Skip to main content
NCBI home page
JAMA Network logoLink to JAMA Network
. 2020 Mar 23;3(3):e200607. doi: 10.1001/jamanetworkopen.2020.0607

Incidence and Lethality of Suicidal Overdoses by Drug Class

Ted R Miller 1,2,✉, David I Swedler 1, Bruce A Lawrence 1, Bina Ali 1, Ian R H Rockett 3, Nancy N Carlson 4, Jennifer Leonardo 5
PMCID: PMC7090840  PMID: 32202643

This cross-sectional study investigates the epidemiology of drug poisoning suicides in US state census and national sample data.

Key Points

Question

What is the association between classes of drugs consumed and the lethality of suicidal drug overdoses?

Findings

In this cross-sectional study of state censuses and national samples that included 421 466 medically identified suicidal drug overdoses, the risk that an overdose would be fatal was highest if an opioid or barbiturate was involved. Lethality increased with age, whereas youth overdoses often involved toxins with low lethality.

Meaning

These findings suggest that drugs that are lethal in overdose when combined should be stored securely in homes, and lethal drugs should be blister packed.

Abstract

Importance

Prior lethality analyses of suicide means have historically treated drug poisoning other than alcohol poisoning as a lumped category. Assessing risk by drug class permits better assessment of prevention opportunities.

Objective

To investigate the epidemiology of drug poisoning suicides.

Design, Setting, and Participants

This cross-sectional study analyzed censuses of live emergency department and inpatient discharges for 11 US states from January 1, 2011, to December 31, 2012, as well as Healthcare Cost and Utilization Project national live discharge samples for January 1 to December 31, 2012, and January 1 to December 31, 2016, and corresponding Multiple Cause of Death census data. Censuses or national samples of all medically identified drug poisonings that were deliberately self-inflicted or of undetermined intent were identified using diagnosis and external cause codes. Data were analyzed from June 2019 to January 2020.

Main Outcomes and Measures

Distribution of drug classes involved in suicidal overdoses. Logistic regressions on the state data were used to calculate the odds and relative risk (RR) of death for a suicide act that involved a drug class vs similar acts excluding that class.

Results

Among 421 466 drug poisoning suicidal acts resulting in 21 594 deaths, 19.6% to 22.5% of the suicidal drug overdoses involved benzodiazepines, and 15.4% to 17.3% involved opioids (46.2% men, 53.8% women, and <0.01% missing; mean age, 36.4 years). Opioids were most commonly identified in fatal suicide poisonings (33.3%-47.8%). The greatest RR for poisoning suicide completion was opioids (5.20 times the mean for suicide acts that did not involve opioids; 95% CI, 4.86-5.57; sensitivity analysis range, 3.99-6.86), followed by barbiturates (RR, 4.29; 95% CI, 3.35-5.45), antidepressants (RR, 3.22; 95% CI, 2.95-3.52), antidiabetics (RR, 2.57; 95% CI, 1.94-3.41), and alcohol (conservatively, because 30% of death certifiers do not test for alcohol; RR, 2.04; 95% CI, 1.84-2.26). The updated toxin diagnosis coding in International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, used to code the 2016 data revealed that calcium channel blockers also had a high RR of 2.24 (95% CI, 1.89-2.61). Translated to attributable fractions, approximately 81% of suicides involving opioids would not have been fatal absent opioids. Similarly, 34% of alcohol-involved suicide deaths were alcohol attributable.

Conclusions and Relevance

These findings suggest that preventing access to lethal means for patients at risk for suicide should extend to drugs with high case fatality rates. Blister packing and securely storing lethal drugs seems advisable.

Introduction

Suicide researchers and prevention specialists use the phrase means matter to highlight that the mechanism used strongly influences a suicidal act’s lethality.1 Through this line of research and practice, means restriction has become a public health focus for population-level suicide prevention.2 Analyses of means, however, usually lump almost all drug overdoses into a single category.2,3,4,5,6,7 The partial exception is alcohol; a study of fatal and hospital-admitted cases8 estimated that people were 5.1 times more likely to attempt suicide when consuming alcohol than when sober. Another small but robust study9 estimated that people were 9.6 times more likely to attempt a medically treated suicide during hours when they were intoxicated. Even these studies did not assess how alcohol involvement affected lethality. A few studies on suicide prevention by means focused on restricting access to specific, easily accessible toxins,10 such as pesticides, barbiturates, paracetamol (acetaminophen), or antidepressants. Evaluations found that restricting access to these lethal means through prescription practices and sales regulation was associated with declining suicide rates.10,11,12 Thus, drug suicide prevention efforts will benefit from a systematic investigation of suicide lethality by drug class that assesses the separate and interactive risks of varied drugs.

This study examined the association of drug classes consumed with the lethality of intentionally self-inflicted drug overdoses. We estimated what proportion of people who died of suicidal acts involving specific drugs may have survived their overdose if their drug mix had omitted a given drug class. For example, we analyzed what portion of completed suicides that involved opioids might not have been fatal if opioids were not involved.

Methods

This cross-sectional analysis combined mortality and hospital data. Its centerpiece is a tabular and regression analysis of hospital data we previously purchased for unrelated research.13 We used mortality data matched by state and year from deidentified, restricted-access Multiple Cause of Death (MCOD) data sets14 that recorded the state where the individual died. We restricted analyses to individuals 6 years and older because determining suicidal intent in younger children is often difficult. Most medical examiners do not classify a child’s death as a suicide if the child is younger than 10 years15; however, suicide cases have been identified in children aged 5 to 9 years.16 This secondary analysis of deidentified data was exempted by the institutional review board of the Pacific Institute for Research and Evaluation, and the need for informed consent was waived. We followed the reporting requirements of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

We had censuses of deidentified hospital inpatient and emergency department (ED) live discharges from January 1, 2011, through December 31, 2012, from Arizona, Iowa, Nebraska, and Utah for 2012 only; California for 2011 only; and Florida, Kentucky, New Jersey, New York, North Carolina, and Rhode Island for both years. These states accounted for 37% of US residents, 34% of total suicides,17 and 39% of drug poisoning suicides in 2012.18

To assess the consistency of the 11 states with the nation, we compared their tabular findings vs 2012 Healthcare Cost and Utilization Project (HCUP) National Inpatient Sample (NIS) and Nationwide Emergency Department Sample (NEDS) tables. To ensure that our findings were current, we compared them with 2016 NIS and NEDS tables (the most recent).

Radically different sampling strategies and the absence of state identifiers in HCUP national samples preclude regressions on pooled MCOD, NIS, and NEDS data. The HCUP NIS builds on a 20% sample of discharges from every community hospital in 46 states plus Washington, DC. The HCUP NEDS 2016 sampled discharges from 953 hospitals in 36 states plus Washington, DC; the HCUP NEDS 2012 sampled discharges from 950 hospitals in 30 states. Neither NIS nor NEDS reveals a case’s state of origin. It would be impossible to calculate standard errors (SEs) accurately in a pooled data set regression.

We largely restricted our analysis to deaths and live discharges in which drug poisoning was the mechanism of injury and intent was purposefully self-inflicted or, except in sensitivity analysis, undetermined (hereafter collectively labeled suicidal). International Classification of Diseases (ICD) intent codes do not differentiate suicidal from nonsuicidal self-inflicted intentional injury. Including drug poisonings of undetermined intent was a conservative correction for documented differential undercounting of drug poisoning suicides across states.19 The opioid epidemic is severely stressing underresourced emergency health care and medicolegal death investigation systems across the nation. Drug poisoning suicides are especially difficult for death certifiers to ascertain relative to suicides by more behaviorally and forensically overt methods, such as gunshot and hanging.20,21 Undetermined is the manner-of-death category most susceptible to obscuring these suicides.22,23

We used diagnosis data to determine what drugs, if any, were present in each hospital-treated or fatal poisoning. Presence of alcohol and other drugs was determined using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) for 2011-2012 hospital data, International Statistical Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) for 2016 hospital data, and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) for deaths. We identified drug poisonings of suicidal or undetermined intent by any listed diagnosis/intent code in ICD-10 or external cause code in ICD-9-CM. Except in sensitivity analysis, we then excluded cases that also included another suicide mechanism (eg, fall). We classified the drugs involved into the following categories: 4-aminophenol derivative (eg, acetaminophen), alcohol, antiallergic or antiemetic (a single ICD code), antidepressant, antipsychotic (ie, tranquilizers other than benzodiazepines), barbiturate, benzodiazepine, cocaine, hallucinogen, opiate, other nonsteroidal anti-inflammatory drug (eg, ibuprofen), muscle relaxant, psychostimulant, salicylate (eg, aspirin), other specified drug, and unknown drug. Because testing only detects components of combination drugs such as acetaminophen and hydrocodone bitartrate, acetaminophen and oxycodone hydrochloride, or cough medicine with codeine, these drugs were counted both as an opioid and as 1 or more other drug classes. The 2016 analysis differentiated additional drug classes identifiable only in ICD-10. eTable 1 in the Supplement lists the codes included in each drug class. A maximum of 8 classes were involved in 1 case. Data management and descriptive statistics were conducted with SAS, version 9.4 (SAS Institute Inc) and RStudio, version 1.2.5033 (RStudio Inc).

Drugs involved in suicide acts were not always tested or identifiable. Fewer than 10% of nonfatal poisonings but almost 45% of fatal poisonings included unknown drugs. If the record identified any drugs, we ignored the unknown drugs. For fatalities with all drugs unknown, we used a published regression method24 to infer the presence of frequently involved drug classes, that is, opioids, benzodiazepines, and stimulants (with probabilities >0.33 counted as present). We followed a similar strategy for nonfatal cases but could only use victim sex, age group, and rurality as explanators.

We tabulated the data. By time period, we computed the percentage of cases that were fatal (ie, the case fatality rate) and the risk of suicide due to drug poisoning involving each drug class relative to drug poisonings that did not involve that class (ie, the incidence rate ratio or the risk ratio).

Statistical Analysis

Data were analyzed from June 2019 to January 2020. Our 2011-2012 state fatality and hospital data were censuses suitable for regression analysis after pooling. Using Stata, version 15.0 (StataCorp LLC), we ran a series of logistic regressions on them. The regressions estimated the odds that a poisoning suicidal act was fatal, with fatal cases equal to 1 and nonfatal cases equal to 0.

Covariates included victim age group (detailed in Table 1) by sex, 3-category urbanicity (large metropolitan/fringe, medium/small metropolitan, and micropolitan/noncore), and state. The HCUP data include urbanicity; we generated comparable categories from state and county information in the MCOD. The ED data offered no other demographic characteristics.

Table 1. Case Counts of Suicidal Drug Poisoning Overdoses by Characteristics and Yeara.

Characteristic 2011-2012 Events in 11 States 2012 National Events 2016 National Events
Deaths (n = 5081) Admissions (n = 120 623) ED discharges (n = 120 658) Deaths (n = 9693) Weighted admissions (SE) (n = 225 609) Weighted ED discharges (SE) (n = 207 365) Deaths (n = 10 525) Weighted admissions (SE) (n = 155 610) Weighted ED discharges (SE) (n = 139 160)
Male 2696 50 439 52 452 5190 99 115 (538) 85 095 (501) 5660 59 618 (419) 52 970 (405)
Age, y
6-14 9 2460 6105 12 10 552 (223) 4765 (153) 26 11 409 (222) 4240 (143)
15-20 114 13 839 25 098 225 45 657 (429) 24 325 (328) 274 39 757 (375) 20 160 (294)
21-25 249 12 000 16 364 486 31 338 (368) 20 470 (304) 571 20 787 (290) 14 885 (258)
26-30 305 11 389 13 480 698 26 261 (341) 19 475 (297) 783 17 046 (266) 13 090 (244)
31-39 668 19 539 19 145 1431 36 508 (393) 38 495 (396) 1687 23 989 (266) 23 225 (311)
40-49 1287 26 345 19 931 2446 36 145 (392) 42 270 (410) 2207 20 582 (290) 24 955 (320)
50-59 1445 21 556 13 615 2746 23 900 (324) 36 440 (387) 2726 15 705 (258) 23 045 (310)
≥60 1004 13 413 6675 1649 11 769 (231) 20 730 (305) 2251 7334 (180) 15 560 (263)
Undetermined intent 1211 27 954 50 234 2952 93 556 (538) 44 055 (417) 3809 31 981 (346) 15 730 (264)
With comorbidities 760 112 017 73 080 NA NA NA NA NA NA
Drug classes per case (excluding alcohol), mean (SE) 1.37 (0.01) 1.38 (0.002) 1.04 (0.002) 1.30 (0.01) 1.06 (0.002) 1.41 (0.004) 1.40 (0.01) 1.21 (0.003) 1.45 (0.01)
4-Aminophenol derivative 278 15715 9435 393 16 159 (273) 26 950 (342) 387 14 228 (246) 19 355 (289)
Alcohol 550 8621 7535 1123 11 038 (231) 14 535 (260) 1429 7153 (178) 11 080 (226)
Antiallergic/antiemetic 373 5218 4238 602 8461 (207) 10 580 (224) 813 8610 (195) 9225 (208)
Antidepressant 851 16 567 10 160 1615 20 696 (311) 32 445 (370) 1714 21 519 (297) 24 240 (316)
Antidiabetic 55 2318 845 87 1603 (85) 4400 (147) 117 1277 (78) 3670 (134)
Antiepileptic 282 6802 4580 471 8942 (212) 12 145 (239) 688 9665 (206) 13 960 (251)
Antiparkinsonian 81 2881 2449 140 3933 (136) 4775 (153) 123 2117 (99) 3635 (133)
Barbiturate 95 1146 307 150 555 (50) 1865 (96) 157 268 (35) 865 (66)
Benzodiazepine 808 34 291 20 372 1245 39 026 (414) 58 700 (459) 1551 25 283 (316) 33 020 (355)
Cocaine 130 7010 2520 228 4495 (144) 8200 (198) 470 2014 (95) 3585 (132)
Hallucinogen 13 1917 2426 28 4652 (146) 3075 (123) 28 2478 (106) 1990 (99)
Muscle relaxant 3 2707 1280 7 2646 (113) 5505 (164) 219 2077 (99) 3115 (124)
Opioid 1981 19 749 16 148 3226 35 350 (381) 33 250 (374) 4386 26 196 (319) 21 635 (302)
Other analgesic 5 380 426 7 746 (62) 510 (50) 4 535 (50) 380 (44)
Other antipsychotic 282 10 818 6070 500 11 202 (233) 19 250 (295) 653 11 978 (228) 16 970 (273)
Other NSAID 26 4796 7117 39 13 515 (248) 8680 (204) 36 14 956 (251) 8515 (200)
Psychostimulant 116 3436 2996 218 6138 (171) 7040 (184) 417 5843 (163) 5635 (164)
Salicylate 58 3237 1936 100 3397 (128) 6085 (172) 69 2812 (114) 3830 (136)
Other named drug class 546 22 435 15 955 904 31 055 (366) 40 905 (405) 722 18 357 (276) 21 665 (302)
Unknown drug only 704 3523 15 066 1470 28 001 (353) 9915 (217) 1522 19 115 (279) 6920 (181)
Open in a new tab

Abbreviations: ED, emergency department; NA, not applicable; NSAID, nonsteroidal anti-inflammatory drug.

a

Unweighted counts are given in eTable 1 in the Supplement. Unless otherwise indicated, data are expressed as number of cases.

Regressions included the presence of comorbidities as a covariate. We searched the record for the 30 Elixhauser comorbid conditions.25 Hospitals assiduously code comorbidities, which can affect treatment choices or require maintaining a medication regimen or diet. Conversely, death certifiers only record comorbidities that contributed to the death.26 Therefore, we treated comorbidity as an absent-present variable rather than a count.

Three overarching regression models covered all drug classes. Model 1 included binary variables indicating the drug classes involved. Models 2 and 3 were sensitivity analyses that paralleled model 1. Model 2 omitted cases in which none of the drug classes involved were known or inferable. Model 3 omitted cases of undetermined intent. Because exploratory data tabulations suggested youths used different drug mixes than adults, we also ran model 1 separately for those 21 years or older and younger than 21 years.

Additional regressions further probed the lethality of the 9 classes of drugs with at least 100 deaths in the 2011-2012 sample. These regressions (model 4) added a set of interaction terms between the binary indicator variable for the drug class of interest and the indicator variables for the other drug classes. We calculated the increased lethality if a focal drug class was involved in a suicide act in 2 steps.27 First, we summed the regression coefficients for the drug class indicator and its interaction terms multiplied by the percentage of a drug class’ cases that involved each interaction. Then we exponentiated the sum to generate the adjusted odds ratio. Model 4 provides our best estimates. Model 1 is best for drug classes with too few deaths to calculate model 4 reliably. By including cases with only unknown drugs and cases of undetermined intent, these models should reduce the biases resulting from differential coding in the fatal and nonfatal data.

Model 5 paralleled model 1 but included suicidal acts from all mechanisms, with binary variables indicating the primary mechanism. The findings appear in eTable 2 in the Supplement.

We converted the odds ratios to relative risks (RRs) using a correction formula that derives directly from the definitions of the 2 terms.28 That formula is apropos because we have a census of cases. For RRs exceeding 1.00, we computed the fraction of deaths involving the focal drug class that would not have been fatal if no drugs in that class had been present. We calculated this drug-attributable fraction among the exposed by subtracting the base risk of 1.00 from the RR and then dividing by the RR.29,30

Results

A total of 421 466 drug poisoning suicidal acts resulted in 21 594 deaths (46.2% men, 53.8% women, and <0.01% missing; mean age, 36.4 years). Table 1 and Table 2 as well as eTable 3 in the Supplement describe the suicidal acts and drug classes analyzed in the 3 populations: 246 362 events in the 2011-2012 state censuses (including 5081 deaths), 102 946 events (unweighted) in the 2012 national data (including 9693 deaths, with 48 340 nonfatal events and 1211 deaths also in the state censuses), and 73 369 events in the 2016 national data (including 10 525 deaths). In 2016, a nonfatal act treated in the ED involved a mean (SE) of 1.45 (0.01) drug classes; an admitted act, 1.21 (0.003) drug classes; and a fatality, 1.40 (0.004) drug classes, excluding alcohol. Across the 3 measurements, benzodiazepines were involved in 19.6% to 22.5% of the suicidal overdoses (Table 2). They were the drug class most often involved in nonfatal acts. Among fatalities, opioids (33.3%-47.8% of cases), antidepressants (16.3%-16.7%), and benzodiazepines (12.8%-15.9%) were most often involved.

Table 2. Distribution of Suicidal Acts Between Categories, Distribution of Drugs Involved in Fatal Cases, and Unadjusted RRs of Death by Characteristics and Year.

Characteristic Suicidal acts, % RR (95% CI)
State cases 2011-2012 National cases 2012 National cases 2016 State deaths 2011-2012 National deaths 2012 National deaths 2016 2011-2012 2012 2016
Male 42.9 42.8 38.7 53.1 53.5 53.8 1.51 (1.43-1.59) 1.60 (1.54) 1.94 (1.87-2.02)
Age, y
6-14 3.5 3.5 5.1 0.2 0.1 0.2 0.05 (0.02-0.08) 0.04 (0.02-0.06) 0.05 (0.03-0.07)
15-20 15.9 15.9 19.7 2.2 2.3 2.6 0.12 (0.10-0.15) 0.13 (0.11-0.15) 0.11 (0.10-0.13)
21-25 11.6 11.8 11.9 4.9 5.0 5.4 0.39 (0.35-0.44) 0.40 (0.37-0.44) 0.44 (0.41-0.48)
26-30 10.2 10.5 10.1 6.0 7.2 7.4 0.56 (0.50-0.63) 0.69 (0.63-0.73) 0.74 (0.69-0.79)
31-39 16.0 17.3 16.0 13.1 14.8 16.0 0.80 (0.73-0.86) 0.86 (0.81-0.90) 1.04 (0.99-1.09)
40-49 19.3 18.3 15.6 25.3 25.2 21.0 1.42 (1.33-1.51) 1.55 (1.48-1.62) 1.49 (1.42-1.56)
50-59 14.9 14.3 13.6 28.4 28.3 25.9 2.28 (2.14-2.42) 2.44 (2.33-2.55) 2.31 (2.21-2.41)
≥60 8.6 7.7 8.2 19.8 17.0 21.4 2.63 (2/46-2.81) 2.51 (2.38-2.65) 3.14 (3.01-3.29)
Undetermined intent 32.2 31.8 16.9 32.2 30.5 36.2 0.66 (0.62-0.70) 0.97 (0.93-1.01) 2.91 (2.80-3.02)
With comorbidities 85.0 NA NA 15.0 NA NA 0.06 (0.05-0.06) NA NA
4-Aminophenol derivative 11.2 9.9 11.1 5.5 4.1 3.7 0.50 (0.44-0.57) 0.40 (0.36-0.44) 0.32 (0.28-0.35)
Alcohol 6.8 6.1 6.4 10.8 11.6 13.6 1.67 (1.53-1.82) 2.09 (1.97-2.22) 2.37 (2.24-2.50)
Antiallergic/antiemetic 4.0 4.4 6.1 7.3 6.2 7.7 1.91 (1.71-2.11) 1.46 (1.34-1.58) 1.33 (1.24-1.43)
Antidepressant 11.2 12.4 15.5 16.7 16.7 16.3 1.60 (1.49-1.71) 1.45 (1.38-1.53) 1.10 (1.05-1.16)
Antidiabetic 1.3 1.4 1.7 1.1 0.9 1.1 0.83 (0.62-1.05) 0.66 (0.53-0.81) 0.69 (0.56-0.82)
Antiepileptic 4.7 4.9 8.0 5.6 4.9 6.5 1.18 (1.04-1.33) 1.02 (0.92-1.11) 0.84 (0.78-0.90)
Antiparkinsonian 2.2 2.0 1.9 1.6 1.4 1.2 0.72 (0.57-0.88) 0.73 (0.62-0.86) 0.62 (0.52-0.73)
Barbiturate 0.6 0.6 0.4 1.9 1.5 1.5 3.01 (2.43-3.60) 2.75 (2.34-3.19) 3.69 (3.14-4.25)
Benzodiazepine 22.5 21.4 19.6 15.9 12.8 14.7 0.65 (0.60-0.70) 0.53 (0.50-0.56) 0.74 (0.70-0.78)
Cocaine 3.9 2.9 2.0 2.6 2.4 4.5 0.64 (0.53-0.76) 0.82 (0.72-0.93) 2.39 (2.18-2.60)
Hallucinogen 1.8 1.8 1.5 0.3 0.3 0.3 0.14 (0.08-0.23) 0.17 (0.11-0.23) 0.18 (0.12-0.26)
Muscle relaxant 1.6 1.8 1.8 0.1 0.1 2.1 0.04 (0.00-0.08) 0.04 (0.01-0.07) 1.22 (1.06-1.38)
Opioid 15.4 16.1 17.3 39.0 33.3 47.8 3.52 (3.32-3.71) 2.64 (2.53-2.75) 4.55 (4.38-4.73)
Other analgesic 0.3 0.3 0.3 0.1 0.1 0.0 0.30 (0.06-0.60) 0.26 (0.07-0.48) 0.13 (0.03-0.26)
Other antipsychotic 7.0 7.0 9.7 5.6 5.2 6.2 0.78 0.69-0.88) 0.74 (0.68-0.81) 0.64 (0.59-0.69)
Other NSAID 4.8 5.0 7.7 0.5 0.4 0.3 0.10 (0.06-0.14) 0.08 (0.05-0.10) 0.04 (0.03-0.06)
Psychostimulant 2.7 3.0 3.9 2.3 2.2 4.0 0.86 (0.70-1.01) 0.75 (0.65-0.86) 1.05 (0.95-1.16)
Salicylate 2.1 2.2 2.2 1.1 1.0 0.7 0.53 (0.40-0.67) 0.48 (0.39-0.60) 0.30 (0.23-0.38)
Other named drug class 15.8 16.0 13.4 10.7 9.3 7.5 0.64 (0.59-0.70) 0.54 (0.50-0.57) 0.54 (0.51-0.58)
Unknown drug only 7.8 8.9 7.9 3.6 15.2 14.5 1.89 (1.74-2.04) 1.87 (1.77-1.98) 1.77 (1.67-1.86)
Open in a new tab

Abbreviations: NA, not applicable; NSAID, nonsteroidal anti-inflammatory drug; RR, relative risk.

The case fatality rate (eTable 3 in the Supplement) measures lethality. It was higher for males (2.6%-4.8%) than females (1.7%-2.6%) and increased with age group (6-14 years, 0.1%-0.2%; ≥60 years, 4.8%-9.0%). It consistently was highest for opioids (4.5%-9.5%), followed by antidepressants (2.9%-3.6%) and barbiturates (5.8%-12.2%). eTable 4 in the Supplement shows the percentage of cases within a given drug class for which that drug class was the only class identified for fatal and nonfatal cases. Among fatalities, antidiabetics (81.8%) and salicylates (60.3%) were the 2 drug classes that often were taken alone.

Opioids (RR, 5.20; 95% CI, 4.86-5.57) and barbiturates (RR, 4.29; 95% CI, 3.35-5.45) consistently had the highest unadjusted RRs in all periods. That pattern persisted in the regression-adjusted RRs shown in Table 3. The risk of a suicide act resulting in death was highest if opioids were involved, with a best estimate that death was 5.18 (95% CI, 4.81-5.58) times more likely. Across sensitivity analyses, this estimate ranged from 3.99 to 6.86. As Table 3 shows, this estimate translates as 75% to 87% of the suicide deaths involving opioids would not have been fatal if opioids were not involved.

Table 3. Relative Risks and Attributable Fractions Derived From Logistic Regressions.

Variable RR (95% CI) Sensitivity analysis range, estimate from models 2, 3, and 5b
Linear model 1 Cross-product model 4a
4-Aminophenol derivative 0.94 (0.82 to 1.09) 0.86 (0.71 to 1.03) 0.63 to 1.06
Alcohol 2.04 (1.84 to 2.26) 1.50 (1.32 to 1.72) 2.00 to 2.49
Antiallergic/antiemetic 3.94 (3.48 to 4.45) 1.17 (0.98 to 1.39) 2.69 to 4.72
Antidepressant 3.22 (2.95 to 3.52) 2.75 (2.48 to 30.5) 1.95 to 3.82
Antidiabetic 2.57 (1.94 to 3.41) NA 2.13 to 3.06
Antiepileptic 1.20 (1.05 to 1.39) 0.91 (0.75 to 1.10) 0.83 to 1.43
Antiparkinsonian 0.67 (0.53 to 0.87) NA 0.64 to 0.78
Barbiturate 4.29 (3.35 to 5.45) NA 4.26 to 4.64
Benzodiazepine 0.71 (0.66 to 0.78) 0.44 (0.39 to 0.49) 0.43 to 0.85
Cocaine 1.30 (1.08 to 1.58) 1.10 (0.85 to 1.40) 1.09 to 1.64
Hallucinogen 0.20 (0.11 to 0.35) NA 0.18 to 0.23
Muscle relaxant 0.04 (0.01 to 0.14) NA 0.05 to 0.05
Opioid 5.20 (4.86 to 5.57) 5.18 (4.81 to 5.58) 3.99 to 6.86
Other analgesic 0.47 (0.18 to 1.16) NA 0.42 to 0.60
Other antipsychotic 1.50 (1.31 to 1.73) 1.24 (0.85 to 1.40) 0.99 to 1.74
Other NSAID 0.24 (0.16 to 0.35) NA 0.16 to 0.26
Psychostimulant 1.44 (1.17 to 1.77) 1.26 (0.98 to 1.63) 1.22 to 1.72
Salicylate 1.59 (1.20 to 2.10) NA 1.16 to 1.85
Other named drug class 1.18 (1.07 to 3.18) NA 1.05 to 1.42
Unknown drug only 2.92 (2.64 to 3.20) NA 1.90 to 4.28
Attributable fractionc
4-Aminophenol derivative NA NA 0.06 to 2.72
Alcohol 0.51 (0.46 to 0.56) 0.34 (0.24 to 0.42) 0.50 to 0.61
Antiallergic/antiemetic 0.75 (0.71 to 0.78) 0.14 (−0.02 to 0.28) 0.63 to 0.80
Antidepressant 0.69 (0.66 to 0.72) 0.64 (0.60 to 0.67) 0.49 to 0.75
Antidiabetic 0.61 (0.48 to 0.71) NA 0.53 to 0.69
Antiepileptic 0.17 (0.05 to 0.28) −0.10 (−0.33 to 0.09) NA
Antiparkinsonism NA NA NA
Barbiturate 0.77 (0.70 to 0.82) NA 0.77 to 0.80
Benzodiazepine NA NA NA
Cocaine 0.23 (0.07 to 0.37) 0.09 (−0.17 to 0.28) 0.08 to 0.40
Hallucinogen NA NA NA
Muscle relaxant NA NA NA
Opioid 0.81 (0.79 to 0.82) 0.81 (0.79 to 0.82) 0.75 to 0.87
Other analgesic NA NA NA
Other antipsychotic 0.33 (0.24 to 0.42) 0.20 (0.05 to 0.32) 0.44 to 101.0
Other NSAID NA NA NA
Psychostimulant 0.30 (0.14 to 0.44) 0.21 (−0.02 to 0.39) 0.18 to 0.43
Salicylate 0.37 (0.16 to 0.52) NA 0.13 to 0.47
Other named drug class 0.15 (0.06 to 0.69) NA 0.05 to 0.30
Unknown drug only 0.66 (0.62 to 0.69) NA 0.47 to 0.78
Open in a new tab

Abbreviations: NA, not applicable; NSAID, nonsteroidal anti-inflammatory drug; RR, relative risk.

a

A cross-product model was not run if fewer than 100 fatalities were in the drug class. No cross-product model was run on unknown drugs.

b

Odds ratios and associated 95% CIs underlying the RRs in the sensitivity analyses appear in eTable 2 and eTable 5 in the Supplement.

c

Calculated if the RR exceeded 1.00.

Relative to the fatality risk for all other drugs combined, antiallergics or antiemetics (RR, 3.94; 95% CI, 3.48-4.45), antidepressants (RR, 3.22; 95% CI, 2.95-3.52), and alcohol (RR, 2.04; 95% CI, 1.84-2.26) had high RRs in the linear model (model 1), but their association was attenuated in a cross-product model (model 4) that accounted for the presence of other drugs. Too few cases involved barbiturates to test whether its high RR would attenuate. The RR for antidiabetics almost certainly would not attenuate because more than 80% of those deaths involved no other drugs. Besides opioids (81%), the attributable fraction of deaths involving another drug only exceeded 0.5 for barbiturates (77%), antidepressants (64%), and antidiabetics (61%). Among alcohol-involved deaths, 34% were alcohol attributable.

eTable 2 and eTable 5 in the Supplement provide the full regression models that produced the results in the sensitivity analysis column in Table 3. eTable 6 in the Supplement displays the regressions underlying the interaction models.

Table 4 as well as eTable 7 and eTable 8 in the Supplement compare drug suicide patterns of youths and adults. After controlling for drugs chosen, lethality was lower for youths than adults. The drug mixes chosen by youths and adults contrasted markedly. Youth poisoning suicide acts were more likely to include nonopioid pain relievers (eg, acetaminophen, aspirin, or ibuprofen), antidepressants, and antiallergic or antiemetic drugs. In contrast, attempts in adults were more likely to include alcohol, benzodiazepines, cocaine, opioids, and drugs to treat chronic conditions (eg, epilepsy, Parkinson disease). The adult regression yielded almost identical odds ratios by drug to the all-age model. The youth regression was hampered by sample size. With a broad 95% CI, the RRs of death for opioids (8.36 [95% CI, 5.36-13.00] vs 5.07 [95% CI, 4.73-5.42]) and barbiturates (9.43 [95% CI, 1.26-62.26] vs 4.22 [95% CI, 3.31-5.36]) were larger for youths than adults. Conversely, alcohol involvement only was associated with increased risk among adults (RR, 2.06 [95% CI, 1.86-2.29] vs 0.25 [95% CI, 0.03-1.85]).

Table 4. Youth and Adult Toxin Distributions, RRs, and Attributable Fractions for 2011-2012 State Data.

Drug Distribution, No. (%) RR (95% CI)
Youth cases (n = 47 625) Adult cases (n = 198 737) Youth deaths (n = 123) Adult deaths (n = 4958) Youth admissions (n = 16 299) Adults admissions (n = 104 324) Youths Adults Adjusted youths Adjusted adults
4-Aminophenol derivative 7658 (1.6) 54 834 (17.4) 6 (4.9) 272 (5.5) 3182 (23.4) 11 893 (11.4) 0.27 (0.08-0.51) 0.59 (0.52-0.66) 0.43 (0.19-1.01) 0.97 (0.84-1.12)
Alcohol 1723 (3.6) 17 751 (8.9) 1 (0.8) 549 (11.1) 576 (3.5) 8044 (7.7) 0.22 (0.00-0.73) 1.52 (1.39-1.66) 0.25 (0.03-1.85) 2.06 (1.86-2.29)
Antiallergic/antiemetic 2906 (6.1) 6911 (3.5) 11 (8.9) 362 (7.3) 1246 (7.6) 3968 (3.8) 1.51 (0.71-2.55) 2.18 (1.96-2.41) 2.13 (1.07-4.27) 3.98 (3.51-4.51)
Antidepressant 5841 (12.3) 21 730 (10.9) 17 (13.8) 834 (16.8) 2670 (16.4) 13 893 (13.3) 1.15 (0.61-1.83) 1.64 (1.53-1.76) 2.43 (1.38-4.27) 3.21 (2.94-3.51)
Antidiabetic 361 (0.8) 2856 (1.4) 1 (0.8) 54 (1.1) 218 (1.3) 2100 (2.0) 1.07 (0.00-3.60) 0.75 (0.56-0.95) 2.78 (0.37-19.95) 2.55 (1.92-3.37)
Antiepileptic 1156 (2.4) 10 507 (5.3) 2 (1.6) 280 (5.6) 473 (2.9) 6329 (6.1) 0.66 (0.00-1.73) 1.07 (0.95-1.20) 0.77 (0.19-3.14) 1.21 (1.05-1.40)
Antiparkinsonian 419 (0.9) 4987 (2.5) 1 (0.8) 80 (1.6) 173 (1.1) 2705 (2.6) 0.92 (0.00-3.13) 0.64 (0.50-0.78) 0.61 (0.08-4.47) 0.69 (0.54-0.87)
Barbiturate 101 (0.2) 1447 (0.7) 1 (0.8) 94 (1.9) 65 (0.4) 1081 (1.0) 3.86 (0.00-13.07) 2.63 (2.12-3.16) 9.43 (1.26-62.26) 4.22 (3.31-5.36)
Benzodiazepine 4565 (9.6) 50 896 (25.6) 10 (8.8) 798 (16.1) 1951 (12.0) 32 336 (31.0) 0.83 (0.34-1.45) 0.56 (0.52-0.60) 0.71 (0.36-1.40) 0.72 (0.66-0.78)
Cocaine 381 (0.7) 9320 (4.7) 0 130 (2.6) 163 (1.0) 6833 (6.5) NA 0.55 (0.45-0.64) NA 1.33 (1.10-1.60)
Hallucinogen 1489 (3.1) 2863 (1.4) 1 (0.8) 12 (0.2) 390 (2.4) 1525 (1.5) 0.25 (0.00-0.87) 0.55 (0.45-0.64) 0.18 (0.03-1.33) 0.20 (0.11-0.37)
Muscle relaxant 434 (0.9) 3555 (1.8) 0 3 (0.1) 227 (1.4) 2480 (2.4) NA 0.03 (0.00-0.08) NA 0.04 (0.01-0.14)
Opioid 3318 (7.0) 34 548 (17.4) 45 (36.6) 1936 (39.0) 1248 (7.7) 18 496 (17.7) 7.70 (5.23-10.95) 3.04 (2.88-3.21) 8.36 (5.36-13.00) 5.07 (4.73-5.42)
Other analgesic 3203 (0.9) 13 957 (7.0) 6 (4.9) 276 (5.6) 1522 (9.3) 9294 (8.9) NA 0.36 (0.07-0.71) NA 0.49 (0.19-1.21)
Other antipsychotic 1823 (7.0) 4712 (2.4) 6 (4.9) 110 (2.2) 669 (4.1) 2763 (2.6) 0.71 (0.22-1.39) 0.78 (0.69-0.88) 1.11 (0.47-2.61) 1.51 (1.31-1.74)
Other NSAID 5593 (0.5) 6340 (3.2) 0 26 (0.5) 1791 (11.0) 3003 (2.9) NA 0.16 (0.10-0.22) NA 0.27 (0.18-0.40)
Psychostimulant 1723 (6.7) 14 979 (7.5) 1 (0.8) 549 (11.1) 576 (3.5) 8044 (7.7) 1.29 (0.38-2.49) 0.93 (0.76-1.11) 1.31 (0.55-3.09) 1.43 (1.16-1.77)
Salicylate 2040 (3.8) 3189 (1.6) 4 (3.3) 54 (1.1) 1031 (6.3) 2204 (2.1) 0.75 (0.17-1.61) 0.67 (0.50-0.86) 1.61 (0.57-4.47) 1.58 (1.17-2.10)
Other named drug class 8246 (17.3) 30 642 (15.4) 17 (13.8) 529 (10.7) 3471 (21.3) 18 950 (18.2) 0.77 (0.42-1.21) 0.65 (0.60-0.71) 1.06 (0.62-1.84) 1.17 (1.06-1.30)
Unknown drug only 4843 (10.2) 14 423 (7.3) 11 (8.9) 693 (14.0) 639 (3.9) 2833 (2.8) 0.87 (0.43-1.47) 2.07 (1.91-2.24) 0.98 (0.48-1.97) 2.98 (2.71-3.28)
Open in a new tab

Abbreviations: NA, not applicable; NSAID, nonsteroidal anti-inflammatory drug; RR, relative risk.

The 2012 patterns for the 11 states and the nation shown in Table 2 were similar. The 2012 and 2016 national data were similar as well, except opioid-involved fatalities increased from 33.3% to 47.8%. Probing the data revealed that this resulted exclusively from deaths involving fentanyl and other synthetic opioids more than tripling from 516 to 1798 deaths.

A few drugs frequently involved in suicide acts are identifiable in ICD-10 but not ICD-9-CM. Relative risks for those drugs, based on 2016 data unadjusted for demographic characteristics and comorbidities, appear in eTable 9 in the Supplement. The attributable fraction of involved deaths for calcium channel blockers was high at 55% (RR, 2.24; 95% CI, 1.89-2.61).

Discussion

This analysis found that including opioids or barbiturates in a suicidal drug mix increased the risk that a suicide attempt would prove lethal. The increase in lethality of suicidal opioid overdoses from 2012 to 2016 paralleled the rise in unintentional overdose deaths, which resulted from the spread of nonpharmaceutical fentanyl and analogues.31

Adults were much more likely than youths to include highly lethal drugs in their suicidal mix. The descriptive data suggest that youths took whatever was at hand, including nonopioid analgesics, allergy medications, or anything easily accessible.32 This finding emphasizes the importance of locking up drugs prescribed to any family member if they can be lethal in overdose.33 Parents of children with depression or who have multiple risks probably would be wise to compile a list of all their family’s medications and ask their pharmacist to check for drugs that are likely to be lethal if combined in an overdose. Disposing of leftover drugs may also help prevent a fatal suicide attempt.34

Limiting access to lethal means is a proven way to prevent suicides.35,36,37,38 Smaller pack size and blister packaging proved protective against suicide by paracetamol.11,39 To reduce suicide risk, we recommend blister packing all opioids and other potentially lethal drugs.40,41 Suicidal acts often are impulsive, and once underway, the individual is impatient.42 Youths, in particular, are likely to simply choose a different bottle rather than patiently empty blister packs.

Adults used more considered drug mixes than youths. Their mixes increased lethality. It is unclear why adults had higher mortality rates than youths when they included alcohol in their mix. Possibly they took more drugs that interacted adversely with alcohol. Conversely, alcohol use disorder is a known risk factor for suicide.43,44,45 Therefore, alcohol involvement among adults might serve as a marker for alcohol use disorders that focused attempters on more lethal drug mixes. Antidepressant involvement might provide a similar signal.

The opioid-involved suicide rate more than doubled from 1999 and 2009, then stabilized.46 In reporting the death toll from the opioid epidemic, analysts routinely include these opioid-involved suicides.47,48,49 Our analysis confirms the wisdom of including them.47 It suggests that 75% to 87% of these deaths would not have occurred if the decedent had lacked access to opioids. Suicide acts often are impulsive and use readily accessible, often previously ideated means, so those who intentionally overdosed on opioids would be likely to try a drug overdose even if opioids were not readily available.35,36 However, without opioids in the mix, it appears that they probably would survive.

Reducing opioid involvement could save lives beyond the short term. A meta-analysis of predominantly drug poisoning cases found only 2.1% of suicide survivors died by suicide in the following 2 years.50 Indeed, means matter.

Limitations

This study has some limitations. Our analysis was hampered by coding deficiencies in mortality and hospital data. First, only a subset of drug self-intoxication cases coded as undetermined intent were suicidal.47 Of the fatalities in Table 1, 27.6% were of undetermined intent, whereas none of the drugs involved were identified in 14.6% of cases. For nonfatal cases in Table 1, these percentages were 27.1% and 8.5%, respectively. Furthermore, 30% of death certifiers would not have tested for alcohol in drug poisonings, so our alcohol-attributable fraction is probably an underestimate.51 Our regression-based estimates of the attributable fractions omitting cases with undetermined intent and unknown drug were similar to estimates including them. Second, we unavoidably omitted fatal and nonfatal suicide acts that were deliberately miscoded as unintentional to avoid stigma or insurance problems. Third, we unavoidably omitted suicide acts among the 8.3% of ICD-9-CM drug poisoning cases that lacked external cause codes. Although poisoning intent cannot be omitted in ICD-10-CM, coders are encouraged to code as unintentional rather than undetermined. The net effect of these omissions and overinclusions is unclear. Nevertheless, some readers may prefer to view the results omitting undetermined intent (model 3) as primary.

Differences in drug coding between ICD editions and sparse counts also limited our analysis. The ICD-9-CM, which underpinned our regression-adjusted estimates, lumped other drugs with cannabis, β-blockers, and calcium channel blockers. We had too few cases involving most drugs to use cross-product regressions that adjusted for drug interactions. Consequently, we could only provide 2016 unadjusted RR estimates for those drugs. For some other drugs, we saw dramatic declines in RR from unadjusted or linear regression to cross-product regression estimates. Regressions using pooled 2016 and 2017 state census data would improve our estimates.

Our data also lack recency. Comparing 2012 and 2016 data, however, suggested that the risk pattern has been stable for all drugs except opioids.

Our regression results may not be nationally representative. We used 2011-2012 data from a convenience sample of states that made affordable hospital discharge censuses available. Although we lack data from the Appalachian states that were hard hit by the opioid epidemic, we have state data from the coasts and heartland, as well as populous and sparsely populated states. Furthermore, the data sets lacked important demographic explanators, including race, ethnicity, and income. Our concerns about representativeness are softened by the consistency of the RRs in the 2011-2012 data before regression adjustment with the national RRs for 2012.

Conclusions

The suicide prevention concept of “means matter” extends to drug class within poisoning suicides. This study appears to reemphasize the need to control access to drug classes that increase the risk of dying of a suicidal overdose. This study suggests that lethal drug access is particularly an issue for youths because they rarely take a targeted set of drugs, seemingly opting for whatever is accessible.

Supplement.

eTable 1. Drug Codes for Poisonings of Suicidal and Undetermined Intent, by Coding System

eTable 2. Regression Model for 2011-2012 Data Including All Suicide Mechanisms (n = 415 595)

eTable 3. Case Fatality Rates (CFR), Case Counts of Suicidal Intent-Only Drug Poisoning Overdoses, and Unweighted Case Counts for 2012 and 2016 NIS/NEDS Data

eTable 4. Percent of Cases in the 2011-2012 Overdoses Where a Given Drug Class Was the Only Drug Class Identified

eTable 5. Complete Regression Output for Model 1 and Sensitivity Analyses for Models Excluding Unknown Drugs and Excluding Cases of Undetermined Intent

eTable 6. Regression Models and Coefficients for Drug Interactions Used to Generate Combined Odds Ratios

eTable 7. Drug Frequencies by Setting for Adults (n = 198 410) and Youths (n = 47 625) in the 2011-2012 Data Used to Generate the Percentages and Regression in Table 4

eTable 8. Regression Models and Odds Ratios for the Population Divided Into Youth (Age 6-20) and Adult (Age 21 and Older)

eTable 9. Frequency of All Drugs in 2016 Available to be Assessed in the 2016 Data With the Additional Precision Available Using ICD-10 and ICD-10-CM Codes

Click here for additional data file. (643.4KB, pdf)

References

  • 1.Harvard School of Public Health Means matter: lethal means counseling. Published 2008. Accessed August 2019. https://www.hsph.harvard.edu/means-matter/lethal-means-counseling/
  • 2.Yip PS, Caine E, Yousuf S, Chang SS, Wu KC, Chen YY. Means restriction for suicide prevention. Lancet. 2012;379(9834):-. doi: 10.1016/S0140-6736(12)60521-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Spicer RS, Miller TR. Suicide acts in 8 states: incidence and case fatality rates by demographics and method. Am J Public Health. 2000;90(12):1885-1891. doi: 10.2105/AJPH.90.12.1885 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Shenassa ED, McCaffery JM, Swan GE, et al. Intergenerational transmission of tobacco use and dependence: a transdisciplinary perspective. Nicotine Tob Res. 2003;5(suppl 1):S55-S69. doi: 10.1080/14622200310001625500 [DOI] [PubMed] [Google Scholar]
  • 5.Miller M, Azrael D, Hemenway D. The epidemiology of case fatality rates for suicide in the northeast. Ann Emerg Med. 2004;43(6):723-730. doi: 10.1016/j.annemergmed.2004.01.018 [DOI] [PubMed] [Google Scholar]
  • 6.Elnour AA, Harrison J. Lethality of suicide methods. Inj Prev. 2008;14(1):39-45. doi: 10.1136/ip.2007.016246 [DOI] [PubMed] [Google Scholar]
  • 7.Chen VC, Cheng AT, Tan HK, et al. A community-based study of case fatality proportion among those who carry out suicide acts. Soc Psychiatry Psychiatr Epidemiol. 2009;44(12):1005-1011. doi: 10.1007/s00127-009-0021-9 [DOI] [PubMed] [Google Scholar]
  • 8.Miller TR, Spicer RS. Hospital-admitted injury attributable to alcohol. Alcohol Clin Exp Res. 2012;36(1):104-112. doi: 10.1111/j.1530-0277.2011.01593.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Borges G, Cherpitel CJ, MacDonald S, Giesbrecht N, Stockwell T, Wilcox HC. A case-crossover study of acute alcohol use and suicide attempt. J Stud Alcohol. 2004;65(6):708-714. doi: 10.15288/jsa.2004.65.708 [DOI] [PubMed] [Google Scholar]
  • 10.Sarchiapone M, Mandelli L, Iosue M, Andrisano C, Roy A. Controlling access to suicide means. Int J Environ Res Public Health. 2011;8(12):4550-4562. doi: 10.3390/ijerph8124550 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Nordentoft M, Qin P, Helweg-Larsen K, Juel K. Restrictions in means for suicide: an effective tool in preventing suicide: the Danish experience. Suicide Life Threat Behav. 2007;37(6):688-697. doi: 10.1521/suli.2007.37.6.688 [DOI] [PubMed] [Google Scholar]
  • 12.Zalsman G, Hawton K, Wasserman D, et al. Suicide prevention strategies revisited: 10-year systematic review. Lancet Psychiatry. 2016;3(7):646-659. doi: 10.1016/S2215-0366(16)30030-X [DOI] [PubMed] [Google Scholar]
  • 13.Lewin Group The Value of Poison Control Centers for Medicare Beneficiaries. Measuring the Impact of Poison Center Intervention: Final Report to Health Resources and Services Administration. Health Resources and Services Administration; 2017.
  • 14.Centers for Disease Control and Prevention Multiple Cause of Death data. Reviewed February 10, 2020. Accessed July 19, 2018. https://wonder.cdc.gov/mcd.html
  • 15.Crepeau-Hobson F. The psychological autopsy and determination of child suicides: a survey of medical examiners. Arch Suicide Res. 2010;14(1):24-34. doi: 10.1080/13811110903479011 [DOI] [PubMed] [Google Scholar]
  • 16.Hedegaard H, Curtin SC, Warner M. Suicide mortality in the United States, 1999-2017. NCHS Data Brief. 2018;(330):1-8. Accessed February 18, 2020. https://www.cdc.gov/nchs/data/databriefs/db330-h.pdf [PubMed] [Google Scholar]
  • 17.National Vital Statistics System Data access—vital statistics online. Updated December 20, 2018. Accessed December 11, 2019. https://www.cdc.gov/nchs/data_access/vitalstatsonline.htm#Mortality_Multiple
  • 18.Centers for Disease Control and Prevention Injury prevention and control: welcome to WISQARS. Reviewed September 18, 2019. Accessed December 11, 2019. https://www.cdc.gov/injury/wisqars/index.html
  • 19.Rockett IRH, Hobbs GR, Wu D, et al. Variable classification of drug-intoxication suicides across US states: a partial artifact of forensics? PLoS One. 2015;10(8):e0135296. doi: 10.1371/journal.pone.0135296 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Stone DM, Holland KM, Bartholow B, et al. Deciphering suicide and other manners of death associated with drug intoxication: a Centers for Disease Control and Prevention consultation meeting summary. Am J Public Health. 2017;107(8):1233-1239. doi: 10.2105/AJPH.2017.303863 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Rockett IRH, Caine ED, Connery HS, et al. Discerning suicide in drug intoxication deaths: paucity and primacy of suicide notes and psychiatric history. PLoS One. 2018;13(1):e0190200. doi: 10.1371/journal.pone.0190200 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.National Research Council Reducing Suicide: A National Imperative. National Academies Press; 2002. [PubMed] [Google Scholar]
  • 23.Huguet N, McFarland BH, Kaplan MS. A comparison of suicides and undetermined deaths by poisoning among women: an analysis of the National Violent Death Reporting System. Arch Suicide Res. 2015;19(2):190-201. doi: 10.1080/13811118.2014.915275 [DOI] [PubMed] [Google Scholar]
  • 24.Ruhm CJ. Corrected US opioid-involved drug poisoning deaths and mortality rates, 1999-2015. Addiction. 2018;113(7):1339-1344. doi: 10.1111/add.14144 [DOI] [PubMed] [Google Scholar]
  • 25.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8-27. doi: 10.1097/00005650-199801000-00004 [DOI] [PubMed] [Google Scholar]
  • 26.Hoyert DL, Minino AM, Munson ML, Anderson RN. Medical Examiners’ and Coroner’s Handbook on Death Registration and Fetal Death Reporting. 2003 ed National Center for Health Statistics; 2013. Document PHS-2003-PHS-1110. [Google Scholar]
  • 27.Hosmer DW, Lemeshow S, Sturdivant RX. Applied Logistic Regression. 3rd ed John Wiley & Sons; 2013. doi: 10.1002/9781118548387 [DOI] [Google Scholar]
  • 28.Zhang J, Yu KF. What’s the relative risk? a method of correcting the odds ratio in cohort studies of common outcomes. JAMA. 1998;280(19):1690-1691. doi: 10.1001/jama.280.19.1690 [DOI] [PubMed] [Google Scholar]
  • 29.Porta M. A Dictionary of Epidemiology. 6th ed Oxford University Press; 2014. doi: 10.1093/acref/9780199976720.001.0001 [DOI] [Google Scholar]
  • 30.Bhopal RS. Concepts of Epidemiology: Integrating the Ideas, Theories, Principles and Methods of Epidemiology. 3rd ed Oxford University Press; 2002. [Google Scholar]
  • 31.Hedegaard H, Warner M, Miniño AM. Drug Overdose Deaths in the United States, 1999–2016. National Center for Health Statistics; 2017. [Google Scholar]
  • 32.Spiller HA, Ackerman JP, Smith GA, et al. Suicide attempts by self-poisoning in the United States among 10-25 year olds from 2000 to 2018: substances used, temporal changes and demographics. Clin Toxicol (Phila). Published online October 6, 2019. doi: 10.1080/15563650.2019.1665182 [DOI] [PubMed] [Google Scholar]
  • 33.McManus BL, Kruesi MJ, Dontes AE, Defazio CR, Piotrowski JT, Woodward PJ. Child and adolescent suicide attempts: an opportunity for emergency departments to provide injury prevention education. Am J Emerg Med. 1997;15(4):357-360. doi: 10.1016/S0735-6757(97)90124-8 [DOI] [PubMed] [Google Scholar]
  • 34.Links PS, Balchand K, Dawe I, Watson WJ. Preventing recurrent suicidal behaviour. Can Fam Physician. 1999;45:2656-2660. [PMC free article] [PubMed] [Google Scholar]
  • 35.Barber CW, Miller MJ. Reducing a suicidal person’s access to lethal means of suicide: a research agenda. Am J Prev Med. 2014;47(3)(suppl 2):S264-S272. doi: 10.1016/j.amepre.2014.05.028 [DOI] [PubMed] [Google Scholar]
  • 36.Daigle MS. Suicide prevention through means restriction: assessing the risk of substitution: a critical review and synthesis. Accid Anal Prev. 2005;37(4):625-632. doi: 10.1016/j.aap.2005.03.004 [DOI] [PubMed] [Google Scholar]
  • 37.Florentine JB, Crane C. Suicide prevention by limiting access to methods: a review of theory and practice. Soc Sci Med. 2010;70(10):1626-1632. doi: 10.1016/j.socscimed.2010.01.029 [DOI] [PubMed] [Google Scholar]
  • 38.Hawton K, Bergen H, Simkin S, et al. Long term effect of reduced pack sizes of paracetamol on poisoning deaths and liver transplant activity in England and Wales: interrupted time series analyses. BMJ. 2013;346:f403. doi: 10.1136/bmj.f403 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Hawton K, Simkin S, Deeks J, et al. UK legislation on analgesic packs: before and after study of long term effect on poisonings. BMJ. 2004;329(7474):1076. doi: 10.1136/bmj.38253.572581.7C [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Programs for suicide prevention. In: Goldsmith SK, Pellmar TC, Kleinman AM, et al; Institute of Medicine Committee on Pathophysiology and Prevention of Adolescent and Adult Suicide, eds. Reducing Suicide: A National Imperative . National Academies Press; 2002. [PubMed] [Google Scholar]
  • 41.Mann JJ, Apter A, Bertolote J, et al. Suicide prevention strategies: a systematic review. JAMA. 2005;294(16):2064-2074. doi: 10.1001/jama.294.16.2064 [DOI] [PubMed] [Google Scholar]
  • 42.Simon TR, Swann AC, Powell KE, Potter LB, Kresnow Mj, O'Carroll PW. Characteristics of impulsive suicide attempts and attempters. Suicide Life Threat Behav. 2001;32(1)(suppl):49-59. doi: 10.1521/suli.32.1.5.49.24212 [DOI] [PubMed] [Google Scholar]
  • 43.Wilcox HC, Conner KR, Caine ED. Association of alcohol and drug use disorders and completed suicide: an empirical review of cohort studies. Drug Alcohol Depend. 2004;76(suppl):S11-S19. doi: 10.1016/j.drugalcdep.2004.08.003 [DOI] [PubMed] [Google Scholar]
  • 44.Esang M, Ahmed S. A closer look at substance use and suicide. Am J Psychiatry Resid J. Published online June 1, 2018. doi: 10.1176/appi.ajp-rj.2018.130603 [DOI] [Google Scholar]
  • 45.Pompili M, ed. Suicide: A Global Perspective. Bentham Science; 2012. doi: 10.2174/97816080504991120101 [DOI] [Google Scholar]
  • 46.Braden JB, Edlund MJ, Sullivan MD. Suicide deaths with opioid poisoning in the United States: 1999-2014. Am J Public Health. 2017;107(3):421-426. doi: 10.2105/AJPH.2016.303591 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Olfson M, Rossen LM, Wall MM, Houry D, Blanco C. Trends in intentional and unintentional opioid overdose deaths in the United States, 2000-2017. JAMA. 2019;322(23):2340-2342. doi: 10.1001/jama.2019.16566 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Davis GG; The National Association of Medical Examiners and American College of Medical Toxicology Expert Panel on Evaluating and Reporting Opioid Deaths . National Association of Medical Examiners position paper: recommendations for the investigation, diagnosis, and certification of deaths related to opioid drugs. Acad Forensic Pathol. 2013;3(1):77-83. doi: 10.23907/2013.011 [DOI] [Google Scholar]
  • 49.Hedegaard H, Bastian BA, Trinidad JP, Spencer M, Warner M. Drugs most frequently involved in drug overdose deaths: United States, 2011-2016. Natl Vital Stat Rep. 2018;67(9):1-14. [PubMed] [Google Scholar]
  • 50.Carroll R, Metcalfe C, Gunnell D. Hospital presenting self-harm and risk of fatal and non-fatal repetition: systematic review and meta-analysis. PLoS One. 2014;9(2):e89944. doi: 10.1371/journal.pone.0089944 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Conner KR, Huguet N, Caetano R, et al. Acute use of alcohol and methods of suicide in a US national sample. Am J Public Health. 2014;104(1):171-178. doi: 10.2105/AJPH.2013.301352 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement.

eTable 1. Drug Codes for Poisonings of Suicidal and Undetermined Intent, by Coding System

eTable 2. Regression Model for 2011-2012 Data Including All Suicide Mechanisms (n = 415 595)

eTable 3. Case Fatality Rates (CFR), Case Counts of Suicidal Intent-Only Drug Poisoning Overdoses, and Unweighted Case Counts for 2012 and 2016 NIS/NEDS Data

eTable 4. Percent of Cases in the 2011-2012 Overdoses Where a Given Drug Class Was the Only Drug Class Identified

eTable 5. Complete Regression Output for Model 1 and Sensitivity Analyses for Models Excluding Unknown Drugs and Excluding Cases of Undetermined Intent

eTable 6. Regression Models and Coefficients for Drug Interactions Used to Generate Combined Odds Ratios

eTable 7. Drug Frequencies by Setting for Adults (n = 198 410) and Youths (n = 47 625) in the 2011-2012 Data Used to Generate the Percentages and Regression in Table 4

eTable 8. Regression Models and Odds Ratios for the Population Divided Into Youth (Age 6-20) and Adult (Age 21 and Older)

eTable 9. Frequency of All Drugs in 2016 Available to be Assessed in the 2016 Data With the Additional Precision Available Using ICD-10 and ICD-10-CM Codes

Click here for additional data file. (643.4KB, pdf)

Articles from JAMA Network Open are provided here courtesy of American Medical Association

RESOURCES