Nurse Education Today

Nurse Education Today 30 (2010) 85 97 Contents lists available at ScienceDirect Nurse Education Today journal homepage: www.elsevier.com/nedt Do calculation errors by nurses cause medication errors in clinical practice? A literature review Kerri Wright * University of Greenwich, Avery Hill Road, Eltham SE9 2UG, United Kingdom article info abstract Article history: Accepted 15 June 2009 Keywords: Drug calculation errors Medication errors Administration of medicine This review aims to examine the literature available to ascertain whether medication errors in clinical practice are the result of nurses miscalculating drug dosages. The research studies highlighting poor calculation skills of nurses and student nurses have been tested using written drug calculation tests in formal classroom settings [Kapborg, I., 1994. Calculation and of drug dosage by Swedish nurses, student nurses and physicians. International Journal for Quality in Health Care 6(4): 389 395; Hutton, M., 1998. Nursing Mathematics: the importance of application Nursing Standard 13(11): 35 38; Weeks, K., Lynne, P., Torrance, C., 2000. Written drug dosage errors made by students: the threat to clinical effectiveness and the need for a new approach. Clinical Effectiveness in Nursing 4, 20 29]; Wright, K., 2004. Investigation to find strategies to improve student nurses maths skills. British Journal Nursing 13(21) 1280 1287; Wright, K., 2005. An exploration into the most effective way to teach drug calculation skills to nursing students. Nurse Education Today 25, 430 436], but there have been no reviews of the literature on medication errors in practice that specifically look to see whether the medication errors are caused by nurses poor calculation skills. The databases Medline, CINAHL, British Nursing Index (BNI), Journal of American Medical Association (JAMA) and Archives and Cochrane reviews were searched for research studies or systematic reviews which reported on the incidence or causes of drug errors in clinical practice. In total 33 articles met the criteria for this review. There were no studies that examined nurses drug calculation errors in practice. As a result studies and systematic reviews that investigated the types and causes of drug errors were examined to establish whether miscalculations by nurses were the causes of errors. The review found insufficient evidence to suggest that medication errors are caused by nurses poor calculation skills. Of the 33 studies reviewed only five articles specifically recorded information relating to calculation errors and only two of these detected errors using the direct observational approach. The literature suggests that there are other more pressing aspects of nurses preparation and of medications which are contributing to medication errors in practice that require more urgent attention and calls into question the current focus on calculation and numeracy skills of pre registration and qualified nurses (NMC 2008). However, more research is required into the calculation errors in practice. In particular there is a need for a direct observational study on paediatric nurses as there are presently none examining this area of practice. Ó 2009 Elsevier Ltd. All rights reserved. Introduction Medication errors have been a focus in healthcare since reports both in United States and the United Kingdom (UK) have highlighted the large number of errors occurring in hospitals (Institute of Medicine (IOM), 1999; Department of Health (DH), 2000). There has also been a recent focus on numeracy as a result of the Moser report (Department for Education and Skills (DfES, 1999) which identified that in the UK, seven million adults are unable to read and write at an 11 year old level, with even more lacking basic * Tel.: +44 020 8331 8965. E-mail address: k.wright@gre.ac.uk numeracy skills. This report led to the government s Skills for Life Framework (DfES, 2001) and raised awareness of this skills deficit in the workplace. In nursing internationally there has been growing concern regarding the numeracy skills of nurses in relation to drug calculations, fuelled by high profile research studies such as Gladstone (1995) and a proliferation of research studies examining the calculation skills of nurses and student nurses (Bliss-Holtz, 1994, Kapborg, 1994; Hutton, 1998; Weeks et al.,. 2000; Wilson, 2003; Wright, 2005, 2006, 2007). The Nursing and Midwifery (NMC) published new standards for pre-registration competencies in 2004 and included a competency relating to use of literacy and numeracy in nursing practice (NMC, 2004). Concerns regarding nurses 0260-6917/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.nedt.2009.06.009

86 K. Wright / Nurse Education Today 30 (2010) 85 97 Table 1 Medline, BNI and CINAHL. Search term No. of articles found Amended search term Medication error (title) 235 Medication error (title) AND nursing (all text) Drug error (title) 44 Drug error (title) AND nursing (text) and calculations (text) Drug calculation error (all text) 1 Adverse drug events (title) 350 Adverse drug events (title) AND (all text) Adverse drug events (Title) AND 3 calculations (all text) Adverse drug events (title) AND 3 nursing (all text) No. of articles found Amended search term 113 Medication error (title) AND nursing (all text) AND calculations (all text) 4 109 Adverse drug events (title) AND (all text) AND nursing (text) No. articles found 18 17 calculation skills have continued with high profile media reports (Hall, 2006; The Daily Mail, 2006) and further research studies relating to nurses mathematical skills and have cumulated in the NMC releasing the Essential Skills Clusters (NMC, 2007) stipulating agreed standards of numeracy, literacy and calculations skills before student nurses are deemed competent to be registered nurses. In addition to this in the UK many local National Health Service (NHS) trusts and Primary Care Trusts (PCT) have begun to test the calculation and numeracy skills of their registered nurses and many now require newly qualified nurses to complete a numeracy and literacy test when applying for jobs. The research studies highlighting poor calculation skills of nurses, along with reports of medication errors in practice, appear to have fuelled the concern regarding calculation skills and led to an assumption that increased errors in practice are a result of nurses poor calculation skills (Wright, 2009). The solution to reducing medication errors in practice seems to be focussed on improving the calculation skills of nurses. The research studies highlighting poor calculation skills of nurses and student nurses have used written drug calculation tests in formal classroom settings (Kapborg, 1994; Hutton, 1998; Weeks et al.,. 2000, Wright 2004, 2005). There have been no reviews of the literature on medication errors in practice that specifically look to see whether the medication errors are caused by nurses poor calculation skills. This review aims to examine the literature available to ascertain whether medication errors in clinical practice are the result of nurses miscalculating drug dosages. Method The databases Medline, CINAHL, British Nursing Index (BNI), Journal of American Medical Association (JAMA) and Archives and Cochrane reviews were searched for research studies or systematic reviews which reported on the incidence or causes of drug errors in clinical practice. The studies needed to either focus on Table 2 Journal American Medical Association (JAMA) and Archives. Search terms Medication error 31 Drug error 6 Calculation error 415 Table 3 Cochrane Reviews. Search terms Medication error (title, abstract or keywords) 7 No. of articles No. of articles nurses or capture nurses practice through a whole system approach. The search terms used can be seen in Tables 1 3. Limits for both searches were articles published from and including 1999. The year 1999 was when the Institute of Medicine Report was published in United States which reported on incidence of adverse drug events and in the United Kingdom (UK) the Department of Health (2000) An organisation with a memory, report of an Expert Group on Learning from Adverse Events in the NHS was published. Studies published after this date were felt to be more directly relevant to current health care policies in relation to adverse drug events. The search was also limited to English language, but studies in other countries were included to compare incidence of the problem and gain insights into common difficulties with drug calculations in practice. All highlighted article titles in bold were initially scanned (see Tables 1 3). Articles that were clearly not research studies and commentary only; those focussing on disciplines other than nursing; studies which focussed on evaluation of strategies solely; and studies which were not based in practice, but simulations were discarded. Articles which reported specifically on drug/medication errors and their causes and those which investigated the whole medication system, due to their potential to capture data relating to nursing, were included. In total 325 article titles from Medline, BNI and CINAHL, 452 from JAMA and Archives and seven from Cochrane were scanned and compared to the criteria. Those that clearly did not meet the criteria for the literature review were discarded. This reduced the articles to 54 (Medline, BNI, CINAHL), 8 (JAMA) and 2 (Cochrane). The abstracts of these 64 articles were then reviewed. Using the same criteria as above the articles were reduced to 20 (Medline), 4 (JAMA) and 1 (Cochrane). These articles were then obtained and read in depth to ensure they met the criteria. In addition a manual search of previously obtained articles on this subject area yielded a further 10 articles which did not appear through the initial literature searches. The reference lists of articles obtained were also reviewed yielding a further 19. Of these an abstract review allowed a further 4 to be discarded. The remaining 15 were obtained for detailed review (see Fig. 1 for details of search). In total 33 articles met the criteria for this review. Findings There were no studies that solely examined nurses drug calculation errors in practice. As a result studies and systematic reviews that investigated the types and causes of drug errors were examined to establish whether miscalculations by nurses were the causes of errors. Four systematic reviews met the criteria for this review. Three focussed on reviewing literature relating to the whole medication

K. Wright / Nurse Education Today 30 (2010) 85 97 87 Medline, BNI and Cinahl 325 JAMA and Archives 452 Cochrane 7 Reviews of reference lists 19 Manual search 10 Title review 784 54 8 2 Abstract review 83 20 4 1 15 Article review 50 33 Fig. 1. Summary of literature search and numbers of articles obtained and reviewed at each stage. system and thus did not offer any new insights into calculation errors (Kanjanarat et al., 2003; Wong et al., 2004; Miller et al., 2007). The third reviewed literature specifically related to nurses and drug errors to ascertain factors that contribution to medication errors (O Shea, 1999). This review included information relating to mathematical skills of nurses as a factor contributing to medication errors. The studies reviewed and cited however, involved the testing of mathematical and calculation skills of both student nurses and nurses using written tests in formal settings. None of the studies identified actually tested calculation skills in the clinical setting or reviewed errors in practice that could be attributed to calculation errors. The validity of written tests as a measure of calculation skills in clinical practice has been questioned and the risk of assuming that poor test results indicate poor clinical practice has also been highlighted in literature (Wright, 2007, 2009). The review concluded that mathematical skills of nurses are a contributing factor to medication errors, based on the literature reviewed. Errors across whole medication system Many studies examined the whole medication process in order to detect errors. The whole medication process involved the medication ordered by the physician, through the whole hospital process, to the of this medication to the patient. The studies reviewed were undertaken across eight different countries, many of which had different medication systems in place (see table 4). For example, in the United States once the physician has ordered a medication this is then transcribed, either by a nurse or secretary, into the patient s medical chart for (for example Kaushal et al., 2001, Marino et al., 2000; Anselmi et al., 2007; Tissot et al., 2003). In some hospitals nurses collect medication already dispensed for the patient from a pharmacy (Greengold et al., 2003) and in others, such as in most UK hospitals, the nurse dispenses the medication from stock and then administers this (for example Bruce and Wong, 2001; Han et al., 2005; Taxis and Barber, 2003). This made it difficult to ascertain information relating specifically to the nurses role and also to compare studies examining the whole medication process. The studies and descriptions of the medication system in use within the hospital being studied needed to be read carefully to ascertain which part of the medication system was the main responsibility of the nurses. In most studies this involved the preparation, and dispensing stages. In order to detect calculation errors by nurses, therefore, the and dispensing stages of the medication process were reviewed. Prospective chart review Studies which examined the whole medication process used either a prospective chart review methodology, where all medication charts and documentation relating to the medication process was reviewed for errors, or a retrospective review of error reports (see table 5). The methodology used resulted in a focus on specific areas of the medication process and could explain why some studies found increased errors in particular areas compared to others. Prospective chart reviews involved a trained observer examining all the charts and documentation relating to the medication process. This often involved a research pharmacist, nurse or physician examining the process, beginning with the medication order through to the medication charts. Nurses are most often involved in the dispensing and of medication using the information written on the medication charts. Any errors as a result of calculations would be as a result of medication dosages being administered that were different from those prescribed on the medication chart. These errors would not be documented and would therefore be difficult to detect using this methodology. Some studies did examine the progress notes which detailed the general observations and patient s progress by the nurses. Details of sudden changes in a patient s condition were examined in these studies to ascertain if this was the result of drug errors and the documentation tracked back to identify the possible cause. Although no studies explicitly discuss this process in detail, it is possible that some errors in drug by nurses could be detected using this method.

Table 4 Research summaries. Authors Country Methodology Adult/ Paediatric Allen and Jollis Anselmi et al. (2007) Barker et al. (2002) Speciality Population Sample Area of medication process studied USA (United States of USA) Chart review Adult Cardiology 1 hospital 54 months, 14,983 pharmacy interventions Brazil Direct observation Adult Various 3 hospitals 35 days 1391 doses observed USA Direct observation Adult Accredited (6), nonaccredited (6) and skilled nursing facilities (6) 36 hospitals 18 wards, 3216 dosages Preparation and Error rate All 19 errors per 100 admissions Preparation and 2.4 9.3% error 19% (n = 605) Bates (1999) USA Not stated Adult General medical wards 2 hospitals 247 ADEs, 70 Preventable ADEs, 194 PADEs All N = 264 preventable ADE Bruce and Wong (2001) United Kingdom (UK) Direct observation Adult Acute admissions ward 1 hospital 1 ward, 107 dosages Preparation and (Parenteral 25.2% n.b defs of error Calabrese et al. (2001) Cousins et al. Cowley et al. (2001) Fanikos et al. (2007) Greengold et al. Han et al. Herout et al. (2004) Hicks and Becker (2006) Holdsworth et al. Kaushal et al. (2001) Kopp et al. (2006) Lisby et al. Marino et al. (2000) Ross et al. (2000) USA Observational evaluation Adult ICU 5 ICU 851 patients over 3 months. 5744 observations. UK, Germany and France USA Direct Observation Adult Various; including general, surgical and ITU Restrospective search using MER and MedMARx systems Paediatric All All USAn hospitals using US MER and MedMARx reporting system USA Retrospective review of ADEs reports Adult Cardiac and cardiac surgery USA Comparison of error rate between dedicated medication nurses (MN) and general nurses (GN) using direct observation Adult Medical and surgical wards 3 hospitals 6 departments, 824 dosages prepared and 798 administered 1 hospital (71 beds cardiac, 51 cardiac surgery) 3818 records (MER); 333 paediatric. 43,287 (MedMarx);1969 paediatric 547 ADEs (274 adverse drug reactions, 273 medication errors) 2 hospitals 5792 opportunities for error (MN), 3661 (GN) Australia Direct observation Adult Surgical wards 1 hospital 3 wards. 687 dosages observed USA Observational Adult Surgical ICU 1 hospital 1 month study period. 206 infusions USA Retrospective review of error reports Adult and Various Various Error reports from 1/1/ Paediatric 2000 to 31/12/2004 n = 73, 769 errors relating to IV USA Prospective review of medical records and staff interviews Paediatrics General and ICU 1 hospital 1 ward and 1 ICU. 1197 patient admissions over 8 months USA Prospective review of charts and verbal Paediatric all specialities 2 hospitals 10,778 medication orders All reports including, ICU, NICU USA Direct observation Adult ICU 1 hospital 185 incidents evaluated All Denmark USA UK Direct Observation, Chart reviews and unannounced control visits Medication orders reviewed from prescribing through to patient Retrospective review of standard errorreporting forms Adult Medical and Surgical 1 hospital 2 wards. 2467 opportunities for error Paediatric ICU, medical and 1 hospital 3312 medication orders surgical written, 11,978 doses passed medication only) All 3.3% errors (n = 187) Preparation and Not calculated n.b defs of errors (intravenous medication only) All 9% (n = 333) MER 5% (n = 1969) MedMARx All Administration Administration (Intravenous infusions only) All (continuous IV Infusions only) All Not calculated 11.2% MN 6.9% GN 18% 105.9 per 1000 patient days. Not reported All ADEs = 6/100 admissions PADE 8/100 admissions All plus discharge summaries All 43% errors 784 medication errors Paediatric All specialities 1 hospital 195 error reports All Not calculated 88 K. Wright / Nurse Education Today 30 (2010) 85 97

K. Wright / Nurse Education Today 30 (2010) 85 97 89 All 35.4% errors (n = 322) Thailand Restrospective review of error reports Paediatric All 1 hospital 32,105 hospital admissions Sangtawesin et al. All All 500 facilities 154,816 error reports All Potential errors 9.7% Actual errors 91.3% USA Retrospective review of error reports submitted to MedMARx Santell et al. USA Retrospective error reports Paediatric Emergency department 1 department 33 reports All Not calculated Selbst et al. (1999) 249 errors Preparation and (Intravenous only) UK Direct observation Adult Various 2 hospitals 10 wards. 430 observations Taxis and Barber 48% (n =58 errors). 1 hospital 2 wards. 121 observations Preparation and Germany Direct Observation Adult Surgical ward and Surgical Intensive care Taxis and Barber (2004) 6.6% (n = 132 errors) France Direct Observation Adult Intensive care 1 hospital 3 units, 2009 observations Preparation and Tissot et al. (1999) 14.9% Preparation and 1 hospital 2 wards. 523 opportunities for errors France Direct observation Adult Older people and Cardiothoracic Tissot et al. 44.6% errors (n = 104) Preparation and Netherlands Direct observation Adult ICU 2 hospitals 2 departments. 233 opportunities for errors Van den Bemt et al. (2002) Not calculated Preparation and USA Prospective error reports Adult Various 1 hospital 8 clinicians. 321 medication error reports Winterstein et al. (2004) 88 preparation errors (26%) 93 errors (34%) Preparation and (Intravenous only) 337 drug preparations observed, 278 drug s Direct observation Adult Not specified 3 hospitals (1 UK, 2 Germany) UK and Germany Wirtz et al. Studies using the chart review methodology may therefore be detecting large numbers of errors in a particular part of the process, not because there are more errors in this process, but because the methodology lends itself to detecting these errors more readily than others. For example, many studies identify that prescribing errors are the most common cause of drug errors compared to errors (Allen and Jollis, 2003; Kaushal et al., 2001; Kanjanarat et al., 2003). An example of this is a study by Kaushal et al. (2001) who carried out a prospective chart review of all paediatric patients admitted to two hospitals during a six week period to identify rates of medication errors and adverse events (ADEs) or potential ADEs. The study identified that 74% (n = 454) errors were due to physician ordering and 13% (n = 78) errors were due to nurse. The large number of errors detected in the prescribing stage could be a result of the methodology being more readily able to detect these errors through review of medication documentation and less able to detect errors. In addition the use of a pharmacist in some studies could result in more errors in the prescribing stage being detected due to their pharmacology expertise compared to studies which used nurses as observers. This was especially noticeable in Allen and Jollis study where a specialist cardiology pharmacist was used to identify cardiology errors on cardiology wards and found much larger numbers of errors in this stage (Allen and Jollis, 2003). Studies using a chart review methodology which detect small numbers of errors in the nurse administrative stage may not therefore be a valid representation of the actual errors occurring in practice and is likely to be under reporting these errors. Retrospective review of errors In contrast to the chart review and a possible bias on detecting errors in the prescribing stage the retrospective review of error reports arguably has a bias towards the stage. Studies using a retrospective review of error reports, review the errors already reported either through local hospital reporting systems (Fanikos et al., 2007; Ross et al., 2000; Winterstein et al., 2004) or through national reporting systems such as the United Statesn MedMARx system in United States (Cowley et al., 2001; Hicks and Becker 2006; Santell et al., 2003). Some studies reviewed very large numbers of error reports (Cowley et al., 2001; Santell et al., 2003) to ascertain causes of error, increasing the validity of the causes and types of errors found. However, there are difficulties with using error reports. Reports rely on the practitioner involved in the error firstly recognising that there has been an error and secondly reporting this error. The errors reported are therefore not necessary a valid representation of the errors that occur in practice (Ross et al., 2000). For example, as already discussed, many studies that use chart review of the medication process by pharmacists have identified that a large number of prescriptions are made in error, frequently the wrong dose for the health status of that particular patient (Allen and Jollis, 2007, Winterstein et al., 2004) and often caused by insufficient knowledge by the physician involved (Bates, 1999, Hicks and Becker, 2006). Unless these prescriptions are reviewed by a pharmacist or the error actually caused the patient harm, and this was attributed to the prescribing error, then it is unlikely that the physician themselves will identify and report the error. Thus leading, potentially, to under reporting by this group. In addition to this, some health care professionals may be more likely to report errors due to professional cultural practices, for example nurses, which could lead to an over representation of errors in the stage of medication process they are most involved in (Selbst et al., 1999). Nurses are also most often involved in of medication to the patient and are thus the last check in the medication system before the medication is given to the

90 K. Wright / Nurse Education Today 30 (2010) 85 97 Table 5 Methodology. Methodology Authors Prospective chart review Allen and Jollis, Holdsworth et al., Kaushal et al. (2001), Marino et al. (2000) Retrospective review error reports Bates (1999), Cowley et al. (2001), Fanikos et al. (2007), Hicks and Becker (2006), Ross et al. (2000), Santell et al., Winterstein et al. (2004) Direct observation Anselmi et al. (2007), Barker et al. (2002), Bruce and Wong (2001), Calabrese et al. (2001), Cousins et al., Greengold et al., Han et al., Herout and Erstad (2004), Kopp et al. (2006), Taxis and Barber, Tissot et al., Wirtz et al. Mixed methodology Lisby et al. patient. It is possible that nurses may be reporting errors in the administrative stage that they have identified, which are actually attributable to a physician s prescription or a transcribing error, not the nurses (Selbst et al., 1999). This could account for higher numbers of administrative errors in studies using this methodology as opposed to errors being causes by nurses in this stage (Ross et al., 2000; Santell et al., 2003). Causes of drug errors Even if the limitations of studies examining the whole system are taken into account there are still difficulties with these studies in ascertaining the actual cause of errors being made in practice and the personnel responsible. Due to the whole system approach the studies generally reported the causes of errors, but not the stage of the medication process that these errors occurred in or the personnel involved. This made it very difficult to tease out the role of the nurses in the errors being reported. For example in the Kausal et al. study discussed previously, the most common error detected was wrong dose errors which accounted for 28% (n = 175) of all errors detected. However, because the stage at which these errors occurred or the personnel involved are not detailed, it is not clear whether these errors are due to the physician prescribing the wrong dose or the nurse administering a different dose to that prescribed. Thus it is difficult to identify the extent that nurses are potentially administering the wrong doses as a result of miscalculation. This is similar to other studies reviewed which gave details of types of error detected, but did not specify what personnel was involved (Calabrese et al., 2001; Cowley et al., 2001; Hicks and Becker 2006; Fanikos et al., 2007). As a consequence of the methodological difficulties in studies involving chart review or retrospective error reporting, these studies were not able to give evidence regarding drug calculations causing errors in practice. The exception to this are two retrospective reviews of errors (Ross et al., 2000; Cowley et al., 2001) which reported the errors detected in more detail. The Ross et al., review was carried out on standard error-reporting forms, submitted over a 5 year period in one paediatric hospital, with the aim of determining the incidence and type of medication errors occurring. The review found that 59% of errors were reported by nurses. The Cowley et al. (2001) study reviewed errors reported to a central error-reporting system in the United States over a 1 year period. There are difficulties with these studies however, for example both studies do not specify whether nurses were actually responsible for the errors reported and as discussed previously the large number of error reports by nurses is not necessarily representative of the actual errors occurring in practice. However, the studies are important as they specifically mention drug calculations. The error reports in Ross et al. s study detailed that 8% (n = 15) of incidents involved a 10-fold dosage error, for example administering 0.8 mg instead of 0.08 mg, and attributed five of these errors directly to a miscalculation of a dose, despite clear prescribing. In the Cowley et al. (2001) study the causes of errors were identified as due to calculation errors in 4% of errors (n = 121). The studies are useful as they give some evidence that miscalculations do occur in practice, through physician and nursing errors, as well as highlighting that errors as a result of miscalculations can also have multiple causes, for example due to unclear prescriptions (Ross et al., 2000). This will be discussed in more detail later. Errors in nurse drug The other studies reviewed used a direct observational methodology to detect medication errors. The direct observation approach involved trained observers observing particular aspects of the medication process to detect medication errors. Thus errors detected could be attributed directly to the personnel involved, the stage of the medication process and in most studies some details of the cause of the error ascertained. The observational studies generally focussed on specific medications that nurses were involved in. Only five studies observed nurses preparing and administering all medications for patients. These included, for example oral, topical, injections and intravenous medications. Most of the studies observed the nurses role in preparing and administering intravenous (IV) medications. Some studies were more specific and focussed on continuous intravenous infusions or specific intravenous drugs which were known to be error prone (see Table 6). The study areas examined nurses working in general medical or surgical areas or focussed specifically on more acute areas such as intensive care units (Table 4). All of the observational studies were focussed on adult nursing and no studies looked specifically at paediatric nursing (Table 7). The studies reviewed, which used the direct observational approach, focussed on observing nurses preparing and administering medications directly to the patient. This involved the nurses reviewing the patent s medication chart to gain information regarding the medication; dose strength, dose form, time, route and dosage specified for a particular patient; preparing this medi- Table 6 Medication observed. Medication Authors All medications Barker et al. (2002), Van den Bemt et al. (2002), Lisby et al., Kopp et al. (2006), Tissot et al. Oral medications Greengold et al. Error prone medications Calabrese et al. (2001) All intravenous Anselmi et al. (2007), Bruce and Wong (2001), Cousins et al., Taxis and Barber, Wirtz et al. Intravenous continuous infusions (excluding bolus dosages) Han et al., Herout and Erstad (2004)

K. Wright / Nurse Education Today 30 (2010) 85 97 91 Table 7 Reviewed studies summarised according to adult or paediatric focus. Adult or paediatric Prospective chart review Retrospective error reports Direct observation Mixed methodology Adult Allen and Jollis Bates (1999), Fanikos et al. (2007), Winterstein et al. (2004) Paediatric Holdsworth et al., Kaushal et al. (2001), Marino et al. (2000) Adult and Paediatric Cowley et al. (2001), Ross et al. (2000), Sangtawesin et al., Selbst et al. (1999) Hicks and Becker (2006), Santell et al. Anselmi et al. (2007), Barker et al. (2002), Bruce and Wong (2001);Calabrese et al. (2001), Cousins et al., Greengold et al., Han et al., Herout and erstad (2004), Kopp et al. (2006), Taxis and Barber, Taxis and Barber (2004), Tissot et al. (1999);Tissot et al., Van den Bemt et al. (2002), Wirtz et al. Lisby et al. cation; and administering it to the patient. In order to detect any errors as a result of a miscalculation the study results were examined in detail. No studies had a specific category for calculation error. The categories of errors which could potentially be as a result of miscalculations were examined. These were wrong dosage and wrong rate of. Generally these were either specified in the preparation stage when the medication dosage was being prepared or in the stage when the medication was being given to the patient. However, studies varied in whether they separated errors into preparation and stage errors or collated errors across these two stages together. In addition to this, one study separated their observation into dispensing and stages (Lisby et al., 2005). This meant that it was not always possible to compare studies and error rates across categories. Definitions of drug error One of the main difficulties in ascertaining error rates for wrong dose or wrong rate is the definitions used in the different studies. Some studies used a standard definition of wrong dose error as a discrepancy between the dose administered and the dose prescribed (Anselmi et al., 2007; Barker et al., 2002; Bruce and Wong 2001; Greengold et al., 2003; Han et al., 2005, Lisby et al., 2005). Others expanded on this by detailing definitions of wrong dose for different medication preparations. For example Tissot et al. (1999) used the previous definition for oral medication, but specified an error rate of ±10% for injections and intravenous medications. This means that studies using the +/10% have a bigger margin of error and will thus detect fewer errors than studies using smaller margins of error. In some of the studies focussing on intravenous medication, the error definitions are more expansive and include errors in standard medication procedures and policy guidelines (Cousins et al., 2005; Taxis and Barber 2003; Tissot et al., 2003; Wirtz et al., 2003). For example Taxis and Barber define errors as any medication administered or prepared that deviates from prescription chart, the manufacturer s instructions and deviations from hospital policy (see Table 8). These variations in definitions mean that errors as a result of wrong dose and wrong rate of can be attributed to a wider variety of causes and makes teasing out the role of miscalculations in the errors even more problematic. Wrong dose errors Most studies had an error category for wrong dose or incorrect dose. The details of the rate of errors reported in the studies can be seen in Table 9. From these studies it is apparent that nurses are administering a large number of medications of a different dose to that specified by the physician, constituting a wrong dose error. However, it is not clear whether the wrong dose has been administered because of miscalculations or because of some other cause. For example, in some studies examining IV infusions, wrong dose is caused by nurses not allowing sufficient time for the drug to dissolve in an antibiotic vial when reconstituting and drawing up and therefore injecting a reduced dosage. This was a common error noted in Wirtz et al. study in one particular study site. Other causes identified for wrong dose include illegible prescriptions. This was indicated in the studies by Ross et al. (2000) and Tissot et al.. Some studies did observe wrong dosage errors which were due to discrepancies in the dose administered compared to the dose prescribed. For example Lisby et al. gave details of wrong dosage errors in the dispensing and stages of the medication process and found 17 errors in the dispensing stage of which five were due to wrong dosages being prepared out of 1067 errors and Kopp et al. (2006) identified four wrong dose errors in the dispensing and 20 in the stages out of 132 errors detected. Although wrong dose errors are not necessarily due to miscalculations, these results do give some indication of the scope of errors where miscalculations could be involved. Wrong rate errors A common error observed in most studies was wrong rate in relation to intravenous medications (for example Cousins et al., 2005; Hans et al., 2005; Taxis and Barber 2003; Tissot et al., 2003). Nurses are responsible for ensuring that any intravenous infusion prescribed is set up at the required rate so that a continuous medication is received over the infusion duration. This often requires a calculation, if gravity flow infusion devises are being used, to determine the amount of infusion required for a specific period, usually per minute, or calculations to determine the rate to set technological devises at, such as syringe drivers or volumetric pumps. Also included in this error category, in studies with a broader definition on drug errors, are bolus intravenous medications. Bolus medications are administered usually using a syringe or through adding the medication to a small infusion bag to administer intravenously. The rate of bolus infusions is not specified by the prescription, but by manufacturer s guidelines and hospital policies. As a consequence errors in bolus s are not a result of calculation errors, but non-adherence to medication guidance and policies. When wrong errors are examined therefore, the studies needed careful scrutiny to ascertain how many of the errors potentially could be attributable to miscalculations. For those studies with a wider definition of drug errors which included hospital policy and manufacturers guidelines most errors in bolus dose s were as a result of these being given faster than manufacturers recommendations and hospital policy, not due to miscalculations (Taxis and Barber, 2003; Tissot et al., 2003; Wirtz et al., 2003). Studies examining continuous infusions identified that these were being administered slower than prescribed by

Table 8 Details of most common error categories and definition of errors for direct observational studies. Author Definition of error Most common errors Related to calculation Notes Anselmi et al. (2007) Barker et al. (2002) Bruce and Wong (2001) Calabrese et al. (2001) Cousins et al. Greengold et al. Han et al. Herout et al. (2004) Kopp et al. (2006) Lisby et al. Taxis and Barber 2003 Dose prepared and/or administered by nursing personnel different from that prescribed by the physician and on the patient s records A dose administered differently than as ordered on the patient s medical record A dose of medication that deviated from the physician s order as written on the patient s chart. A medication error was defined as any preventable event that may cause or lead to inappropriate medication use or patent harm whilst the medication is in the control of the health care professional, patient or consumer A deviation in the preparation or of a medicine from a doctor s prescription, hospital intravenous procedures or the manufacturer s instructions Discrepancies based on observers record vis a vis the physicians orders Medication errors (MAE)=Deviation from the prescribers legal prescription Chart inconsistencies = any variation (including omissions) between the recorded information on the flow chart and what was actually infusing into the patient. Non weight based infusions = difference in dose recorded or prescribed versus the dose infusing Weight based infusions = >5% difference between dose infusing and the dose prescribed or recorded. A medication error was an error occurring during the medication use process, regardless of whether an injury occurred or the potential for injury was present. Dispensed = Dispensed medication is concordant with prescribed drug in nurse medication chart Administration = the right medication to the right patient in the right way and at the right time Deviation in the preparation or of a drug from a doctor s prescription, the hospitals intravenous policy or the manufacturer s instructions First Second P = Omission dose 2.9 A = Omission (0.8 3.4%) Wrong dose No further detail of causes of wrong dose errors given 11% A = Wrong dose 1.3 5.7% P = wrong dose (0.9 7.4%) Omission 6% n = 183 Wrong dose 3% n = 103 Wrong dose Any dose of preformed dosage units (e.g. Tablets) that contained the wrong strength or number. If injectable than any dose that was ±10% or different from the correct dosage. If any other dosage form, than any dose that was ±17% Wrong time errors n = 5 Wrong preparation No errors related to dosages or (out of 107 technique n =3 rate opportunities for error) Dose omission 14.4% n =27 UK wrong rate 48% n = 132) Germany (GER) labelling error 99%, n = 421 France (FRA) labelling error 20%, n = 20) Administration technique mean 6.4% (n = 607) Wrong rate 79.3% n = 100 (of all MAE errors) Chart inconsistencies: Drug omitted from flow sheet 58% (n = 15/26) Dosing errors Weight based 10.3% n = 9/87) Improper dose 11.7% (n = 22) UK labelling error 43% (n = 118) Germany WRONG diluents 49%, n = 208 France Wrong diluents 18%, n =18 Dose preparation 1.4% mean, n = 134 Dose omission 11.9% n = 15 (of all MAE errors) Chart inconsistencies: Incorrect or no concentration reported 19% (n = 5/26) Dosing errors non weight based 5.9% (n = 7/119) Improper dose and wrong infusion rate 40.1% n = 75 Wrong strength/concentration 2.3% n =4 Wrong dose or infusion volume: UK = 1% (n = 1); GER = 2% (n = 7); FRA = 5% (n =5) Wrong admin rate: UK = 48% (n = 132); GER 21% (n = 34); FRA 5% (n =5) Dose 0.8% mean, n = 73 Wrong rate 100/687observations Dosing errors in 7.8% of infusions (n = 16/206) Incorrect infusion rate 15% (4/ 26 chart inconsistencies) Omission 23% n = 30 Wrong dose 20% n = 26 In the stage of medication process wrong dose 12% (n = 5/42) D = Omission of dose (7 out of 17 errors) A = lack of identity control (150 out of 166 errors) P = Errors in solvent/ diluents 8% of 430 observations (n = 36) A = fast bolus dose peripheral line 30% (n = 127) D = wrong dose and unordered drug (both 5 out of 17 errors) A = wrong time (18 out of 166 errors) P = wrong dose and omission both 3% of 430 observations (n = 12) A = fast bolus central line 8% (n = 36) Wrong dose in dispensing stage = 5 out of 419 opportunities Wrong dose in preparation stage 3% Three specific errors mentioned which were caused by miscalculation This referred to medication left unlabelled prior to administering This referred to the of medication at a faster rate than recommended No further details re: cause of dosage errors given Caused by free flow IV infusions running too slowly mostly through nurses not evaluating and regulating rate throughout infusion period. No details given about dosing errors being the result of prescribing errors or nursing errors. Proximal causes of errors in stage were 40% n = 17 slips and memory lapses and 26% n =11 rule violations as most frequently reports causes of errors. Definitions wrong dose: to the patient of a medication of a dose that is greater than, or less than the amount prescribed (±10%) 92 K. Wright / Nurse Education Today 30 (2010) 85 97

K. Wright / Nurse Education Today 30 (2010) 85 97 93 Dose error 7.2% (n = 41) Wrong dose 38 out of 568 observations Definitions given for each error type only Physiochemical incompatibility 18.6% (n = 19) Tissot et al., 1999 Definitions given for each error type only Wrong time 26%, n = 20 Wrong rate 19%, n = 15 Wrong dose 12% (n = 9) Definitions wrong dose: to the patient of a medication of a dose that is greater than, or less than the amount prescribed (±10%) Wrong rate 26% (n = 20) Wrong rate: inappropriate rate of by intravenous route of a medication to the patient, whatever the technique (direct intravenous, Tissot et al. perfusion by gravity or infusion) No further details given about causes of this error Wrong dose error: Hospital 1 = 8.8% (n =5) Hospital 1 = wrong technique 38.6% (n = 22) Hospital 1 = wrong time error 38.6% (n = 22) Any error in the preparation and of drugs by nurses: deviation from written, printed or verbal medication orders; a deviation from the drug information sheets provided by the manufacturer or from the information in a handbook on injectable drugs or from general nursing procedures used in the hospitals Van den Bemt et al. (2002) Hospital 2 = 5.4% (n =4) Hospital 2 = wrong dose preparation 21.6% (n = 16) Hospital 2 = wrong time error 40.5% (n = 30) Wrong dose mainly due to undissolved drug remaining in the vial Calculation errors 12% n = 6 (out of 51 preparations where a calculation was required) P = omission (1 20% across 3 study sites) P = wrong dose 10% n = 34 of 337 observations Any deviation of preparation or of an IV dose from the original prescription or any act in the preparation or which deviated from the manufacturer s instructions or hospital drug policy Wirtz et al. A = compatibility errors 10% n = 20 out of 209 doses A = wrong rate error 88% n = 73 (out of 83 injections) nurses (Bruce and Wong, 2001; Han et al., 2005). On further examination however, the slow rates were not as a result of miscalculations when the infusion was being set up, but through not monitoring gravity infusions and not readjusting the rate to take account of changes in gravity as the patient altered their position. However, some studies do indicate that some infusion rate errors are attributable to calculating the wrong infusion rates. Calabrese et al. (2001) observed medication in five intensive care units in United States and found 75 errors out of 187 were due to wrong infusion rates. The study indicates that at least three of these errors were due to miscalculation of the infusion rate required, although no further details are given. Although it is difficult to quantify error rates that are directly attributable to miscalculations, there is some indication from the studies undertaken therefore that miscalculations are causing a number of infusions to be set up in error. Drug calculation errors Only five studies actually mention miscalculations as a cause of medication errors. The first study discusses calculation errors as a cause of errors, but does not quantify or evidence this from the study results (Selbst et al., 1999). The second study is Ross et al. (2000) retrospective review of error reports in a paediatric hospital which mentioned 15 tenfold errors with 5 of these being a result of miscalculations. Unfortunately the study does not specify whether any other errors reported were as a result of miscalculations which were not tenfold errors. The study relies on voluntary error reports. As a result errors are likely to be underreported and the actual number of errors in this hospital could potentially be higher. The number of errors attributable to miscalculations could therefore be more of a problem than evidenced here. This is similar to the third study by Cowley et al. (2001) who reviewed error reports in childcare and found 4% errors (n = 121) caused by calculation errors, although the stage of the medication system that the calculation error occurred in is not stipulated nor the health professional responsible. It is possible therefore that the calculation errors could also be attributable to doctor s prescribing errors not solely errors in nurses. This is consistent with other similar studies that found that tenfold errors were attributable to dosing errors by physicians (Selbst et al., 1999). The fourth study is Wirtz et al. s observational study across three wards at three teaching hospitals, one in Britain and one in Germany. The study observed nurses preparing and administering intravenous medications using a disguised approach over six consecutive days on each ward. The study observed 337 preparations, and identified 88 (26%) errors and 278 s identifying 93 (34%) errors. Of the 337 preparations observed, 51 required a calculation and 6 of these calculations led to dose errors. The fifth study which mentions calculation errors, but does not quantify these in their study was in Calabrese et al. (2001) observational study carried out in five ICU units in United States. This study mentions three miscalculations that resulted in incorrect rates, but does not indicate whether these were the only errors or quantify this cause of error further. Low incidence of calculation errors It is interesting to notes that out of the 15 studies that focussed exclusively on observing nurses preparing and administering medications in a range of settings from general nursing to specialist areas such as intensive care only two of these mention calculations as a contributor or a cause of errors. It is possible that these observational studies did not have the capacity to break down categories, such as wrong dose, into smaller categories into calculation errors and thus the results found only detail the broader categories

94 K. Wright / Nurse Education Today 30 (2010) 85 97 Table 9 Causes of errors identified in the direct observational studies. Causes of errors Statistics Authors and date Incorrect intravenous infusion rate 15.8% (n = 32) Ross et al. (2000) a 19% (n = 15) Tissot et al. 20 27% Wirtz et al. Peripheral line 30% (n = 127) Taxis and Barber 2003 Central line 8% (n = 36) 79.3% (n = 100) of all errors Han et al., 2005 0.2% (n = 19) Greengold et al. 48% (n = 132) Cousins et al., 2005 0% Bruce and Wong 2001 40.1% (n = 75) Calabrese 2001 4/26 chart inconsistencies Herout et al., 2004 6% (n = 29) Tissot et al., 1999 Incorrect dose 14.8% (n = 30) Ross et al. (2000) a 12% (n = 9) Tissot et al., 2003 Preparation 3 21% Wirtz et al. 3% (n = 12) Taxis and Barber 2003 0.8% (n = 73) Greengold et al. 5/17 Lisby et al., 2005 1% (n = 1) Cousins et al., 2005 0% Bruce and Wong 2001 2 7.4% Anselmi et al. (2007) 12% (n =5) Kopp et al. (2006) 11.7% (n = 22) Calabrese (2001) 7.8% (n = 16) Herout et al. (2004) 3% (n = 103) Barker et al. (2002) 38 out of 568 Tissot et al., 1999 Dose omitted 12.3% (n = 25) Ross et al. (2000) a 16% (n = 13) Tissot et al. 1 20% Wirtz et al. 3% (n = 12) Taxis and Barber 11.9% (n = 15) of all errors Han et al. 1/166 errors Lisby et al. 0.9% (n = 81) Greengold et al. 0% Bruce and Wong (2001) 2.9 11% Anselmi et al. (2007) 48% (n = 20) Kopp et al., 2006 14.4% (n = 27) Calabrese (001) 6% (n = 183) Barker et al. (2002) 0% Tissot et al. (1999) Incorrect drug/infusion given 12.3% (n = 25) Ross et al. (2000) a 0% Taxis and Barber 3.2% (n = 4) of all errors Han et al., 2005 0% Cousins et al., 2005 0 0.2% Anselmi et al. (2007) 0% Kopp et al. (2006) 0.6% (n = 1) Calabrese (2001) Unauthorised drug/ Extra dose given 16% (n = 13) Tissot et al. 0 2% Wirtz et al. 0% Taxis and Barber 2003 1/166 errors Lisby et al. 0.1% (n = 7) Greengold et al. 0% Bruce and Wong (2001) 13.8% (n = 28) Ross et al. (2000) a 4% (n = 5) of all errors Han et al. 14% (n =6) Kopp et al. (2006) 1% (n = 22) unauthorised/0%, n = 10 extra dose Barker et al. (2002) 3 out of 568 Tissot et al. (1999) Labelling error/lack of identity control 9.9% (n = 20) Ross et al. (2000) a 43% (n = 118) Cousins et al. 1/27 Bruce and Wong (2001) 150/166 errors Lisby et al., 2005 Incorrect route of 4.4% (n =9) Ross et al. (2000) a 0% Lisby et al. 0.6% (n = 55) Greengold et al. 1% (n = 1) Cousins et al. 0% Bruce and Wong (2001) 0% Kopp et al. (2006) 0% Calabrese (2001) 0% (n = 6) Barker et al. (2002) Incorrect patient 3.9% (n =8) Ross et al. (2000) a 1.6% (n = 2) of all errors Han et al. 0% Anselmi et al. (2007) 0% Calabrese (2001)