Appendix F

Medicaid Analytic eXtract Methods

Design and Method

Study population

The study population comprised all Medicaid-eligible youth aged 3 to 17 in a selected subset of states for the years 2001-2010. Enrollee age, for the purposes of inclusion in analyses, was computed as of July 1 of each year. Eleven months (not necessarily consecutive) of Medicaid eligibility in the year were required for inclusion in the analytic population. The twenty states selected for inclusion in the study were those that we and others found, in internal analyses as well as in published work, to provide relatively complete diagnosis and treatment detail. The overall study population meeting the age, eligibility, and state inclusion criteria ranged from 5.2m in 2001 to 8.2m in 2010.

Data sources

The primary data source for this study consisted of Medicaid enrollment, claims, and prescription drug fill data from the Medicaid Analytic Extract (MAX). The MAX data provide a set of research files constructed from regular mandated data submissions from the state Medicaid programs and compiled and processed by the Centers for Medicare and Medicaid Services (CMS). Enrollment data include information on beneficiary characteristics such as age, sex, and race/ethnicity; as well as Medicaid eligibility; managed care participation; and other details that affect how care is organized and paid for. Claims are provided separately for inpatient, long-term care, and other (primarily outpatient) services, and include details on diagnoses received and services delivered. A prescription drug file provides records for each filled prescription, allowing researchers to identify which medications were received, when the prescriptions were filled, how much of the drug was provided, and for how long. The prescription drug records were linked to the First Data Bank National Drug Data File (NDDF), which provides the means to look up drug details using the National Drug Codes (NDCs) provided in the MAX data. Prescription medications were extracted from the MAX data by their generic names.

Database construction and analysis

Our preliminary MAX data analyses included 44 states and the District of Columbia. We then identified a subset of 20 states that in 2009 either (a) had predominantly fee-for-service (FFS) youth Medicaid populations or (b) had been identified as having relatively complete and usable

managed care encounter data for that population.¹ Key variables included in these data sets were: sociodemographic characteristics (age, sex, race/ethnicity), month-by-month eligibility data, diagnoses recorded, services received, and prescriptions filled. An eligibility threshold, requiring a minimum of 11 Medicaid-eligible months in the year, was also established following preliminary analyses of the MAX enrollment data. Only data for youth age three to under 18 were included in analyses. The few youth with dual Medicaid-Medicare eligibility were excluded from analyses because records from Medicare, which is the first payer for many services, were not observable.

Enrollees were assigned to one of several basis-of-eligibility (BoE) groups following examination of their eligibility records: SSI/Disability, Foster care, and Other. Analyses were stratified or subset by this grouping variable. We used each enrollee’s last observed BoE category in each year to assign her/him an overall status for the year. This was necessary in order to establish mutually exclusive eligibility categories. Preliminary analyses revealed substantial consistency throughout the year for the focal SSI/Disability group, suggesting that our method of assigning BoE categories on the basis of the last observation of the year did not distort these enrollees’ eligibility histories.² The mappings from MAX BoE categories to our three groupings are details in Appendix A. Where the tables, charts, or text below refer to the full population or to subpopulations (e.g., SSI/Disability), it is implied that enrollees also met the age, eligibility, and state inclusion criteria.

Presence of ADHD and other diagnoses was identified using the multiple diagnosis fields in the MAX claims files, excluding the long-term care claims. A threshold of one or more inpatient claims, or two or more outpatient or other non-inpatient claims on different dates, was used to establish the presence of the following conditions for each enrollee for each year of analysis (via ICD-9-CM codes):

1. ADHD (both alone and complicated by other conditions)

_________________

¹ See https://www.cms.gov/Research‐Statistics‐Data‐and‐Systems/Computer‐Data‐and‐Systems/MedicaidDataSourcesGenInfo/Downloads/MAX_IB_15_AssessingUsability.pdf and http://mathematicampr.com/publications/pdfs/health/MAX_IB14.pdf

² We found that 96% of enrollees assigned to the SSI/Disability group on the basis of their last observation of the year had no other BoE at any point earlier in the year.

2. Conduct disorder
3. Emotional disturbances
4. Oppositional defiant disorder
5. Mood disorders (Depression)
6. Mood disorders (Bipolar disorders)
7. Anxiety disorders
8. Autism spectrum disorders
9. Intellectual disorders
10. Speech and language disorders
11. Hearing disorders (control)
12. Learning disorders (control)

Two subcategories of ADHD complicated by (1) any of the non-control conditions and (2) an externalizing disorder (conduct disorder or oppositional defiant disorder) were also identified. Cerebral palsy and asthma were identified as control conditions. The ICD-9-CM codes used to establish diagnoses are listed in Appendix B.

Pharmacological treatments were identified via generic drug names associated with filled prescriptions in the linked MAX/NDDF data and reported using the following categories of medications: ADHD medications, antipsychotic medications, antidepressants, anxiolytic/hypnotic medications, and mood stabilizers. The generic drug names for each category are listed in Appendix C. Non-pharmacological treatments were identified via procedure codes recorded in the MAX claims (using Current Procedural Terminology [CPT] codes in Appendix D).

Once enrollees’ eligibility, sociodemographic characteristics, diagnoses, prescriptions, and services were identified, analyses were performed to track diagnosis and treatment trends over the 2001-2010 period.

Project conducted by:

Center for Health Services Research on Pharmacotherapy, Chronic Disease Management, and

Outcomes, and Center for Education and Research on Mental Health Therapeutics

Institute for Health, Health Care Policy, and Aging Research

Rutgers University

Part A: Study categorizations (left) of MAX BoE groupings (right)

Part B: Diagnostic groupings (ICD-9-CM)

Part C: Medication groupings, using generic names

ADHD medications

amphetamine, dextroamphetamine, dexmethylphenidate, lisdexamfetamine, methamphetamine, methylphenidate, modafinil, atomoxetine, clonidine, guanfacine, bupropion, imipramine, nortriptyline, desipramine

Antipsychotic medications

aripiprazole, asenapine, benperidol, chlorpromazine, clopenthixol, clotiapine, clozapine, droperidol, flupenthixol, fluphenazine, fluspirilene, haloperidol, iloperidone, levomepromazine, loxapine, lurasidone, mesoridazine, molindone, olanzapine, paliperidone, pericyazine, perphenazine, perphenazine, pimozide, pipotiazine, promazine, quetiapine, risperidone, thioridazine, thiothixene, trifluoperazine, triflupromazine, zuclopenthixol, and combined agents amitriptyline/perphenazine and olanzapine/fluoxetine

Antidepressants

amitriptyline, bupropion, citalopram, clomipramine, desipramine, desvenlafaxine, doxepin, duloxetine, escitalopram, fluoxetine, fluvoxamine, imipramine, isocarboxazid, levomilnacipram, milnacipram, mirtazapine, nortriptyline, paroxetine, phenelzine, selegiline, sertraline, tranylcypromine, trazadone, trimipramine, venlafaxine, and combined agents amitriptyline/perphenazine and chlordiazepoxide/amitriptyline

Anxiolytic/hypnotic medications

alprazolam, atenolol, bromazepam, buspirone, chlordiazepoxide, clonazepam, clorazepate, diazepam, flurazepam, lorazepam, oxazepam, pregabalin, propranolol, temazepam, triazolam, and combined agent chlordiazepoxide/amitriptyline

Mood stabilizers

carbamazepine, gabapentin, lamotrigine, lithium, oxcarbazepine, topiramate, valproic acid/valproate/divalproex, zonisamide

Part D: Non-pharmacological treatments (Current Procedural Terminology, CPT)*

Project conducted by:

Center for Health Services Research on Pharmacotherapy, Chronic Disease Management, and

Outcomes, and Center for Education and Research on Mental Health Therapeutics

Institute for Health, Health Care Policy, and Aging Research

Rutgers University