Medication errors: not just a few “bad apples”

Abstract

Objective: The purpose of this study is to describe the distribution of medication errors among physicians.

Design: Data on a cohort of internal medicine physicians were gathered prospectively over a seven-month period.

Setting: The study was conducted in four adult primary care practices.

Participants: Twenty-four physicians from Boston, MA participated in this study. A total of 661 patients agreed to participate in the study. The researchers surveyed patients to identify possible adverse drug events (ADEs). This information was supplemented by a chart review for each participating patient.

Measurements: The principal measurement collected in this study was total medication errors per physician. The number of prescriptions written during the study period was also collected.

Results: Twenty-two of the 24 physicians made at least one error. Although there was one outlier, the error rate among this cohort of physician was evenly distributed.

Conclusion: The wide distribution of errors among this group of physicians undermines the argument that the majority of medication errors are due to a “few bad apples,”

Introduction

Leaders in the patient safety movement have argued that systems, and not providers, should be the main target of error reduction approaches. This view holds that all providers make errors and that our efforts should be aimed at activities that promote cultures of safety and use tools that represent systems changes, such as computerized physician order entry (CPOE) [1]. However, many attorneys and others who oppose reform of the current malpractice environment, have a different view [2]. They subscribe to the “Bad Apples Theory,” [3] which claims that a minority of physicians cause the majority of errors, and thus these bad apples should be pursued through litigation. To our knowledge, no published data are available to support or refute this hypothesis. Thus, using data gathered during a previous study of prescribing errors, we conducted a secondary data analysis concerning the distribution of medication errors among physicians [4].

Methods

The data available for this analysis was limited in scope to errors involving medication prescriptions. The original researchers prospectively gathered data on a cohort of primary care physicians and their patients. The physicians were recruited from four different primary care practices in Boston, MA. The researchers had solicited participation from all patients who received a prescription from one of the participating doctors during a four-week enrollment period. For each practice, the four-week enrollment period began on a different date. Patients were excluded if their doctors thought they were too ill to participate or had a hearing impairment that would interfere with participation or if the patients did not speak English or Russian. For each patient who agreed to participate, the researchers attempted to identify all adverse drug events (ADEs) that occurred during the three months that followed that patient’s enrollment. Potential ADEs were identified by patient surveys and chart reviews. Two physicians reviewed these potential ADEs to determine if indeed the event could be classified as adverse. The researchers then categorized actual ADEs as non-preventable, preventable, or ameliorable (i.e., the duration or severity of the event could have been reduced if alternative actions had been taken). For the purposes of our analysis, we considered all preventable and ameliorable ADEs to be medication errors. Due to the asynchronous commencement of enrollment periods at each practice, the total length of the study was seven months (September 1999 to March 2000).

Results

A total of 24 physicians participated in this study (13 men and 11 women). The mean age of participants was 41 years old. On average the physicians had been practicing for 12 years prior to the study. One thousand two hundred two patients of these 24 physicians qualified for the study, 661 of whom agreed to participate. For these 661 patients, the physicians wrote a total of 2134 prescriptions during the study period. The mean number of prescriptions written by a physician was 89. Of the 24 participants, 22 made at least one prescribing error over the seven-month period (92%) that led to a preventable or an ameliorable ADE. The ADEs identified ranged in their degree of seriousness from minor events such as sleep disturbances to serious events such as gastro-intestinal bleeding.

Figure 1 shows the rate of error for each of the 24 physicians. The median percentage of prescriptions written in error was 3.0 (interquartile range: 1.3 - 4.2%). The sample had one outlier whose error rate was more than 3 standard deviations above the mean. The number of prescriptions written by this outlier was roughly equal to the mean. The three physicians with the highest error rates accounted for only 34% of the errors.

Figure 1.

Rate of error for 24 physicians.

Conclusion

Our data offer conflicting evidence regarding the “bad apples” hypothesis. On one hand, we found that almost all adult primary care physicians in our sample made medication errors. On the other hand, three clinicians accounted for one-third of all errors. This latter finding is compatible with the argument that a small proportion of physicians account for a disproportionate share of errors. Our data do not allow for a closer examination of the outliers’ characteristics such as the types of patients they treat, their work conditions or other factors that might explain the results. Nevertheless, two thirds of errors were made by others, so it would not be possible to substantially reduce the total number of errors without introducing systems change that affect all providers. This small sample supports the view of leaders in the patient safety movement [1] that medication errors are widely distributed among physicians and other providers. Although the health care system has an obligation to identify and deal with substandard providers, we believe that solutions that focus solely on weeding out the worst offenders will not have a large impact on the overall safety of health care. Instead, we advocate the adoption of system-wide approaches that foster a safety culture as well as systemic changes (such as CPOE) that fundamentally alter the way care is delivered.