Results: Errors were multifactorial, with a median of 4 (range, 2–5) different types of performance-influencing factors per error. Lack of drug knowledge was not the single causative factor in any incident. The factors in new-prescribing errors included team, individual, patient and task factors. Factors associated with errors in represcribing were environment, task and number of weeks into the term. Defences against error, such as other clinicians and guidelines, were porous, and supervision was inadequate or not tailored to the patient, task, intern or environment. Factors were underpinned by an underlying culture in which prescribing is seen as a repetitive low-risk chore.
Adverse events caused by medication have been estimated to harm 1% to 2% of patients admitted to hospitals in the United States, United Kingdom and Australia.1-3 Most incidents that result in harm to patients originate in the prescribing process.4 Human error is a frequent factor.5
The intern year is a time for consolidating medical school education through continued learning and acquisition of knowledge and skills under direct supervision. More prescribing errors occur in the first year after graduation than in all other years.6 In an American study, 45% of interns reported having made at least one clinical error; 29% were prescribing errors, of which 15% were fatal.7
Although a particular action or omission may be the immediate cause of an incident (described as an active failure or error),8 closer analysis usually reveals a series of events and departures from safe practice, known as an error chain.5 The notion of a single root cause, although widespread, is an oversimplification.8,9 A structured process, which uses human psychology methods, is available for analysis of errors.8,10 This analysis method considers the individual’s working environment, including the team and the organisation.
In this prospective qualitative study, we aimed to identify and analyse the factors underlying prescribing errors made by interns in order to identify multifaceted interventions that would reduce the risk of similar events recurring.
A clinically meaningful prescribing error occurs when, as a result of a prescribing decision or prescription writing process, there is an unintentional significant (1) reduction in the probability of treatment being timely and effective or (2) increase in the risk of harm when compared with generally accepted practice.11
All prescribing was handwritten on an inpatient medication chart. Discharge prescriptions were generated from a triplicate carbon copy system. All 34 interns at the hospital were informed that the study was taking place, that the goal was to introduce safety-improvement strategies, and that deidentified errors would be reviewed by the medication safety team and used subsequently in prescribing training for medical students. Interns rotated among clinical units every 10 weeks. Following local ethics approval, all interns interviewed gave signed consent. Only interns who were involved in incidents and contactable within 72 hours of the incident were interviewed, to ensure adequate recall of events.
Semistructured face-to-face interviews incorporating a questionnaire were conducted to assess the causes of the prescribing error. The process was adapted, with permission, from the methods of Vincent et al8 and Dean et al.12 Interviews took between 15 and 115 minutes (average, 44 minutes) and used the process outlined in Box 1.
We (a senior physician and two senior pharmacists) analysed and coded the transcripts independently, using Vincent et al’s framework of error-producing factors.8 Consensus was achieved through discussion.
Errors were classified as either “new prescribing” or “represcribing.” New prescribing involved a decision to start, stop or change a drug, or a drug’s form, route or dose. Represcribing was any continuing therapy, and included prescriptions written when patients were admitted to hospital, transferred or discharged.
Twenty-six errors were not analysed because prescribers were unable to be identified (8), declined to be interviewed (7; with 3 having been interviewed before), or did not attend (3), or the errors were reported more than 72 hours after the event (8).
The 21 errors that were analysed are described in Box 2. They occurred on admission to hospital (3), during the inpatient stay (9), and on discharge (9). Antithrombotic and antibiotic drugs accounted for half the errors.
In 10 of the incidents where prescribing was directed by a senior doctor, the intern determined the route, form, dose, frequency or duration of the drug and had to consider the previous adverse drug reactions or renal impairment. In only one case of new prescribing was the intern solely responsible for the decision to initiate a medication. In three cases, the intern was prescribing for another doctor’s patient, two when on call.
In 13 incidents, the interns had previously prescribed a drug, but they often mentioned that they did not know the dose. In seven cases, they admitted to having made the same or a similar error previously.
Causes of the errors, presented in accordance with Reason’s model of accident causation, are shown as holes in layers of Swiss cheese to indicate the four levels in which failures occur, enabling an accident to penetrate barriers and defences, resulting in harm to the patient (Box 3).5
We identified at least one active failure, or error, in each incident (Box 2).
Interns cited two or more underlying or influencing factors contributing to each error (Box 2). In new-prescribing errors, a median of 5 (range, 3–5) different factors were mentioned. In represcribing errors, the median was 3 (range, 2–5). Details of different components of the error-producing factors were identified from analysis of the transcripts (Box 4).
Major factors involved with new-prescribing errors were team (11), individual (11) and patient (10). For represcribing errors, they were environment (10), task (9) and the intern’s time on the term (8).
Environment factors were the most frequent factors in 19 incidents, and included the office area, staffing levels and workload (17) (ie, being either busy or working longer than rostered hours). Twelve interns felt they were being pressured to get things done, most frequently when prescribing for patients being discharged. The pressure was often further exacerbated by the working environment being “cramped”, “noisy”, “busy”, “hectic” or “distracting”.
Task factors were associated with 16 incidents. The design of the regular or long-stay medication chart was identified by interns as leading to slips related to eight represcribing errors and one new-prescribing error. The medication chart layout and its location on the ward were key themes in nine of 21 incidents. The design of discharge prescriptions and Pharmaceutical Benefits Scheme requirements were mentioned in four incidents. Interns assumed that if they were directed to prescribe by registrars or copying a senior doctor’s order, it would be correct, and guidelines need not be checked. In two incidents, guidelines were ambiguous, leading to dosing errors.
Individual factors, including physical or mental wellbeing and lack of skills or knowledge, were mentioned in 17 incidents, including all new-prescribing incidents. In 17 incidents, interns reported that they did not know the dose of the drug, had never prescribed the drug before, or had never had to modify the drug choice or dose. In 10 incidents, interns indicated that they lacked experience; six incidents occurred within 3 weeks of starting the 10-week term. In 11 incidents, interns were distracted. The comments “not thinking”, “doing four discharges at once”, “trying to get lunch” were mentioned. Nine interns cited physical issues of tiredness, hunger and thirst, such as “I was anuric each day for a week”. Eight of the interns had received no prescribing training as undergraduates, and four reported minimal training. Seven had received some instruction during their induction program. Six interns mentioned that they were “down” or had low morale, stating that they felt like a “clerk” or “secretary”.
Team factors were present in 16 incidents. They were associated with supervision, communication and responsibility. Poor supervision was a primary theme in nine new-prescribing errors but only three represcribing errors. Interns frequently (in 14 errors) mentioned communication about medication, with comments such as “dialogue is very one-way” and “there’s not much discussion or opportunity to learn”.
A major underlying theme emerged: that interns did not have the skills or knowledge to follow the instructions given and were not prepared to question or seek clarification. Implicit trust and an assumption that the registrar must have been correct were frequently mentioned (eg, “my registrar is really good; he has been prescribing for years”). Interns were often (11 errors) unclear as to who was responsible for the different stages of prescribing. Senior staff invariably decided to initiate therapy (eg, “the team decided to start the diltiazem”). However, in other incidents, the complex decisions to continue or change therapy were left to the intern. In four cases of errors in new prescriptions, the registrar gave instruction over the telephone. One intern received a pager instruction to discharge four patients. Ambiguities in other doctors’ prescribing contributed to three errors.
Patient factors were mentioned for 13 incidents, the most frequent being the complexity or acuity of the case (10 new-prescribing and 3 represcribing). Other patient factors included belonging to other teams (4), being seen out of hours (2) and inability to communicate (3) because of language difficulty, sedation or a neurosurgical complication.
Defences against errors, in the form of other staff, played a major role in preventing harm. In 13 incidents, a pharmacist identified the error and contacted the prescriber before drug administration. On one occasion, the intern was awaiting information from a pharmacist when a nurse administered the wrong drug. One nurse identified an error and refused to administer the order. In four incidents, the drugs were administered before the error was identified by a pharmacist. An adverse drug event (harm to a patient) was detected in at least two of these incidents. On 10 occasions, interns assumed that another senior doctor would have checked the drug order before administration of the drug to the patient. Self-initiated defences by four interns included checking their own orders (usually for completeness, not clinical safety) before signing them. The interns described being distracted from their checking process, resulting in a lapse or an attention slip. On two occasions, ambiguous dose-range information in guidelines, such as the Australian medicines handbook, led to errors.
Quotations illustrating specific factors are shown in Box 5.
Represcribing was commonly mentioned as requiring little thought and having low risk or importance. In four cases, simultaneous multiple prescribing tasks contributed to errors. In general, prescribing was mentioned as a “job” or a “chore”, as in “I just copied it out”. Lack of training in drug knowledge and prescribing skills was another latent factor. Staffing numbers and expected patient throughput affected workloads, which led to mental and physical fatigue, stress and distraction. This included forced long working hours, such as early morning and early evening ward rounds. The need for interns to admit specialist patients out of hours was a result of management decisions and work practices.
All errors were associated with a varying combination of environment, team, individual, task, patient and latent factors, in a system with porous defences. We found that, while almost all errors were influenced by environment factors, factors associated with new-prescribing errors and represcribing errors were different, which was not the case in previous studies of error causation.1,12,13
New-prescribing errors involved inexperienced interns, who were tired, hungry, and distracted, prescribing for patients with complex disorders. Team factors, in particular lack of supervision, were also more frequently associated with new-prescribing errors. Represcribing errors were often related to the task, including the design and location of the medication chart. Therefore, the view that a single intervention in isolation will prevent most prescribing errors is simplistic.14
Our results confirm those of others that improving drug knowledge may decrease the risks of new-prescribing errors.14-18 However, a lack of drug knowledge was a partial cause of only one represcribing incident. Interns need to be able to apply drug knowledge to allow them to tailor therapy to an individual patient. Safe-prescribing skills and awareness of medication errors is required by all members of the health care team,19,20 and should be a core component of undergraduate and postgraduate training programs, as outlined in the new curriculum framework for junior doctors.21 Practical safe-medication practice training that improves the safety of medical students’ prescribing should be a core component of prescribing education.14,20
The primary focus of the intern year is to produce competent, independent practitioners through an apprenticeship, with training in a range of supervised posts.22 In this study, there was an assumption by interns that senior staff would check their prescriptions. Often, this did not happen. Our findings reinforce those from the UK, where a culture exists in which new prescribing is seen by senior and junior staff as focused on drug selection, and represcribing as a low-risk chore for which training or supervision is not required.12,17
The complexity of prescribing is not appreciated by the novice, and the novice’s lack of deeper understanding does not appear to be understood and supported by supervisors. Interns often have insufficient knowledge to appreciate when they need to seek advice. Deference to a hierarchical structure is a well recognised risk in all complex teams, and junior staff need to have the skills and feel able to confirm and clarify directions.23
Supervision should take account of all prescribing risk factors, including the patient’s complexity, the intern’s competence, the specific medications being prescribed and the availability of guidelines. This must be within a culture in which prescribing is seen as an important, high-risk intervention. Institutions must develop an environment in which prescribing errors can be constructively discussed and analysed, and learning from errors should occur at an individual, team and organisation level.12
Electronic prescribing with decision support offers a partial solution,24,25 but an effective system is not currently widely available in Australia.26 Standardised medication charts and systems incorporating decision support and forcing functions have been developed and should be implemented to reduce prescribing errors.27 With a standard chart in place, students can be trained to use one system, and the risks of error due to unfamiliarity with chart design can be reduced.17
Nursing staff provide a critical defence by reviewing medications before administration, but training in safe medication is also required.28 Pharmacists detect errors and improve the safety of prescribing.29 Their role of reviewing prescriptions and contributing to prescribing decisions must be further developed.13
Guidelines and drug information must be readily available to prescribers. The Australian medicines handbook has reworded dosing instructions in response to the findings of this study, but similar risks remain unless guidelines and protocols are evaluated for such risks.
Changes to workload and staffing levels remain a risk for all health care professionals. Strategies such as split shifts need to be considered, but must ensure clinical handover of patients between doctors. The environments in which interns prescribe must not be distracting. The location of medication charts, ideally at the bedside, has already been addressed at the study site.
Our study, using a sample of convenience, cannot estimate the incidence of errors. Also, there may have been a degree of social desirability in responses, and the 26 incidents not investigated may have provided additional or different perspectives. However, we believe our findings are representative, and have raised important issues, which may lead to significant interventions.
The prescribing errors identified have happened before and will happen again unless changes at many levels are made. Prescribing skills and awareness of medication errors must be developed through training. Standardised medication charts reduce errors and are being implemented across Australia, and guidelines should be redesigned and readily available. A cultural shift, in which prescribing is seen as important, must occur, with continual senior review and tailored supervision of interns in an atmosphere that encourages clarification and learning.
1 Interview to identify causes of prescribing errors by interns
Background to the interview
The interviewer (I D C) discussed with the pharmacist the error, the patient outcome, and whether the pharmacist had discussed the error with the prescriber. Medication charts and medical records were reviewed.
The intern was contacted and interviewed at a location of his or her choice away from the ward. The interview identified the intern’s role in the incident and whether the error was in a new prescription or a represcription, and any instruction or supervision.
This was based on Vincent et al’s framework of contributory factors8 and was used to identify and systematically explore any contributory factors. These factors included:
Staffing levels, skill mix and workload
Layout of workplace, ward office
Administrative and managerial support
Poor design of equipment such as medication chart
Availability, clarity and use of protocols
Availability and accuracy of test results
Verbal and written communication
Supervision and seeking help
Team structure (consistency and leadership)
Closing of interview
Interviews were closed by asking the interns if they had any questions, what they would do differently with the benefit of hindsight, and if they had any suggestions for systems improvements.
2 Prescribing errors and performance-influencing factors from interviews with interns and thematic analysis
3 Incident analysis framework*
AMH = Australian medicines handbook.
4 Factors involved in prescribing errors*
5 Examples of error-producing conditions
I have about 12 general med patients to look after today, sometimes one to 20 patients. On the day I made the mistake I think we had up to 30, my registrar was doing their exams and there was a floating covering consultant, who changed each week. (11)
I felt like I had just made a mistake — I take total responsibility — I was told what to prescribe, but I didn’t know about not using it [enoxaparin] in renal impairment, I’m not sure I would have worked out her renal function from the creatinine anyway. I have not done it before. (6)
I had just made a mistake — I knew what I had to prescribe — but I was busy and missed it. When writing these scripts (discharges) you worry about the PBS [Pharmaceutical Benefits Scheme] quantities and number of tablets etc . . . but don’t necessarily link it to the patient . . . I knew he was complicated but just missed writing them up. (18,19)
Latent underlying themes: prescribing as a task or chore
I made the decision to rewrite all of the charts that would have been about to expire in the next few days. During rewriting approximately 10 medication charts in total, I was doing these two patients’ medication charts at the time. I attached patient ID stickers to the blank charts but mixed them up, and the wrong patient’s medication was prescribed on one chart and vice versa. It’s a kind of faceless system as all the charts are kept in a folder away from the patients and other patient information. (14)
* Numbers in parentheses are incident numbers (Box 3).