• Decrease font size
  • Return font size to normal
  • Increase font size
U.S. Department of Health and Human Services


  • Print
  • Share
  • E-mail

Appendix C

Table of Contents: Managing the Risks From Medical Product Use: Creating a Risk Management Framework

Previous Section : Appendix B

Cooperative Agreements and Collaborative Resources



The FDA is a long-time user of IMS Health products and services. This includes IMS's National Disease and Therapeutic Index (NDTI) and National Prescription Audit (NPA).

FDA primarily uses IMS's NDTI and the NPA for postmarketing surveillance activities. The NPA provides national estimates of prescription drug use in the United States based on data collected from over 20,000 computerized retail, chain, grocery, and mail-order pharmacies nationwide. These data can be used to track prescribing trends over calendar time and for calculation of adverse drug reaction (ADR) reporting rates. The NDTI database provides information on the age and gender distribution of patients using outpatient drug products, as well as information on the duration of treatment course and indication for use. These large population-based databases have application in a variety of areas, including hypothesis testing, signal refinement and/or confirmation of signals emerging from spontaneous case reports. In addition, they may also be used to assemble case-series and retrospective cohort and/or case-control studies to estimate relative risks and identify important risk factors.

Drug exposure is reported as number of outpatient prescriptions. It can be trended over calendar time, and stratified by age, gender, and indication for use. When used in conjunction with supplemental data obtained from population-based claims or record-linked databases, it is possible to estimate the actual number of patients exposed to a drug product. These data are used in association with spontaneous case reports data to understand the context within which ADRs occur. Potential uses of this data could include patterns of drug usage, patterns of concomitant drug usage, and/or trends in drug usage.

Examples of Cooperative Agreement Utility


Routine collaborative use has been made of IMS Health to support regulatory actions by the Center for Drugs and a variety of other users.

  • Used extensively in the review of terfenadine prior to its withdrawal from the market
  • Used routinely to evaluate effect of prescription-to-OTC switches
  • Used to evaluate Diprivan (propofol), a generic drug marketed with a different preservative from the innovator product
  • Used to provide data regarding schedule II and other drugs with abuse potential for use in determining manufacturing quotas
  • Used to provide sales figures to support cost/benefit arguments for PDUFA II. Demographic data has been used to provide support decisions concerning the Medicare formulary
  • Used in providing demographic patterns of use in support of congressionally-mandated initiatives to encourage study in various subpopulations.


The Center for Biologics Collaborative Resources


The Vaccine Safety Datalink (VSD) - This is a large linked database that studies vaccine safety issues. The availability of large healthcare databases with medical interventions linked with outcomes provides the potential for substantial improvements in the sensitivity of postmarketing safety surveillance programs. The Centers for Disease Control and Prevention (CDC) has contracted with four large health maintenance organizations on the West Coast to provide such databases for the investigation of arising vaccine safety issues. FDA staff have served in an advisory capacity for this project and the Agency has contributed funding when available.


VSD has been used to address a variety of concerns, some of which have arisen from the Vaccine Adverse Event Reporting System (VAERS) reports. As an example, an FDA review of adverse events reported in infants following receipt of hepatitis B vaccine revealed an apparent difference between the two different brands of this vaccine with regard to reporting rate (number of reports divided by number of doses distributed). Nothing in the product content or manufacturing processes provided a likely explanation for this difference. Because of all the limitations of data in spontaneous reporting systems like VAERS, it was felt essential to study this issue further before concluding that the difference was real. Data from VSD sites that had used both vaccines were reviewed; these data, which were more reliable than those from VAERS in that they could provide a true event rate in a defined population, did not suggest any increased reactogenicity of the suspect vaccine brand.


Limitations of the VSD include the fact that not all outcome data are automatically captured. For example, deaths occurring at home or outside the HMO system are not generally recorded. Substantial effort is required to prepare an analysis dataset from the raw data provided by the clinics to study a specific question, so rapid response to a pressing issue is usually not feasible. The VSD population is probably not sufficiently representative of the U.S population, which has implications for investigation of issues that may be specific to a geographic region. For example, the VSD does not provide an optimal population for follow-up studies of Lyme disease vaccines. Many of these problems can be diminished in principle, and the CDC has been actively pursuing the additional resources required to enhance the capability of the VSD.


The Center for Drugs Collaborative Resources

Cooperative Agreement Program


The Office of Post-Marketing Drug Risk Assessment (OPDRA) leverages its internal expertise and surveillance data with formalized access to extensive databases and epidemiologists (principal investigators) experienced with the use of each database. The following are the current agreement holders working with FDA under the Cooperative Agreement Program:

1. United Health - Serves more than 40 million individuals through a broad continuum of healthcare products and specialty services. Current and historical medical and pharmaceutical data from 12 health plans are included in the research database, covering about 2.1 million lives. This database comprises more than 13.1 million member years of data since 1990.

2. Harvard Pilgrim Health Care - The Joint Pharmacoepidemiology Program combines populations and resources from three HMOs -- Harvard Pilgrim Health Care, Boston; HealthPartners, Minneapolis; and Fallon Community Health Plan, Worcester, Massachusetts. Each HMO maintains automated record linkage systems, including drug dispensing information, coded diagnoses for ambulatory and inpatient care, and access to medical records. Combined current membership is in excess of 2 million.

3. Vanderbilt University - This database contains data from Tennessee Medicaid, a joint federal-state program that finances medical care for qualifying low income patients. In 1997, Tennessee Medicaid had approximately 1.4 million enrollees, 500,000 children under 15 years of age, 150,000 persons over age 65, 400,000 African-Americans, and 35,000 births annually. Computerized files of the Medicaid program - which include enrollment, pharmacy, hospital, outpatient, and nursing home files - define the population and patient time, provide a measure of drug exposure, facilitate rapid disease ascertainment, and provide some information on confounders.

4. Boston Collaborative Drug Surveillance Program - The General Practice Research Database (GPRD) provides information derived from computerized general practices in the United Kingdom and Group Health Cooperative (GHC) of Puget Sound (recently merged with Kaiser Northwest, creating potential access problems with current data). GPRD provides relevant information on 4,000,000 patients in the United Kingdom. GHC has maintained computer files for 325,000 patients since July 1976, as well as discharge diagnoses in excess of 400,000 from hospitalizations since 1972.

5. Johns Hopkins University - The Johns Hopkins AIDS service is the largest care provider for HIV/AIDs infected persons in Maryland. In 1992 a comprehensive longitudinal, observational database was launched through funding from the Maryland Department of Health and Mental Hygiene and the Agency for Health Care Policy and Research (AHCPR). Currently the database represents over 3,500 patients from January 1990 through the present and is fully linkable with other Johns Hopkins Health Systems.


The Center for Devices Collaborative Resources

  1. IMS America Hospital Supply Index
  2. This database contains estimates of medical surgical products in 1500 product categories purchased by 350 hospitals in the United States. This data can be used to derive comprehensive estimates of product purchases and device sales. However, the data do not allow analysis of user characteristics, and devices purchased but not used are not accounted for.

  2.  Nationwide Inpatient Sample (NIS)

  1. This database is a stratified probability sample that approximates 20 percent of the U.S. community hospitals. Records include all inpatient hospital stays, patient demographic characteristics, diagnosis, procedures, length of stay, and patient status. Information from this database can be used for national estimates and is particularly good for research on implantable devices. These data are not current and cover inpatient hospital stays only.

  3.  Medicare Data

  1. Medicare data are available from the Health Care Financing Administration (HCFA). This hospital insurance information contains data on inpatient hospitalizations, hospice, home health, and skilled nursing care. The supplementary medical insurance data files contain information on physician services, outpatient visits, durable medical equipment, and home health visits. These large databases contain healthcare information on the majority of the elderly population in the United States. Medical procedures and diagnoses are recorded instandard coding conventions. These data are a good source for obtaining denominator and descriptive information for the use of;certain devices, and cost and use data for the elderly population. However, there is no information on younger age groups, and detailed claims data are not available on managed care plan members. 

  4. Harvard Pilgrim Health Care

  1. This is a fully automated medical record system with computerized pharmacy records. This HMO has 300,000 active members with healthcare records maintained since 1990. The enrollment size limits study of rare events, but information on use for some devices is good because brand and model information is available. 

  5.Managed Care Data Sources

  1. This is a large, relatively inexpensive database, with easy access to medical records and access to hospital discharge and diagnoses information. There is incomplete information on confounders and one may not be able to generalize the results.

  6. National Mortality Followback Survey

  1. This includes data from a population-based survey of individuals aged 15 years or older who died in the U.S. during 1993. The sample used for the survey was 22,957 death certificates. The survey sample was selected by age, race, and gender, with some oversampling of certain groups. Included are a wide range of decedent characteristics, such as occupation, income, activities of daily living, associated medical devices, and alcohol consumption.

  7. National Health and Nutrition Examination Survey

  1. This survey of the health of the general population includes results of physical examination and diagnostic tests. The data useful to the Center for Devices is restricted to commonly used devices, but can be used to make national estimates.

Next Section : Appendix D