Suspected Heroin Overdoses in US Emergency Departments, 2017–2018

Abstract

Objectives. To describe changes in suspected heroin overdose emergency department (ED) visits.

Methods. We analyzed quarterly and yearly changes in heroin overdoses during 2017–2018 by using data from 23 states and jurisdictions (including the District of Columbia) funded by the Centers for Disease Control and Prevention Enhanced State Opioid Overdose Surveillance program. The analyses included the Pearson χ² test to detect significant changes.

Results. Both sexes, all age groups, and some states exhibited increases from quarter 1 (Q1) 2017 to Q2 2017 and significant decreases in both quarters from Q3 2017 to Q1 2018 in heroin overdose ED visits. Overall, there was a significant yearly decline of 21.5% in heroin overdose ED visits. Three states had significant yearly increases (Illinois, Indiana, and Utah), and 9 states (Kentucky, Maryland, Massachusetts, New Hampshire, Ohio, Pennsylvania, Rhode Island, West Virginia, and Wisconsin) and the District of Columbia had significant decreases.

Conclusions. We identified decreases in heroin overdose ED visits from 2017 through 2018, but these declines were not consistent among states. Even with the possibility of a stabilization or slowing of this epidemic, it is important that the field of public health and its partners implement strategies to prevent overdoses and target emerging hot spots.

The leading cause of injury death in the United States is drug overdose.¹ Overdose deaths involving opioids (i.e., opioid pain relievers and illicit opioids such as heroin or illicitly manufactured fentanyl) have quadrupled since 1999, and in 2017 68% of drug overdose deaths involved opioids.² Monitoring is difficult as a result of time lags in reporting of overdose deaths; however, emergency department (ED) data can be used to rapidly detect changes in overdose trends.^3,4

In 2016, the Enhanced State Opioid Overdose Surveillance (ESOOS) program (https://www.cdc.gov/drugoverdose/foa/state-opioid-mm.html) was created to increase the timeliness and comprehensiveness of nonfatal and fatal opioid overdose reporting. We analyzed ED data from ESOOS-funded states to assess changes in suspected heroin overdoses from 2017 to 2018.

METHODS

We analyzed available aggregate data from January 1, 2017, to June 30, 2018, in 23 of 33 ESOOS states and jurisdictions (including the District of Columbia; hereafter “states”). Twenty-one states shared syndromic ED data, and 2 shared hospital ED billing data. States used case definition guidance provided by the Centers for Disease Control and Prevention (CDC) on suspected heroin overdoses. For states sharing data with CDC’s National Syndromic Surveillance Program (NSSP), a suspected heroin overdose query was built into the system for querying ease. States not using the NSSP or using hospital discharge data used the CDC case definition guidance to build their own definition (see the CDC’s guidance on the case definition tab at https://www.cdc.gov/drugoverdose/data/nonfatal.html). To protect confidentiality, states shared aggregate data with the CDC. Data from the 23 ESOOS states included all ED visits among patients aged 11 years or older and captured more than 80% of all ED visits to facilities not including federal hospitals.

Billing data definitions included the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM; Geneva, Switzerland: World Health Organization; 1980) and 10th Revision, Clinical Modification (ICD-10-CM; Geneva, Switzerland: World Health Organization; 2015) diagnosis codes for initial ED encounters resulting from unintentional and undetermined heroin poisonings (965.01, E850.0, T40.1X1A, and T40.1X4A). We excluded diagnosis codes for suicide and assault so that we would be able to detect changes in unintentional overdoses. Syndromic ED definitions included ICD-9-CM and ICD-10-CM codes, heroin-related SNOMED codes (https://www.nlm.nih.gov/healthit/snomedct), and chief complaint text. To be included, overdose and heroin terms had to be present in chief complaint text. Withdrawal or detoxification terms were excluded to eliminate nonoverdose visits.³ We used both discharge codes and chief complaint text to capture suspected heroin overdose visits and decrease false positives.

Data on suspected heroin overdoses (the dependent variable) were derived according to the earlier-mentioned definition. Information on sex and age groups was captured; age categories were 11 to 24 years, 25 to 34 years, 35 to 54 years, and 55 years or older. With respect to time period (the independent variable), we included 4 quarters of 2017 and 2 quarters of 2018: January 1 to March 31, 2017 (quarter 1 [Q1] 2017); April 1 to June 30, 2017 (Q2 2017); July 1 to September 30, 2017 (Q3 2017); October 1 to December 31, 2017 (Q4 2017); January 1 to March 31, 2018 (Q1 2018); and April 1 to June 30, 2018 (Q2 2018).

We calculated rates of suspected heroin overdoses (e.g., Q1 heroin overdose ED visits divided by Q1 total ED visits, multiplied by 10 000) for each quarter and demographic group.⁴ Quarterly and yearly percentage changes (i.e., Q2 2017 to Q2 2018) were calculated and analyzed by sex, age group, and state. The two-tailed Pearson χ² test was used to determine significant changes; SAS version 9.4 (SAS Institute, Cary, NC) was used in conducting this analysis.

RESULTS

Of the 81 131 400 overall ED visits, 100 734 were classified as involving suspected heroin overdoses. All percentage changes are presented in Table 1 by sex, age group, and state. Overall, there was a significant yearly decrease of 21.5% in heroin overdose ED visits in the 23 ESOOS states; however, there were fluctuations between quarters, with an overall significant increase from Q1 to Q2 2017 (12.6%), a stabilization from Q2 to Q3 2017 (−1.3%; nonsignificant), significant decreases from Q3 to Q4 2017 (−13.3%) and from Q4 2017 to Q1 2018 (−16.1%), and a significant increase from Q1 to Q2 2018 (9.4%).

TABLE 1—

Quarterly Percentage Changes in Suspected Heroin Overdoses in 23 Enhanced State Opioid Overdose Surveillance Program States, by Sex, Age Group, and State: January 1, 2017–June 30, 2018

	Change, %
Variable	Q1 2017–Q2 2017	Q2 2017–Q3 2017	Q3 2017–Q4 2017	Q4 2017–Q1 2018	Q1 2018–Q2 2018	Q2 2017–Q2 2018
Overall	12.6*	−1.3	−13.3*	−16.1*	9.4*	−21.5*
Sex
Male	10.1*	−1.4	−12.4*	−15.8*	8.3*	−21.2*
Female	15.7*	−1.9	−14.1*	−16.2*	9.3*	−22.8*
Age group, y
11–24	7.2*	−6.3*	−12.3*	−17.6*	11.9*	−24.3*
25–34	7.7*	−3.9*	−6.7*	−16.5*	6.8*	−20.0*
35–54	18.0*	−0.9	−13.4*	−13.5*	8.5*	−19.4*
≥ 55	18.2*	13.1*	−25.6*	−19.0*	17.5*	−20.0*
State or jurisdiction
California	28.9*	4.7	−21.2*	3.6	12.7*	−3.8
Delaware	37.4*	22.9*	−10.1	−7.0	10.6	13.7
District of Columbia	64.3*	−22.1*	−30.0*	−18.3	−27.2	−67.5*
Illinois	−5.1	51.1*	−19.0*	−9.9*	6.0	16.9*
Indiana	1.7	15.7*	14.9*	−20.7*	5.7	11.5*
Kentucky	−7.9	−8.0	−7.2	−9.7*	−8.8	−29.7*
Maine	22.0	16.2	6.7	−28.8*	−4.6	−15.8
Maryland	13.9*	−21.7*	−12.2*	−4.8	4.9	−31.4*
Massachusetts	4.3	21.7*	−28.8*	−8.6*	10.4*	−12.6*
Missouri	10.1	0.9	−11.4*	−1.7	9.0	−4.2
Nevada	54.7*	−4.7	7.4	0.9	10.9	14.4
New Hampshire	33.6*	−17.0*	−11.1	−13.7	−21.6*	−50.1*
New Mexico	−16.4	3.2	−4.3	−30.4*	45.6*	0.2
North Carolina	25.4*	24.4*	−19.9*	−5.1	3.3	−2.3
Ohio	20.8*	−33.5*	−8.7*	−35.7*	8.7*	−57.6*
Pennsylvania	27.1*	−7.3*	−18.5*	−23.9*	29.6*	−25.5*
Rhode Island	6.0	−6.8	−2.0	−38.1*	22.5	−30.7*
Tennessee	−9.7	−14.9*	40.8*	−10.3	3.1	10.9
Utah	25.5	−2.3	37.3	28.5	−4.5	64.6*
Vermont	−22.9	26.2	9.1	−26.8	47.1	48.3
Virginia	10.4	−6.1	−4.5	−19.0*	31.3*	−4.6
West Virginia	0.1	−31.5*	−14.1	−28.7*	−2.9	−59.2*
Wisconsin	0.3	0.4	−34.9*	−3.1	20.5*	−23.7*

Note. Q = quarter.

^*P < .05.

Ten states witnessed significant yearly decreases ranging from 12.6% (Massachusetts) to 67.5% (District of Columbia). Significant decreases occurred mostly in eastern states (District of Columbia, Kentucky, Maryland, Massachusetts, New Hampshire, Ohio, Pennsylvania, Rhode Island, West Virginia, and Wisconsin). Conversely, 3 states witnessed significant yearly increases: Indiana (11.5%), Illinois (16.9%), and Utah (64.6%).

DISCUSSION

Our results suggest a slowing of suspected heroin overdose visits to EDs, as indicated by declines overall, declines for both sexes, and declines in all age groups from April 1, 2017, through June 30, 2018. The declines were driven primarily by substantial decreases from Q3 2017 to Q1 2018 that offset increases earlier in 2017. These results are consistent with a stabilization of heroin deaths in the United States from 2016 to 2017.² Still, significant increases occurred from Q1 to Q2 2018 in many states.

Although it is difficult to interpret these findings without additional data, several explanations can be hypothesized. First, states are intervening earlier and more often to prevent overdoses, especially locally.⁵ Decreases in states such as Kentucky and West Virginia are consistent with this explanation and published data on trends in suspected opioid overdoses.³

Second, surges in the availability of illicitly manufactured fentanyl mixed with heroin or cocaine or pressed into counterfeit pills have been associated with sharp increases in overdose deaths.⁶ Four states (Kentucky, Maryland, Massachusetts, Ohio) with yearly decreases were among the first reporting overdose outbreaks or synthetic opioid overdose deaths from 2013 to 2014.⁷ Thus, decreases may indicate a stabilization or decline in the supply of heroin and heroin adulterated with fentanyl. Indeed, the rate of fentanyl law enforcement submissions (drugs seized, purchased by undercover agents, etc) stabilized in the final 6 months of 2017,⁸ whereas the rate of heroin submissions has continued to decline since 2015.⁶ These changes should be interpreted with caution because, with increased naloxone availability, bystanders and emergency medical service providers may be treating overdose victims who never visit an ED.

Conversely, states reporting yearly increases (i.e., some western and midwestern states) have been more recently affected by fentanyl and should be monitored for changes in their illicit drug supply. The widespread distribution of fentanyl and a growing number of outbreaks associated with fentanyl analogs⁹ suggest an increase in fatal overdoses.

In addition, in comparison with 2016, 2017 data showed increases in treatment rates for heroin addiction and a lower rate of initiation of heroin use.¹⁰ Integration of ED data with other sources such as emergency medical service, mortality, and treatment data and data from law enforcement drug submissions is needed to test the hypotheses outlined here.

One of the limitations of our study is that the syndrome definitions may have underestimated or overestimated overdoses as a result of issues related to data quality, data completeness, and coding differences or misclassification (e.g., preliminary chief complaint text may have been incomplete). As updates to visits arrive, relevant overdose discharge diagnosis codes or revised chief complaint texts may be received, which may change rates. These data are sent in near real time and are not finalized on the basis of toxicological results; thus, the overdoses are considered “suspected” and not confirmed. Syndromic surveillance data should not be interpreted as exact counts.

We also lacked a granular data structure, prohibiting use of more complex analytic techniques. Seasonal factors may drive quarterly changes; however, we are unaware of studies that have uncovered seasonal variations in opioid use or overdose. Data from previous years were not available for all states; therefore, we cannot determine whether random variations introduced bias. Our ED data may not have covered an entire state, and some state systems do not include all hospitals. Consequently, we used total ED visits as the denominator instead of total state populations. Using total ED visits as the denominator takes into account that we do not have ED data from all hospitals within state systems. Finally, our findings are not generalizable to non-ESOOS states (additional limitations are reported on the CDC Injury Center Web site; see https://www.cdc.gov/drugoverdose/data/nonfatal.html).

In summary, our study showcases the importance of using ED data as an early warning system for communities to better monitor overdoses. To maximize the benefit of this system, public health departments need support in developing rapid surveillance of local trends, and the CDC’s standardized national definition needs to be refined and validated to allow more consistent tracking. With these data, public health practitioners can provide community members with real-time information on overdoses to guide prevention implementation, including distributing naloxone,¹¹ linking patients to peer navigators,¹² and initiating medication-assisted treatment¹³ for opioid use disorder. Our data demonstrate the local and dynamic nature of this epidemic and the need for timely regional, state, and local information.