Naloxone Prescriptions Following Emergency Department Encounters or Opioid Use Disorder, Overdose, or Withdrawal

Abstract

Objective

To determine the rate at which commercially-insured patients fill prescriptions for naloxone after an opioid-related ED encounter as well as patient characteristics associated with obtaining naloxone.

Methods

This is a retrospective cohort study of adult patients discharged from the ED following treatment for an opioid-related condition from 2016 to 2018 using a commercial insurance claims database (Optum Clinformatics® Data Mart). The primary outcome was a pharmacy claim for naloxone in the 30 days following the ED encounter. A multivariable logistic regression model was examined the association of patient characteristics with filled naloxone prescriptions, and predictive margins were used to report adjusted probabilities with 95% confidence intervals.

Results

21,700 patients had opioid-related ED encounters during the study period, of which 1743 (8.0%) had encounters for heroin overdose, 8825 (40.7%) for overdose due to other opioids, 5400 (24.9%) for withdrawal, and 5732 (26.4%) for other opioid use disorder conditions. 230 patients (1.1%) filled a prescription for naloxone within 30 days. Patients with heroin overdose (2.6%; 95%CI 1.7 to 3.4), recent prescriptions for opioid analgesics (1.4%; 95%CI 1.1 – 1.7), recent prescriptions for buprenorphine (1.9%; 95%CI 1.0 to 2.9), and naloxone prescriptions in the prior year (3.3%; 95%CI 1.8 to 4.8) were more likely to obtain naloxone. The rate was significantly higher in 2018 [1.9% (95%CI 1.5 to 2.2)] as compared to 0.4% (95%CI 0.3 to 0.6) in 2016.

Conclusions

Few patients use insurance to obtain naloxone by prescription following opioid-related ED encounters. Clinical and policy interventions should expand distribution of this life-saving medication in the ED.

Introduction

Wider distribution of naloxone can reduce deaths from opioid overdose.¹ Recent campaigns have sought to increase awareness of naloxone as an essential harm-reduction strategy, such as the Surgeon General’s Advisory on Naloxone and Opioid Overdose.² Overdose education and naloxone distribution (OEND) programs are effective and cost-effective interventions for at-risk patients, peers, and families.³ In addition to OEND programs that directly distribute naloxone in the community, prescriptions are strongly recommended for patients with increased risk for overdose, including those with history of overdose or high-risk prescription medications.⁴ States have also expanded access to naloxone through laws that permit third-party prescribing.²

Emergency department (ED) visits are an important opportunity to provide naloxone and education to patients at risk for opioid overdose, whether via prescription or direct distribution. Given the high risk for subsequent fatal overdose, the American College of Emergency Physicians recommends that naloxone be prescribed during an ED encounter for opioid poisoning or intoxication.^5,6 ED overdose education programs that provide take-home naloxone have demonstrated feasibility.⁷ However, these programs are not widely available.⁸ Prior studies suggest that patients at risk for opioid overdose rarely fill naloxone prescriptions but have not specifically examined the critical window following an ED encounter.^9,10

This study sought to determine the rate at which commercially-insured patients fill prescriptions for naloxone after an ED encounter for opioid use disorder, overdose, or withdrawal. We also sought to describe patient and prescription characteristics associated with filled prescriptions for naloxone.

Methods

We conducted a retrospective cohort study of patients discharged from the ED following treatment for opioid-related conditions from 2016 to 2018. We used the Optum de-identified Clinformatics® Data Mart, an administrative claims database that includes ED and pharmacy claims for patients across the United States with commercial and Medicare Advantage health plans. The Institutional Review Board at __ determined that this study was exempt from review. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

ED encounters were identified using validated International Classification of Disease, Tenth Revision, Clinical Modification diagnosis codes and Current Procedure Terminology codes (Supplement).¹¹ Conditions included opioid abuse, misuse, and dependence; opioid withdrawal; and non-fatal opioid overdose. We excluded ED encounters for patients less than 18 years and those resulting in inpatient hospital admission or observation. We also excluded encounters occurring fewer than 30 days from insurance enrollment or termination. For patients with multiple qualifying ED encounters, we included the first encounter in which the patient obtained a naloxone prescription within 30 days. If the patient did not obtain naloxone after any ED encounter, we then included the first qualifying ED encounter during the study period.

The primary outcome was a pharmacy claim for naloxone in the 30 days following the ED encounter. Naloxone claims were identified using National Drug Codes and included any formulation. Medications including naloxone as a secondary agent, such as buprenorphine/naloxone, were excluded. Data was only available for medications that were dispensed. Data was not available for medications that were prescribed but not filled, and the data do not indicate whether patients obtained the prescription during the ED encounter.

We examined patient characteristics including age, sex, race/ethnicity, and insurance type. The claims database reports race/ethnicity by self-report. We classified the ED encounter according to the year and specific diagnosis (heroin overdose, other opioid overdose including prescription and synthetic agents, opioid withdrawal, and other visits for opioid use disorder which encompass diagnosis codes for abuse, misuse, and dependence). We reviewed pharmacy claims within the preceding 30 days to identify recent claims for prescription opioid analgesics, buprenorphine, benzodiazepines. We identified prescriptions for naloxone within the preceding 1 year. No data were available on methadone maintenance therapy.

First, we described the rate at which patients filled naloxone prescriptions. We then used a multivariable logistic regression model to examine the association between the outcome and patient characteristics. We used predictive margins to report the adjusted probability of obtaining naloxone with 95% confidence intervals. Adjusted probabilities are derived from adjusted odds ratios that result from logistic regression, allowing for direct comparison of the probability that patients are estimated to fill naloxone prescription between categories of patient characteristics.^12,13 Finally, we described characteristics of naloxone claims provided in the claims database including the timing of the claim, formulation, status as initial fill or refill, and medication co-pay. Analyses were performed with Stata software, version 15 (StataCorp).

Results

The total cohort consisted of encounters for 21,700 unique patients, of which 1743 (8.0%) were for heroin overdose, 8825 (40.7%) for other opioid overdose, 5400 (24.9%) for opioid withdrawal, and 5732 (26.4%) for other opioid use disorder conditions. 4766 patients had more than one encounter during the study period. The mean age was 54.1 years (SD 18.6) and there were 11,139 (51.3%) female patients. There were 9550 (44.0%) encounters for patients greater than 60 years of age. There were 13780 (63.5%) patients of Non-Hispanic White race and ethnicity, 2759 (12.7%) non-Hispanic Black, 1795 (8.3%) Hispanic, 218 (1.0) Asian, and an additional 3148 (14.5%) with unspecified race/ethnicity.

Of all patients in the cohort, 230 (1.1%) filled a prescription for naloxone in the 30 days following the opioid-related ED encounter (Table). In the adjusted analysis, patients older than 60 years had 0.7% (95%CI 0.5 to 0.8) probability of obtaining naloxone compared to 1.9% (95%CI 1.3 to 2.5) of patients aged 18–39. Patients with heroin overdose were significantly more likely to obtain naloxone (2.6%; 95%CI 1.7 to 3.4). Patients with recent claims for prescription opioid analgesics (1.4%; 95%CI 1.1 – 1.7) and buprenorphine (1.9%; 95%CI 1.0 to 2.9) were significantly more likely to obtain naloxone. Patients who had filled naloxone prescriptions in the preceding 1 year had significantly increased probability of obtaining naloxone again (3.3%; 95%CI 1.8 to 4.8). Finally, the adjusted probability of obtaining naloxone was significantly higher in 2018 [1.9% (95%CI 1.5 to 2.2)] as compared to the reference year, 2016 [0.4% (95%CI 0.3 to 0.6)].

Table.

Patient characteristics and naloxone prescription claims following opioid-related emergency department encounters, with adjusted probability of obtaining prescribed naloxone

Patient Characteristics		All patients No. (%) N = 21700	Naloxone claim within 30 days No. (%) N = 230	No naloxone within 30 days No. (%) N = 21470	Adjusted Probability, 95% CI	P
Age, years
	18 – 39	5496 (25.3)	78 (33.9)	5418 (25.2)	1.9 (1.3–2.5)	-
	40 – 59	6654 (30.7)	78 (33.9)	6576 (30.6)	1.2 (0.9–1.4)	0.02
	> 60	9550 (44.0)	74 (32.2)	9476 (44.1)	0.7 (0.5–0.8)	<0.001
Sex
	Male	10561 (48.7)	121 (52.6)	10440 (48.6)	1.1 (0.9–1.3)	-
	Female	11139 (51.3)	109 (47.4)	11030 (51.4)	0.9 (0.8–1.2)	0.37
Race/Ethnicity
	Non-Hispanic White	13780 (63.5)	150 (65.2)	13630 (63.5)	1.1 (0.9–1.3)	-
	Non-Hispanic Black	2759 (12.7)	26 (11.3)	2733 (12.7)	1.0 (0.6–1.4)	0.61
	Hispanic	1795 (8.3)	17 (7.4)	1778 (8.3)	0.9 (0.5–1.4)	0.56
	Asian	218 (1.0)	2 (0.9)	216 (1.0)	1.0 (0.0–2.4)	0.88
	Unspecified	3148 (14.5)	35 (15.2)	3113 (14.5)	0.9 (0.6–1.2)	0.32
Year
	2016	7557 (34.8)	32 (13.9)	7525 (35.1)	0.4 (0.3–0.6)	-
	2017	7653 (35.3)	78 (33.9)	7575 (35.3)	1.0 (0.8–1.2)	<0.001
	2018	6490 (29.9)	120 (52)	6370 (29.7)	1.9 (1.5–2.2)	<0.001
Type of insurance
	Commercial	7700	88 (38.3)	7612 (35.5)	0.9 (0.6–1.1)	-
	Medicare Advantage	14000	142 (61.7)	13858 (64.6)	1.2 (1.0–1.5)	0.05
Type of ED Visit
	Heroin overdose	1743 (8.0)	43 (18.7)	1700 (7.9)	2.6 (1.7–3.4)	<0.001
	Other opioid overdose	8825 (40.7)	93 (40.4)	8732 (40.7)	1.1 (0.9–1.3)	0.32
	Opioid withdrawal	5400 (24.9)	40 (17.4)	5360 (25.0)	0.7 (0.5–0.9)	0.21
	Other opioid use disordera	5732 (26.4)	54 (23.5)	5678 (26.5)	0.9 (0.7–1.2)	-
Recent prescription opioid analgesic, 30d
	No	10884 (50.2)	106 (46.1)	10778 (50.2)	0.8 (0.7–1.0)	-
	Yes	10816 (49.8)	124 (53.9)	10692 (49.8)	1.4 (1.1–1.7)	0.001
Recent buprenorphine, 30d
	No	21009 (96.8)	216 (93.9)	20793 (96.9)	1.0 (0.9–1.2)	-
	Yes	691 (3.2)	14 (6.1)	677 (3.2)	1.9 (1.0–2.9)	.03
Recent benzodiazepine, 30d
	No	16524 (76.2)	170 (73.9)	16354 (76.2)	1.0 (0.9–1.2)	-
	Yes	5176 (23.9)	60 (26.1)	5116 (23.8)	1.1 (0.8–1.4)	0.58
Prior claim for naloxone, 1 year
	No	21356 (98.4)	212 (92.2)	21144 (98.5)	1.0 (0.9–1.1)	-
	Yes	344 (1.6)	18 (7.8)	326 (1.5)	3.3 (1.8–4.8)	<0.001

^aIncludes ICD-10 diagnosis codes for opioid abuse, misuse, and dependence

The median number of days from the ED encounter to obtaining naloxone was 4 days (IQR 1 to 13), with 66 (28.7%) patients filling the prescription on the same day or next day after the ED encounter. Among all prescriptions, 191 (83.0%) were for intranasal naloxone, 27 (11.7%) were for prefilled naloxone auto-injector syringes, and 12 (5.2%) were for single dose vials of naloxone. 189 (82.2%) prescriptions were filled for the first time. The median co-pay was $8.25 (IQR $0.00 to $31.05), with 73 patients having no co-pay and 6 patients paying more than $50.00.

Discussion

Although guidelines recommend that patients who seek emergency care for opioid use disorder, overdose, or withdrawal should be prescribed naloxone, we found that only 1% of commercially-insured patients fill a prescription within 30 days following an ED encounter. We observed a small, yet statistically significant increase from 2016 to 2018.

In this population, insurance and prescription drug coverage are not barriers to filling prescriptions. Instead, several possibilities may explain this low rate. One is that neither ED clinicians nor follow-up providers prescribe this medication. A second possibility is that patients do not fill prescriptions due to barriers such as cost, education, motivation, or stigma. Third, high-risk patients may already have naloxone in their possession. For this possibility, it is important to note that few patients in this study filled a prescription in the preceding year. However, the distribution of naloxone by public health programs would not be captured through claims data. The prevalence and relative proportions of directly distributed naloxone and prescribed naloxone are not known, but OEND programs are present in many communities.¹⁴

Regardless of these possibilities, these results suggest that ED prescriptions are not a common source of naloxone for patients with opioid use disorder. Clinical interventions should urgently address this deficit, given that patients have high risk for overdose in the days following discharge as well as low rates of follow-up treatment.⁶ EDs may adopt default prescribing options in the electronic medical record for appropriate patients. Additional policy approaches include development and reporting of ED quality measures, over-the-counter availability of naloxone, and lowering out-of-pocket costs.

If patients are prescribed naloxone but do not ultimately fill the prescription, there is greater urgency to deploy ED-based OEND programs. OEND protocols overcome patient barriers by directly dispensing naloxone and providing important education. Pharmacist, nursing, and peer-led interventions may increase uptake and efficiency. Although demonstrated to be feasible in the emergency setting, most EDs have not implemented OEND programs.⁷

Half of the patients in this study obtained a prescription opioid analgesic in the 30 days preceding the ED encounter. In this study, it is not possible to distinguish patients who misuse prescription opioid medications from those who take medications as prescribed but experience complications requiring emergency care. Nonetheless, it is recommended that patients who take prescription opioids are prescribed naloxone if they have increased risk for overdose, history of substance use disorder, or concurrent benzodiazepine use.⁴ Therefore, many patients in this population meet criteria for naloxone prescription, including older adults who were found to obtain naloxone at a lower rate than younger adults in this study.

This study has several limitations. These data cannot determine whether prescriptions originated in the ED, were written but not filled, or whether patients obtained naloxone directly through take-home naloxone or other public health interventions. In addition, this study excludes patients who declined to use their insurance and paid out-of-pocket. Another limitation which may limit generalizability is that this study was conducted in a population with commercial insurance, although many patients with opioid use disorder may be uninsured or have other types of insurance. However, we hypothesize that patients with commercial insurance are likely to have fewer barriers to filling outpatient prescriptions and that this study may overestimate this rate for other populations. This study did not directly account for mortality, but a sensitivity analysis excluding patients who did not file any medical or pharmacy claim beyond 30 days from the ED encounter demonstrated equivalent results. Finally, we examined previous naloxone claims for the preceding year although patients may not have been insured over that period.

In summary, patients rarely use insurance to fill naloxone prescriptions following ED encounters for opioid use disorder, even though that encounter signals the importance of having naloxone for themselves, peers, and family members in case of future overdose. Interventions including default prescribing and ED-based OEND programs are needed to improve adoption of this crucial strategy to combat the opioid epidemic.