Medicaid Databases

| Home | | Pharmacovigilance |

Chapter: Pharmacovigilance: Overview of North American Databases

The US Medicaid Program is a health insurance system created in 1965 to provide access to medi-cal care for economically disadvantaged and disabled persons.


The US Medicaid Program is a health insurance system created in 1965 to provide access to medi-cal care for economically disadvantaged and disabled persons (Hennessy et al., 2005). It is supported jointly by federal and state funds, and managed by states with federal oversight. Benefits are available for members of three groups: (1) low-income preg-nant women and families with children; (2) persons with chronic disabilities; and (3) low-income elderly, including those receiving benefits from the federal Medicare (65 years and older) program. In addi-tion to these categories for eligibility, individual states may set up their own programmes for specific groups of persons who do not qualify for federally supported programs. Services provided by the states under the federal Medicaid programme include inpa-tient hospital services, outpatient hospital services and physician services. All states provide outpatient prescription drugs for at least some categories of enrollees, even though this coverage is not federally mandated. Rather than serving as a direct provider of health care services, Medicaid functions as a payer for eligible services provided by participating physi-cians, hospitals and pharmacies. Of the US popu-lation 16%, or 51 million persons, received health care services through Medicaid in 2002, serving as the largest health insurance programme in the United States (Iglehart, 2003). Compared with the overall US population, the Medicaid population has a dispropor-tionate number of children, females and non-whites. Income and disability status are also not representa-tive of the total population. These are the populations that are often under-represented in randomized trials.

The Medicaid programme is administered by the Centers for Medicare and Medicaid Services (CMS), which has established a mechanism for researchers for obtaining data that have been received from the indi-vidual states and have undergone editing and range checks. A lag-time of 4 years currently exists for the availability of the cleaned Medicaid Analytic Extract (MAX) files; crude data from the Medicaid Statis-tical Information System (MSIS) are also available. Support for the process of obtaining files and technical assistance in the use of the data is supplied through a contract with the University of Minnesota’s Research Data Assistance Center (ResDAC), instituted in its School of Public Health. ResDAC’s description of the CMS data and of its services is publicly avail-able through its website: Data can also be obtained through a commercial data vendor, a common source of Medicaid data in the past (Hennessy et al., 2005).

Five types of MAX files are available for CMS Medicaid data, separately by year and by state: personal summary, inpatient, prescription drug, long-term care and other therapy. The personal summary file contains one record per person enrolled in the specific state’s Medicaid programme for any part of the specific year. It includes demographic data, namely date of birth, sex, race and zip code of resi-dence, and identifies the months in which the person was enrolled in the plan. The inpatient file contains information on hospitalizations, including admission and discharge dates, discharge status, up to nine diagnoses, up to six procedures, and payment informa-tion. Drugs used during hospitalization are not avail-able in this file. The prescription drug file contains records for drugs reimbursed for outpatient or nurs-ing home prescriptions. NDC Codes provide informa-tion on the manufacturer and the name, strength and dosage form of the drug. Data elements include date and quantity dispensed, whether the drug was new or a refill, and cost information. The long-term care file contains information on care provided by skilled nursing, intermediate care and independent psychi-atric facilities. Data elements include type of facility, dates of service, diagnosis and discharge status. The other therapy file contains records for physician, labo-ratory, radiology and clinic services. Date, type of service, diagnosis and procedure codes (where appli-cable) are recorded. Although the types of laboratory and radiology testing are recorded, their results are not reported. Medicaid data have been linked to other databases, such as Medicare data (for persons eligible for both programmes), the National Death Index and state vital statistics registries.

The quality of the Medicaid database has been evaluated for six states. Results suggest the need for macro-level descriptive analyses of the parent dataset, with a particular focus on the number of medical and pharmacy claims over time, checking for gaps, assessing the validity of markers for hospitalization and the accuracy of diagnostic and demographic data (Hennessy et al., 2003).

The strengths of the Medicaid databases are their large size, permitting the study of infrequently used drugs and rare outcomes, and the accuracy of the drug data. More than 10 states have over a million Medi-caid recipients each; prescriptions for the top medica-tion dispensed numbered 9.3 million prescriptions (for albuterol) for the total Medicaid programme in 2001. Far down the list, ranked at number 50, were prescrip-tions for trimethoprim/sulfamethoxazole, accounting for 2.4 million prescriptions (Hennessy et al., 2005).

As a claims database (similar to most of the other databases described), information is lacking on vari-ables often needed to control for confounding, such as smoking, environmental exposures, illicit drug use, alcohol use, occupation, family history and use of over-the-counter drugs.

The International Classification of Disease Ninth Revision – Clinical Modification (ICD-9-CM) is the coding scheme for diagnoses. Together with factors such as the level of accuracy of the clinical diagnosis and need for information on poten-tial confounding variables, experience suggests that investigators should obtain medical records in at least a sample of outcomes to confirm the diagnosis and characterize the severity of the disease, in addition to obtaining information on potential confounding variables. Although a mechanism exists through the recently implemented Health Insurance Portability and Accountability Act (HIPAA) for requesting hospi-tal records of specific patients without patient contact, the willingness of hospitals to use this mechanism has yet to be gauged. Studies where primary record confirmation is less important are those which focus on drug-to-drug relationships, or studies which can use drugs or procedures as markers of diagnoses.

Contact Us, Privacy Policy, Terms and Compliant, DMCA Policy and Compliant

TH 2019 - 2022; Developed by Therithal info.