A Look at Substance Use Disorders (SUD) Among Medicaid Enrollees

Deaths due to substance use disorder (SUD) have risen sharply during the pandemic, highlighting longstanding gaps—such as under-identification and undertreatment of SUD. Even before the pandemic, SUD contributed to a large and growing share of deaths. For example, alcohol was listed as a contributing factor in 20% of deaths in adults between the ages of 20 and 49 from 2015 to 2019 and drug overdose deaths increased by 35% over the same period. Yet in 2019, only 1 in 10 people (12 and older) with past year SUD received any treatment, including specialty treatment or self-help groups. The Medicaid population may be particularly impacted, as 21% have mild, moderate, or severe SUD, compared to 16% of commercially insured. In its role as a public program and the single largest payer of behavioral health services in the country, Medicaid is particularly positioned to implement policy to improve the delivery, quality, and effectiveness of behavioral health services. The detailed and comprehensive claims data available for Medicaid can help answer questions and inform policy.

Efforts are being made at the state and federal levels to increase SUD awareness, coordination, and treatment access—including some provisions that were recently passed as a part of the Consolidated Appropriations Act enacted in December 2022, such as expanding providers who are able to prescribe buprenorphine for treating OUD. In light of recent efforts to expand access to SUD treatment services, we examine the share of enrollees with SUD using both Medicaid claims and data from the National Survey on Drug Use and Health (NSDUH).

What are the rates and characteristics of Medicaid enrollees with identified SUD?

In 2019, 7.3% of Medicaid enrollees ages 12 to 64 had at least one clinically-identified SUD in Medicaid claims data. Opioid use disorder was identified in 3.3% of Medicaid enrollees; alcohol use disorder in 2.5%; cannabis use disorder in 1.9%; stimulant use disorder (including cocaine or other stimulants) in 1.7%; and 1.7% of Medicaid enrollees had some other type of substance use disorder. These groups were not mutually exclusive, and while we did not look at the share of people with multiple substance use disorders, it is not uncommon for substance use disorders to co-occur. For this analysis, any diagnosis or prescription code that suggests the presence of a SUD is flagged as a “clinically-identified SUD.” This is not a measure of overall prevalence of SUD because not everyone is screened and diagnoses are not always recorded, but it does provide some insight into how often SUD is recognized and possibly treated in clinical settings (Figure 1).

People with clinically-identified SUD are more likely to be male, White, over 25 years old, and qualify for Medicaid based on a disability or through Medicaid expansion. At least one clinically-identified substance use disorder is found in Medicaid claims data for 8.9% of males, 10.0% of White people, and 11.6% of individuals aged 35 to 49. Medicaid beneficiaries who qualify as a result of a disability or through Medicaid expansion have higher rates of clinically-identified SUD than other groups. This pattern holds across most types of SUD, with only a few exceptions. For example, alcohol use disorder is most commonly identified among people aged 50 to 64, while cannabis use disorder is the most commonly identified among people aged 26 to 34. Clinically-identified SUD rates are highest among White people for all substance types except cannabis, where Black people have similar or slightly higher rates (2.4% versus 2.1%, respectively) (Figure 2).

Rates of clinically-identified SUD vary widely by state. Vermont has the highest share of any clinically-identified SUD, with 13.3% of Medicaid enrollees having a clinically-identified SUD, while Arkansas has the lowest rate, with only 3% of Medicaid enrollees having at least one clinically-identified SUD (Figure 3). Rates of clinically-identified SUD vary across states not only because of prevalence, but also because of other factors, such as provider screening behavior and variation in Medicaid coverage of SUD services.

What are implications of findings from claims data?

Other data sources generally suggest that SUD rates from Medicaid claims are undercounts—but even national survey data may undercount SUD for a variety of reasons. National prevalence of SUD is estimated through surveys such as the National Survey of Substance Use and Health (NSDUH), which uses questions based on diagnostic criteria to identify individuals with SUD, including those who haven’t already been diagnosed. We analyzed NSDUH data and found that the prevalence estimates of SUD among Medicaid enrollees is generally higher in national survey data than in Medicaid claims data, and that difference is greatest among adolescents, young adults, and Hispanic people. Even national survey data may undercount SUD due to underreporting of substance use and exclusion of unhoused, incarcerated, and institutionalized people—which are populations where SUD may be more prevalent. Research that adjusts for these undercounts estimates that alcohol and opioid use disorders are at least four times more prevalent than the NSDUH estimates.

National recommendations instruct providers to screen for substance use and conduct brief interventions for adults 18+, yet there may be gaps between SUD screening and referral. Research found that while most patients with an alcohol use disorder were screened for alcohol use, only 14.6% received brief interventions from their providers, and even fewer–about 6%–were referred to treatment.  Despite the serious health consequences and comorbidity of physical health and SUD conditions, most doctors do not receive much training in substance use disorders. Even within psychiatric residency programs, only 2% of training time is dedicated to substance use disorders. According to fourth-year medical students in Massachusetts, fewer than one-fifth report feeling very prepared to screen for opioid use disorders and/or refer patients with related symptoms to treatment.

Other factors–such as patient privacy concerns or few healthcare visits–may also play a role in low identification of SUD. Even when providers ask about substance use, patients may feel uncomfortable disclosing their use or may be worried about stigma or legal consequences if they do. Even if a SUD is identified, providers may be hesitant to record it due to concerns about whether recording the SUD violates the privacy rules that add additional protections for people receiving SUD treatment. Other reasons may be population specific. For example, people who are younger and generally healthier may have fewer health care appointments and therefore fewer opportunities for providers to identify SUD. In at least one state, school-based screenings are required for younger populations. Growing drug overdose deaths among adolescents and people of color may suggest disparities in the identification and treatment of SUD.

There is broad variation in Medicaid policy and coverage of SUD services across states. Medicaid coverage of SUD services, as well as utilization management policies, such as prior authorization, can vary widely across states (and even across managed care organizations within states). Although more comprehensive coverage of SUD services has been linked to higher Medicaid acceptance by SUD treatment facilities, as of 2018, only 12 states covered the full continuum of SUD services. People experiencing symptoms of a substance use disorder may find it difficult to navigate this complex landscape, and difficulty accessing treatment is likely exacerbated in areas with workforce shortages.

Looking Ahead

In response to the growing number of overdose deaths and longstanding challenges accessing SUD treatment, state and federal governments have taken action to address ongoing gaps in SUD care—from identification of SUD to treatment. For example, many Medicaid programs have expanded coverage of SUD services and extended benefits to new eligibility groups; increased provider reimbursement rates for SUD services; and/or permanently adopted or continued pandemic-era telehealth expansions for behavioral health services.

At a federal level, HHS has issued notices of proposed rulemaking that may result in improved coordination of SUD services (42 CFR part 2) and expanded access to methadone for opioid use disorder treatment. Congress passed the Consolidated Appropriations Act (CAA) in December 2022, with funding to improve SUD awareness, prevention, treatment, coverage, and increase workforce. For example, the CAA added at least 100 new residency positions dedicated to psychiatry and required that all prescribers of controlled substances undergo training in managing and treating patients with SUD. CAA also lifted some administrative barriers to expand access to medications to treat opioid use disorder including the removal of additional registration requirements for prescribing buprenorphine (X-waiver) and reduced barriers to opioid treatment programs. Recent legislative efforts may lessen some longstanding barriers to SUD care, which could lead to better identification, referral, and treatment of SUD.

Methods

Medicaid Claims (T-MSIS) and State Exclusion Criteria

This analysis uses the following 2019 T-MSIS Research Identifiable claims files: demographic eligibility base (DE) and header and line files from inpatient (IP), long-term care (LT), other services (OT), and prescription (RX) claim files.

We use 48 states and D.C. in the main analysis and 29 states in our analyses that include race and ethnicity. We evaluated states’ claims data using the DQ Atlas criteria and by comparing SUD estimates from T-MSIS to NSDUH. Specific DQ Atlas measures used to determine state data quality include the restricted benefits code, Medicare benefits code, OT claims/encounter volume. We excluded Alabama because the Medicare coverage code was missing for more than 10% of enrollees and the T-MSIS SUD rates were 93 percent lower than NSDUH estimates. We excluded Colorado because the OT file encounter data volume was below 50% of the national median and the T-MSIS SUD rates were 156 percent lower than NSDUH estimates. For analyses involving race/ethnicity, the following states rated as “high concern” or “unusable” data by the DQ Atlas were also excluded:  AL, AZ, AR, CO, CT, DC, HI, IA, KS, LA, MD, MA, MO, MT, NY, OR, RI, SC, TN, UT, WV, WY.

T-MSIS Enrollee Sample Selection

Our sample includes nonelderly Medicaid and CHIP enrollees between the ages of 12 and 64 that have at least one day of enrollment in 2019. Enrollees were excluded if they did not have full or comprehensive Medicaid, had Medicare coverage, or were enrolled for less than one month in 2019. These exclusions are similar to those reported in the CMS SUD data book technical specifications, but the CMS data book included people with Medicare coverage. After enrollee and data quality exclusions, our main sample includes 46,967,389 enrollees from 48 states and the District of Columbia.

Identification of SUD in T-MSIS

We linked header and line files using MSIS_ID and CLM_ID and linked claims files to the DE file using MSIS_ID (see the T-MSIS User Guide for information on linking variables) for fee-for-service and encounter claims. We identified ICD-9 and ICD-10 diagnosis codes and National Drug Codes (NDCs) from an adapted version of reference codes used in the 2019 CMS SUD data book. Modifications to CMS SUD reference codes include (1) exclusion of tobacco in our definition of “any SUD”; (2) removal of NDC codes primarily used to treat pain rather than opioid use disorder; and (3) removal of methadone NDC codes as these codes are thought to represent pain treatment, rather than opioid use disorder. One or more occurrence of a SUD diagnosis code or NDC code in OT, IP, LT, or RX files is coded as an SUD. Following CMS data book technical specifications, naltrexone—used to treat alcohol use disorder and opioid use disorder—is coded as alcohol use disorder when an alcohol use disorder diagnosis code is present, but coded as opioid use disorder in all other instances. “Any Substance Use Disorder” includes enrollees with at least one opioid, alcohol, cannabis, stimulant, or other SUD. “Other Substance Use Disorder” includes diagnosis codes for sedative, hallucinogen, caffeine, inhalant, unknown, or other SUD.

National Survey on Drug Use and Health. The National Survey of Drug Use and Health (NSDUH) asks respondents 12 and older about substance use and symptoms of substance use disorders, and those who exceed certain thresholds are classified as having a SUD. This analysis uses 2018/19 NSDUH data and includes Medicaid enrollees between the ages of 12 and 64. Although the NSDUH collects nationally representative data and asks each respondent about their substance use and symptoms, it may still underestimate SUD prevalence. NSDUH only collects data from people with an address–excluding those who are unhoused, institutionalized, or incarcerated – which is relevant because these populations may have higher rates of substance use disorders.