Outcomes of Pharmaceutical Outcomes Research: Impact of Studies Funded

Key Impacts and Findings by Project

Pharmaceutical Care: The Patient Role

Principal Investigator: Betty A. Chewning, Ph.D., M.S.
Grant Number: HS07773
Project Period: March 1993-February 1998

Overview

The purpose of this project was to look at the patient's perspective on the process and outcomes of arthritis drug regimen decisions. It considered:

  • How patients' perceptions about quality of life, symptoms and medication benefits affect adherence to prescribed treatments and over-the counter self-care use.
  • The effects of patient medication-taking behaviors on arthritis outcomes, including quality of life, clinical symptoms and health care utilization.
  • How patients' perceptions of quality of life vary with their clinical symptoms.

The project has revealed interesting findings related to the patient's role in decisionmaking and in providing information useful to ongoing treatment planning. The research suggests that the patient's role may be particularly important for diseases such as arthritis where patient-reported symptoms are a primary determinant of regimen (as opposed to blood sampling or blood pressure tests for diseases such as diabetes or hypertension). Analysis of project results are not yet complete and additional information about this finding as well as others will be forthcoming.

Impact on Further Research (Level I)

Information generated by this project has been presented at numerous meetings of national and international organizations, including events sponsored by the American Association of Pharmaceutical Scientists, the American Pharmaceutical Association, the American College of Clinical Pharmacy, the American Association of Colleges of Pharmacy, and the American Public Health Association.

Tools for Patient Management

  • Comparison of Two Quality-of-life Measures for Arthritis: The study found that the two health quality-of-life measures reviewed add important information to that gathered by traditional clinical status measures. Some of this information unique to arthritis would be lost if the shorter, more widely used Medical Outcomes Study Shortform (SF-36) were administered instead of the Arthritis Impact Measurement Scale (AIMS2) subscales, as this sacrifices the chance for comparison across health/disease conditions.70
  • Role of Patient in Medication Management: The study found a high level of prescription and over-the-counter (OTC) medications taken as needed (PRN, according to patient judgement). "Approximately 37 percent of the rheumatoid arthritis patients and 43 percent of the osteoarthritis patients had a regiment that included a prescription or OTC drug to be taken PR.... While health care is conceptualized typically as prescribed by a provider or as self-care, these data help to describe a more complex medication management role of patients. Especially for patients who have symptoms (such as pain) which they can monitor personally, PRN medication judgements become part of their daily care. The PRN category is interesting not only for what it implies about patient arthritis medication management roles, but equally important for what it suggests about physicians' roles in relation to patients, acknowledging that patients are partners in deciding when certain medications are needed."71

    In addition, the study found a high degree of change in medication regimens, including re-calibration of doses, changes in brand, and changes in drug. "Nearly half of the rheumatoid arthritis patients had their regimens change along the four dimensions across one year...more than half of the osteoarthritis patients showed changes in their regimens." This, along with the importance of patient-reported symptoms and patient assessment, suggests the need for paying greater attention to preparing patients as partners in decisionmaking. This also suggests a focus on developing tools and approaches to prepare, facilitate, and cue patients to provide the right information and data. The development of new techniques to facilitate patient documentation of drug side effects and other symptoms can be important to other diseases as well, such as cancer.71,72

  • Influences on Patient Choice of Drugs: The study found that side-effects are one of the most important factors influencing patient choice of drugs, while out-of-pocket costs influence adherence to a drug regimen. This indicates a need for better information about toxicity and side-effects on the one hand (which the study will also look at), and on the other hand, raises issues related to the importance of drug coverage.72

Return to Contents

Use of Record Linkage to Study Outcomes of Drug Therapy

Principal Investigator: Richard Platt, M.D., M.S.
Grant Number: HS07821
Project Period: August 1993-July 1998

Overview

By use of a new microelectronic technology, the project aimed to achieve four objectives with reference to hypertension medication compliance:

  • Developing a standard definition of adherence.
  • Validating other alternative methods of adherence assessment such as patient reports, pharmacy dispensing records or pill counts.
  • Evaluating the relationship between adherence and a target medical condition (blood pressure).
  • Examining the predictors of adherence.

Impact on Further Research (Level I)

This study has led to AHRQ funded research in other related areas.

Tools for Patient Management

  • Monitoring Patient Compliance with Drug Regimens: The study found that there is considerable overlap of monitored adherence across levels of patient-reported adherence. Nonetheless, patient-reported forgetfulness of doses was qualitatively informative and predictive of adherence to dose number and timing. Patient-reported non-adherence when coupled with measures of drug effect or levels may be useful in guiding therapeutic decisions.

    Factors associated with adherence included: younger age, increased number of children in household, lower vitality on the Medical Outcomes Study Short Form (SF-36), tendency to forget, and side effects of medication. To varying degrees these factors may be modifiable or addressable. Adherence is lower for timing than for number of doses; where dose timing is important, electronic monitoring should be considered. The relationship between adherence and outcomes (e.g., blood pressure) needs to factor in minimum thresholds for effect, and long- versus short-acting nature of treatment.73,74

Return to Contents

A Health Status Measure to Evaluate Drug Therapy for PCP

Principal Investigator: Albert W. Wu, M.D., M.P.H.
Grant Number: HS07824
Project Period: September 1993-August 1995

Overview

As the first study in developing and using a health status measure in a clinical trial of an AIDS-related complication, this project was designed to demonstrate the reliability, validity, and usefulness of such a brief health status measure for acute Pneumocystis carinii pneumonia (PCP). Unlike traditional methodology studies in the field, this project added two new dimensions, namely, responsiveness and clinical utility, in the evaluation of health status measurement. By doing so, the project has enhanced the traditional validation approach that was based exclusively on reliability and validity of the instrument and set a new model for future studies in the field.

Impact on Further Research (Level I)

In addition to the application of new dimensions to health status measurement for this AIDS-related complication, this project influenced related publications that provide guidelines to researchers in understanding and selecting among quality-of-life measurement tools for different research purposes.

Treatment Effectiveness

  • Differentiating Among Treatment Regimens Using Quality-of-life Measures: In one study that compared the tolerability and efficacy of three oral regimens for the treatment of patients with AIDS and Pneumocystis carinii pneumonia, it was found that the rates of dose-limiting toxicity, therapeutic failure, and survival did not differ across the three drugs. With respect to quality of life, at day seven the median increase in patient-reported health status scores was greatest for patients assigned to receive clindamycin-primaquine. This difference was particularly marked in comparison with patients receiving trimethoprim-sulfamethoxazole. By day 21, health status scores in all patients had improved further, and differences among the groups were less evident.75
  • Incorporating Quality of Life in Treatment Decisions for HIV/AIDS: This study included a descriptive assessment of different quality of life instruments to serve as a reference for investigators selecting an instrument for use. "Medical Outcomes Study (MOS) data have added importantly to understanding how treatments affect health. In some studies the Quality of Life (QOL) findings were discordant with the results suggested by other outcomes variables (e.g., cytomegalovirus retinitis retreatment trial: combination vs. monotherapy). Application of the measures results in a more comprehensive definition of the efficacy of treatment and a tangible description of the impact on patients' health. MOS measures have increased our understanding of how other factors affect the health of patients with HIV."76

Return to Contents

Cognitive Impairment and Medication Appropriateness

Principal Investigator: Joseph T. Hanlon, Pharm.D.
Grant Number: HS07819
Project Period: March 1993-May 1995

Overview

Built on a series of secondary data analyses using the longitudinal data from a representative sample of community-dwelling elderly, this project evaluated the relationship between cognitive status and medication use. Focused on two classes of commonly used medications in the elderly, the study demonstrated, in a dose-response fashion, that current benzodiazepine use is associated with memory impairment while nonsteroidal antiinflammatory drug (NSAID) use is not. The study also provided nonsteroidal antiinflammatory drug some clues to the pattern of medication use after the elderly develop cognitive impairment. In general, those elderly were less likely to use over-the-counter medications and analgesics than cognitively intact community-dwelling elderly.

Impact on Further Research (Level I)

This project provided information that can inform treatment planning for the elderly. In addition, it addressed issues that can be incorporated into educational programs for clinicians regarding the special needs of the elderly who are cognitively impaired, and those who are African American.

Special Needs of Target Populations

  • Prescription Drug Use by Elderly Blacks: This longitudinal study of prescription and nonprescription drug use among community-residing elderly showed that the use of prescription drugs increases significantly with age. The study identified characteristics that predict change in the use of prescription and nonprescription drugs over a period of 3 years. Blacks were less likely than whites to become prescription drug users or to increase the number of prescription drugs used. This may reflect poorer communication between physicians (typically white) and their black patients, greater reliance of black patients on non-traditional methods (e.g., folk remedies, faith healing), or voluntary noncompliance, or, it may represent under-treatment or inappropriate treatment of disease. While cross-sectional studies consistently indicate that prescription drug use increases with age, the present data suggest that while age is an important predictor of use and increase in use, it is not as important as race or health status.

Public Health and Prevention, Chronic and Persistent Diseases and Conditions

  • Cognitive Impairment in the Elderly: This study sought to determine whether medication use differs by cognitive status among community-dwelling elderly. This is of concern because medication is a common risk factor for cognitive impairment, and cognitive individuals may be over- or under-medicated, with each possibility having potential adverse outcomes. Given these issues and the growing number of drugs under investigation for treatment of dementia, documentation of the prescribing and self-medication patterns of cognitively impaired elderly is needed to guide health policy and future clinical research.

    Cognitively impaired individuals were less likely to be users of prescription or over-the-counter medications than those who were cognitively intact. Those who were cognitively impaired were less likely to take analgesic medications, specifically NSAIDs, than those who were cognitively intact, but more likely to take central nervous system (CNS) medications. High use of CNS medications may reflect appropriate use of psychotropics, although comparative clinical trials have shown greater efficacy with selected alternative medicines as compared to benzodiazopine. Also, higher use of CNS medications may cause or exacerbate cognitive impairment.77

    In a related study of the medication use patterns among demented, cognitively impaired and cognitively intact community-dwelling elderly people, researchers found an increasing level of cognitive dysfunction was associated with decreased use of OTC, cardiovascular and analgesic medications, and the use of fewer prescription medications.78

Return to Contents

Preference Assessment for Pharmaceutical Evaluation

Principal Investigator: Alan Garber, M.D.
Grant Number: HS07818
Project Period: March 1993-August 1996

Overview

The underlying economic principles for the preference assessment, which serves as the foundation for quality-of-life (QOL) measurement, are somewhat complicated. Aimed at increasing a subject's understanding of the preference assessment instrument and detecting inconsistencies in response, this project developed a computer-based multimedia presentation of health states and preference elicitation to provide support for the preference assessment process. The project demonstrated the validity, reliability, and usefulness of this multimedia presentation. In addition, the project also found that when a subject assigns a preference to a health state, that preference appears to be affected by the subject's current health state.

Impact on Further Research (Level I)

Tools for Patient Management

  • Multi-media Presentation to Assess Patient Preference: This study found that a multimedia (MM) presentation of multi-attribute health states helped respondents provide quality-adjustment weights for cost-utility analysis better than paper-based text presentations. "The results suggest that: 1) MM presentation results in better recall and recognition, indicating better transfer of information; 2) MM presentation appears to result in better definition of preferences (a smaller preference interval) and 3) recall and recognition testing of a health state description can identify material in the description that has an unintended impact on the respondents. In order to perform rating tasks on a health state, subjects must first have an accurate understanding of the health state being presented." The study found that multimedia presentation leads to greater recall and recognition memory of the health state than does a text-only presentation.79

    Another article describes the software construction of IMPACT (Interactive Multimedia Preference Assessment Construction Tool). "Validation studies show that preference assessments performed using IMPACT have high test-retest reliability. Future work with IMPACT will focus on additional validation of the preference elicitation procedures implemented in the program. We are developing methods to assess the internal consistency of valuations across different methods and within a given method of preference assessment [as well as] developing interface designs that will allow us to apply these tests in an interactive fashion during computer interviews."80

  • Patient Cost-utility and Preference Analysis: Although cost-utility analyses were designed to provide a common metric to compare competing health care interventions, the method of eliciting utilities and the population surveyed may so influence utility estimates that they make it difficult or impossible to compare the results of different studies. "We have shown that preferences depend on both the method of elicitation and the population surveyed." Further, this work suggests that future cost-utility analyses which attempt to elicit preferences for hypothetical health states from the general population should consider subjective rating of a respondent's own health status in determining representative population groups.81

    A related article described the effort to determine whether different procedures for establishing preferences for health states among various scaling models produced different results, and whether repeated testing led to a convergence of utility values towards a single true estimation: Do the effects of search procedures diminish with repeated testing or do they persist indefinitely? This project sought to ascertain numerical, patient-derived quality weights ("utility values") using:

    • The visual analog scale (VAS) rating quality of life with a given health condition on a linear scale of 1-100.
    • The time tradeoff (TTO), the amount of time in perfect health deemed equivalent to a longer period of time in poor health.
    • The standard gamble (SG), which finds maximum risk of immediate death or some other adverse health condition that the subject would accept to avoid the health condition in question.

    The project found that "the search procedure (using either the TTO or the SG method) can influence utility values as much as does the rating task. The search procedure effects were surprisingly large. While it is well known that different scaling methods yield different results, our results suggest that the exact implementation of the utility-elicitation task is as important as the underlying conceptual valuation framework." There was no difference in utility values between VAS and SG. "The fact that subtle differences in procedures for utility elicitation result in large differences in estimated preferences suggests that estimated utility values, like other preferences, are constructed during the process of elicitation."82

Return to Contents

Pharmaceutical Cost, Use and Outcome Among Insured Elderly

Principal Investigator: Laura B. Gardner, M.D., Ph.D., M.P.H.
Grant Number: HS08217
Project Period: September 1993-August 1996

Overview

The purpose of this study was to describe and analyze pharmaceutical and medical care utilization and costs for a group of older Americans over a period of time during which the patient's share of the cost of a prescription increased and changed several times. Whereas initially in 1988 enrollees were responsible for $5.00 of the charge of each prescription, this was increased to $8.00 for generics and $10.00 for brand name drugs in 1989, and again to $10.00 for generics and $15.00 for brand name in 1993. Then in 1994, the structure of the pharmaceutical benefit was radically modified so that the patient's co-payment became 50 percent of the allowed prescription charge, and capped at $50.00 per prescription. The study database was a group of 19,350 retirees of the Oregon Public Employees Retirement System (PERS) who were demographically representative of the general Medicare population. The database maximized its epidemiological potential by linking with Medicare claims data of those PERS individuals.

Impact on Policies (Level II)

This study had a direct impact on the Oregon Public Employees Retirement System (PERS) in demonstrating that the 50 percent co-pay was both a feasible and meaningful co-payment system. PERS has maintained this co-payment structure, in part because the study showed that it creates sufficient incentive for enrollees to consider generic drugs.

In addition, the study demonstrated the potential value of a database that links claims and clinical information. PERS has expanded the database so that International Classification of Disease (ICD-9) codes for visits can be linked to the drugs that are prescribed. This has allowed PERS to look at specific disease categories (e.g., osteoporosis) and the potential impact of concomitant and unrelated prescriptions.

Impact on Further Research (Level I)

The study is important in the context of policy discussions about a Medicare Prescription Drug benefit because it provides a case description of the relationship between utilization and a co-payment structure for a representative population.

Cost and Economics

  • Pharmaceutical Costs Among Insured Elderly: The study found that none of the prescription co-payment options except the 50 percent prescription co-payment, affected utilization of and expenditures for pharmaceuticals and their generic substitutions. There was a dramatic shift to greater use of generics with the 50 percent co-payment structure, but no perceived negative impact on the quality of care provided—none of the co-payment options produced observable changes on subsequent use and expenditures of other Medicare services among those retirees.83

Research Tools and Translating Research for Clinical Care

  • Linked Databases and Total Health Services Use: This research involved linking multiple data sources to provide a picture that extended beyond drug use to total medical care use. Presentations at AHRQ and the Academy for Health Services Research (AHSR) led to positive acknowledgement of the level of sophistication of this large-scale time-series database that incorporated adjustments for bias.84

Return to Contents

Statistical Methods for Quality-of-Life Outcome Research

Principal Investigator: Marcia Anne Testa, Ph.D.
Grant Number: HS07767
Project Period: March 1993-February 1997

Overview

In this project the investigators:

  • Evaluated and analyzed current and potential measurement and statistical techniques used in quality-of-life-related patient outcomes through a review of the literature.
  • Refined existing analytical and statistical methodology so that it was appropriate for evaluation of pharmacologic therapies.
  • Developed data analysis demonstration projects within major analytical areas by applying selected methods to existing quality-of-life clinical trials databases of hypertension, diabetes, and human immunodeficiency virus.

Impact on Further Research (Level I)

This project illustrated through demonstration projects on key clinical conditions how quality-of-life data can provide substantive additional information that may alter treatment choices of individuals.

Treatment Effectiveness

  • Benefits from Improved Glycemic Control in Patients with Type 2 Diabetes Mellitus: This analysis involved a double-blind, randomized, placebo-controlled parallel trial, which demonstrated that improved glycemic control of type 2 diabetes mellitus (DM) is associated with substantial short-term symptomatic, quality-of-life (QOL) and health economic benefits. These findings are among the first to link improved control of DM with self-reported productivity and ability to function at work. While baseline rates of employment, group productive capacity, health-related absenteeism, bed-days and days of restricted activity were similar for the placebo and active treatment group, by week 15, improved glycemic control for patients in the active therapy group was associated with greater improvement in overall work and disability outcomes compared with patients in the placebo group. Active therapy patients experienced higher retention of employment (97 percent versus 85 percent) and greater retained productive capacity (99 percent versus 87 percent) compared with the placebo group, even beyond what could be attributed simply to the difference in the dropout rate. Changes in absenteeism, bed-days and days of restricted activity also were statistically different between the two groups.85
  • Hypertension Treatment and Quality of Life: This study examined the differences in quality of life produced by two once-daily calcium channel blockers using different delivery systems: nifedipine gastrointestinal therapeutic system (GITS) and amlodipine. Results "suggest compound-specific effects on quality of life that may be due to differences in the delivery system. Nifedipine GITS is short-acting (2-hour half-life) and is delivered continuously over a 24 hour period, while amlodipine has a half-life of 40 hours, which may produce more sustained low-level effects. While a more beneficial profile was observed for nifedipine, amlodipine demonstrated positive effects on cognitive functioning."86
  • Sensitivity of Quality-of-life Measures: "General measures of quality of life may be too crude and insensitive to capture the important gains in health outcomes due to new therapeutic interventions and programs in diabetes. Quality-of-life evaluations for diabetes are at risk of favoring inferior programs with lower costs simply because gains or losses in health outcomes go undetected."

    Specifically, this study demonstrated that current health state levels influence how patients value decrements and improvements in health. The health states rating clearly revealed that losses in function at relatively asymptomatic states have a greater negative utility than do losses at more severe states. It further notes that "our health states analysis demonstrated that people with mild-to-moderate diabetes substantially value smaller health improvements." The study makes the following recommendations: (1) enhance sensitivity of generic instruments rather than using disease-specific QOL instruments; (2) a QOL outcome measure must incorporate the subjective nature of preference by providing a comprehensive evaluation that focuses on patient self-perceptions of symptoms and health.87

  • Quality-of-life Measurement and Zidovudine Treatment: "Recent reports suggesting that the initiation of ziduvudine treatment when patients are asymptomatic provides no survival benefit have called into question the value of early intervention." This study explored whether integrating outcomes related to quality of life with traditional clinical end points may clarify this controversy. The study investigated quality-of-life considerations and found that, "even for the patient who valued the time after a severe adverse event four times more than the time after the progression of disease, the quality-of-life-adjusted time gained with 500 mg of ziduvudine was less than 1 week during a period of 18 months." The study concludes that, "for asymptomatic patients treated with 500 mg of zidovudine, a reduction in the quality of life due to severe side effects of therapy approximately equals the increase in the quality of life associated with a delay in the progression of HIV disease.... These findings suggest that clinicians should attempt to incorporate patients' preferences into treatment recommendations, instead of making such decisions solely on the basis of the CD4+ cell count and published guidelines.... Future research may establish utility values that are based on patients' preferences, so that information on the quality of life can be integrated into traditional efficacy trials. In addition, a direct assessment of quality of life should be included in future clinical trials of treatment for HIV infection."88

Tools for Patient Management

  • Measuring Quality of Life for AIDS-associated Wasting: This study reported findings from four focus groups conducted on patients with AIDS-associated wasting and highlights conceptual and measurement issues important for evaluating the therapeutic impact of pharmacological agents. "Generic measures of quality of life, while covering the appropriate domains, fail to provide the sensitivity and depth required to evaluate the major issues raised by the focus group participants with respect to the specific impact of HIV wasting on their lives.... Treatments for AIDS-associated wasting [i.e., involuntary loss of more than 10 percent of pre-morbid weight] have led to development of new treatments for enhancing weight gains.... However, the functional and quality-of-life impacts of these treatments have not been well studied.... It was clear that simply measuring increases in lean body mass or exercise endurance are inadequate metrics of improved health and quality of life for the evaluation of new therapeutic agents for AIDS-associated wasting.... Health outcomes should therefore focus not only on the biomedical parameters which signify therapeutic efficacy, but on [other] areas [including negative feelings about appearance, shame, embarrassment, social isolation, loss of appetite, loss of libido, grief, fear, loss of self-worth/self-esteem, hopelessness, cognitive dysfunction and difficulty sleeping] as well." Finally, "semi-structured focus groups of patients with specific conditions are a useful method for translating the best interests of the patient into therapeutic research and goals."89

Research Tools and Translating Research for Clinical Care

  • Comparison of Randomized Clinical Trials and Primary Care Settings: The study examines whether discontinuation rates for anti-hyperlipidemic drugs in two health maintenance organizations (HMOs) differs from the discontinuation rates reported in clinical trials published from 1975 through 1993. "We evaluated the risks of discontinuation in patients enrolled in two HMOs and compared them with discontinuation rates for the same drugs reported in long-term clinical trials. "The risk of discontinuing anti-hyperlipidemic drugs in the two HMOs 1 year after the start of therapy ranged from 15 percent for lovastatin to 46 percent for niacin. For the bile acid sequestrants, gemfibrozil, and niacin, the risks observed in the HMOs were substantially higher than the summary estimates of discontinuation rates reported in randomized clinical trials." The study reports that the discontinuation rates in randomized clinical trials may not reflect the rates actually observed in primary care settings. The study also found that the principal reasons for discontinuation of drug therapy were adverse effects (18 percent), therapeutic ineffectiveness (10 percent), and noncompliance (2 percent).90
  • Patient vs. Physician Report: This parallel randomized, double blind multi-center study found that, of the symptoms to which the typical patient confessed, only about one-quarter were also reported by the physician. By one measure, fewer than 7 percent on average also appeared in the physician's adverse event report. On average and across different prompted report processes, patients responded "yes" to adverse symptoms 50 percentage points more often than did their physicians. Differences are attributed to: "objectivity," "seriousness" of symptoms (such that physicians are more likely to report more serious symptoms than less serious ones), "clinical relevance," and "sensitivity" (reflecting the tendency to underreport embarrassing symptoms).91

Return to Contents

Community-based Pharmaceutical Care: A Controlled Trial

Principal Investigator: Morris Weinberger, Ph.D.
Grant Number: HS09083
Project Period: June 1996-May 2000

Overview

The purpose of this study is to develop algorithms to facilitate pharmaceutical care for the treatment of asthma. A randomized controlled trial, with 460 patients in a pharmacy intervention group and two control groups, will be undertaken. The pharmacy intervention group will be provided with patient-specific clinical information displayed on their computer workstations when filling prescriptions. This study will take advantage of Indiana University's long-standing project to develop an electronic medical record. Patient records will be available to the pharmacist from six hospitals and 234 free-standing clinics. Pharmacy records will also be made available to these linked providers.

Findings and publications related to this study are forthcoming.

Return to Contents

Impact of Prospective Drug Use Review on Health Outcomes

Principal Investigator: Frank M. Ahern, Ph.D.
Grant Number: HS09075
Project Period: August 1996-July 2000

Overview

The purpose of this study is to compare two different models of outpatient prospective drug utilization review (ProDUR) programs in the state of Pennsylvania, one of which allows active participation by a pharmacist in the process while the other does not. The study aims to:

  • Conduct a descriptive epidemiological analysis of psychotropic drug use, prescribing patterns, and yield of ProDUR interventions, and compute estimates of drug-related outcome measures.
  • To evaluate the independent effects of two different ProDUR procedures on health outcomes.

The investigators will use data from the Program of All Inclusive Care for the Elderly (PACE) and Medicaid. Outcomes to be studied include hospitalization and utilization of other health-related resources.

Findings and publications related to this study are forthcoming.

Return to Contents
Proceed to Next Section

Current as of October 2001
Internet Citation: Outcomes of Pharmaceutical Outcomes Research: Impact of Studies Funded. October 2001. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/research/findings/final-reports/pharmaceutical-outcomes-research/findings3.html