US Emergency Department Visits for Acute Harms From Prescription Opioid Use, 2016–2017

Abstract

Objectives. To estimate the number of US emergency department visits for prescription opioid harms by patient characteristics, intent, clinical manifestations, and active ingredient.

Methods. We used data from medical record–based surveillance from a nationally representative 60-hospital sample.

Results. Based on 7769 cases, there were 267 020 estimated emergency department visits annually (95% confidence interval [CI] = 209 833, 324 206) for prescription opioid harms from 2016 to 2017. Nearly half of visits (47.6%; 95% CI = 40.8%, 54.4%) were attributable to nonmedical opioid use, 38.9% (95% CI = 32.9%, 44.8%) to therapeutic use, and 13.5% (95% CI = 11.0%, 16.0%) to self-harm. Co-implication with other pharmaceuticals and concurrent illicit drug and alcohol use were common; prescription opioids alone were implicated in 31.5% (95% CI = 27.2%, 35.8%) of nonmedical use visits and 19.7% (95% CI = 15.7%, 23.7%) of self-harm visits. Unresponsiveness or cardiorespiratory failure (30.0%) and altered mental status (35.7%) were common in nonmedical use visits. Gastrointestinal effects (30.4%) were common in therapeutic use visits. Oxycodone was implicated in more than one third of visits across intents.

Conclusions. Morbidity data can help target interventions, such as dispensing naloxone to family and friends of those with serious overdose, and screening and treatment of substance use disorder when opioids are prescribed long-term.

A national epidemic of drug use resulting in poisoning-related deaths began in the 1990s, with sharp increases in deaths from opioid analgesics.¹ In 2016, more than 17 000 deaths were directly attributed to prescription opioids, mostly from unintentional poisonings, and these numbers are likely underestimates.^2–4 Deaths attributed to prescription opioids show some signs of plateauing; however, they remain elevated at more than 4 times the rate in 1999.³ In response to the opioid overdose epidemic, the US Department of Health and Human Services declared a public health emergency in 2017.⁵

To better gauge the full magnitude of the epidemic and to effectively target interventions to prevent nonfatal as well as fatal events, data on morbidity from prescription opioid use is helpful. Therapeutic use of prescription opioids is a common cause of emergency department (ED) visits for adverse drug events, and opioids have been identified as a high-priority medication class in the National Action Plan for Adverse Drug Event Prevention.⁶ Visits to the ED for opioid overdoses are increasing in most jurisdictions; however, since discontinuation of the Drug Abuse Warning Network after 2011, detailed national data describing morbidity from prescription opioids across all reasons for use are limited.⁷

We used a newly expanded, nationally representative public health surveillance system to (1) estimate numbers of ED visits for prescription opioid harms by patient characteristics and intent of use and (2) identify clinical manifestations and specific implicated opioids.

METHODS

We based national estimates of ED visits for harms from prescription opioids on data from the National Electronic Injury Surveillance System–Cooperative Adverse Drug Event Surveillance (NEISS-CADES) project, a collaboration of the Centers for Disease Control and Prevention (CDC), the US Consumer Product Safety Commission, and the US Food and Drug Administration. NEISS-CADES is an active public health surveillance system based on a nationally representative, stratified probability sample of 60 hospitals in the United States and its territories with at least 6 beds and a 24-hour ED.^8,9

Trained data abstractors review medical records of every ED visit to identify harms (adverse events) from therapeutic pharmaceutical use and, beginning in 2016, harms from pharmaceuticals used for any reason based on clinician documentation.⁹ Abstractors in the 60 participating hospitals record up to 4 implicated pharmaceutical products, patient demographics, intent of pharmaceutical use, narrative descriptions of the event (including clinical manifestations, precipitating circumstances, and concurrent illicit drug or alcohol use), clinician diagnoses, laboratory testing, treatments administered in the ED or by emergency medical services, and discharge disposition. Clinical manifestations are coded according to the Medical Dictionary for Regulatory Activities (MedDRA), version 9.1 (International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use, September 2006).

Definitions

For this analysis, cases included ED visits from January 1, 2016, to December 31, 2017, for harms attributed to use of prescription opioids. Prescription opioids included single-ingredient or combination products. In some cases, prescription opioids were identified based solely on toxicology testing (e.g., when patients were unable or unwilling to provide a drug use history). We excluded cases in which the only implicated opioid was an opioid-containing cough medication or an unspecified opioid (e.g., diagnosis of opioid overdose but no indication if prescription opioid or heroin). We did not include cases involving inadequate therapy, drug withdrawal, detoxification treatment, medical clearance, occupational exposures, harms from ED treatment, and deaths.

We classified intent of pharmaceutical use as therapeutic (e.g., adverse effects, allergic reactions, medication errors, and unsupervised pediatric ingestions),^8,9 self-harm (administration of pharmaceuticals to injure or kill oneself), or nonmedical use. Nonmedical use included abuse (documented by clinician diagnosis or description of recreational use; e.g., “to get high”), therapeutic misuse (documented therapeutic intent, but not used as directed; e.g., taking someone else’s prescription medication for pain, intentionally taking larger doses than prescribed), and opioid overdoses without indication of therapeutic intent, abuse, or self-harm (e.g., patients found unresponsive by paramedics and unable or unwilling to provide description of circumstances or intent). Although concern has been raised that the term “abuse” may contribute to stigma,^10,11 it is employed here because the term remains commonly used by clinicians in the medical documentation.

Statistical Analysis

To calculate national estimates of ED visits for prescription opioid harms, the US Consumer Product Safety Commission weights cases based on the inverse probability of selection, adjusted for nonresponse and poststratified to adjust for changes in the total number of hospital ED visits.¹² State- or local-level estimates cannot be calculated. We calculated national estimates of ED visits and corresponding 95% confidence intervals (CIs) by using the SURVEYMEANS procedure in SAS version 9.4 (SAS Institute, Cary, NC) to account for sample weights and complex sample design. We calculated annual national estimates by dividing the cumulative NEISS-CADES estimates for the 2-year period from 2016 to 2017 by 2. We considered estimates based on fewer than 20 cases or total estimates fewer than 1200 to be statistically unstable and these are not shown. Estimates with a coefficient of variation greater than 30% may be statistically unstable and are noted. We calculated age-specific population rates by using intercensal estimates from the US Census Bureau,¹³ which were considered free of sampling error.

RESULTS

Based on 7769 unweighted cases, there were an estimated 267 020 (95% CI = 209 833, 324 206) ED visits for harms from prescription opioids in the United States annually from 2016 to 2017. Nearly half (47.6%; 95% CI = 40.8%, 54.4%) were visits attributable to nonmedical opioid use (hereafter “nonmedical use visits”), 38.9% (95% CI = 32.9%, 44.8%) were visits following therapeutic opioid use (hereafter “therapeutic use visits”), and 13.5% (95% CI = 11.0%, 16.0%) were visits after self-harm involving opioids (hereafter “self-harm visits”; Table 1). One half of nonmedical use visits were for abuse (51.8%; 95% CI = 46.5%, 57.0%), 38.5% (95% CI = 34.0%, 43.1%) were for overdoses without indication of intent, and 9.7% (95% CI = 7.4%, 12.0%) were for therapeutic misuse.

TABLE 1—

National Estimates of Emergency Department Visits for Harms From Prescription Opioids, by Patient Characteristics and Intent: United States, 2016–2017

	Nonmedical Usea		Therapeutic Useb		Self-Harm
Patient Characteristics	Cases, No.	Annual Estimate, No. (%; 95% CI)	Cases, No.	Annual Estimate, No. (%; 95% CI)	Cases No.	Annual Estimate, No. (%; 95% CI)
Patient age, yc
< 10	0	. . .	206	4 456 (4.3; 2.6, 6.0)	0	. . .
10–24	502	17 097 (13.4; 11.3, 15.6)	228	7 814 (7.5; 6.1, 9.0)	283	7 893 (21.9; 18.0, 25.8)
25–34	1 000	34 169 (26.9; 23.6, 30.1)	314	11 781 (11.4; 9.3, 13.4)	195	6 799 (18.9; 14.5, 23.2)
35–44	749	24 804 (19.5; 17.4, 21.6)	342	11 424 (11.0; 9.4, 12.6)	178	6 441 (17.9; 14.7, 21.0)
45–54	711	23 136 (18.2; 15.1, 21.3)	443	15 288 (14.7; 12.7, 16.7)	196	7 206 (20.0; 17.0, 23.0)
55–64	615	19 893 (15.6; 12.8, 18.5)	558	20 396 (19.7; 17.5, 21.8)	122	4 418 (12.3; 9.2, 15.3)
65–74	193	6 830 (5.4; 4.4, 6.4)	436	16 595 (16.0; 13.5, 18.5)	42	2 294d (6.4d; 1.9, 10.9)
> 74	33	1 182 (0.9; 0.4, 1.5)	393	16 033 (15.4; 12.2, 18.7)	26	947d (2.6; 1.2, 4.1)
Patient sex
Female	1 499	52 800 (41.5; 39.0, 44.0)	1 723	61 052 (58.8; 55.3, 62.3)	641	22 419 (62.2; 59.5, 64.9)
Male	2 307	74 377 (58.5; 56.0, 61.0)	1197	42 734 (41.2; 37.7, 44.7)	402	13 638 (37.8; 35.1, 40.5)
Dispositione
Hospitalizedf	1 574	54 183 (42.6; 31.6, 53.6)	804	26 228 (25.3; 22.0, 28.5)	880	29 780 (82.6; 76.5, 88.6)
Not hospitalized	2 232	72 994 (57.4; 46.4, 68.4)	2 115	77 546 (74.7; 71.5, 78.0)	162	6 234 (17.3; 11.3, 23.3)
Number of implicated pharmaceuticalsg
1	1 999	68 356 (53.7; 49.1, 58.4)	2 056	72 302 (69.7; 67.0, 72.3)	386	13 215 (36.7; 31.4, 41.9)
2	1 324	42 029 (33.0; 29.4, 36.7)	601	22 494 (21.7; 19.6, 23.7)	337	12 325 (34.2; 29.8, 38.5)
3	358	12 078 (9.5; 7.6, 11.4)	178	6 233 (6.0; 4.8, 7.2)	194	6 296 (17.5; 14.4, 20.5)
4	125	4 714 (3.7; 2.4, 5.0)	85	2 757 (2.7; 1.9, 3.4)	126	4 222 (11.7; 9.6, 13.9)
Total	3 806	127 177 (100.0)	2 920	103 786 (100.0)	1 043	36 057 (100.0)

Notes. CI = confidence interval. Data are from the National Electronic Injury Surveillance System–Cooperative Adverse Drug Event Surveillance project, Centers for Disease Control and Prevention.

^aIncludes pharmaceutical abuse, therapeutic misuse (use other than as directed by a clinician), and opioid overdoses without indication of intent.

^bIncludes therapeutic adverse drug events (e.g., adverse effects, allergic reactions, medication errors, unsupervised ingestions by children). An estimated 85.0% (95% CI = 78.4%, 91.6%) of visits by children younger than 10 years involved unsupervised ingestions.

^cMissing for 3 cases of nonmedical use and 1 case of self-harm.

^dCoefficient of variation > 30%.

^eMissing for 1 case of therapeutic use and 1 case of self-harm.

^fIncludes inpatient admissions, observation admissions, and transfers to other hospitals.

^gIncludes prescription and over-the-counter medications, dietary supplements, homeopathic products, and vaccines.

Nonmedical use (59.8%; 95% CI = 54.1%, 65.6%) and self-harm (58.6%; 95% CI = 53.0%, 64.2%) visits more commonly involved patients younger than 45 years, whereas therapeutic use visits more commonly involved patients aged 45 years or older (65.8%; 95% CI = 61.7%, 70.0%). The estimated proportion of visits for nonmedical use was highest among adults aged 25 to 34 years (64.8%; 95% CI = 58.6%, 71.0%) and the proportion of visits for therapeutic use was highest among adults aged 65 years or older (74.4%; 95% CI = 69.8%, 78.9%). Most visits by children younger than 10 years were for unsupervised opioid ingestions (85.0%; 95% CI = 78.4%, 91.6%). More than half (58.5%) of nonmedical opioid use visits were made by male patients, whereas therapeutic use visits (58.8%) and self-harm visits (62.2%) were more commonly made by female patients. The proportion of visits resulting in hospitalization was highest for self-harm visits (82.6%) and was significantly lower for nonmedical use (42.6%) and therapeutic use visits (25.3%; Table 1). Multiple pharmaceutical products were more commonly implicated in self-harm visits (63.3%; 95% CI = 58.1%, 68.6%) than in nonmedical use visits (46.3%; 95% CI = 41.6%, 50.9%) or therapeutic use visits (30.3%; 95% CI = 27.7%, 33.0%). Benzodiazepines were the pharmaceutical class most commonly co-implicated with opioids in ED visits across intents (nonmedical use visits: 32.9%, therapeutic use visits: 6.1%, and self-harm visits: 34.0%; Table A, available as a supplement to the online version of this article at http://www.ajph.org).

Estimated population rates of ED visits for prescription opioid harms varied by age and intent. The population rate for nonmedical use visits was highest among patients aged 25 to 34 years (7.6 per 10 000 population; 95% CI = 5.5, 9.7) and declined with age to 1.6 per 10 000 population (95% CI = 1.2, 2.0) among those aged 65 years or older (Figure A, available as a supplement to the online version of this article at http://www.ajph.org). In contrast, the population rate for therapeutic use visits was highest among patients aged 65 years or older (6.5 ED visits per 10 000 population; 95% CI = 4.2, 8.8) and declined with age to 1.1 per 10 000 population (95% CI = 0.7, 1.5) among patients younger than 10 years. Among patients aged 10 to 24 years, the rate was the same for therapeutic use (1.2 per 10 000 population; 95% CI = 0.8, 1.6) and self-harm (1.2 per 10 000 population; 95% CI = 0.9, 1.6) visits.

Concurrent use of illicit drugs was most commonly documented in ED visits for nonmedical opioid use (33.8%), and use of alcohol was most commonly documented in ED visits for self-harm (26.3%; Table A). Among nonmedical opioid use visits, marijuana was the most common illicit drug documented (16.0%). Nonmedical opioid use visits involving marijuana without other illicit drugs was most common among patients aged 10 to 24 years (23.5%; 95% CI = 17.2%, 29.7%), and was less than 10% for all other age groups. Prescription opioids were the only documented substance (licit or illicit) used in 31.5% (95% CI = 27.2%, 35.8%) of nonmedical use visits, 73.8% (95% CI = 70.8%, 76.9%) of therapeutic use visits, and only 19.7% (95% CI = 15.7%, 23.7%) of self-harm visits (Table 2).

TABLE 2—

Substances Involved in Emergency Department Visits for Harms From Prescription Opioids, by Intent: United States, 2016–2017

Concurrent Substance Usea	Nonmedical Use,b Annual Estimate, No. (%; 95% CI)	Therapeutic Use,c Annual Estimate, No. (%; 95% CI)	Self-Harm, Annual Estimate, No. (%; 95% CI)
Prescription opioids with illicit drugs (other than marijuana)	30 505 (24.0; 20.9, 27.1)	. . .	4 014 (11.1; 7.7, 14.6)
Prescription opioids with marijuana	12 498 (9.8; 7.8, 11.8)	1 107 (1.1; 0.5, 1.6)	3 371 (9.3; 6.7, 12.0)
Prescription opioids with alcohol	13 141 (10.3; 8.7, 11.9)	2 533 (2.4; 1.6, 3.3)	7 701 (21.4; 18.1, 24.6)
Prescription opioids with unspecified drugsd	3 944 (3.1; 2.3, 3.9)	. . .	905 (2.5; 1.6, 3.4)
Prescription opioids with other pharmaceuticals	27 038 (21.3; 19.0, 23.6)	23 082 (22.2; 19.8, 24.7)	12 965 (36.0; 31.3, 40.6)
Multiple prescription opioids	2 810 (2.2; 1.6, 2.8)	7 232 (7.0; 5.5, 8.5)	. . .
Single prescription opioid	37 243 (29.3; 25.3, 33.3)	69 388 (66.9; 64.0, 69.7)	6 624 (18.4; 14.5, 22.2)
Total	127 177 (100.0)	103 786 (100.0)	36 057 (100.0)

Note. CI = confidence interval. Data are from the National Electronic Injury Surveillance System–Cooperative Adverse Drug Event Surveillance project, Centers for Disease Control and Prevention. Estimates based on < 20 cases or total estimates < 1200 are considered statistically unstable and are not shown (. . .).

^aSubstances involved in emergency department visits were categorized in a mutually exclusive and hierarchical manner from “prescription opioids with illicit drugs (other than marijuana)” to “single prescription opioid.” For example, a case in which oxycodone and alprazolam were implicated and concurrent alcohol use was documented would be categorized as “prescription opioids with alcohol.”

^bIncludes pharmaceutical abuse, therapeutic misuse (use other than as directed by a clinician), and opioid overdoses without indication of intent.

^cIncludes therapeutic adverse drug events (e.g., adverse effects, allergic reactions, medication errors, unsupervised ingestions by children).

^dIncludes opioids and unspecified amphetamines for which there was not enough information to determine if the substance was a prescription product or illicit substance (e.g., heroin or illicit methamphetamine).

The most common clinical manifestations varied by intent of pharmaceutical use (Table 3). For nonmedical use visits attributable to prescription opioids, an estimated 30.0% of patients were unresponsive or experienced cardiorespiratory failure, and an additional 35.7% had altered mental status. For therapeutic use visits, the most common manifestations were gastrointestinal effects (30.4%), altered mental status (23.0%), and allergic reactions (12.2%; 9.1% mild-to-moderate and 3.1% severe reactions with respiratory effects). The most common gastrointestinal effects included constipation (66.1%; 95% CI = 58.8%, 73.4%) and abdominal pain (29.6%; 95% CI = 20.8%, 38.5%). For self-harm visits, 68.7% had only elevated drug levels or did not have manifestations documented; 10.0% of patients were unresponsive or experienced cardiorespiratory failure, and 15.3% had altered mental status.

TABLE 3—

National Estimates of Emergency Department Visits for Harms From Prescription Opioids, by Clinical Manifestation and Intent: United States, 2016–2017

Clinical Manifestationa	Nonmedical Use,b Annual Estimate, No. (%; 95% CI)	Therapeutic Use,c Annual Estimate No. (%; 95% CI)	Self-Harm, Annual Estimate, No. (%; 95% CI)
Unresponsive or cardiorespiratory failure	38 090 (30.0; 22.5, 37.4)	11 880 (11.4; 9.2, 13.7)	3 619 (10.0; 7.7, 12.4)
Severe allergic reaction	. . .	3 178 (3.1; 2.1, 4.0)	. . .
Altered mental status	45 375 (35.7; 28.6, 42.8)	23 834 (23.0; 19.5, 26.4)	5 503 (15.3; 10.9, 19.6)
Injection-related infection or reaction	6 319d (5.0d; 1.4, 8.5)	. . .	. . .
Fall or injury	3 591 (2.8; 2.2, 3.4)	2 669 (2.6; 1.8, 3.3)	. . .
Presyncope, syncope, or dyspnea	4 351 (3.4; 2.3, 4.5)	9 215 (8.9; 6.8, 10.9)	. . .
Psychiatric or other central nervous system effect	5 247 (4.1; 2.7, 5.5)	4 041 (3.9; 3.0, 4.8)	. . .
Cardiovascular effect	2 476 (1.9; 1.4, 2.5)	1 745 (1.7; 1.3, 2.1)	. . .
Mild-to-moderate allergic reaction	. . .	9 402 (9.1; 7.2, 10.9)	. . .
Gastrointestinal effect	3 177 (2.5; 1.5, 3.5)	31 596 (30.4; 26.1, 34.8)	816 (2.3; 1.1, 3.4)
Other or unspecified effect	18 141 (14.3; 9.8, 18.7)	6 184 (6.0; 4.3, 7.6)	24 784 (68.7; 62.3, 75.1)
Total	127 177 (100)	103 786 (100)	36 057 (100)

Note. CI = confidence interval. Data are from the National Electronic Injury Surveillance System–Cooperative Adverse Drug Event Surveillance project, Centers for Disease Control and Prevention. Estimates based on < 20 cases or total estimates < 1200 are considered statistically unstable and are not shown (. . .).

^aClinical manifestations were categorized in a mutually exclusive and hierarchical manner based on severity (e.g., a case involving depressed consciousness and constipation would be classified as altered mental status based on the depressed consciousness).

^bIncludes pharmaceutical abuse, therapeutic misuse (use other than as directed by a clinician), and opioid overdoses without indication of intent.

^cIncludes therapeutic adverse drug events (e.g., adverse effects, allergic reactions, medication errors, unsupervised ingestions by children).

^dCoefficient of variation > 30%.

Oxycodone was the most commonly implicated opioid ingredient across intents, implicated in an estimated 39.0% of nonmedical use, 35.6% of therapeutic use, and 38.0% of self-harm visits (Table 4). For nonmedical use visits, the next most commonly implicated opioid ingredients were methadone (20.4%) and buprenorphine (14.6%). For therapeutic use and self-harm visits, the next most commonly implicated opioid ingredients were hydrocodone (21.8% and 26.3%, respectively) and tramadol (10.8% and 15.8%, respectively).

TABLE 4—

Most Commonly Implicated Prescription Opioid Drug Products in Emergency Department Visits for Pharmaceutical Harm, by Intent: United States, 2016–2017

Opioid Product	Annual Estimate, No. (%; 95% CI)
Nonmedical usea (total estimate = 127 177 visits)
Oxycodone-containing product	49 609 (39.0; 32.7, 45.3)
Methadone	25 926b (20.4; 8.2, 32.6)
Buprenorphine-containing product	18 529 (14.6; 7.1, 22.0)
Hydrocodone-containing product	14 901 (11.7; 6.1, 17.3)
Morphine-containing product	7 814 (6.1; 4.0, 8.3)
Tramadol-containing product	4 250 (3.3; 2.4, 4.2)
Fentanylc	4 105 (3.2; 2.4, 4.0)
Unspecified prescription opioid	3 917 (3.1; 1.9, 4.2)
Hydromorphone	3 392b (2.7b; 0.7, 4.6)
Codeine-containing product	1 801 (1.4; 0.6, 2.2)
Oxymorphone	1 565b (1.2b; 0.2, 2.3)
Therapeutic used (total estimate = 103 786 visits)
Oxycodone-containing product	36 997 (35.6; 29.3, 42.0)
Hydrocodone-containing product	22 647 (21.8; 15.7, 27.9)
Tramadol-containing product	11 206 (10.8; 9.4, 12.2)
Morphine-containing product	8 175 (7.9; 6.2, 9.5)
Methadone	7 570b (7.3b; 2.0, 12.6)
Codeine-containing product	5 919 (5.7; 4.0, 7.4)
Fentanylc	5 393 (5.2; 3.6, 6.8)
Unspecified prescription opioid	4 558 (4.4; 2.5, 6.3)
Buprenorphine-containing product	4 243 (4.1; 2.3, 5.9)
Hydromorphone	3 946 (3.8; 3.2, 4.4)
Self-harm (total estimate = 36 057 visits)
Oxycodone-containing product	13 707 (38.0; 30.7, 45.3)
Hydrocodone-containing product	9 478 (26.3; 19.8, 32.8)
Tramadol-containing product	5 687 (15.8; 11.7, 19.8)
Methadone	2 480 (6.9; 3.6, 10.1)
Morphine-containing product	2 102 (5.8; 4.3, 7.4)
Buprenorphine-containing product	1 607b (4.5b; 1.4, 7.5)
Codeine-containing product	1 425b (4.0b; 1.1, 6.9)
Hydromorphone	675 (1.9; 0.9, 2.9)

Note. CI = confidence interval. Data are from the National Electronic Injury Surveillance System–Cooperative Adverse Drug Event Surveillance project, Centers for Disease Control and Prevention.

^aIncludes pharmaceutical abuse, therapeutic misuse (use other than as directed by a clinician), and opioid overdoses without indication of intent.

^bCoefficient of variation > 30%.

^cIncludes prescription products only. Does not include illicitly manufactured fentanyl or unspecified fentanyl (i.e., not enough information to determine if the fentanyl was a prescription product or illicit substance).

^dIncludes therapeutic adverse drug events (e.g., adverse effects, allergic reactions, medication errors, unsupervised ingestions by children).

The most common clinical manifestations for nonmedical use also varied by implicated opioid product. The proportion of estimated nonmedical use visits in which patients were unresponsive or experienced cardiorespiratory failure was lowest for buprenorphine (11.4%; 95% CI = 6.8%, 16.1%) and higher for fentanyl (47.3%; 95% CI = 33.2%, 61.5%), hydromorphone (46.4%; 95% CI = 22.3%, 70.5%), oxycodone (38.8%; 95% CI = 29.1%, 48.6%), morphine (35.8%; 95% CI = 24.0%, 47.6%), and hydrocodone (31.0%; 95% CI = 21.7%, 40.3%). One fifth (21.6%; 95% CI = 10.7%, 32.4%) of nonmedical visits for buprenorphine were for buprenorphine injection-related infections or reactions; an estimated 78.6% (95% CI = 64.7%, 92.5%) of these visits were attributable to use of buprenorphine/naloxone. Concurrent illicit drug or alcohol use was similar for the 4 opioids most commonly implicated in nonmedical use visits (range = 42.1%–47.1%).

DISCUSSION

From 2016 to 2017, there were more than a quarter million estimated ED visits each year for prescription opioid harms, and nearly half of these visits were attributable to nonmedical use. Overall, the population rate of ED visits for prescription opioid harms was similar across adult age groups; however, there were differences in intent of use, involvement of concurrent substances, and clinical manifestations by age, with important implications for tertiary, secondary, and primary prevention efforts.

Efforts to expand availability of naloxone for tertiary prevention of opioid overdose deaths could initially target patients who receive emergency care for serious overdose symptoms and their families or friends. In April 2018, the US Surgeon General called for expanding naloxone availability to family members and contacts of those at high risk of overdose.¹⁴ The CDC Guideline for Prescribing Opioids for Chronic Pain also recommends offering naloxone for individuals at increased risk for opioid overdose, such as those with a history of substance use disorder, high opioid dosages, concurrent benzodiazepine use, and those previously treated for opioid overdose,¹⁵ as they have increased risk of subsequent overdose.^16,17 According to these national surveillance data, 30% of patients treated in EDs following nonmedical prescription opioid use were unresponsive or experienced cardiorespiratory failure; however, naloxone is not routinely available to overdose patients on discharge from the ED.¹⁸ Patients who have had such overdoses and their family or friends could be priority candidates for naloxone and training on its use at ED discharge.^7,19,20

The ED setting also provides an opportunity to connect patients with high risk for overdose with secondary prevention programs. Patients treated for nonfatal overdoses have increased risk of repeat overdose and, according to this study, many patients treated in EDs for harms from nonmedical prescription opioid use were young (40% were aged < 35 years). Initiation of brief motivational interviewing in the ED can help reduce overdose risk behaviors²¹ and overdose risk.²² Connecting patients to peer navigators for follow-up and linkage to medication-assisted treatment programs could also be part of postoverdose protocols.^7,23 Although some EDs have implemented such programs,²⁰ implementation is not widespread. An estimated 5% of nonmedical use visits overall, and 22% that involved buprenorphine products, involved injection-related infections and other complications. Further characterization of infections could aid prevention; nonetheless, the ED setting can provide opportunities to link patients with harm reduction interventions such as syringe services programs.^24,25

Primary harm prevention can also be targeted on the basis of patient characteristics. Multiple substances were involved in more than two thirds of ED visits for nonmedical use, and the rate of nonmedical use visits was nearly 5 times higher among those aged 25 to 34 years compared with older adults. Screening to identify patients at risk for nonmedical substance use could be used during initiation and continuation of long-term opioid therapy²⁶ and could help prescribers assess whether opioid benefits are likely to outweigh risks.¹⁵ Among older adults, nearly three quarters of ED visits for prescription opioid harms involved therapeutic use, which most commonly involved gastrointestinal effects (e.g., constipation). Particularly for older adults, clinicians should recommend bowel management regimens when opioids are prescribed.¹⁵ One third of therapeutic use visits involved patients that were unresponsive or experienced cardiorespiratory failure, or had altered mental status, reinforcing the recommendation that clinicians exercise caution in prescribing opioids and prescribe the lowest effective dose and duration of therapy.¹⁵ Self-harm visits were most often made by patients younger than 45 years (59%) and by female patients (62%). Some suggest that clinicians screen patients with chronic pain or substance use disorder for suicide risk and ensure that treatment of mental health conditions is optimized.^15,27

Passive primary prevention strategies to limit the amount of opioids prescribed could reduce overdose risk across patient ages and intents of use. Most opioid pills prescribed after surgeries remain unused and become a potential source of misuse or unintentional exposure.²⁸ Drug-specific numbers of doses standardized by indication could be prepopulated in computerized physician order entry systems,²⁹ and manufacturers could make these doses available in packages of predetermined sizes (e.g., unit-dose blister packs).³⁰ These dose packs could help avoid prescription of excess doses, allow patients to better track their remaining doses, and reduce unintended access by young children.^31,32

Although prescribed less frequently than other opioids, buprenorphine or methadone was involved in approximately one third of ED visits for nonmedical use. Notably, these medications are prescribed to patients at higher risk for opioid overdose because they are often used to treat opioid use disorder. Nonetheless, substance use treatment with these medications is associated with substantial reductions in overdoses and all-cause mortality.³³ Indeed, compared with other commonly implicated opioids, ED visits involving buprenorphine products less frequently involved unresponsiveness or cardiorespiratory failure (11% vs 31%–47%).

Recent US data on morbidity from prescription opioids are limited and other existing systems do not identify specific implicated products. Other ED-based surveillance systems have different definitions and means of identifying cases, which makes direct comparison of estimates challenging. From July 2016 through September 2017, CDC’s Enhanced State Opioid Overdose Surveillance identified 142 500 ED visits for suspected unintentional or undetermined opioid overdoses by using International Classification of Diseases codes and text searches of chief complaints in 52 jurisdictions in 45 states. These data did not differentiate prescription opioids from heroin or other illicit opioids, and the method to identify overdoses varied across states.⁷ Similarly, International Classification of Diseases–code based data from the Healthcare Cost and Utilization Project estimated approximately 80 000 ED visits and hospitalizations for opioid poisoning (excluding documented heroin) in 2014, but this estimate does not include visits for self-harm, adverse effects of therapeutic use, or nonmedical use that were not diagnosed as poisonings (e.g., falls, infections, or gastrointestinal effects).³⁴

Limitations

Public health surveillance data used in this study have limitations. First, the overall burden of opioid-related morbidity is underestimated by NEISS-CADES as only acute harms from prescription opioid use that are treated in EDs were included. Patients treated in other health care settings or non–health care settings (e.g., bystander naloxone administration), patients whose harms were not acute effects of active prescription opioid use (e.g., withdrawal, seeking detoxification and substance use disorder treatment, and violence-related injuries), patients for whom a drug history could not be obtained, and deaths in or en route to the ED were not included. In addition, 1749 cases (representing nearly 55 000 estimated nonmedical use visits and 6700 self-harm visits annually) were excluded because an unspecified “opioid” was the only opioid implicated and these cases could have involved prescription or illicit opioids (e.g., heroin). Similarly, cases in which the only opioid used was an illicit opioid were excluded.

Second, ED physicians and chart abstractors may make recording errors and omissions. In particular, harms that did not contribute to the patient’s acute presentation (e.g., chronic hepatitis from injection drug use may not be documented in the ED record of an unconscious patient with an acute overdose) and harms that might require extensive evaluation to diagnose (e.g., endocarditis) are not reliably included. Nonetheless, nonfatal national data on acute harms of prescription opioid use are important to help focus prevention efforts and monitor progress.

Third, there may be misclassification in the patient’s intent of drug use. Nonmedical use may be overestimated by including opioid overdoses without indication of intent (e.g., unresponsive patients), as some could have involved therapeutic use or self-harm.^27,35 On the other hand, patients who had used drugs nonmedically or for self-harm may have reported a therapeutic purpose, leading to overestimation of therapeutic use.

Fourth, some pharmaceuticals and illicit substances were identified on the basis of laboratory testing alone, which favors identification of those included on standard toxicology screens. In addition, there is potential for false-positives on toxicology screens.

Finally, when patients are hospitalized from the ED, the type of unit (general medical, intensive care, or psychiatric) is not systematically reported. Nonetheless, leveraging an existing surveillance system such as NEISS-CADES can be a cost-efficient strategy to help assess patterns of harms by specific pharmaceuticals and concurrent use of other substances and could be expanded to include ED visits attributable to illicit substances alone.

Public Health Implications

Visits to the ED for immediate evaluation and treatment provide opportunities for targeted interventions including dispensing naloxone for future use, if needed, and linking patients to medication-assisted treatment and harm-reduction services (e.g., syringe services programs). Additional targeted interventions include screening for substance use and mental health disorders among patients on long-term opioid therapy, recommending bowel regimens, and using the lowest effective dose and duration. To address this national public health emergency, a multifaceted approach will be needed and will need to address substance use beyond prescription opioids.