OUP user menu

Detection and prevention of medication misadventures in general practice

Ka Wae Tammy Tam , Hung Kon Kwok , Yuen Man Cecilia Fan , Kwok Biu Tsui , Kwok Keung Ng , King Yip Anthony Ho , Kam Tong Lau , Yuk Chun Chan , Ching Wan Charmaine Tse , Cheuk Man Lau
DOI: http://dx.doi.org/10.1093/intqhc/mzn002 192-199 First published online: 13 March 2008


Background Adverse drug events are leading categories of iatrogenic patient injury. Development of preventive strategies for general practice setting depends on effective detection of events.

Objective The aim of the study is to compare the strengths and weaknesses of voluntary reporting, chart review and patient survey in measuring medication misadventures in general practice and to analyze the events by severity and preventability, drug groups and patients' and doctors' characteristics, for the formulation of preventive strategies.

Method In the 2-month study period, we applied voluntary report, chart review and patient survey to collect data related to medication misadventures and compared their detection rate.

Results The chart review demonstrated the highest yield for detecting overall medication misadventures (2.03% medication orders), followed by patient survey (1.46% medication orders) and voluntary reporting (0.52% medication orders). Chart review and patient survey were better than voluntary reporting in uncovering preventable adverse drug events. However, voluntary reporting was pivotal in capturing sentinel events. Beta-blocker, diuretic, angiotensin-converting enzyme inhibitor, aspirin and non-steroidal anti-inflammatory drugs had caused 82.0% of all adverse drug events. These events were more common with advanced age of patients, greater number of consultation problems and prescribed drug items. Additional resources implicated were minimal.

Conclusion We suggested a complementary approach using chart review and voluntary reporting in measuring and monitoring medication misadventures in general practice. Close monitoring of the events was necessary for older patients, multiple medical problems and poly-pharmacy and for patients using beta-blocker, diuretic, angiotensin-converting enzyme inhibitor, aspirin or non-steroidal anti-inflammatory drugs on a long-term basis.

  • adverse drug event
  • chart review
  • medication error
  • patient survey
  • voluntary reporting


An adverse drug event (ADE) refers to any injury that is caused by a medication (or lack of an intended medicine) [1, 2]. The incidence of adverse drug events was estimated to be 6.5 to 27.4% in different clinical settings [2, 3]. They are also one of the leading categories of iatrogenic patient injury in the Harvard Medical Practice Study, accounting for 19% of all adverse events [4]. The economic consequences are equally compelling. The attributable cost per adverse drug event was estimated at $2000–2500, largely resulting from increased hospital admissions, prolonged hospital stay and time away from work [57].

Clearly, medication safety is an essential component in risk management for a standard healthcare system. The humanistic and economic benefits of reducing preventable drug related morbidity are great. However, medication errors are difficult to measure because of their great variability. No method in isolation can study every aspect of the problem; therefore, error detection requires a multipronged approach [8]. Furthermore, more detailed analyses of the drug events, by classes of drugs responsible and by associated risk factors, are necessary for development of cost-effective strategies to better detect, prevent and manage these problems [9].

As most of the current published findings were related to hospital inpatients only, to evaluate the problem of medication misadventure in general practice, due to the differences in patient spectrum and operating structure, the approach of selecting the appropriate methods for event detection and monitoring would vary accordingly. It is prudent for primary care physicians to be aware of the most effective method for their own setting before improvement strategies can be appropriately and effectively implemented and reviewed.

In this study, we sought to compare the strengths and weaknesses of the three commonly used methods—voluntary reporting, chart review and patient survey—in detecting medication misadventures in general practice. We also described the type and nature of the drug events detected, the drug groups involved, and the associated factors observed in these events.


We conducted the study in four primary care clinics over a 2-month period from May 2006 to July 2006. The four clinics shared similar practice setting and were operated by the same department.

In these study sites, doctors prescribed medications by manual but there was a computer system installed in Pharmacy for keeping prescription records and printing drug labels. During the study period, we used three methods—voluntary report, chart review and patient survey—to collect data related to medication misadventures from different sources of people including doctors, nurses, pharmacists and patients.

Definition of terms (Table 1)

Medication misadventures consist of the sum of medication errors and adverse drug events [10]. Medication errors are specific types of errors in that they are preventable events that can occur at any stage in the medication use process that lead to patient harm or inappropriate medication use [11]. Minor medication errors are those having little or no potential for harm, e.g. missing the frequency for a non-critical medication. If the medication error has resulted in actual patient harm, it is also categorized as an adverse drug event. An adverse drug event can be either preventable or non-preventable, and is due to a medication error. Potential adverse drug event is a medication error that has the potential to harm a patient but which does not actually cause any harm (near-miss) [2]. It is further categorized as intercepted or non-intercepted. An intercepted potential adverse drug event is a drug order that is intercepted before it reaches the patient, whereas a non-intercepted potential adverse drug event has reached the patient but does not cause injury because the patient has sufficient physiological reserves [2].

View this table:
Table 1

Definition of terms

To determine the relationship between the drug and the incident, we referred to the rating method developed by Naranjo et al. [12] to decide how likely the observed event was related to the drug.

Voluntary reporting

We devised two types of incident report forms (Appendix I and II) for recording ‘Medication errors’ and ‘ADEs and potential ADEs’. The severity of events was graded according to the scale significant, serious, life-threatening or fatal [13]. To avoid duplicated reporting of the same event using both report forms, for a medication error that was also an ADE, we reported it as an ADE and therefore used the ‘ADE report form’. In this way, medication error that had no potential for patient injury was reported using ‘Medication error report form’, while medication error that had already caused or had the potential to cause patient injury was reported using ‘ADE report form’.

All ranks of health professionals were encouraged to report all the drug events that they had encountered during the study period. Clinic meetings were arranged to introduce the objective of project and method of reporting. There was a one-week pilot reporting period before commencement of the study. The name of the reporting subject was kept anonymous, and we emphasized on the creation of a non-punitive culture to maximize data capture at the study sites.

Chart review

A chart-review panel comprising of eight doctors, two from each clinic, was responsible for charting the drug events from case notes and prescription sheets. Panel meetings were arranged before the study for standardization of charting, and then every 2-months thereafter to discuss the problems in charting so arisen.

The two representatives from each clinic worked as a pair. They selected the records and the date of consultation to be assessed by random selection from the corresponding prescription sheets issued within the study period, and followed a series of preset classification tables to determine whether a drug event had occurred. The basic preset criteria were derived from a reference sample published in 2004 and which was shown to be feasible and reliable for use in various clinical settings to measure medication safety [13]. The chart-review panel modified the reference sample according to the characteristics of our own practice setting and drug formulary. In each clinic, individual doctor worked independently on charting all the records for that clinic, but the paired doctors would compare their results after completing the process and settle any discrepancy in the ratings among themselves. Any further disagreement would be discussed in the panel meetings for consensus.

For every medication error identified, the ‘Medication error report form’ would be completed; for every adverse drug event/potential adverse drug event identified, the following data would be recorded: classification and summary of the event, drug group(s) involved and the route of administration, severity of the event, and any additional resources required, in terms of number of consultations, investigations, medications, hospital admissions and referrals to the Emergency Department. Besides, we also considered the underlying reasons for every doctor's prescription of undesirable or contraindicated medications to target our improvement strategies.

Patient survey

A team of eight nurses was responsible for conducting the telephone interview to collect the drug events from 600 subjects, whom we identified by random selection from all the prescription sheets filed in the study period. The nurses followed an identical set of structured questions (Appendix III) and standardized the way they asked the questions. The format of data collection and entry was the same as the chart review method.

Statistical analysis

Our sample-size calculation was based on the results of previous studies of the prevalence of adverse drug events. Each detection method included more than 1000 drug orders and had 90% power at the 0.01 significance level to test the hypothesis that chart review would reveal twice as many cases of adverse drug events as voluntary reporting.

We compared the frequency of medication misadventures identified by each method. The data were analysed by using the Statistical Package for the Social Sciences version 10.0 (SPSS Inc. Chicago [IL], USA). Frequencies were described as percentages and the descriptive data as mean or median, if appropriate. We performed t-test for comparing parametric data, chi-squared test for norminal categorical variables and somers'd test for ordinal categorical variables. Mulitvariate logistic regression was then applied to adjust for confounding relationship between different explanatory variables. A two-sided P-value of less than 0.05 was considered to indicate statistical significance.


During the study period, the four clinics issued a total of 73 117 medication orders from 27 339 prescription sheets, giving an average of 2.67 drug items per prescription sheet. Of all the medication orders issued, voluntary reporting had identified 250 medication errors (0.34% medication orders; 95% confidence interval, 0.30–0.38%) and 132 ADEs (0.18% medication orders; 95% confidence interval, 0.15–0.21%). Over the same period, chart review for 2056 medical records or 5466 medication orders had captured four medication errors (0.07% medication orders; 95% confidence interval, 0–0.14%) and 107 ADEs (1.96% medication orders; 95% confidence interval, 1.59–2.33%). Out of the 600 subjects surveyed by telephone, for whom 1438 medication items had been prescribed, six medication errors (0.42% medication orders; 95% confidence interval, 0.09–0.75%) and 15 ADEs (1.04% medication orders; 95% confidence interval, 0.54–1.56%) were elicited. As there was minimal overlap between the three methods, the overall frequency of medication misadventures in the study period was estimated to be around 0.64% medication orders (95% confidence interval, 0.58–0.70%), of which half of them were non-preventable occurrences.

Table 2 shows the comparison of the yield of the three strategies in detecting medication errors, adverse drug events and overall medication misadventures. Chi-squared test showed that chart review had produced the highest yield of detection of adverse drug events and overall medication misadventures among the three strategies (P < 0.0001); patient survey was the intermediate (P < 0.0001), while voluntary reporting had produced the lowest yield of detection (P < 0.0001).

View this table:
Table 2

Frequency of medication misadventures detected by voluntary reporting, chart review and patient survey

Medication errors

Of the 260 medication errors detected during the study, 252 were prescription errors, five were dispensing errors, and three were administration errors, due to non-compliance of patients with given instructions. Most of the prescription errors were discovered by pharmacists (92.8%), followed by patients (4.0%), doctors (2.8%) and nurses (0.4%). The dispensing errors were detected mostly by patients (80.0%).

Adverse drug events

Of the 72 potential adverse drug events captured, most of them were intercepted (95.8%) in voluntary reporting, but almost all of them were non-intercepted when detected by chart review (97.9%) or patient survey (100.0%).

For adverse drug events (Table 3), most of them were non-preventable occurrences, but a larger fraction of them noted by chart review (15.0%) and patient survey (14.3%) were preventable, when compared with voluntary reporting (4.6%). Among all 182 adverse drug events detected, 159 (87.3%) were graded as significant in severity, with no obvious differences among the three strategies; 22 (12.1%) were graded as serious, and one (0.6%) was graded as life-threatening, which was detected by voluntary reporting and was due to allergic reaction from a non-steroidal anti-inflammatory drug. Doctors (62.2%) had discovered more of the adverse drug events than patients (24.8%), pharmacists (6.7%) or nurses (6.3%).

View this table:
Table 3

Distribution of adverse drug event and potential adverse drug event detected by voluntary reporting, chart review and patient survey

Table 4 shows the analysis of the 122 adverse drug events captured by chart review and patient survey. We found that a significant proportion of adverse drug events (57.4%) were related to side effects of medications. There were only four events caused by drug allergy, but three of which were serious. Nine out of 122 events (7.4%) were related to the use of contraindicated drugs. Concerning the drug groups involved, anti-hypertensive medications including beta-blockers, diuretics and angiotensin-converting enzyme inhibitors had accounted for a major proportion (70.5%) of the adverse drug events, followed by aspirins and non-steroidal anti-inflammatory drugs (11.5%). For the 2656 patients recruited in chart review and patient survey, adverse drug events had resulted in an additional use of 8 consultations, 12 investigations, 11 medications and 2 referrals to hospital/Emergency department, at a rate of 0.004 consultation, 0.006 investigation, 0.005 medication and 0.001 hospital/emergency referral, per patient. Results from bivariate analysis indicated that adverse drug events were more common in patients of older age (P < 0.0001), male patients (P = 0.001), greater number of drug items prescribed (P < 0.0001), and more diagnostic codes per consultation (P < 0.0001) (Table 5). But the effect of male gender was excluded from the multivariate logistic regression model after adjustment for confounding variables. Higher professional qualification was not associated with less adverse drug events.

View this table:
Table 4

Summary of adverse drug events captured by chart review and patient survey

View this table:
Table 5

Factors noted in adverse drug events in bivariate analysis


Medication misadventures are common and significant contributing factors to patient morbidity and mortality in general practice [24, 1416]. The reported rates of medication errors and adverse drug events have varied depending on the criteria used for defining events and the methods by and intensity with which they are sought [17]. A chart-review study on adult inpatients by Bates et al. detected medication errors in 5% of drug orders [1]. Another report from 36 tertiary institutions in Georgia and Colorado showed that 19% of medication administrations contained error [18].

In the current study, the estimated frequency of medication misadventures was around 0.64% medication orders, of which half of them were non-preventable occurrences. The figure was much lower than that reported by other studies. But the variations in patient settings and reporting criteria did not allow a valid comparison of our findings with those from other countries.

Among the three methods, voluntary reporting is the least costly, chart review is the most personnel intensive, while patient survey is the intermediate between the two. From our results, voluntary reporting demonstrated the lowest power for detecting adverse drug events—in particular, non-intercepted potential adverse drug events and preventable adverse drug events, which had the greatest implication for informing improvement strategies—and overall medication misadventures among the three strategies. But its role was unique in providing data quick enough about any sentinel event emerged. Voluntary reporting had captured the most significant adverse drug event—life-threatening drug reaction from a non-steroidal anti-inflammatory drug—for this project, and had enabled timely implementation of targeted measures for practice improvement.

A well-designed report form that is anonymous, simple and concise was an important factor for facilitating and stimulating the reporting culture and habit and for increasing the yield of event detection.

For chart review, we exhausted a series of standardized classification tables to extract data on adverse drug events intensively, with the aim to provide more information for formulating preventive strategies. A comprehensive set of rules served as ‘identification triggers’ for possibility of occurrences of drug events. This difference in emphasis from the other two methods might have explained why chart review had produced the highest yield for adverse drug events among the three, and had provided us with the most comprehensive data for risk management in medication safety.

Nonetheless, chart review was the least effective among the three methods in detection of medication errors. One factor might be that in our practice, most prescribing errors had already been discovered by pharmacists during the filing of prescriptions, and the errors would have been corrected on day of prescription and not able to appear in the retrospective reviewing of records. Considering also the high demand in manpower involved, chart review was probably not cost-effective for detecting medication errors.

The strength of patient survey in general practice lay in its power to uncover dispensing errors, as suggested by our results that four out of the five dispensing errors disclosed in the study were being identified by patients. To improve the detection of dispensing and administration errors, some institutions have recommended the method of direct observation [19]. Nevertheless, the application of direct observation in our pharmacy may not be as feasible and acceptable for our population. In our practice, patient survey might be considered a practical method for investigating dispensing errors.

From our data, we recommended a complementary approach by use of an effective incident reporting system and regular chart reviews for detection and monitoring of medication misadventures in general practice.

As for the adverse drug events detected, most of them were not serious, and over half were constituted by side effects of medications, frequently electrolyte disturbances. Drug allergy and use of contraindicated medications, in contrast to other types of adverse drug events, have a relatively higher potential to cause serious adverse clinical outcomes and medico-legal consequences. The chart review analysis showed that three out of four unexpected allergic reactions were classified as serious. Proactive measures to prevent any known drug reaction should be one of the major focuses in medication safety. To minimize harm to patients from inappropriate use of medications, studies in the USA had promoted the use of a computer-based decision support system—Computerized Physician Order Entry—for prescription [20, 21]. They showed promising results in demonstrating its ability to prevent the occurrence of medication errors that have a high likelihood of leading to adverse drug events [20, 21]. For practices using manual prescription system, improvement strategies such as manual reminders, guidelines, audit and educational outreach were evaluated to be effective [2225].

We also observed that more adverse drug events had occurred in patients who were older, had multiple medical problems and had multiple drug prescriptions. Understandably, elders were more susceptible to adverse drug reactions because they had a lower physiological reserve, were more likely to have organ impairment, had a lower symptom response to detect early side effects of medications, and were less able to communicate their symptoms to case physicians. The need to handle multiple medical problems and prescribe multiple medications in a consultation probably added to time constraint in our busy setting and posed psychological stress on doctors, contributing to the increased errors in prescription. Poly-pharmacy increased the risk for untoward effects of medications, drug interactions and patient non-compliance.

For the groups of drugs involved, we found that beta-blockers, diuretics and angiotensin-converting enzyme inhibitors had constituted 70.5% of all adverse drug events, followed by aspirins and non-steroidal anti-inflammatory drugs (11.5%). These medications were mostly prescribed for chronic illnesses such as hypertension, cardiac failure and coronary artery disease, which were also more prevalent in the elders. Despite the heterogeneity of adverse drug events, our results suggested that a few types of drugs had accounted for a substantial proportion of these occurrences. Targeting these high-priority areas could have successfully reduced the overall frequency of adverse drug events.

As for the limitations of this study, first, the preset classification tables for chart review were complicated. The doctors had to be very familiar with the preset criteria to be able to elicit accurate findings from the records. The intensive training and standardization process, pairing up of doctors and regular panel meetings were there to minimize the biases caused by human errors. Secondly, for telephone survey, the interviewees' recall power of drug events might have diminished by the time they were interviewed. A good interviewing technique would be essential for prompting the individuals to recall the events out of their fading memories.


Medication misadventures are significant contributing factors to patient morbidity and mortality in general practice. We proposed the use of chart review complementary with an effective incident reporting system for detection and monitoring of medication misadventures in general practice. Close monitoring for adverse drug events was necessary for older patients, patients with multiple medical problems and poly-pharmacy, and for patients using beta-blocker, diuretic, angiotensin-converting enzyme inhibitor, aspirin or non-steroidal anti-inflammatory drug on long-term basis. Targeting these high-priority areas could have successfully reduced the overall frequency of medication misadventures.


We would like to express our deepest appreciations to Ms. Edborough Lai, Ms. Chung Lai Ngo, Joan, Ms. Kwong So Ying, Ms. So Wai Yi, Libby, Ms. Hung Hoi Ying, Wasa, Ms. Wong Siu Yuk, Mr. Kwok Chi Kong, Richel, Ms. Ho Yuk Ling, Priscilla and Ms. Tong Lai Yee, Anna for their dedication and effort in coordination and conduction of this project.

Ethical approval: A formal ethical approval was not sought, as the activities involved in this project were in fact an intrinsic component of our customer service for quality improvement.

Appendix 1

Medication Error Report Form

1. Patient's Gum Label

 I.D. Number:


2. Date of medication error:

3. Description of the medication error:

 a. Prescription error (can tick more than one)

 □ Unclear/wrong drug name□ Illegible handwriting
 □ Wrong abbreviation□ Missing whole item claimed by patient and confirmed with doctor
 □ Wrong/missing dosage□ Drug not available in formulary*
 □ Wrong/missing drug strength□ Wrong patient
 □ Wrong/missing frequency□ Double entries
 □ Wrong/missing duration/quantity□ Wrong/missing date of prescription
 □ Wrong route□ Others:
 □ Missing doctor's signature

 b. Dispensing error (can tick more than one)

 □ Wrong drug□ Double dispensing
 □ Wrong strength/dosage□ Drug omission
 □ Wrong quantity□ Expired drug issued
 □ Wrong patient□ Others:
 □ Wrong label information

 c. Patient factor**

4. a. Error detected by (one tick only) b. Error reported by (one tick only)

 □ Medical□ Medical
 □ Nursing□ Nursing
 □ Dispensary□ Dispensary
 □ PatientOthers:

* Excluding drug stock insufficiency

** For example, compliance problem

Appendix 2

Adverse Drug Event Report Form

1. Patient's Gum Label

  I.D. Number:


2. Date of the Adverse Drug Event (ADE) I:

3. Classification of the ADE II (one tick only)

 Potential ADEADE
 □ Intercepted□ Preventable
 □ Non-intercepted□ Non-preventable

4. Brief summary of the ADE III

5. Severity of the ADE IV (one tick only)

 □ Significant

 □ Serious

 □ Life-threatening

 □ Fatal

6. a. ADE detected by (one tick only)b. ADE reported by (one tick only)
 □ Medical□ Medical
 □ Nursing□ Nursing
 □ Dispensary□ Dispensary
 □ Patient□ Others:

I. ADE: an injury resulting from medical intervention related to a drug. i) Preventable ADE: ADE caused by prescription or dispensing errors, e.g. a coma due to an overdose of a sedative. ii) Non-preventable ADE: non-preventable occurrences, e.g. an antibiotic causing Clostridium difficile infection, unpredictable drug rashes and side effects of medications.

II. Potential ADE: drug errors that have the potential to harm patient (near-misses). i) Intercepted potential ADE: the drug order is withheld before it reaches the patient. ii) Non-intercepted potential ADE: the drug order reaches the patient but does not cause injury because the patient has sufficient physiological reserves.

III. For example: 1) augmentin being prescribed to a patient with known penicillin allergy, 2) napoxen causing epigastric pain.

IV. Example: drug allergy to an antimicrobial: i) Significant—a rash associated with the antimicrobial. ii) Serious—the antimicrobial causes hives. iii) Life-threatening—the antimicrobial causes anaphylaxis.

Appendix 3

Questions for patient survey

Q1. Introduction of yourself and objective of the telephone survey.

Q2. In that medical consultation, did you notice anything wrong with the prescribed or dispensed drug items, for example with drug name, dosage, frequency, duration, quantity, missing items by doctors, drug omission by pharmacists, wrong label information, expired drug, or any occasion under which the pharmacist had asked you to go back to see the doctor again for clarification of the drug prescription?

Q3. Had you followed the instructions given by doctor and pharmacist when using the medications?

Q4. If there was any amendment ever made to your prescription sheet, was it made before or after you had taken the medications?

Q5. Did you feel unwell after taking the medications which you thought may be related to the prescribed medications e.g. side effects or allergic reactions?

Q6. Were you ever called back to see the doctor again after that medical consultation for reasons related to medications, e.g. change drug, or adjust dosage?

Thank you very much.


View Abstract