Objective To examine the effect of a multifaceted educational intervention on the incidence of medication preparation and administration errors in a neonatal intensive care unit (NICU).
Design Prospective study with a preintervention and postintervention measurement using direct observation.
Setting NICU in a tertiary hospital in the Netherlands.
Intervention A multifaceted educational intervention including teaching and self-study.
Main outcome measures The incidence of medication preparation and administration errors. Clinical importance was assessed by three experts.
Results The incidence of errors decreased from 49% (43–54%) (151 medications with one or more errors of 311 observations) to 31% (87 of 284) (25–36%). Preintervention, 0.3% (0–2%) medications contained severe errors, 26% (21–31%) moderate and 23% (18–28%) minor errors; postintervention, none 0% (0–2%) was severe, 23% (18–28%) moderate and 8% (5–12%) minor. A generalised estimating equations analysis provided an OR of 0.49 (0.29–0.84) for period (p=0.032), (route of administration (p=0.001), observer within period (p=0.036)).
Conclusions The multifaceted educational intervention seemed to have contributed to a significant reduction of the preparation and administration error rate, but other measures are needed to improve medication safety further.
Statistics from Altmetric.com
Pharmacotherapy in the neonatal intensive care unit (NICU) is complex. There is a lack of evidence for the use of a lot of medications in neonates. Dosing has to be individualised based on age, gestational age, weight with or without body surface area. Changes in weight and length, maturation of the enzyme system and renal function require frequent adaption of medication dosages and administration intervals. Furthermore, there is only a limited range of licensed medication available in appropriate dosage forms. Consequently, the process of prescribing, dispensing, preparation and administration of medicines to neonates involves frequent calculations and dilutions of medicines presenting a large number of opportunities for error. What is more, small errors may have serious consequences in this very vulnerable patient group.1,–,4
What is already known on this topic
Medication preparation and administration errors occur frequently in adult intensive care units and possibly also in neonatal intensive care units.
Lack of knowledge of practical procedures and deviations from guidelines were frequent causes of medication preparation and administration errors.
What this study adds
About one in four administered doses was associated with a potentially clinically relevant medication preparation or administration error in neonatal intensive care.
A multifaceted educational intervention seemed to contribute in reducing the overall frequency of errors, but other measures are needed to improve medication safety further.
The incidence and nature of medication errors in the NICU have been reviewed.1 A US-based study4 found an overall rate of 6% medication errors using a combination of chart review and voluntary reporting for data collection. But so far, few studies investigated in depth the process of drug preparation and administration in the neonatal intensive care setting using the ‘gold standard’ of direct observation.5 Two studies using observation methods, one from the USA and one from Malaysia, showed error rates of 20%6 and 31%,7 respectively. Another observation-based study on paediatric patients from the UK reported medication administration error rates of around 15% for the two NICUs.8 The majority of other studies1 ,9 reported much lower error rates using different types of incident-reporting methods. But incident reporting is known for under-reporting.10 Related studies on administration errors in adult intensive care units11 and intravenous medication errors12 showed that preparation and administration of medication is one of the most risky steps in the whole medication management process.
Lack of knowledge of practical procedures, deviations from guidelines (violations) and nurses' experience were shown to be frequent causes of medication preparation and administration errors.13,–,16 This would suggest that educational interventions could be useful to improve medication safety, but this has not been studied extensively. In adult intensive care, simulation-based teaching successfully reduced medication administration errors, whereas no error reduction was observed when the same material was presented in a traditional-style lecture.17 In paediatric patients, a combination of written material, lectures and practical teaching sessions was successful in reducing medication administration errors.18 In NICU patients, the study from Malaysia7 investigated the impact of an educational programme consisting of lectures and educational posters on medication administration errors. Errors in correct time of medication administration could be reduced from 31% to 15% and adherence to follow procedures such as correct documentation and labelling medication could be improved. But clinical significance of the errors was not studied. Similar studies from the USA or Europe carried out in the NICU setting seem absent.
Therefore, we investigated the effect of a multifaceted educational intervention on the incidence and clinical importance of medication preparation and administration errors in a NICU in a tertiary hospital in the Netherlands.
Patients and methods
The study was conducted in a 14-bed NICU in a tertiary hospital in the Netherlands. Patients were admitted from 25 weeks postconceptional age. Medical management of patients was performed by neonatologists, residents and fellows. They rounded each morning on the patients. The majority of decisions were made during this time. Computerised physician order entry was not available in the NICU during the study period. Doctors prescribed the medication in the medical notes and nurses transcribed the orders onto the medication administration records. They prepared the medication using guidelines available on the ward. There was no separate preparation room; medication was prepared on the ward on a dedicated bench. Several commonly used medications were stocked on the ward. The remainder of the medications was dispensed for individual patients daily from the central pharmacy. Other clinical pharmacy services included prescription review (about twice a week) and participation in medical (teaching) rounds (about once a week).
This was a prospective study with a preintervention and postintervention measurement using direct observation.
The intervention was an educational programme consisting of five theoretical teaching session (1 h each, repeated about four times to reach all nurses), one individual practical teaching session for each nurse lasting about 30 min covering the preparation and administration of all commonly on the ward used medication and a short-guided tour around the pharmacy department. The theoretical programme was repeated three times between March 2006 and May 2008. Teaching included calculation, reconstitution, compatibilities, administration rate and aseptic technique of drug preparation and administration. The content of the programme was based on an analysis of the preintervention data. The programme was developed and implemented by a clinical pharmacist responsible for the NICU in cooperation with a multidisciplinary team including neonatologists and nurses specialised in neonatology and infection control. Teaching aids included a PowerPoint presentation, a video presentation, both also available on the hospital's intranet as well as a poster with important recommendations for safe preparation and administration which was placed in the preparation area. Furthermore, during the intervention period, all guidelines outlining drug preparation and administration were updated and available on the ward.
Data for the preintervention period were collected in February 2006 and for the postintervention period in June 2008. To obtain a representative set of medication errors, data were collected for 10 consecutive days in each period for 24 h per day. Four pharmacy students, two per period, used the observation method.19 Before data collection, all four pharmacy students were carefully trained by the experienced clinical pharmacist (IC) about the preparation and administration of all medication used on the ward. This included pilot observations on the study ward to ensure that similar definitions were used. Furthermore, all observations were discussed daily between observers and clinical pharmacist (IC). A medication error was defined as any deviation in preparation or administration of the medication or both from the doctor's prescription, the hospital's intravenous policy or the manufacturer's instructions (table 1). Deviations from aseptic procedures including hand-washing or use of protective gloves were not included as errors. The observers collected details of each medication preparation and administration on a standardised data collection form. Particular attention was paid to the following: drug, dose, choice and volume of solvent/diluent, administration time, technique, route, rate of administration, and physical and chemical compatibilities. Enteral feedings, parenteral nutrition and blood-derived products were excluded from observation. The observer's notes were compared with the original physician prescription, the hospital guidelines, recommendations of the manufacturers and data available in the literature to identify errors. Some basic characteristics (gender, age and weight) were recorded for each included patient.
We informed the ethics committee about our study, but formal approval was not required. The study was approved by ward managers. Permission to observe was obtained from each individual nurse. Nursing personnel were not aware of the objective of this study. We presented the study to staff at ward level as a research project investigating common problems of medication preparation and administration. Observers were instructed to intervene in a discreet and non-judgemental manner if an error could result in substantial patient discomfort or harm. These incidents were still included as medication errors. Personnel requesting medication information from the observers were referred to the clinical pharmacist.
At the time of planning the study, there was little information on medication preparation and administration error rates from NICU setting.1 Based on studies from other areas, we assumed to find an error rate of about 10%.4 ,11 Sample size calculation, using formulae (3.19) and (3.20) from,20 suggested we had to observe 383 drug preparations and administrations in each period to be able to determine an error reduction (one-sided) to 5% (α=0.05; β=0.2).
The error rate was calculated as the number of observations with one or more errors divided by the total number of observations plus dose omissions. One observation was defined as the complete process from preparation to administration of a medicine. Observations were excluded from analysis if only part of this process had been observed, with an exception to continuous infusions. The rate of infusions was checked at regular intervals, but for practical reasons the end of infusions could not always be observed.
The clinical importance of medication errors was determined using an established method which we validated for use in the Netherlands.21 ,22 Briefly, clinical importance of each medication error was assessed independently by three specialists (a pharmacist specialised in paediatrics, a neonatologist and a neonatology nurse, none involved in the data collection process) on a visual analogue scale between 0 (no harm) and 10 (death). The mean score was determined. Scores between 0 and 3 were classified as minor, 3 and 7 as moderate and above 7 as severe errors.
Proportions of medication errors with 95% coincidence intervals (CI), using the formulae (1.26) and (1.7) from,20 were calculated for each period first. Then a univariate approach on the binary outcome variable of all medication errors was conducted to investigate which variables had an effect on the medication error individually. Pearson's χ2-statistic was used for this. The same test statistic was applied to investigate if the proportions of levels of particular variables were the same over both periods and whether patient characteristics between periods were similar. In the third and final statistical analysis, a generalised linear mixed model with logit link function was applied to the binary outcome of all medication errors using ‘type of preparation’, ‘route of administration’, ‘observer within period’ and ‘period’ as explanatory variables. Estimation was performed using generalised estimating equations (GEE) and the cluster variable was taken equal to days within period. Exchangeability was selected for the working correlation matrix. The generalised score statistic was used to test the effects of the individual variables separately after correction of the other variables (type 3 tests). The statistical analyses were performed with SAS, version 9.2, using procedures FREQ and GENMOD.
Twenty patients were admitted during the preintervention period and 22 patients were admitted during the postintervention period. Patient characteristics were similar for both periods (table 2). In total, 311 (43%) doses were observed of 718 administered doses in the preintervention period and 284 (23%) of 1221 doses in the postintervention period. The reason for missing observations was that several nurses prepared and administered medication at the same time and could not all be followed by one observer. The most common types of observed medication classes were anti-infective medication, analgesics and caffeine. The majority of observed medications were administered parenterally (n=242, 78% in the preintervention period and 152, 54% in the postintervention period). Type of preparation and route of administration were significantly different between periods of observation (table 2).
In the preintervention period, we observed an error rate of 49% (151 medications with one or more errors of 311 observations) (43; 54%); 0.3% (0; 2%) medications contained severe errors, 26% (21; 31%) moderate and 23% (18; 28%) minor errors. In the postintervention period, we observed an error rate of 31% (87/284) (25; 36%); none was considered severe 0% (0; 2%), 23% (18; 28%) medications contained moderate and 8% (5; 12%) minor errors.
Table 3 gives details of observed errors for both observation periods. In the preintervention period, 159 errors (31 preparation and 128 administration errors) were observed in 151 doses. The most frequent types of errors were not mixing, wrong administration rate and expired shelf-life (continuous intravenous preparations used more than 24 h). The observers intervened three times, twice during preparation and once during administration of the medication. One of these errors was classified severe and two were classified moderate. In the postintervention period, 104 errors (34 preparation and 70 administration errors) were observed in 87 doses. The most frequent types of errors were not mixing, wrong administration rate, not flushing intravenous lines and incompatibilities. The observers did not intervene during this period. We did not observe any of the errors being related to the transcribing of orders. Specific examples of observed errors are illustrated in table 4.
In the univariate analysis (table 5), all variables (days in preintervention period, type of preparation and type of administration) seemed to be associated with medication errors individually (ie, uncorrected for other variables) at the significance level of 0.05, except for the variable observers and days within postintervention period. The proportions of the different levels of the two variables ‘type of preparation’ and ‘route of administration’ were not consistent over periods (table 2). Since they seemed to be associated with the medication errors, this implied or indicated the need for a more elaborate statistical analysis that corrected for this imbalance in proportions of these variables to evaluate an effect between periods.
The type 3 generalised score statistic for the effects of the variables (type of preparation, route of administration, observers, period) in the GEE multiple logistic regression analysis (with an exchangeable correlation working matrix) demonstrated that the route of administration (p=0.001), the observer within period (p=0.036) and the period (p=0.032) were all significant at the level of significance of 0.05 (when corrected for the other variables). The type of preparation of medication was not considered significant (p=0.083). The OR for period was determined at 0.49 (0.29–0.84), indicating that the proportion of errors had been reduced between the two periods.
We observed a significant reduction of the preparation and administration error rate, but the error rate remained alarmingly high, with about one in four doses having a potentially clinically significant error (classified as moderate or severe). Our overall error rates are higher than in previous observation-based studies carried out in the NICU setting6,–,8 and in paediatric and adult intensive care units,11 but similar to studies investigating errors in parenteral medication only.12 ,16 ,23 However, studies are difficult to compare due to variations in methods of data collection and definitions.5 For example, medical students collected data in the study by Raja Lope et al7 and pharmacy students collected our data. Knowledge and background of observers may have an impact on the problems that are recognised.
Previous studies evaluating educational interventions in the NICU setting also showed improvements in medication safety although not all of them assessed medication preparation and administration errors. In an Argentinian paediatric hospital, Otero et al24 found that an educational intervention focusing on medication safety culture reduced medication errors significantly with the largest decrease on the NICU wards (8% preintervention, 3% postintervention of all dose administrations). Overall, a larger decrease was observed for prescribing errors than administration errors, but data were collected reviewing nurse documentation rather than direct observation. Raja Lope et al7 found a considerable reduction in drug administration errors, but our results are difficult to compare as they did not report much details about the content of the educational measure. Campino et al25 showed that a comprehensive educational strategy, mainly carried out by a hospital pharmacist (similar to our study), was successful in reducing prescribing error rates from 21% to 3% (of all reviewed prescriptions), but effects on administration error rates have not been measured. Recently, it has been shown that the use of simulation techniques during teaching sessions reduced medication error rates considerably.17 Although we included ward-based practical teaching sessions and not just written materials and lectures, it may be useful to include a human patient simulator during teaching sessions.
We found that our educational intervention was insufficient to achieve an adequate level of medication safety. What would be other strategies to improve medication safety in the NICU? Education in combination with wearing designated aprons during drug rounds has been shown to reduce the number of interruptions during drug rounds.26 A recent pharmacoeconomic analysis suggested that the use of ready-to-use syringes could be a cost-effective solution.27 But so far, medications for injections are not always available in ready-to-use forms from the manufacturers and many injections need to be prepared or diluted before they can be administered. In practice, such services are gradually taken over by the pharmacy department.28 It has also been suggested to use standard medication concentrations. However, this may be complicated because of the neonate's need for careful fluid balance. Furthermore, infusion pump speed may have to be changed frequently in acutely ill neonates thereby introducing new risks for medication errors.29 ,30 Another strategy which has been shown to reduce prescribing error rates in the intensive care setting as well as in paediatrics is the use of computerised physician order entry,31 but limited data are available for the NICU setting and effects on administration error rates have rarely been studied. In an observation-based study, Taylor et al6 found a significant reduction of administration error rates, in particular wrong route errors, after the introduction of computerised physician order entry (CPOE) on a NICU. The addition of a barcode medication administration to a CPOE has also been shown to improve medication safety in the NICU.32 But overall, more research is needed to find strategies to improve medication preparation and administration in the NICU setting.
Our study has several limitations. One limitation is that a change in medication error rates between the two periods could have been caused by the educational intervention, and also by other variables, such as observers. Indeed, different observers were used in the preintervention and postintervention periods, and the observers within the same period did observe different proportions of medication errors. It is unknown what has caused this difference, but it could have been related to the fact that they observed different nurses or shifts in these periods. As described earlier, observers were trained carefully in the same manner in both periods, and observation remains the ‘gold standard’ in research about preparation and administration errors.19 Of course, also other (non-measured) variables which changed between the two periods could have eliminated or increased the effect of the intervention. One example was the introduction of a new type of intravenous line which made it possible to administer more than one intravenous drug administration at the same time through the same line. This may explain the higher number of incompatibilities in the postintervention period as nurse administered potentially incompatible medication through the same line. Seasonal events have been reported to affect medication error rates in another study using a time series design.33 Other variables remained relatively constant. For example, there were minimal changes in staff over the study period. There were also no annual cycles of staff training which may have resulted in differences in the data collected in February (preintervention) and June (postintervention).We did not achieve the sample size we originally aimed for within the periods of 10 study days. This was partly because more than one nurse prepared medication at the same time and therefore could not be followed by one observer. However, we found a much higher error rate than expected, and therefore the study did have sufficient power to detect a significant change. Although an effect of the intervention is not yet proven definitely, the study does suggest that education could be a method to diminish medication errors, since it remained significant after correction for several confounding variables. This supported our belief that education improved the procedures for preparing and administering medication. Finally, this study was carried out in a single institution in a tertiary hospital in the Netherlands. The findings may not be applicable to other hospitals or other countries, especially hospitals using different kinds of services, for example, using centralised intravenous additive services.
The multifaceted educational intervention seemed to have contributed to a significant reduction of the preparation and administration error rate, but other measures are needed to improve medication safety further.
The authors are grateful to all NICU medical and nursing personnel for their collaboration and willingness to participate in this study, and especially to the neonates admitted to the unit during the performance of this study, our final stakeholders and the pharmacy students.
Funding This study was partly funded by the Netherlands association of hospital pharmacists (NVZA). The sponsor did not have a role in the study design; in the collection, analysis and interpretation data; in the writing of the report; and in the decision to submit the paper for publication.
Competing interests KT received funding from Astra Zeneca for a project unrelated to this work. KT also received speaker's fees from Hospira and BBraun.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement We could provide access to the primary data.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.