The Institute of Medicine report To Err Is Human shone a spotlight on preventable medical errors and since the release of the report patient safety has become the pre-eminent issue for health care.¹ With our understanding of the problems and solutions for patient safety growing daily, it has become clear that the prescribing, dispensing, and administration of medications represent a substantial portion of the preventable medical errors that occur with children and that children are more at risk for medication errors than adults.^2,3

Despite the awareness that children are at increased risk for medication errors, little is known about the epidemiology of these errors and where the gaps remain in our present knowledge. We conducted a systematic literature review to synthesise all the peer reviewed knowledge on medication errors for children published since the release of the To Err Is Human report.¹ Our scope included all care settings and all types of medications. In addition, we synthesised all the recommendations to improve paediatric medication safety from national entities and evaluated the paediatric evidence provided to support these recommendations for effectiveness, efficacy, cost effectiveness, feasibility, appropriateness in different settings and institutional barriers.

METHODS

Study inclusion criteria for systematic literature review on medication errors

Articles eligible for inclusion in our synthesis had to report peer reviewed English language original data on the epidemiology of medication errors in children published between 1 January 2000 and 30 April 2005. Medication errors were defined as any preventable error in the medication administration process starting from prescribing and including preparing, dispensing, administering, monitoring the patient for effect, and transcribing (eg, medication administration record (MAR)). We only included adverse drug events (ADEs) that were described by the studies as either preventable or having significant potential for harm to the patient (fig 1).²

Search strategy for systematic literature review on medication errors

We completed searches of PubMed, Embase and Cinahl in April 2005. The search strategy combined terms for the population (eg, paediatric) and terms to identify articles dealing with medication errors (eg, medication errors as Medical Subject Heading, preventable adverse event) (Appendix 1, available at http://qshc.bmj.com/supplemental). References for all eligible articles were also reviewed. The search results were tracked in a database created in the bibliographic software ProCite (ISI, Berkley, California, USA).

Two independent non-blinded reviewers screened the title and abstract of each article to determine eligibility. At the full-text level, two non-blinded reviewers screened articles and, if the article was eligible, extracted relevant information in a sequential fashion so that the second reviewer was able to see the extraction results from the first reviewer. All reviewers had either clinical degrees or health services research degrees with experience in systematic reviews. All reviewer pairings at all stages of this effort included at least one clinician, and the reviewer pairings for the abstracts were kept for the full text review. Discrepancies were resolved by consensus between the two reviewers after discussion. We developed and pilot tested forms to extract information such as duration of the study, type of study, the incidence of medication errors and information about the medication errors studied, such as the type and severity (Appendix 2, available at http://qshc.bmj.com/supplemental). Evidence tables summarising the information from the articles were created from the spreadsheets and we qualitatively synthesised the literature since no articles included comparable numerators, denominators, or definitions for medication error that would have permitted quantitative synthesis of the articles.

Synthesis of recommendations to reduce medication errors for children

Working in conjunction with the Institute of Medicine, we identified national entities that have created and disseminated recommendations to improve medication safety either specifically for children or more broadly for all patients. These entities were: Institute for Safe Medication Practices, Pediatric Pharmacy Advocacy Group, American Hospital Association, American Academy of Pediatrics/National Initiative for Children’s Healthcare Quality, Institute of Medicine, National Quality Forum, Massachusetts Hospital Association/Massachusetts Coalition for the Prevention of Medical Errors, National Coordinating Council for Medication errors Reporting and Prevention, Agency for Healthcare Research and Quality, and Joint Commission on Accreditation of Healthcare Organizations.^{1,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22} We reviewed all the published recommendations from these bodies and any cited literature to support the recommendations to determine whether this literature included, or was specific for, children.

RESULTS

Literature search for systematic review on medication errors

Our search identified 358 articles. Eight-four (23%) of these articles were deemed eligible through the title and abstract screening. The most common reason for excluding an article from further consideration was lack of original data. A further 52 articles were excluded during the full-text review, and we were unable to retrieve one article, leaving 31 articles for full text data extraction.^23–53 Figure 2 provides an overview of the search and screening process (Appendix 3 lists the articles excluded, available at http://qshc.bmj.com/supplemental).

Systematic literature review on medication errors

Study characteristics

Design

Table 1 details the characteristics of the 31 studies. Twenty-three of the 31 studies occurred within single institutions and 21 included data from a ⩽1 year period, with the minimum time period being 1 week. Twenty-two of the studies evaluated paediatric inpatients, five studies were focused on either ambulatory clinics or emergency departments and three studies evaluated the home setting. Eighteen of the studies evaluated all medications able to be dispensed in that care setting, whereas 13 studies focused on only a subset of medications (table 1).

Overall the numerator data reported was described as or was consistent with “medication errors” in 25 studies, “ADEs” in 1 study, and 5 studies reported both medication errors and ADEs. The types of medication errors reported included prescribing errors (n = 14), dispensing errors (n = 7), administering errors (n = 14), monitoring patient for effects errors (n = 1), MAR errors (n = 7), and overall lumping of all of these error types (n = 14). There was non-uniformity in the definitions used, if they were explicitly stated, for medication errors. For example, one study defined an error as only those orders with a 10-fold dosing error and another defined medication errors as only those with >10% deviation from recommended dose. These differences are detailed in tables 3–8 under numerator and numerator description columns.

Denominator data

The denominator data were equally non-uniform among the studies. The possible denominators included: manual (paper-based) error reports, computerised error reports, manual medication orders, computerised medication orders, patient days, number of admissions, and time periods. Furthermore, many studies had very narrowly defined denominators, such as number of total parenteral nutrition orders or patients <10 years of age who received paracetamol or ibuprofen in the past 24 hours.

Source of data collection

The majority of the overall data was collected by either chart reviews (n = 14) or incident/error reports (n = 11). Although studies that used incident/error reports cannot be used to assess overall epidemiology of medication errors in children, we included them in order to provide insight into the distributional epidemiology of types of medication errors seen in children.

Overall medication error results from systematic literature review

Fourteen of 31 studies reported overall medication error data that included the entire spectrum from prescribing through to monitoring patient for effect. Table 2 gives the results of these studies. Of these, seven reported broad estimates of overall medication error rates in all children based on actual or estimated data using denominators such as patient days, admissions, or orders as opposed to evaluating only medication error reports (Simpson 2004²³; Cimino 2004²⁸; Potts 2004³⁰; Sangtawesin 2003³¹; Holdsworth 2003³⁵; Kaushal 2001⁴²; Marino 2000⁴³). Using the studies, the results showed a range of estimated medication errors per medication orders from 5% to 27% based on three studies with similar numerators and denominators that allow consideration together (Cimino 2004²⁸; Kaushal 2001⁴²; Marino 2000⁴³).

Prescribing error results from systematic literature review on medication errors

Fourteen studies reported medication prescribing errors (table 3). These summarised an estimated prescribing error rate per medication orders of 30%, 20%, and 4% from the three studies that used similar numerators and denominators (Potts 2004³⁰; Fontan 2003³⁴; Kaushal 2001⁴²). The first two studies in this estimate appeared to have broader definitions of medication errors, which may explain the higher error rate estimates. For example, both included all types of omissions as a medication error, such as omissions of weight and prescriber’s name. Three of the studies reported overall prescribing errors as rates per patient. The estimates from these studies are prescribing errors for each patient of 4–400 per 1000 patients (Sangtawesin 2003³¹; Kozer 2002³⁹; Kaushal 2001⁴²).

Dispensing error results from systematic literature review on medication errors

Seven studies reported dispensing errors, although the design of the studies was very different (table 4). Looking at the greatest commonality between these studies—namely, those studies based on error reports, the estimates of the percentage of reported errors that are related to dispensing vary widely from 5% to 58% (France 2004²⁴; King 2003³³; Frey 2002³⁸).

Administration error results from systematic literature review on medication errors

Fourteen studies reported administration errors for children (table 5). Six of these studies were medication-specific with two focused on vaccines and four focused on paracetamol and/or ibuprofen only. Of the three studies which were global in scope, nevertheless, variation among the studies in numerator and denominator definitions and methods of data collection made comparisons difficult. Using the one study that defined “total opportunities for administering errors” as the global denominator, the distributional epidemiology of administration errors shows that the majority of these errors involved either dose omissions (42%) or wrong time of administration (50%) (Fontan 2003³⁴).

MAR/documentation error results from systematic literature review on medication errors

Seven studies evaluated documentation errors among children (table 6). The estimate of transcription errors from these studies varies from <1% of orders to 20% of orders having a transcription error.

Monitoring the patient for effect error results from systematic literature review on medication errors

Only one study, listed in table 7, reported monitoring a patient for effect errors and estimated, via chart review, that the incidence was four errors per 1000 patients (Kaushal 2001⁴²).

Distributional epidemiology of medication error from error reports

Four studies provided data that can be synthesised to understand the distributional epidemiology of medication errors in paediatrics based on error reports (France 2004²⁴; King 2003³³; Lesar 2002³⁶; Frey 2002³⁸). Such syntheses are difficult because each study location undoubtedly has different safety culture climates. The safety culture will clearly influence who completes error reports, how often they complete error reports, and what types of event are reported. Little is known about how bias in reporting influences the distributional epidemiology of medication errors. Two of these studies provided data on all medications relative to prescribing, dispensing, administering, and documentation errors (King 2003;³³ Frey 2002³⁸). Between these two studies, the distributional epidemiological estimates of the relative percentages of error types are: prescribing 3–37%, dispensing 5–58%, administering 72–75%, and documentation 17–21%.

Estimates of the severity of medication errors for patients

Only 11 studies categorised medication errors by severity of outcome for the patient (Simpson 2004²³; France 2004²⁴; Cimino 2004²⁸; Sangtawesin 2003³¹; Holdsworth 2003³⁵; Frey 2002³⁸; Kozer 2002³⁹; Kaushal 2001⁴²; Marino 2000⁴³; Upperman 2005⁴⁷; Cowley 2001⁴⁸). Among these studies, however, at least four different scales were used to rank error severity from scales with two categories to scales with nine categories.

Synthesis of recommendations to reduce medication errors for children

We identified a total of 26 unique recommendations for strategies to reduce medication errors from national entities. The recommendations ranged from equipment/software tools, representation of personnel on groups making decisions on paediatric medications, training and competency of personnel, policies, clear labelling, continuous quality improvement efforts, clear and accurate documentation, standardisation, patient education, and teamwork improvement. Table 8 summarises these recommendations and the paediatric specific evidence behind. In short, none of these recommendations was based on published evidence of effectiveness in children. The vast majority of recommendations were based on expert opinion (n = 22), with the remainder being based on studies in adult populations (n = 4). No recommendation had supporting paediatric specific evidence on efficacy, cost effectiveness, feasibility, appropriateness in different settings, and institutional barriers or risks.

CONCLUSIONS

Since the Institute of Medicine To Err Is Human report¹ a significant amount of research has been done on medication errors in children, and a significant number of recommendations have been made by various entities on how to make medication administration safer for children. There can be no doubt, based on this evidence, that medication errors are a significant percentage of medical errors in children. Our review estimates that 5–27% of medication orders for children contain an error somewhere along the spectrum of the entire delivery process involving prescribing, dispensing, and administering based on three studies (Cimino 2004²⁸; Kaushal 2001⁴²; Marino 2000⁴³). Our review also estimates that there are 100–400 prescribing errors per 1000 patients and highlights that the majority of research to date has focused on the prescribing step of medication delivery (Kozer 2002³⁹; Kaushal 2001⁴²).

Looking at error reporting systems, it is clear that each step of the medication process is error prone, although the majority of research has focused on prescribing errors. Our evidence based estimates at the overall “share of the pie” that each step contributes to the overall rate of medication errors among children are the following: prescribing 3–37%, dispensing 5–58%, administering 72–75%, and documentation 17–21% (King 2003³³; Frey 2002³⁸).

Overall, our depth of understanding of the epidemiology of paediatric medication errors remains poor. Our systematic literature review on medication errors highlights the fact that estimates of the incidence of medication errors in children are severely hampered by the lack of uniform definitions of medication errors (numerator data) and study population (denominator) among studies and by the different means of data collection used to identify errors.

Also of importance, many studies did not explicitly define medication errors. A recently published report looking only at prescribing errors in children highlighted the great difficulty in defining what a medication error is.⁵⁴ Barriers to defining medication errors in children and to then being able to measure the epidemiology of medication errors include: off-label use of medications with dosage ranges extrapolated from adult literature, different recommendations for dosing ranges for the same medication from different sources, and unclear rules as to when adult doses may be appropriate for children. None of the studies looking at all medications with details on prescribing errors stated what the “correct” dosing range was that guided their definitions and data collection.

Focusing on the source of data, the vast majority of studies evaluated in this report relied on either chart review or error reports, with a handful using administrative or registry data. Although each mode of identifying medication errors has strengths and weaknesses and will produce varying results, it seems likely that an ideal error identification system may involve multiple data sources and potentially include triangulation between administrative data, chart review, and voluntary error reports of critical incidents in order to maximise the ability to identify events at each step of the process. Taken alone, each data source has significant limitations for defining the epidemiology of medication errors. Administrative data, as analysed here, are inexpensive, nearly universal, and permit unsolicited identification of potential events, although the depth of clinical information is limited. Chart review, on the other hand, provides in-depth clinical information but is fairly expensive to implement on a large scale and is limited by what is documented in the chart. Lastly, voluntary critical incident reporting depends completely on the compliance of providers with reporting but does provide real time in-depth clinical insights. Interdigitated use of these types of data collection would create a system less likely to produce a biased estimate of the epidemiology.

These significant limitations of the examined studies—namely, differing definitions of the numerators and denominators, lack of consistent definition of medication error, less robust and/or narrowly focused methodologies, and the aforementioned short time frames of data collection and single institutional experiences in these studies, make it very difficult, if not impossible, to generalise easily the findings to all healthcare settings. Indeed the vast ranges on some of the results for estimates of different types of medication errors bear testament to this difficulty.

Despite the limitations of the available literature, several key findings warrant discussion and suggest a further national agenda for medication errors in children.

First, standardisation of recommended doses for children is an essential step to enable providers, researchers, and developers of technological solutions for prescribing to speak a common and uniform language on what doses are acceptable and what doses are in error for children. For example, a recent review exploring the limitations of recommended doses for children found a nearly twofold difference in recommended doses of oxycodone, a narcotic, among three widely used references while a fourth reference simply listed no weight-based dose recommendation.⁵⁴ In a recently published study on paediatric ambulatory medication errors, one key finding was that no fewer error rates occurred at the one of three sites evaluated that used an electronic prescription writer.⁵⁵ This last finding was probably due to the absence of paediatric-specific dosing logic in the electronic prescription writer. Despite the push for computerised order entry and prescribing, the lack of uniform agreement on standard paediatric doses is at least part of the reason for the usual absence of paediatric-specific dosing tables powering most commercially available computerised order entry tools. Without standard paediatric doses, and requirements that these dosage rules are built into computerised prescribing tools, children will fail to reap the benefit of information technology in the medication delivery process.

Second, standardisation of definitions of medication errors is a clear need at hand. As examples of this problem based on the studies examined here, the range of definitions of medication errors included medications prescribed at >10% of the recommended dose all the way up to medications prescribed at 10-fold the recommended dose. The vast majority of the articles simply did not describe the details of the definition of a medication error that was used. Comparably, looking at the entire medication delivery system, some articles did not include errors that were detected before they reached the patient, whereas other articles counted these events as errors. This ambiguity about what exactly is a medication error also permits a wide range of severity of errors to be lumped together. For example, some of the studies counted as medication errors orders that were lacking a prescriber’s signature. Although this is clearly an error, the magnitude of potential harm to patients is substantially different from that of orders with dosage errors. Without standardised guidance, all these vastly different interpretations of medication errors are lumped together and make elucidation of high priority areas difficult.

Third, despite much work on medication errors in the inpatient setting, our review identified only a handful of research on medication errors in the emergency department, ambulatory care, and home environments. All of these are critical targets for future research.

Fourth, most of the research to date has been skewed on prescribing errors. Our review of error reporting systems’ data clearly shows that the medication process steps of dispensing and administering are as error-prone, if not more so, than prescribing. Understanding the unique risks for children in these two steps is critical in order to understand better which interventions will remedy the risks. Unlike the medication process for adults, these steps for children rely much more heavily on manual compounding of liquid medications and administration to patients who are unable to perform their own medication safety checks. These facts may well make the dispensing and administering of medications more error prone for children than adult patients.

Last, our synthesis of the various medication error reduction strategies recommended by national bodies resoundingly illustrated the lack of paediatric-specific evidence. However, it is inarguable that many items on the list of reduction strategies do not need multiple clinical trials to prove their impact. High cost or high resource problems such as computerised order entry, automated dispensing devices, and use of bar coding for medication administration clearly do need high quality evidence in order to foster use and, perhaps more importantly, need to have paediatric-specific evidence. Many other items, on the other hand, are relatively inexpensive, and many even broach on commonsense based on knowledge of human factors. Strategies in this latter category include: paediatric presence on Formulary committees, appropriate and competent pharmacy personnel and environment, policies on verbal orders, and clear and accurate medication labelling. The lack of paediatric-specific data on these types of recommendations is non-troubling. Perhaps more troubling is the enormous scope of recommendations coming from numerous official bodies. Such a piecemeal recommendation path leaves most providers unclear about which of the recommendations has a greater priority should they be faced with human or monetary resource limitations. National research and efforts to summarise and endorse recommendations could help care givers prioritise safety efforts and ensure that the most promising strategies of those recommended are broadly implemented first.

In summary, our review of the literature on medication errors in children and on medication error reduction strategies highlights without question that we know medication errors occur across the entire spectrum of prescribing, dispensing and administering and are a significant source of concern for paediatric patients. Furthermore, the research also confirms that medication errors are a significant concern across all settings of care, including within the home. There can be no doubt of the need for greater understanding of all the aspects of medication errors discussed here so that effective interventions and policy can be crafted. This desired understanding, however, needs a firm foundation of standardisation for issues such as dose ranges, definitions of medication errors, and even for prioritisation of implementation of medication error reduction strategies.