• Voluntary physician reporting of adverse drug events (ADES) in their patients remains the single most important source of Information on serious and rare ADEs. Yet, substantial underreporting exists and the factors producing it are unclear. We surveyed 3000 randomly chosen physicians by malled questionnaire to determine their knowledge about the reporting system, attitudes toward reporting, and their past behavior in reporting the ADEs they had detected. Responses numbered 1121. Only 57% were aware of the Food and Drug Administration's reporting system. While 418 physicians had detected an ADE in their practices during the previous year, only 21 physicians reported these events directly to the Food and Drug Administration. The physicians appear to appreciate the safety issues involved in prescription drug use and view reporting as a professional obligation; however, the current reporting system is considered Inconvenient.

(Arch Intern Med 1988;148:1596-1600)

The voluntary or spontaneous reporting system (SRS)

for detecting adverse drug events (ADEs) in the United States has been operated by the Food and Drug Administration (FDA) for three decades. The function of the SRS is to collect, sort, and examine all reports of clinically serious events associated with drug use, evaluate the probability of causality, characterize the events, search for

For editorial comment see p 1499.

trends, and identify those safety and toxicity questions that require research and other confirmatory efforts or immediate regulatory action.

These efforts are necessary because at the time a drug is marketed, data examining its safety are limited. Since premarketing trials seldom enroll more than 2000 patients or extend beyond three years, uncommon or delayed effects

Accepted for publication Nov 18, 1987.

From The Johns Hopkins University School of Medicine, Baltimore (Drs Rogers, Smith, and Levine); State of Maryland Department of Health and Mental Hygiene, Baltimore (Dr Israel); The Johns Hopkins University School of Hygiene and Public Health, Baltimore (Dr McBean); The Center for Health Education, Baltimore (Dr Valente); and the Department of Epidemiology and Biostatistics, Food and Drug Administration, Rockville, Md (Dr Faich).

Reprints not available.

1596 Arch Intern Med-Vol 148, July 1988

may not be detected before a drug is released.' Certain effects only emerge when the drug is used in large numbers of people, or special populations, or over long periods of time for the treatment of chronic conditions.

The utility of the SRS is in its ability to generate early warning signals of drug risk. Its capacity to do this is unmatched because it is central, independent of economic concerns, and operates within an agency with regulatory authority. The SRS alone cannot, except on the rare occasion when the signal is strong and unique, provide a totally scientific basis for regulatory action. The SRS data have major limitations. They consist of numerator events only; ie, there is no information on the population of equally exposed individuals from which the adverse events emerged. Therefore, incidence rates and estimates of risk cannot be directly calculated. The inability to derive such estimates from the system is a serious limitation and, when not appreciated, can subject the data to misuse.

However, when these data are used correctly, they provide a monitoring system that may be uniquely efficient in accomplishing its objective. Studies by Rossi and Knapp' suggest that the SRS may be the only affordable mechanism available for the early detection of serious clinical events that occur less frequently than once per 10 000 drug exposures. Furthermore, they demonstrate that the physician-reporting component of the SRS appears to be the most efficient one in defining serious clinical risk.'

The SRS is entirely voluntary. Only professional responsibility obligates a physician to report adverse patient experiences with drug therapy. The rate of ADE reporting per 1000 physicians in the United States in 1982 was 57.6; this is one third to one half the rate in Canada, New Zealand, the United Kingdom, Sweden, and Denmark. While reporting has improved in recent years, there is still a high rate of underreporting. Many hypotheses to explain underreporting have been offered. Inman has attributed underreporting to complacency about drug safety, fear of involvement in litigation, guilt due to patient harm, ambition to publish, ignorance of reporting requirements, diffidence about reporting mere suspicions, and lethargy.

Conversely, factors that do influence reporting of the ADE by physicians are the severity of the reaction, the

Adverse Drug Events-Rogers et a

professional

obligation to report adverse

(1.0)

3.3 (0.9)

1.5

(0.8)

drug events

certainty of the attribution of the reaction to the drug, the mechanism of action of the reaction (idiosyncratic or allergic events are more likely to be reported than pharmacologic ones), and previous publicity about the reaction. Milstien et al surveyed physician reporters to the SRS to determine which factors induced reporting. The most important factor was the severity of the reaction (74%). However, little is known about factors that influence community physicians to respond or not.

The objective of this study was the assessment of Arch Intern Med-Vol 148, July 1988

*Group 1 indicates Baltimore City, Baltimore County +Group 2 indicates Prince Georges and Montgomery counties.

community physician knowledge, attitudes, and reporting behavior relating to the SRS of the FDA.

METHODS

A self-administered questionnaire was mailed to 1500 randomly selected, licensed physicians of the 5139 in Baltimore City and Baltimore County and to 1500 randomly selected, licensed physicians of the 3104 in Prince Georges and Montgomery counties in Maryland surrounding Washington, DC. These counties were selected because they are the most dense in physician and general populations. Undeliverable questionnaires or those returned from self-declared ineligible physicians (due to retirement or no direct patient care) were replaced by other randomly selected physicians from the same area (N=161). Physicians were guaranteed confidentiality and all forms were coded to facilitate second mailings. No follow-up beyond the second mailing was attempted.

Questionnaires sought information in four areas: (1) demographic information about physician responders; (2) knowledge about the FDA-SRS; (3) attitudes toward the FDA-SRS and ADE reporting in general; and (4) behavior relating to the use of FDASRS during the previous calendar year. To assess the representativeness of our sample, we compared study responders with Maryland State Health Department statistics for all physicians licensed in the respective areas for the years 1981 through 1982 and with the statistics for half the physicians licensed in the respective areas for the year 1985. These are the only available data and are generated as part of the biennial licensing process. Comparison factors are limited to sex, medical specialty, and age (1985 only). The one-half sample in 1985 is not a scientific random sample.

Severe ADEs were defined on the questionnaire as those requiring hospitalization or contributing to death of a patient during the previous year, moderate events required an emergency room or office visit.

In the study analysis, we compared responder subsets on demographic factors and knowledge, attitudes, and behavior relating to the FDA-SRS. The attitudes examined were presented as a series of statements (listed in Table 1) to which physicians were asked to respond as follows: strongly agree, slightly agree, slightly disagree, or strongly disagree.

Logistic regression analysis was employed to model the association between selected factors and reporting behavior. xa Tests of association and tests for independent proportions" were employed in the stratified analyses.

Adverse Drug Events-Rogers et al 1597

A total of 1121 questionnaires was returned (response rate, 37%). Of these, 719 were returned in response to the first mailing (24%); the remainder was subsequent to the second mailing. The demographic profile of the study respondents is similar to that derived from available state statistics for the areas involved (Table 2).

Table 1 presents the attitudes of the physician respondents. They supported the importance of the single case, the ability to attribute the reaction to the drug, and inadequate safety information at marketing, but they did so with decreasing certainty. Professionally, they rejected financial reimbursement for a behavior that they clearly considered a professional obligation. Respondents did not perceive reporting as increasing their own liability. There was no great concern about being involved with additional government forms. Respondents tended to disagree that reporting takes too long, but, in general, supported the desirability of an easier reporting method.

Fifty-seven percent of the physician respondents were aware of the FDA-SRS. Of the 1121 respondents, 37% had detected a serious or moderate ADE or both among their patients during the previous year (Figure). There were 144 reports of serious reactions (events leading to hospitalization or death). Only 5% (21/418) of all detected ADEs were reported directly to the FDA-SRS.

Reporter-Nonreporter Comparison

Of the 1121 respondents, 418 had detected a moderateto-severe ADE in their practices during the previous year. Eighteen percent (76/418) reported the ADE; 28 percent (21/76) of these were reports directly to the FDA.

Demographic Factors-Reporters (N=76) differed from nonreporters (N = 342) in that they saw fewer patients per day, wrote fewer prescriptions per day, and had a larger proportion of inpatient practice. All these differences are statistically significant (Table 3). There are no differences in age or time in direct patient care. Reporters tended to be from internal medicine and family/general practice;

1598 Arch Intern Med-Vol 148, July 1988

Majority of patients Inpatients Outpatients

Time in direct

patient care, %

0-25

26-50

51-75

76-100

Medical specialty

Internal medicine
Pediatrics
Surgery

&8

20

80

Family/general practice

12

Obstetrics/gynecology

0

Anesthesiology

Psychiatry

however, the numbers became small and significa ing was not performed.

Attitudes-Reporters did not differ from now. in any attitude score when mean values were However, when responses were categorized, significant differences emerged for four of th tudes (Table 4). No reporter disagreed with sional obligation to report ADEs. Nonrepor disproportionately that reporting increases 1 that they would be more likely to report if an ea were available. Reporters more strongly agre porting takes too long; however, this was on those reporters with greater than 20 patien Except for this last attitude, these differences pendently associated with classification as repc and remained after control by demographic & variables.

Reporting Knowledge and Behavior.-Les in five of those physicians who had detected ADE them. Reporting was broadly defined in the que and some of the 76 physicians who had reported than one of the following methods: FDA-SRS (n manufacturer (n=34), colleagues (n=36), med ture (n=2), and others, eg, hospital committees Nonreporters were asked for the reasons tha

Adverse Drug Events

not report. Some of the 342 physicians who had observed an ADE within the prior year that they had not reported gave one or more of the following reasons for not reporting: did not have reporting forms available (n=188), event already documented (n=89), did not get to it, got busy (n = 56), did not believe it was important (n=47), forms were too much trouble (n=37), minor or expected side effect (n = 37), did not like interacting with the government (n = 16), had liability concerns (n = 13), did not know how to report (n=13), and other reasons that the respondent was asked to specify (eg, undetermined as ADE, not primary physician) (n = 35).

Multivariate Analysis-The probability of reporting an ADE once detected during the past year was assessed and the results of this modeling are presented in Table 5. The only variables to be retained in the model were the three reasons for not reporting (no forms, minor reaction, and documented reaction), inpatient practice, and medical specialty. It is not unexpected that reasons for not reporting should be retained; it is the strength of their influence that is of interest. All three exerted a sixfold to sevenfold influence on not reporting. Inpatient practice remained in the model; this was probably a function of the broad definition of reporting (colleagues and hospital committees were listed as choices and represented approximately 50% of responses). Physicians perceived that patient care and outcome were monitored in the inpatient setting, and processes like reporting were therefore automatically performed. Medical specialty was defined as primary care specialties (internal medicine, pediatrics, and family/general medicine). This characteristic emerged independently of its association with ADE detection since only detectors were examined in this analysia It may be that primary care practitioners are more attuned to reporting because of health department communicable disease requirements.

While this model explains approximately 15% of the probability of reporting, it does so with negative coefficients. Thus, it may not follow that if those factors so strongly associated with not reporting were corrected, reporting would occur.

COMMENT

The response rate in this study was 37%; thus, it is uncertain whether the results can be generalized to the

Arch Intern Med-Vol 148, July 1988

entire physician community. Respondents did not differ from the medical community at large in their respective counties when compared on selected demographic characteristics. This comparison, however, was limited by the data available. It is reassuring that the study results conform to other findings and are intuitively acceptable.

The finding that nearly half of physician respondents stated they were unaware that the FDA maintained a reporting system is disturbing since the SRS is relied on for first-line monitoring. It may indicate ignorance of how the system operates rather than of the existence of the system itself; the questionnaire did not make that distinction. There may also be attitudinal or belief factors, unmeasured in this study, that are operating. In any of these situations, however, the SRS would not be used.

Three processes are involved in ADE reporting: detection, attribution, and actual reporting. Adverse drug events have become the "greatest imitator" of disease, as was

Adverse Drug Events-Rogers et al 1599