and compare the two age groups (15–19 vs. 20–29)
to the DSA terms governing use, protection, and reporting of data and on demographic and precipitating/risk factors and suicide means.
to the confidential data control plan approved by our home institution's Second, we used χ2 to describe and compare rates of marijuana and
review body. In this study, we included suicide decedents aged 15–19 other substance testing and positive toxicologies among those tested
(N = 6002) and aged 20–29 (N = 20,925) whose death occurred be- (H1). Third, we used logistic regression models, with incident years
tween 2005 and 2015 to minimize potential state by time confounding 2005 through 2015 (and 2005 as the reference year), to examine
effects that could result from having fewer states' data available prior to changes in toxicology positive rates for marijuana, BAC ≥ 0.08, anti-
2005. Reflecting increases in both the number of NVDRS participating depressants, amphetamines, and opioids over the study period for both
states and overall suicide rates over time, the number of suicide dece- age groups (H2). Cocaine was excluded given its overall low positive
dents increased every year during the study period. Suicide rates also rates. Fourth, we used age-group separate multivariable logistic re-
increased for both age groups of interest in our study and did not differ gression models to test associations of marijuana-positive toxicology
from each other (p = .175). with the dichotomous incident year variable (2012–2015 vs.
2005–2011), and injury state (H3a), interactions between incident year
2.2. Measures and state (H3b), and suicide means (H4). Logistic regression results are
reported as adjusted odds ratios (AOR) with 95% confidence interval
2.2.1. Outcome variables (CI) for main effects and adjusted ratio of OR (AROR) and 95% CI for
Marijuana-positive toxicology was coded as positive (=1) or nega- interactions. Significance was set at p < .05.
tive (=0) in NVDRS. Other substances were alcohol (in blood alcohol
concentration [BAC] levels), antidepressants, amphetamines, opioids, 3. Results
and cocaine (positive or negative toxicology). Unknown or missing test
results (~1.0% in each age group) were categorized as negative. BAC 3.1. Decedent characteristics
level was first coded as positive (BAC > 0.0 g/dL) or negative
(BAC = 0), and then BAC < 0.08 g/dL or BAC ≥ 0.08 g/dL (generally Table 1 shows that compared to the 20–29 age group, the 15–19 age
considered prima facie evidence of intoxication in the US) among those group included higher proportions of females, racial/ethnic minorities,
with BAC > 0.0 g/dL. Since anticonvulsants, antipsychotics, barbitu- and those who used hanging/suffocation but a lower proportion of
rates, benzodiazepines, and muscle relaxants (central nervous system those who used firearms. A little more than half of decedents in both
depressants) were added in the NVDRS in August 2013, they were ex- age groups were injured in marijuana-legal or decriminalized states.
cluded from our analyses. Higher proportions of those aged 15–19 had family/other relationship
Independent variables were incident year (2005–2015), i.e., the problems, school problems, recent suicide/death of friends or family,
year in which the decedent died of suicide, injury/death state (states and recent crises, while lower proportions had a previous suicide at-
where marijuana use was legalized for medical and/or recreational tempt/intent disclosure, mental health, alcohol, substance use, intimate
purposes or decriminalized as of June 2014 (=1) versus states where partner, and job/finance/housing, legal, and physical health problems
marijuana use was illegal (=0) (Canna Law Group, 2014); and suicide and interpersonal violence.
means (firearms; hanging/suffocation; poisoning from alcohol/drug/
medicine overdose or gas [e.g., carbon monoxide, nitrogen]); falling/ 3.2. Post-mortem toxicology test and results
jumping from heights or other blunt contact injury involving motor
vehicle or train or other objects; and all other [laceration/sharp in- Table 2 shows that 41% and 38% of decedents aged 15–19 and
struments, drowning, fire, hypothermia, electrocution, etc.] based on 20–29, respectively, were tested for marijuana, and about a quarter of
the NVDRS ICD-10–International Statistical Classification of Diseases those tested in each age group had a positive toxicology (with no age
and Related Health Problems, 10th revision–codes for underlying cause group difference). Those aged 15–19 were most likely to test positive
of death). for marijuana followed by alcohol (BAC > 0.0), antidepressants,
463
N.G. Choi, et al. Children and Youth Services Review 100 (2019) 461–467
Table 1
1.76 (1.36–2.27)⁎⁎⁎
1.25 (0.96–1.0.63)
Demographic characteristics and precipitating/risk factors for suicide:
1.39 (1.07–1.82)⁎
1.14 (0.87 (1.51)
0.95 (0.72–1.27)
0.99 (0.75–1.32)
1.09 (0.83–1.43)
0.98 (0.75–1.28)
1.17 (0.90–1.52)
1.26 (0.99–1.61)
Comparison of youth and young adults, 2005–2015.
OR (95% CI)
9799; 43.08;
20–29 years
N (%) 15–19 years 20–29 years P
< 0.001
6002 (22.29%) 20,925
(77.71%)
1.05 (0.52–2.14)
1.22 (0.61–2.42)
1.06 (0.54–2.08)
1.35 (0.70–2.62)
0.90 (0.44–1.85)
1.15 (0.60–2.20)
1.36 (0.71–2.60)
0.99 (0.49–2.01)
0.78 (0.38–1.62)
0.94 (0.50–1.77)
2916; 5.38; 0.86
Gender (%) < 0.001
OR (95% CI)
15–19 years
Male 78.30 82.17
Female 21.70 17.83
Opioids
Race/ethnicity (%) < 0.001
Non-Hispanic White 71.01 72.13
Non-Hispanic Black 10.11 11.14
Hispanic 9.93 8.52
1.77 (1.29–2.43)⁎⁎⁎
0.49 (0.31–0.77)⁎⁎
0.56 (0.37–0.85)⁎⁎
0.56 (0.38–0.84)⁎⁎
1.75 (1.24–2.47)⁎⁎
1.67 (1.18–2.35)⁎⁎
American Indian/Alaska 3.68 2.71
0.58 (0.38–0.88)⁎
0.71 (0.47–1.08)
0.77 (0.51–1.15)
0.97 (0.68–1.40)
Native
9312; 161.26;
OR (95% CI)
Asian/Pacific Islander 1.97 2.30
20–29 years
Two or more races-non- 2.77 2.65
< 0.001
Hispanic
Other/Unknown 0.53 0.54
Precipitating/risk factors (%)
0.96 (0.35–2.62)
1.06 (0.40–2.81)
0.71 (0.26–1.92)
0.39 (0.12–1.31)
0.94 (0.35–2.47)
0.98 (0.39–2.44)
1.40 (0.58–3.38)
1.69 (0.71–4.03)
1.47 (0.61–3.54)
1.85 (0.83–4.10)
Mental health problem 35.09 38.09 < 0.001
Amphetamine
Depressed mood 32.32 32.66 0.623
2833; 18.72;)
OR (95% CI)
15–19 years
Previous suicide attempt or 36.75 38.56 0.011
intent disclosure
Alcohol problem 5.73 14.58 < 0.001
0.04
Other substance use problem 14.26 20.73 < 0.001
Intimate partner problem 28.06 36.00 < 0.001
Family relationship problem 14.10 6.66 < 0.001
1.48 (1.10–1.99)⁎⁎⁎
2.00 (1.51–2.65)⁎⁎⁎
1.83 (1.40–2.39)⁎⁎⁎
1.25 (0.93–1.67)
1.13 (0.84–1.52)
0.94 (0.69–1.28)
1.04 (0.77–1.40)
0.98 (0.73–1.31)
0.98 (0.74–1.29)
1.28 (0.97–1.69)
Other relationship problem 15.41 6.96 < 0.001
School problem 13.68 1.76 < 0.001
OR (95% CI)
7435; 73.95;
20–29 years
Job/financial/housing problem 6.28 15.61 < 0.001
< 0.001
Criminal/other legal problem 10.41 14.05 < 0.001
Recent suicide/death of friends 7.05 6.09 0.007
or family
Physical health problem 3.20 4.89 < 0.001
Interpersonal violence 2.47 4.42 < 0.001
2.03 (1.12–3.67)⁎
1.91 (1.07–3.43)⁎
2.00 (1.15–3.47)⁎
1.17 (0.63–2.20)
1.67 (0.92–3.05)
0.97 (0.52–1.81)
1.00 (0.53–1.88)
0.93 (0.48–1.78)
0.82 (0.44–1.54)
1.20 (0.66–2.21)
perpetrator or victim
Antidepressant
OR (95% CI)
2254; 31.31;
Recent crisis 19.03 16.65 < 0.001
15–19 years
< 0.001
use treatment (%)
Left a suicide note (%) 31.61 26.18 < 0.001
Died in marijuana-legal or 55.21 55.17 0.954
decriminalized state (%)
1.46 (1.22–1.74)⁎⁎⁎
Changes in positive test results among those tested, 2005–2015: Logistic regression results.
1.25 (1.06–1.48)⁎⁎
0.92 (0.76–1.11)
1.11 (0.92–1.34)
1.17 (0.98–1.41)
1.07 (0.89–1.29)
1.08 (0.90–1.30)
1.15 (0.96–1.37)
1.18 (0.99–1.41)
Table 2
Post-mortem toxicology test rate and positive test results among those tested
(%, N).
1.83 (1.39–2.42)⁎⁎⁎
2.49 (1.88–3.28)⁎⁎⁎
3.08 (2.36–4.02)⁎⁎⁎
3.41 (2.65–4.38)⁎⁎⁎
1.51 (1.13–2.02)⁎⁎
1.57 (1.19–2.07)⁎⁎
7946; 326,16;
OR (95% CI)
1.84 (1.14–2.96)⁎
0.83 (0.48–1.46)
0.82 (0.47–1.44)
1.04 (0.63–1.73)
1.15 (0.69–1.94)
1.34 (0.80–2.23)
1.08 (0.66–1.76)
2437; 77.48;
N; LR χ2; P
⁎⁎⁎
⁎⁎
⁎
464
N.G. Choi, et al. Children and Youth Services Review 100 (2019) 461–467
opioids, amphetamine, and cocaine. Those aged 20–29 were most likely 4. Discussion
to test positive for alcohol (both BAC > 0.0 and BAC ≥ 0.08) followed
by marijuana, antidepressants, opioids, cocaine, and amphetamine. Results confirm that marijuana was the most commonly sub-
Additional analyses showed lower odds of post-mortem toxicology stantiated substance in post-mortem toxicology tests among youth
testing for each substance in recent years, with the odds of marijuana suicide decedents (support for H1). During the study period, an average
test rates in 2012–2015 compared to 2005–2011, 0.83 (95% of 25% of youth decedents tested for marijuana tested positive. Among
CI = 75–92) and 0.80 (95% CI = 75–84) for the 15–19 and 20–29 age young-adult decedents, marijuana was the second most substantiated
groups, respectively. substance after alcohol (BAC > 0.0 and BAC ≥ 0.08). Our findings also
show that marijuana-positive rates have increased significantly in re-
cent years, from 2012 for youth (support for H2) and 2010 for young
3.3. Changes in positive toxicology, 2005–2015
adults. The significant increase in marijuana-positive toxicology rates is
not due to higher post-mortem toxicology test rates, as these rates de-
Table 3 shows that the odds of marijuana-positive toxicologies
creased steadily over the study period. (Decreased rates of testing may
among those aged 15–19 did not differ between 2005 and 2011, but the
be attributable to strained resources for autopsies as drug overdose
odds were significantly higher in 2012 (OR = 1.63, 95%
deaths have skyrocketed.) In the case of youth, increasing marijuana-
CI = 1.01–2.63), 2013 (OR = 2.25, 95% CI = 1.39–3.64), 2014
positive toxicologies are also not likely due to increases in substance use
(OR = 1.84, 95% CI = 1.14–2.96), and 2015 (OR = 2.76, 95%
problems, because positive toxicology rates for alcohol, amphetamine,
CI = 1.79–4.26) than in 2005. Among those aged 20–29, odds began to
and opioids did not change over the study period. Higher anti-
increase in 2010 and steadily increased between 2010 (OR = 1.51, 95%
depressant-positive rates in 2013–2015 than previous years may reflect
CI = 1.13–2.02) and 2015 (OR = 3.41, 95% CI = 2.65–4.38).
an overall increase in antidepressant use among children and youth
Among those aged 15–19, the odds of BAC ≥0.08 and ampheta-
(Bachmann et al., 2016). In comparison, among young-adult decedents
mine- and opioid-positive toxicology did not change during the study
in 2013–2015, positive rates for all these other substances increased.
period, but the odds of antidepressant-positive toxicology were higher
The findings also show that marijuana-positive rates were higher in
in 2013–2015. Among those aged 20–29, the odds of a BAC ≥ 0.08 and
marijuana-legal or decriminalized states than in marijuana-illegal states
antidepressant-, amphetamine-, and opioid-positive toxicologies were
for both age groups, controlling for incident year, demographic and
higher in 2013–2015 than in 2005.
precipitating/risk factors, and suicide means (support for H3a).
However, in the interaction model, neither the main effect for state nor
3.4. Associations of marijuana-positive toxicology with incident year, the interaction effect between state and incident year were significant
injured state, and suicide means for youth (no support for H3b). Thus, liberalized marijuana policies do
not appear to have interacted with time to increase marijuana-positive
The first column in Table 4 shows that among those aged 15–19, the toxicologies among youth. The higher odds of marijuana-positive tox-
odds of a marijuana-positive toxicology were 2.21 (95% icologies in states with liberal policies in the main-effect model may be
CI = 1.78–2.73) higher in 2012–2015 compared to 2005–2011 and due to higher marijuana use among youth in these states regardless of
1.46 (95% CI = 1.18–1.79) higher in marijuana-legal or decriminalized marijuana policy. However, interaction effects show that young adults
states compared to states without these laws. The second column shows were twice as likely to have a marijuana-positive toxicology in mar-
that when the incident year × state interaction term was entered in the ijuana-legal or decriminalized states in 2012–2015 than in the same
model, the main effect of incident year remained significant states in prior years and in other states regardless of incident year.
(AOR = 1.86, 95% CI = 1.34–2.59), but the main effect for state and These findings indicate that liberalized marijuana policies do appear to
the interaction effect were not significant. The odds of a marijuana- have had some impact on the higher odds of marijuana-positive tox-
positive toxicology were lower among decedents who used poisoning icologies among young-adult suicide decedents.
than firearms (AOR = 0.35, 95% CI = 0.21–0.57). Increasing odds of marijuana-positive toxicologies in 2012–2015
The third column of Table 4 shows that among those aged 20–29, compared to 2005–2011 among youth suicide decedents, regardless of
the odds of a marijuana-positive toxicology were 2.29 (95% marijuana policy changes, should be understood in the context of
CI = 2.04–2.57) higher in 2012–2015 than in 2005–2011 and 1.39 overall (i.e., national) decreasing marijuana use rates among youth
(95% CI = 1.24–1.56) higher in marijuana-legal or decriminalized (Johnson et al., 2015; Salas-Wright et al., 2015; SAMHSA, 2018). The
states. The fourth column shows that when the incident year × state findings suggest that compared to prior years, recent youth suicide
interaction term was entered in the model, the main effect of incident decedents were more likely to have used marijuana. Since marijuana-
year (AOR = 1.47, 95% CI = 1.23–1.77) and the interaction effect positive youth suicide decedents were reported to be three-times more
(AROR = 2.02, 95% CI = 1.61–2.53) were significant. Follow-up pair- likely to have substance use problems than marijuana-negative youth
wise comparisons for the interaction variables showed no difference decedents, it is also plausible that marijuana-positive decedents were
between marijuana-legal or decriminalized states and marijuana-illegal more likely to have been marijuana (and other substance) users prior to
states in 2005–2011 (AOR = 0.97, 95% CI = 0.83–1.15); however, (and not just at the time of) their suicide. More years of data are needed
decedents in marijuana-legal or decriminalized states were twice as to evaluate if increases in marijuana-positive toxicologies in recent
likely to test positive for marijuana in 2012–2015 (AOR = 1.97, 95% years are an anomaly or the start of a new trend. The increase is wor-
CI = 1.68–2.32). The odds of a marijuana-positive toxicology were risome nevertheless given marijuana's increased availability and ac-
lower among decedents who died by poisoning (AOR = 0.46, 95% cessibility with increasing legalization coupled with research showing
CI = 0.38–0.56) and falling/other blunt injury (AOR = 0.71, 95% increased marijuana-related visits to emergency and urgent care, de-
CI = 0.55–0.91) than firearms. creasing marijuana risk perceptions, and increased use among users in
Table 4 also shows that among both age groups, marijuana-positive marijuana-legal states (Cerdá et al., 2017; Rusby et al., 2018; Wang
toxicology odds were higher among males than females, Blacks than et al., 2018).
non-Hispanic Whites, and those with than without substance use pro- Our findings also show that those who died by more violent/lethal
blems. For those aged 15–19, having school problems was associated means (firearms and hanging/suffocation and also blunt force injury in
with lower odds of a marijuana-positive toxicology. For those aged the case of young adults) were more likely to be marijuana-positive
20–29, Hispanics and those with alcohol problems had lower odds, (support for H4). Choi, DiNitto, and Marti (2017) found among youth
while those with substance use, family relationship, and interpersonal suicide decedents who used firearms, impulsivity was a likely factor
violence problems had higher odds. coupled with easily accessible firearms. Although we could not
465
N.G. Choi, et al. Children and Youth Services Review 100 (2019) 461–467
Table 4
Marijuana-positive versus negative toxicology among those tested: Associations with incident year, injury state, and suicide means, controlling for demographic
characteristics and precipitating/risk factors for suicide.
Marijuana-positive vs. negative toxicology among those aged Marijuana-positive vs. negative toxicology among those aged
15–19 years 20–29 years
AOR (95% CI) AOR or AROR (95% CI) AOR (95% CI) AOR or AROR (95% CI)
⁎
p < .05.
⁎⁎
p < .01.
⁎⁎⁎
p < .001.
determine whether marijuana use directly contributed to using more this study differ from those used in other suicide studies. Caution
violent/lethal means, overall higher rates of substance use problems should therefore be used in comparing our study findings (e.g., suicide
among marijuana-positive decedents may have contributed to im- means) with those of previous studies. Sixth, we did not adjust p values
pulsivity that led to using more violent suicide means. More in-depth for multiple comparisons, and we thus accept a small number of Type 1
investigation of this possibility is needed. The lower odds of marijuana- errors (e.g., 1% of tests at the 0.01 alpha level).
positive toxicologies among poisoning deaths also calls for more re- The study has implications for suicide prevention. First, public
search as it appears that those who took a lethal dose of other drugs/ health approaches should include monitoring marijuana use and its
medicine/gas did not mix these substances with marijuana. impact on mental health and behavior changes and evaluating the
There are study limitations. First, since only about half the dece- impact of marijuana use frequency and severity on suicidal ideation,
dents were tested and CMEs' basis for ordering toxicology testing is suicide attempts, and impulsive and violent behaviors among youth and
unknown, the actual rates of positive toxicologies may be higher. young adults. Second, youth, their parents, and others in their social
Exclusion of anticonvulsants, antipsychotics, barbiturates, benzodiaze- support systems should be educated on the negative impacts of mar-
pines, and muscle relaxants from toxicology test results may also have ijuana use during the formative years. Third, education on marijuana
contributed to underestimating the presence of other substances. should be accompanied by policies and practices that effectively restrict
Second, although marijuana use frequency and quantity influence tox- youth's access to marijuana, especially in states with or considering
icology test results, data on those were not available. Third, information adopting laws that legalize recreational marijuana use. The significant
on substance use habits and use severity would also have allowed more increase in Δ9-THC (tetrahydrocannabinol) concentration over the past
in-depth analysis of how marijuana and other substances may have decade (Chandra et al., 2019; Hasin, 2018) also underscores the im-
contributed to suicide. Fourth, Colorado was the only NVDRS partici- portance of access restriction monitoring. Fourth, the higher odds of
pating state that had legalized marijuana for recreational use during the other substance use and related problems among youth and young-
study period. Examination of the association between marijuana-posi- adult decedents who were marijuana-positive suggest the need for
tive toxicologies and legalization of recreational use is needed. Fifth, substance abuse treatment as a means of suicide prevention. Fifth, ac-
while no previous study has examined post-mortem marijuana tox- cess to firearms and other violent means by youth (and anyone else) at
icologies among suicide decedents, the age groupings (15–19, 20–29) in risk of suicide should also be restricted.
466
N.G. Choi, et al. Children and Youth Services Review 100 (2019) 461–467
467