The effect of prescriber education on medication‐related patient harm in the hospital: a systematic review

Abstract

Aims

Educating prescribers is a strategy to reduce prescription errors in hospitals. The present systematic review gives an overview of original research papers on the education of prescribers and reporting outcomes on (potential) patient harm.

Methods

A search of the databases Embase and Medline, using the Ovid interface, was performed. Research on the effect of physician education in order to prevent medication‐related problems in inpatients, and on reporting original data and outcomes on prescribing errors and/or (potential) patient harm, was included. The assessment of methodological quality and risk of bias was performed using the Methodological Index for Non‐Randomized studies (MINORS) checklist and the suggested risk of bias criteria for Effective Practice and Organization of Care (EPOC) reviews.

Results

Eight studies investigated an intervention on education alone, and in seven studies education was the main part of a multifaceted intervention. All studies were small and had short follow‐up periods. The educational programmes varied and were given to physicians of different specialties and levels of experience. Most studies reported intermediate process parameters as the outcome. The risk of performance and reporting bias were high.

Conclusion

All included studies suffered from poor methodology. The majority, especially studies in which education was part of a multifaceted intervention, reported effectiveness on intermediate outcome markers as prescription errors and potential adverse drug events. However, we found no firm evidence that educating prescribers in the hospital leads to a decrease in patient harm. Further work is needed to develop educational programmes, accompanied by more high‐quality research with outcomes on the improvement of patient care.

Introduction

Although drugs can cure, they can also cause harm. This holds especially true inside the hospital. Vulnerable patients are often admitted because of a transient disease, and this category of patients frequently needs surgical procedures. In these patients, prescribing errors can easily have serious consequences. Several papers have addressed adverse drug reactions in the hospital, showing that up to 20% of admitted patients experience adverse drug reactions during their hospital stay 1.

Risk factors for preventable adverse drug events (ADEs; defined as an injury resulting from the administration of a drug with a causal link to a drug effect 2) are patient age, time since starting new drug, total number of prescription drugs and type of hospital ward. The drug classes most commonly associated with potentially preventable ADEs are antiplatelet drugs, anticoagulants, diuretics (loop and thiazide diuretics), angiotensin‐converting enzyme inhibitors, opioids, antibiotics and antiepileptic drugs 3, 4. Many potentially preventable drug reactions are a consequence of inappropriate prescribing by hospital physicians 5.

Many barriers limit the prescribing process, such as information and communications technology (ICT) shortcomings, high workload, increasingly complex polypharmacy and patient factors, lack of standardization and frequent rotations of inexperienced physicians on the ward 6. It is necessary to understand the causes that contribute to prescribing errors, in order to be able to address these factors.

Various strategies, such as the introduction of computerized physician order entry (CPOE), pharmacist involvement, the introduction of protocols, guidelines, education programmes and support systems for clinical decision making (SSCD) have been studied to improve clinician prescribing in hospitalized patients 7, 8, 9.

Errors – for example, due to incomplete prescriptions or prescriptions that do not satisfy medication order checklists – can probably best be solved by CPOE with decision support or other strategies, although a combination with a form of education may be useful. Prescribing errors defined as ‘irrational, inappropriate and ineffective prescribing, underprescribing and overprescribing’ caused by a lack of clinical pharmacological knowledge might particularly be amenable to a clinical pharmacological educational intervention.

Therefore, a logical strategy to reduce prescribing errors in hospitals is to educate prescribers. It is well known that, in the hospital, the majority of drugs are prescribed by junior doctors, who do not feel adequately prepared to do so 10. However, it is unclear if education programmes actually reduce patient harm. Various education programmes, with different scopes, have been described. Scientific evaluation of these programmes is challenging due to difficulties in blinding and the definition of outcome.

Earlier reviews published in this area describe education as one of several possible interventions to improve prescribing quality, examples of others including the introduction of CPOE and pharmacist involvement in prescribing. In addition, many of the education programmes in the hospital target nurses, or even the patient, rather than the prescribing physician 7, 8, 9.

Other reviews that have reported more specifically on educational interventions have targeted medical students or general practitioners (GPs). Almost every included study has reported on intermediate endpoints, such as an increase in knowledge or measures of self‐assessment instead of improvement of patient care 11, 12, 13. Increase in knowledge in these studies has mainly been measured by written examinations. Practical assessments have been based primarily on written patient scenarios, with a limited number of disease topics 11. Although simulation‐based education has been shown to improve learning outcomes, the contribution to clinical outcomes remains unclear 14. Moreover, a recent study showed that self‐assesment of prescribing skills is poorly correlated with assessed competence 15.

In the present review, we focus on the existing literature on the education of prescribers in hospitals reporting outcomes of (potential) patient harm. We address the scope and form of the education programmes described in the literature and give an appraisal of the scientific merits of the individual studies.

Methods

Search strategy and study selection

A computer‐assisted search of the medical databases Embase and Medline, using the Ovid interface (from 1990 to May 2016), was performed with the aid of a clinical librarian.

A combined search term was constructed as outlined below. The search aimed at finding articles that reported original research data on the prevention of (potential) patient harm due to an intervention involving the pharmacotherapy education of prescribers in a hospital. The search was constructed combining searches according to the PICO (Population: inpatients; Intervention: hospital prescriber education; Comparison: usual care; Outcome: patient harm) model.

The search was performed using MeSH subject headings, combined with keywords to search in the title/abstract and in keyword heading words. The language was restricted to English, Dutch and German.

To retrieve studies on hospitalized patients, the MeSH terms ‘Inpatients’, ‘Adolescent, hospitalized’, ‘Child, hospitalized’, ‘Critical care’, ‘Emergency service, hospital’, ‘critical care’, ‘hospitalization’ and ‘trauma centers’ were used. Keywords to search in the title/abstract and keyword heading words were ‘inpatient*’, ‘hospital*’, ‘emergenc*’, ‘intensive care’ and ‘critical care’.

The MeSH terms ‘physicians’, ‘hospitalists’, ‘surgeons’ ‘physicians/ed’, ‘hospitalists/ed’, ‘surgeons/ed’ ‘education, professional’ or ‘education, medical’, ‘education, medical, continuing’, ‘education, medical, graduate’, ‘internship and residency’ and ‘drug therapy’ were used to retrieve articles that studied the pharmacotherapy education of doctors. Keywords to search in the title/abstract and keyword heading words were ‘educat*’, ‘drug therap*’, ‘medicat*’ and ‘polypharmac*’.

To retrieve articles that reported on patient harm caused by prescribing errors, the MeSH terms ‘Drug‐Related Side Effects and Adverse Reactions’ and ‘Medication Errors’ were used. Keywords to search in the title/abstract and keyword heading words were ‘adverse drug event’, ‘adverse drug reaction’, ‘medication error’, ‘medication related problem/event/error’ and ‘prescribing error’.

Appendix 1 shows the search strategy performed in Ovid.

Two independent reviewers (J.B., C.K.) selected the articles that were retrieved from the search. This selection was based on titles and abstracts. In the case of disagreement, the full text of the article was retrieved. Research on the effect of pharmacotherapy education on doctors in hospitals in order to prevent medication‐related problems in inpatients, and on reporting original data and outcomes on prescribing errors and/or (potential) patient harm was included. The full text of these studies was retrieved. Each of these selected articles was read fully by two authors (J.B., C.K), who independently assessed whether the articles met the inclusion criteria. In the case of disagreement, consensus was achieved in a consensus meeting. In addition, the reference lists of the selected articles were checked for potentially relevant literature. Research was excluded if the education of doctors in hospitals was only a small part of the intervention.

Assessment of methodological quality and risk of bias

The Methodological Index for Non‐Randomized studies (MINORS) checklist, developed by Slim et al. 16 was used for quality assessment of the included studies. This checklist was developed to determine the methodological quality of nonrandomized studies, and consists of 12 methodological items. Eight items are scored for noncomparative studies, and four additional items for comparative studies. The items are scored on a three‐point scale: 0 (not reported), 1 (reported, but not adequate) or 2 (reported and adequate) 16.

To assess the risk of bias, the suggested risk of bias criteria for Effective Practice and Organization of Care (EPOC) reviews were used. These consist of different items regarding risk of bias for interrupted time series (ITS) studies, controlled before‐after (CBA) studies and randomized controlled trials (RCTs) 17.

The assessment of methodological quality and risk of bias was performed by two reviewers (J.B., C.K.). In the case of disagreement, consensus was achieved in a consensus meeting.

Because of the high heterogeneity between studies in terms of study designs and outcome measures reported, a meta‐analysis was not deemed feasible. We therefore provide a descriptive summary of the available evidence.

Results

Search results

The initial literature search strategy yielded 899 articles. A total of 846 articles were excluded after selection, based on title and abstracts, and the full texts of the remaining 53 articles were retrieved. Based on the full text, 15 articles fulfilled the inclusion criteria. The reasons for exclusion of the other articles by the reviewers were that the education was not aimed at prescribers in the hospital (nine articles), no outcome was reported on prescribing errors or (potential) patient harm (20 articles), or no original research was reported (nine reviews and commentaries).

The study characteristics of the selected articles are listed in Table 1. Eight studies investigated an intervention on education of the prescriber alone, and in seven studies the education of the prescriber was the main part of a multifaceted intervention.

Table 1.

Overview of selected studies

Study (year)	Study design	Setting and country	Follow‐up	Sample size	Intervention	Outcome measures	Outcome prescribing error	Outcome (potential) ADE
Intervention on education of prescriber in the hospital
Ajemigbitse et al. 2016 20	Controlled prospective pre–post intervention study	Tertiary hospitals, Nigeria	6 months	3889 prescriptions preintervention 4361, prescriptions postintervention	Audit on prescribing/feedback sessions on prescription errors	Prescribing errors	No change in prescribing error rate overall (5.8% vs. 5.8%; P = 0.984)
Foster et al. 2013 28	Pre–post intervention study	Paediatric ED, USA	3 months	1101 pharmacy interventions preintervention, 1283 pharmacy interventions postintervention	Education on good prescribing/ED medication and prescription errors, followed by daily feedback on patient cases and therapy	Number of pharmacy interventions, ADEs	Significant decrease in specific pharmacy interventions (dose adjustments and order clarifications)	No substantial effect on ADEs
Kozer et al. 2006 24	Prospective controlled cohort study	Paediatric emergency care, Canada	1 month	899 prescriptions	Education on prescribing medication in the ED	Prescribing errors	No difference in prescribing errors between educated and non‐educated residents (12.4% vs. 12.7%, OR 1.07)
Lacombe et al. 2005 29	Prospective pre–post intervention study	University hospital, France	6 months	75 patients preintervention, 52 patients postintervention	Providing guidelines on fluoroquinolone utilization/education session on prescribing errors	Prescribing errors	Decrease in prescribing errors (60.8% vs. 30.8%; P < 0.05)
Peeters et al. 2009 26	Prospective interrupted time series analysis	Internal medicine department, University hospital, USA	7 months	38 275 prescriptions preintervention, 35 441 during intervention, 39 031 prescriptions postintervention	Education on safe prescribing, followed by bi‐weekly feedback sessions on prescription errors	Prescribing errors	Decrease in prescribing errors during intervention (2.3% vs. 1.5%; P < 0.001), postintervention return to baseline (2.3%)
Thomas et al. 2015 31	Controlled pre–post intervention study	Acute medical unit, 3 hospitals, Australia	<2 months	278 patients preintervention, 271 patients postintervention	Good prescribing (online education and high‐intensity education)	Prescribing error	Decrease in some specific prescribing error rates, no difference in low‐ and high‐intensity education
Trivalle et al. 2010 30	Randomized prospective study	Geriatric centre, public hospital, France	2 weeks	576 elderly patients	Education on ADEs and relevant pharmacotherapy in elderly	ADEs		Decrease in ADEs in intervention group (22% vs. 36% P = 0.004)
Intervention on education as main part of a multifaceted approach
Alagha et al. 2011 32	Pre–post intervention study	Paediatric ICU, Egypt	5 months	139 patients preintervention, 101 patients postintervention	Use of medication order chart, education on good prescribing and feedback on prescription errors of physicians, provision of dosing assistance	Prescribing errors	Significant reduction in prescribing error rate (35.2% vs. 78.1%; P < 0.001)
Burmester et al. 2008 21	Pre–post intervention study	Paediatric cardiac ICU, USA	3 years	3648 prescriptions preintervention, 8929 prescriptions postintervention	Implementation of a postcardiac surgery templated physician order, systematic physician education on prescribing errors	Prescribing errors, potential ADE	Significant reduction in total prescribing errors (16.8% to 4.8%; P < 0.001)	No effect on potential ADEs
Campino et al. 2009 22	Pre–post intervention study	Neonatal ICU, Spain	7 months	4182 prescriptions preintervention, 1512 prescriptions postintervention	Education on prescribing errors, several preventive strategies as update of protocols and standardization of medication process	Prescribing errors	Significant reduction in prescribing error rate (20.7% vs. 3.0%; P < 0.001)
Garbutt et al. 2005 23	Pre–post intervention study	Teaching hospital, surgical and medical departments, USA	2 months	2287 prescriptions pre‐ and postintervention	Introduction of grand rounds, interactive education on prescribing errors, promoting safe prescribing behaviours	Prescribing error	Significant decrease in prescribing errors by surgical house staff, an increase by medical house staff
Gazarian et al. 2012 19	Prospective interrupted time series analysis	Paediatric hospital, Australia	4 years	359 patients preintervention, 326 patients postintervention (4y)	Guideline development and dissemination, intensive interactive education, data feedback	(Potential) ADE, prescribing error	Decrease in prescribing error (4.1% vs.2.1%; P < 0.05)	Decrease in potential ADEs (12.3% vs. 4.6% P < 0.05), no decrease in actual or preventable ADEs
Martinez‐Anton et al. 2012 25	Pre–post intervention study	Paediatric ICU, Spain	4 months	2228 prescriptions preintervention, 1791 prescriptions postintervention	Standardization of prescription sources, update of protocols, education on good prescribing	Prescribing error	Decrease in prescribing errors (21.7% vs. 34.2%; P < 0.001)
Rajamani et al. 2011 18	Pre–post intervention study	General ICU, Australia	10 weeks	54 patients preintervention, 58 patients postintervention	Knowledge Translation quality improvement intervention/education and feedback on prescribing errors	Prescribing errors	Decrease in different prescribing error rates
2Thomas et al. 2008 27	Pre–post intervention study	ICU, UK	6 weeks	373 prescriptions preintervention, 974 and 1022 prescriptions postintervention	Development of standards for safe prescribing, education on good prescribing and problems on ICU, audit and feedback	Prescribing error	Significant decrease in prescribing errors. (incident rate ratio 0.25). Wide variation in prescribing error between trainees

ADE, adverse drug event; ED, emergency department; ICU, intensive care unit; OR, odds ratio

Descriptive summary

Study design

All included studies were relatively small. Sample sizes varied from 112 evaluated patients by Rajamani et al. 18 to 685 patients by Gazarian et al. 19. Several studies reported the total number of prescriptions 20, 21, 22, 23, 24, 25, 26, 27 or the number of pharmacy interventions 28 as the sample size. Only two studies made a prospective calculation of the study size using the prescription error rates in the literature or the results of a pilot study 25, 29.

Most studies had a prospective before–after intervention design; two studies had a prospective ITS design 16, 23, two studies had a CBA design 20, 24 and one study was randomized for the intervention 30. In only one study, a blinded assessment of the outcome parameter was performed 24.

The follow‐up of the studies was relatively short, varying from 2 weeks to a few months. Consequently, the sustainability of the effect of the intervention was poorly investigated. Only the studies by Peeters et al. 26 and Gazarian et al. 19 looked at sustainability after 7 months and 4 years, respectively.

Scope and form of the educational intervention

In seven studies, the intervention was aimed solely at education of the prescriber in the hospital 20, 24, 26, 28, 29, 30, 31, and in eight studies the educational intervention was the main part of a multifaceted strategy. In these studies, education was combined with updates and implementation of guidelines and protocols, the introduction of medication order checklists or the performance of audits on the prescribing process 18, 19, 21, 22, 23, 25, 27, 32.

Most of the educational programmes that were studied aimed to improve prescribing skills. The definition of these skills varied from prescribing correctly (prevention of incomplete prescriptions) to safer prescribing behaviour and the prevention of ADEs. In only a few studies, the educational programme aimed primarily to increase the specific pharmacotherapeutic knowledge of the physicians 29, 30.

The method of educating the prescriber also varied between the studies. For example, Thomas et al. 27 compared online education with high‐intensity education sessions.

Educational programmes were offered to different physicians in different settings. Seven studies describe education to paediatricians 19, 21, 22, 24, 25, 28, 32. Three studies were performed on general intensive care units 18, 27, 31; the others were performed on surgical, internal or geriatric wards 20, 23, 26, 29, 30. Garbutt et al. 23 focused their intervention on surgical house staff and medical house staff. Finally, Thomas et al. 27 looked at the education of trainees.

Outcome

Outcomes on potential patient harm were most frequently investigated on intermediate process parameters, with prescribing errors as the main outcome. One study reported the number of pharmacy interventions 28. A few studies reported potential ADEs as an outcome measure 19, 21, 28. Only the study of Trivalle et al. 30 reported ADEs as the main outcome.

Definitions of prescribing errors and (potential) ADEs differed between studies 2. A potential ADE is the result of a prescription error that could possibly have resulted in patient injury if it had reached the patient. However, damage was not assessed as an outcome measure. This might have been one of the reasons why there was a wide variability in reported percentages of prescribing errors and (potential) ADEs (Table 1).

Nine studies reported a significant decrease in prescribing error rates after an education intervention 18, 19, 21, 22, 25, 26, 27, 29, 32. In seven of these, the intervention was part of a multifaceted approach and in two the intervention was education alone.

Three studies reported no change in the overall rates of prescribing errors 20, 24, 31. Kozer et al. 24 and Thomas et al. 27 showed no change in the first month after the intervention, and Ajemigbitse et al. 20 measured the rate of prescribing errors over a 6‐month period after their intervention; in all of these studies, the intervention was education alone. Foster et al. (education alone) reported a significant decrease in specific pharmacy interventions 28. Garbutt et al. (multifaceted approach) found a decrease in prescribing errors by surgical staff, but an increase by medical house staff 23. The study by Trivalle et al. is the only study that reports an effect on actual patient harm 30; a decrease in ADEs was shown in the intervention group (from 36% to 22% P = 0.004).

Three studies (Gazarian et al. 19, Burmester et al. 21, Foster et al. 28) reported on potential ADEs as secondary endpoint. Two (Burmester et al. 21, Foster et al. 28) found no substantial effect, but Gazarian et al. 19 reported a decrease in the potential ADEs from12.3 to 4.6 per 100 patients (P < 0.05).

Peeters et al. reported a return to baseline 7 months postintervention 26. Gazarian et al. showed a sustained long‐term improvement in reducing medication errors after 4 years 19.

Methodological quality and risk of bias

The results of the assessment of methodological quality and risk of bias for EPOC are presented in Figure 1.

Figure 1.

Assessment of methodological quality and risk of bias. MINORS, Methodological Index for Non‐Randomized studies

The most important methodological limitations of the studies included in the present review were the short follow‐ups (less than 3 months), lack of appropriate (equivalent, contemporary) control groups and absence of blinded evaluation of the outcome.

It was generally unclear if the educational intervention was independent of other changes over time and whether the rate of prescription errors was influenced by other confounders during the study period. All together, this resulted in the included studies having a high risk of performance bias. In addition, the risk of reporting bias through selective outcome reporting scored high.

Discussion

The present review showed that only a relatively small number of studies have evaluated the effects of educational programmes for hospital physicians, and of reporting outcomes on (potential) patient harm.

Most studies suffered from poor methodology; the majority of studies were small and the follow‐up of the studies was relatively short. The risk of performance bias and reporting bias were high in all of the selected studies.

In half of the studies, education was the main part of a multifaceted intervention, and all of these studies showed efficacy on intermediate outcome markers as prescription errors and potential ADEs; this was the case in only four out of seven studies in which the intervention was education alone. The content of the educational programmes and the way of providing the education varied considerably. Different definitions of prescribing errors and ADEs were used, contributing to a large variation in the percentages found. This large variation might also be explained by the difference in the study settings and in the prescribers targeted in the intervention.

The restriction of the present review to studies of inpatients, in combination with the restriction to outcome measures on prescribing errors and/or patient harm, resulted in the selection of a limited number of studies. We chose to focus on education in the hospital, especially because of the complex patient categories and the specific care environment. We consider that the effect of education in other settings – for example, nursing homes or GP practice – will be diverse, and measures to improve medication safety, including education, probably need a different approach.

There might have been publication bias on this subject, although we did not find the number of negative studies targeting education alone 20, 24, 27 to be larger than that of positive studies 26, 28, 29, 30. However, it is possible that some initiatives on education in hospital practice do not reach the literature, owing to negative outcomes or methodological challenges.

It has been suggested that the prevention of patient harm is likely to require complex, multifaceted intervention strategies 4. This implies that education alone is not likely to have a large (if any) effect, and that education should be embedded in a broader array of measures aimed at appropriate prescribing. All of the studies in which education was part of a multifaceted approach reported a positive outcome, whereas only four out of seven studies in which the intervention was education alone were positive. This suggests that future research should be targeted at the most optimal combination of measures – for instance, combining education with the implementation of clinical rules 33 and medication reconciliation.

We found considerable methodological limitations in all of the studies, but it should be noted that evaluating education or a multifaceted strategy including education cannot easily be performed in RCTs. Instead, a CBA/ITS design or a cluster RCT must be used, all of which have their inherent methodological limitations and challenges. However, blinded evaluation of clinically relevant outcomes is important, given the subjectivity of the definition of ADEs. In addition, sustainability is an important aspect that should be addressed. In the case of a CBA design or an ITS, a control group without intervention should be included. The outcome of any research regarding this subject should be improvement of patient care, preferably using clinically relevant endpoints.

Although it is plausible that patients will benefit from educating hospital prescribers in pharmacotherapy, the present review showed that further work is needed to develop effective educational interventions and to perform robust evaluations.

The knowledge of how to teach effectively should be combined with optimizing the content of the education. At present, there are no data supporting a specific form of education. A recent article by Franchi et al. 34 showed that an e‐learning educational programme alone failed to improve clinician drug prescription for hospitalized older patients. One of the authors’ suggested explanations for failure was the low level of interactivity of the programme. In addition, they suggested that educational programmes need a follow‐up, to enhance learning retention. The authors also indicated that education should be combined with different strategies in a multifaceted intervention to obtain a real improvement in prescription quality 34.

The qualifications or training of teachers who deliver the educational interventions was rarely addressed in the included studies. This could also have influenced the quality of the intervention and the outcomes.

The present review suggested that educational sessions should be combined with other measures to improve medication safety. In regard to the form of education, there are indications that workplace‐based pharmacotherapy education, using complementary knowledge in interdisciplinary settings, is most effective 35. With continuous audit and feedback on the main pharmacological issues and prevailing guidelines in the workplace of healthcare professionals, there will be a higher likelihood of a sustained effect. Moreover, in our opinion, this educational approach should start early in the programmes of undergraduates. Clinical pharmacological teaching involving students in prescribing in ‘real context’ training programmes, in addition to more classical teaching, has been described to be of great educational value 36. Furthermore, in our view, the content of pharmacology education should be related to known risk factors for medication errors in hospital patients and should focus on the use of high‐risk drugs in high‐risk patients or high‐risk situations. In addition, education should also cover the correct use of the electronic prescribing system and a clinical decision support system.

Conclusion

Taken together, there is currently no firm evidence that educating prescribers in the hospital leads to a decrease in patient harm. However, there is also no sound research showing that education has no effect, and many studies, especially those with the multifaceted interventions, have shown benefit on intermediate outcome parameters. Future research should be targeted at the development and implementation of educational programmes, with outcomes on improvement of patient care, which should be evaluated by high‐quality research. In our view, these programmes should be a part of a multifaceted approach in which education is supported by other measures. It is hoped that this will result in evidence for measures which can be taken to improve medication safety in the hospital.

Competing Interests

All authors declare: no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work; no other relationships or activities that could appear to have influenced the submitted work.

There was no financial support for this study.

Database: Ovid MEDLINE(R) In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE(R) Daily and Ovid MEDLINE(R) <1946 to Present>.

Search strategy:

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐.

1. adolescent, hospitalized/ or child, hospitalized/ or inpatients/ or inpatient?.tw. or inpatient?.kf. or hospital*.tw. or hospital*.kf. or exp critical care/ or hospitalization/ or emergency service, hospital/ or trauma centers/ or emergenc*.tw. or emergenc*.kf. or [(intensive or critical) adj3 care].tw. or [(intensive or critical) adj3 care].kf. (1320173)

2. physicians/ or hospitalists/ or surgeons/ or physicians/ed or hospitalists/ed or surgeons/ed or education, professional/ or education, medical/ or education, medical, continuing/ or education, medical, graduate/ or ‘internship and residency’/ or ed.fs. or educat*.tw. (706878)

3. exp Drug Therapy/ or [(drug adj3 therap*) or medicat* or polypharmac*].tw. or [(drug adj3 therap*) or medicat* or polypharmac*].kf. (1373658)

4. 1 and 2 and 3 (7555).

5. dt.fs. (1855920)

6. 3 or 5 (2637038).

7. 1 and 2 and 6 (9757).

8. ‘Drug‐Related Side Effects and Adverse Reactions’/ or Medication Errors/ or [adverse adj3 drug adj3 (event? or reaction?)].tw. or {[medicat* adj3 error?] or [medicat* related adj3 (problem? or event? or error?)]}.tw. or (prescribing adj3 error?).tw. or [adverse adj3 drug adj3 (event? or reaction?)].kf. or {[medicat* adj3 error?] or [medicat* related adj3 (problem? or event? or error?)]}.kf. or (prescribing adj3 error?).kf. (49627)

9. 7 and 8 (977).

10. (dutch or english or german).la. (22079904)

11. 9 and 10 (928).

12. limit 11 to yr = “1990 ‐Current” (899).

Bos, J. M. , van den Bemt, P. M. L. A. , de Smet, P. A. G. M. , and Kramers, C. (2017) The effect of prescriber education on medication‐related patient harm in the hospital: a systematic review. Br J Clin Pharmacol, 83: 953–961. doi: 10.1111/bcp.13200.