Pharmacologic Treatment Initiation among Medicare Beneficiaries Hospitalized with Alcohol Use Disorder

Background

In the United States, 29 million adults have alcohol use disorder (AUD), and alcohol contributes to over 140,000 annual deaths.¹ Hospitalizations provide an opportunity to promote behavior change by initiating medications for AUD (MAUD) including naltrexone, acamprosate, and disulfiram, which have been approved by the Food and Drug Administration for decades.² Low prescribing rates for hospitalized patients have been observed in single center studies, but there is a lack of nationwide data on post-hospitalization MAUD initiation.³

Objective

Characterize MAUD initiation following AUD hospitalizations using the 20% national sample of Medicare Part D claims from 2015-2017.

Methods

Our sample included Medicare Parts A, B and D beneficiaries with continuous enrollment 12 months prior and following cohort entry (or until death if within 12 months) and excluded Medicare Advantage beneficiaries. We identified acute care hospitalizations in 2016 with a discharge diagnosis of AUD, defined using either primary or secondary (i.e. non-primary) International Classification of Disease – 10^th Edition (ICD-10) discharge diagnosis codes.⁴ We excluded patients with recent MAUD fills or contraindications to both naltrexone and acamprosate.

Discharge MAUD initiation was defined as pharmacy claims for naltrexone, acamprosate, or disulfiram between the day before discharge and 2 days after. The secondary outcome was MAUD initiation up to 30 days post-discharge. We used mixed effect logistic regressions with clustering of repeat hospitalizations to identify predictors of discharge and 30-day MAUD initiation adjusted for sociodemographics, clinical characteristics, and hospitalization factors.

Findings

The cohort included 28,601 AUD hospitalizations representing 20,401 unique patients (Table). Overall, 206 patients (0.7%) initiated MAUD within 2 days of discharge, and 364 patients (1.3%) initiated MAUD within 30 days (Figure). Among patients with a primary AUD discharge diagnosis, 70 (2.3%) initiated MAUD within 2 days of discharge.

Table:

Characteristics of alcohol-related hospitalizations and predictors of medication initiation

Characteristics	No. (%) (N = 28,601)A	MAUD initiationB aOR (95% CI)
		Within 2 days of discharge	Within 30 days of discharge
Age category, years
18-39	2,183 (7.6)	3.87 (1.34-11.16)	5.53 (2.52-4.75)
40-54	6,915 (24.2)	2.82 (1.06-7.50)	3.90 (1.88-8.10)
55-64	7,398 (25.9)	3.29 (1.25-8.61)	3.59 (1.75-7.38)
65-74	8,605 (30.1)	1.91 (0.75-4.84)	2.89 (1.44-5.78)
75+	3,500 (12.2)	Ref	Ref
Female	8,335 (29.1)	1.49 (1.02-2.18)	1.61 (1.25-2.08)
Race/ethnicityC
Non-Hispanic Black	4,999 (17.5)	Ref	Ref
Non-Hispanic White	20,554 (71.9)	1.78 (0.995-3.18)	1.48 (1.01-2.16)
Hispanic	1,989 (7.0)	1.47 (0.59-3.67)	0.98 (0.51-1.89)
Other	1,059 (3.7)	1.61 (0.53-4.94)	1.49 (0.72-3.08)
RegionD
Northeast	6,806 (23.8)	Ref	Ref
Midwest	6,770 (23.7)	0.88 (0.53-1.46)	0.76 (0.55-1.07)
South	10,381 (36.3)	0.90 (0.57-1.43)	0.71 (0.52-0.96)
West	4,644 (16.2)	0.91 (0.49-1.71)	0.62 (0.40-0.97)
Non-Medicaid dual-eligible	10,690 (37.4)	1.52 (0.96-2.41)	1.39 (1.01-1.89)
Primary AUD discharge diagnosisE	3,072 (10.7)	4.75 (2.91-7.74)	3.46 (2.52-4.75)
Elixhauser readmission index,F	NA	0.98 (0.96-0.99)	0.98 (0.97-0.99)
Psychiatric disorderG	21,000 (73.4)	2.41 (1.12-5.16)	2.11 (1.31-3.41)
Absence of opioid use disorderH	23,522 (82.2)	1.39 (0.88-2.21)	1.33 (0.97-1.82)
Inpatient addiction medicine or psychiatryI
None	21,657 (75.7)	Ref	Ref
Psychiatric hospital	3,416 (11.9)	9.80 (5.54-17.34)	3.87 (2.74-5.47)
Psychiatry/addiction medicine at general hospital	3,528 (12.3)	6.23 (3.68-10.57)	2.90 (2.09-4.02)
Medical hospitalization (vs surgical)J	24,994 (87.4)	5.44 (1.20-24.64)	2.57 (1.27-5.23)
Past year primary care visitK	16,844 (58.9)	1.18 (0.82-1.72)	1.08 (0.84-1.40)
Past year mental health visitL	3,243 (11.3)	0.78 (0.49-1.25)	1.04 (0.76-1.43)
Absence of self-directed dischargeM	27,119 (94.8)	18.48 (2.11-162.13)	2.86 (1.36-6.02)
Remote prior MAUD treatmentN	294 (1.0)	11.41 (4.92-26.48)	13.09 (7.15-23.94)

^ACohort included patients with a primary or secondary discharge diagnosis of AUD who were discharged home and excluded those with MAUD fills in the 3 months prior to hospitalization or contraindications to both naltrexone and acamprosate (liver disease defined by Elixhauser ICD-10 codes and renal failure defined by either Elixhauser ICD-10 codes, inpatient hemodialysis, or qualification for Medicare due to end stage renal disease)

^BDefined as any pharmacy claim for naltrexone, acamprosate, or disulfiram

^CSelf-reported and imputed race and ethnicity were determined using the Research Triangle Institute race code

^DDefined by US census bureau. US territories excluded (n=8)

^EDefined by discharge ICD codes (F10.1, F10.2, and F10.9) excluding “in remission” specifiers (F10.11, F10.21, F10.91)⁴

^FIndex score using discharge co-morbidity variables. Median 24, IQR 15, 36

^GIncludes anxiety, depression, post-traumatic stress disorder, psychosis, or bipolar disorder, as defined by the Chronic Conditions Data Warehouse (CCW)

^HDefined by CCW

^IPsychiatric hospital defined by CMS special unit codes. Psychiatry/addiction medicine at general hospital defined as non-psychiatric hospital with involvement either as primary service or consult defined using inpatient billing by provider specialty codes 26, 27, or 79

^JDefined using diagnosis-related groups

^KDefined using CPT and provider specialty codes for primary care (1, 8, 11, 12, 38, 84) or any advanced practice clinician (50 and 97 which are not associated with a specialty)

^LDefined using CPT and provider specialty codes for psychiatry or addiction medicine (26, 27, 79)

^MDefined by administrative claims for “discharge against medical advice”

^NIncludes MAUD dispensing in the year prior to hospitalization but not during 3 months leading to hospitalization

Figure: Initiation of medications for alcohol use disorder following alcohol-related hospitalizations.

AUD = alcohol use disorder

MAUD = medications for alcohol use disorder

The most predictive demographic factor for discharge MAUD initiation was younger age (age 18-39 versus 75+ aOR 3.87; 95% CI 1.34-11.16) (Table). The strongest hospitalization predictors included absence of self-directed discharge (aOR 18.48; 95% CI 2.11-162.13), psychiatric hospital or psychiatry/addiction medicine inpatient care (aOR 9.80; 95% CI 5.54-17.34 and aOR 6.23; 95% CI 3.68-1-.57, respectively) compared to no addiction medicine or psychiatry, and a primary discharge diagnosis of AUD (aOR 4.75; 95% CI 2.91-7.74). Additional predictors included remote MAUD use, female sex, psychiatric disorders, and lower Elixhauser index.

Predictors were similar for the 30-day MAUD initiation outcome, though the south and west geographic regions were associated with reduced likelihood of initiation compared to the northeast (aOR 0.71; 95% CI 0.52-0.96 and aOR 0.62; 95% CI 0.40-0.97, respectively).

Discussion

In this national sample of eligible non-recently treated Medicare Part D beneficiaries hospitalized for AUD in 2016, MAUD was rarely initiated during hospital discharge or follow-up care and was more likely among patients who were younger and with involvement of psychiatry or addiction medicine.

Hospitalizations allow for engagement with healthcare resources such as clinicians and social workers which may be otherwise difficult to access, and health vulnerability experienced by patients during hospitalizations may provide motivation for behavior change. MAUD initiation should involve a long-term treatment plan, and if this is not feasible during hospitalization then referral for outpatient treatment may be a preferred alternative.² However, the low rate of MAUD initiation within 30 days of discharge indicates that initiation on follow-up rarely occurs. Even patients engaged with longer-term care are infrequently prescribed MAUD.⁵ Hospitalizations should be leveraged for interventions to initiate MAUD as a part of a comprehensive treatment plan.

Generalizability of our results to patients enrolled in other insurance plans may be limited. In addition, ICD-10 codes for AUD are unable to discern severity of AUD or severity of withdrawal. We were unable to identify rates of long-acting injectable naltrexone that was initiated during hospitalization, rates of unfilled MAUD prescriptions, or rates of patients declining MAUD. Lastly, prescribing rates in 2016 may not reflect current practice given growth of addiction medicine consult services in this time period.

Our findings highlight missed opportunities for MAUD initiation among the high-risk group of patients with AUD who are hospitalized. Increased access to psychiatry or addiction medicine presents one potential solution, but interventions to increase prescribing by generalists and non-addiction specialists are needed to bridge the MAUD initiation gap.