Supplements Economic Considerations for the Treatment of Opioid and Alcohol Dependence
Extended-Release Naltrexone for Alcohol Dependence: Persistence and Healthcare Costs and Utilization
Objective: Evaluate persistence with treatment, healthcare costs, and utilization in stably enrolled Aetna Behavioral Health members receiving extended-release naltrexone (XR-NTX) for alcohol use dependence compared with oral medications and psychosocial therapy only.
Study Design: Historical cohort study.
Methods: Aetna beneficiaries with stable enrollment (at least 6 months before and after index treatment) who initiated pharmacotherapy with XR-NTX (n = 211), disulfiram (n = 1043), oral naltrexone (n = 1408), acamprosate (n = 2479), or psychosocial therapy only (n = 6374) for alcohol use disorders between January 1, 2007, and December 31, 2008, were extracted and deidentified from Aetna’s nationwide claims and utilization database. Survival analysis compared persistence with XR-NTX versus oral pharmacotherapies. Difference-in-differences analysis compared healthcare costs and utilization among patients receiving XR-NTX versus oral pharmacotherapies and psychosocial therapy only. Multivariate analyses controlled for demographics.
Results: Patients taking acamprosate and disulfiram were more likely to discontinue treatment than patients taking naltrexone, and patients given oral naltrexone were more likely to discontinue treatment than those given XR-NTX. Outpatient behavioral health treatment visits increased in all study groups. Nonpharmacy healthcare costs and utilization of inpatient and emergency services decreased in the XR-NTX group relative to other study groups.
Conclusion: Patients receiving XR-NTX persisted with treatment longer than patients receiving oral alcohol use–disorder medications or psychosocial therapy only, and had decreased inpatient and emergency healthcare costs and utilization compared with those receiving other medications.
(Am J Manag Care. 2011;17:S213-S221)
Reasons for limited adoption of alcohol use-disorder (AUD) medications are numerous, including lack of organizational support, incompatibility with treatment model and philosophy, limited provider exposure to information, provider concerns regarding efficacy and side effects, and reimbursement difficulties.2-4 The 2 most commonly cited barriers to adoption of oral NTX in a 2001 survey of addiction physicians were poor adherence and medication cost.4
Poor adherence is a recognized problem among alcohol dependence pharmacotherapies. Side effects, difficulty “feeling” the effect of medication, and lack of understanding of the need for consistent dosing contribute to discontinuation.5 Acamprosate requires dosing 3 times daily and disulfiram produces unpleasant deterrent effects when alcohol is consumed. Oral NTX has a narrow therapeutic window.6 Approximately 15% to 20% of patients continue to fill prescriptions for oral NTX regularly over 6 months.7-9 Kranzler and colleagues found that persistence (a surrogate for adherence) was associated with lower utilization of expensive inpatient healthcare services.9 Subsequent work reported that patients taking oral NTX decreased healthcare spending relative to control patients with and without AUD diagnoses, even when the cost of treatment was included.10
XR-NTX was developed with improved pharmacokinetic properties and a monthly dosing regimen to address the adherence limitations of oral AUD pharmacotherapies.11 Despite these pharmacokinetic advantages, XR-NTX prescribing remains limited due in part to its high cost, the complexity of delivery (must be administered by a qualified healthcare professional), and lack of information about the medication.1,12 Available data suggest that XR-NTX is associated with fewer heavy drinking days and longer time to first drink than placebo over 12-week13 and 24-week14-16 periods. Studies conducted in primary care clinics17 and privately funded substance abuse treatment clinics12 found that 70% to 75% of patients who initiated treatment with XR-NTX returned for a second injection, and adherence was associated with decreased alcohol consumption.17 Patients who received XR-NTX, moreover, incurred lower healthcare costs and decreased utilization in the 6 months following treatment initiation compared with those who took oral pharmacotherapies.18
This study takes a different analytic approach and uses a different patient population from those employed by Mark and colleagues18 to compare treatment adherence and healthcare costs of XR-NTX with oral medications and psychosocial therapy. We hypothesized that (1) treatment adherence would be greater for XR-NTX, given its unique pharmacokinetics, and (2) XR-NTX would be associated with decreased healthcare costs (excluding the cost of the medication itself) and utilization of inpatient and emergency treatment.
For this historical cohort study, the population of interest included all continuously enrolled Aetna Behavioral Health (Aetna) members who began pharmacologic or psychosocial treatment for AUDs between January 1, 2007, and December 31, 2008. Patient-level data on allowed behavioral, physical, and prescription drug claims are stored in an integrated national database. Aetna uses the database to coordinate management of physical healthcare with services for alcohol, drug, and mental health problems among members who receive physical, behavioral health, and pharmacy benefits from Aetna. Concurrent review of claims and utilization data identified patients with AUD diagnoses.
Patients were eligible if they met all inclusion criteria: (1) claims review flagged the patient as treated for an AUD, and (2) a prescription for AUD pharmacotherapy (XR-NTX, oral NTX, acamprosate, or disulfiram) was filled or psychosocial therapy was initiated. Treatment initiation (the index date) for the psychosocial therapy only group was the date of the first claim with a full psychiatric evaluation (Current Procedural Terminology code 90801) and an AUD diagnosis.
There were 4 exclusion criteria: (1) lack of continuous enrollment for 6 months before and after the index date; (2) single claims over $25,000; (3) prescriptions for AUD pharmacotherapies in the 3 months prior to the index date; or (4) prescriptions for multiple AUD pharmacotherapies during the 6 months following the index date. The exclusion criteria were designed to eliminate loss to follow-up (number 1), remove outliers (number 2), and prevent exposure misclassification (numbers 3 and 4). Patients receiving psychosocial therapy were only excluded if they had taken AUD pharmacotherapy at any time in the past. All pharmacotherapy patients who met selection criteria were included, as well as a random sample of psychosocial therapy only patients.
There were 73,292 Aetna beneficiaries with AUD diagnoses between January 1, 2007, and December 31, 2008, and 12,994 (18%) with at least 1 claim for an alcohol dependence medication: 241 given XR-NTX; 3779 given oral NTX; 6059 given acamprosate; and 2915 given disulfiram. A total of 13,968 patients comprised the random sample of those receiving psychosocial therapy only. After exclusion criteria, the final analytic data set contained 211 patients given XR-NTX, 1408 given oral NTX, 2479 given acamprosate, 1043 given disulfiram, and 6374 given psychosocial therapy only. The Oregon Health & Sciences Institutional Review Board determined that the study was a secondary analysis of deidentified data and qualified for exemption.
Primary outcome variables were (1) persistence with medication; (2) healthcare spending; and (3) healthcare utilization. Spending and utilization data were aggregated over 6 months before and after the index date.
Persistence measured the number of consecutive days the patients had alcohol dependence pharmacotherapy in their possession. Patients were considered to be in possession of AUD pharmacotherapy from the date they filled a prescription until the date the prescription should have been exhausted. Nonpersistence (discontinuation) was defined as the first time after the index date that patients went more than 10 consecutive days without medication in their possession. The 10-day cutoff was determined empirically. Patients who refilled their prescriptions within 10 days tended to continue to fill them regularly, whereas patients who waited 10 days or longer to refill tended to discontinue.
Healthcare spending measured total nonpharmacy healthcare costs recorded in the Aetna claims database during the 6-month pre- and post-index periods, including out-of-pocket and health plan expenses. Healthcare utilization encompassed inpatient admissions, days spent in inpatient treatment, outpatient behavioral health visits, and emergency department (ED) visits. Outpatient behavioral health visits included psychosocial therapy (psychiatrist and therapist visits) and outpatient visits to hospitals/inpatient facilities (intensive outpatient treatment, partial hospitalization).
Primary predictors were (1) time relative to initiation of AUD treatment and (2) medication (study) group. There were 5 distinct study groups: XR-NTX, oral NTX, acamprosate, disulfiram, and psychosocial therapy only. The dichotomous time variable tested the 6-month pre-index period versus the 6-month post-index period.
Covariates included age, sex, beneficiary status, plan type, region, pretreatment period mental health and substance abuse diagnoses, and pretreatment period physical health diagnoses. Age was a 4-level categorical variable (<35 years, 35-44 years, 45-54 years, and >55 years). Region was a 6-level categorical variable (West, Southwest, North Central, Southeast, Mid Atlantic, and Northeast).
Comorbidities included physical health, mental health, or substance use disorders diagnosed in the 6-month pretreatment period. Diagnoses were grouped into mental health and substance abuse (MH/SA) categories by International Classification of Diseases, 9th Revision code following Ettner et al.19 The MH/SA groups represented schizophrenia and other non-mood psychosis, bipolar disorder, major depression, anxiety disorders, and drug use disorders. The Charlson Comorbidity Index (CCI) represented physical health comorbidities.20 Due to low prevalence of physical health comorbidities, the CCI score was collapsed into a 3-level categorical variable (0, 1-2, 3 or more).
Survival analysis compared persistence with XR-NTX versus oral pharmacotherapies. Discontinuation of medication was the “failure event,” and it was defined as allowing 10 days to elapse after a prescription was exhausted without refill. Prescriptions filled after the first episode of discontinuation were not included in the analysis. Patients who remained persistent at 180 days were censored at that time. The primary predictor was study group (XR-NTX, oral NTX, acamprosate, disulfiram). Covariates included all demographic variables and pretreatment comorbidity indicators. Kaplan-Meier survival curves plotted persistence over the 180-day follow-up period. A Cox proportional hazards model compared the risk of discontinuation for XR-NTX versus oral pharmacotherapies.
Difference-in-differences analysis estimated the impact of XR-NTX versus oral NTX, acamprosate, disulfiram, and psychosocial therapy only on healthcare costs and utilization. We compared the change in healthcare costs and utilization that each group experienced in the 6-month period before versus after treatment initiation for XR-NTX and the other therapy groups. Primary predictors were (1) a time dummy variable that had the value of 1 in the posttreatment period and 0 in the pretreatment period and (2) 4 study group dummy variables with XR-NTX as the reference group. Interactions between the time and study group dummy variables were the primary estimands of interest and tested the difference (between XR-NTX and comparison groups) in the differences (between pretreatment and posttreatment).
A 2-part model estimated the difference-in-differences for average spending per patient per half year.21 Logistic regression determined the probability of any spending and utilization, and generalized linear modeling determined average spending conditional on use. Negative binomial regression modeled utilization outcomes. We used the method of recycled predictions, based on the estimated regression model, to calculate the average effects of the predictor variables on spending/utilization. Bootstrapping (500 replications) generated 95% confidence intervals. Demographic variables were included as covariates in all regression models. Pretreatment comorbidities were not included in the cost and utilization analyses due to poor overlap between study groups. All analyses were conducted using Stata/IC version 11.0 (StataCorp, College Station, Texas).