The study was conducted at the RWTH Aachen University Hospital and has been approved by the local ethics committee of the Medical faculty of RWTH Aachen University (EK 126/13).

The type, incidence, and severity of prescribing errors in PN in a tertiary level (single center) NICU/PICU of 18 beds (all equipped for mechanical ventilation) were surveyed prospectively for all PN orders between March 2012 and July 2013 (n = 3,012) by a clinical pharmacist during his daily routine check. Time period was selected related to the presence of the clinical pharmacist. According to the current Federal German pharmacy work rules, there is a requirement that a pharmacist will review the orders of individual formulations like PN before preparation. In our institution, this reviewing is performed by or under direct supervision of a clinical pharmacist. Clinical pharmacists have undergone a 3-year postgraduate specialization and are examined by the German chamber of pharmacists.

Neonatal and pediatric intensive care unit PN orders in our institution are regularly performed using a CPOE system (Visite 2000, Lyomark Pharma GmbH, Oberhaching, Germany). Visite 2000 was introduced in November 2007. It had been established in prescription practice for over 4 years prior to the beginning of this study. Decision support is provided by calculating the total intake of fluids, nutrients, and electrolytes during 24-h period. This takes into account all medication as well as enteral and PN, which has been prescribed based on patient age and weight. Standards for PN formulations and notifications for drug dosing for different patient ages are integrated. Alerts for drug dosing (amino acids, trace elements, electrolytes, and fluids) as well as osmolarity were already integrated in the software upon its introduction and maintenance. Warnings for exceeding osmolarity occur only after the selection of peripheral vein administration by the physician. Warnings for incompatibility or drug–drug interactions are not features of Visite 2000. Alerts for calcium overdose for peripheral vein administration had not been included into the database prior to the beginning of this study.

After approval through the physician, the pharmacist has online access to the CPOE and performs order review. This process comprises the verification of patient data, ingredients, appropriate dose according to age, weight, and clinical condition, adherence to maximum osmolarity for route of administration, as well as compatibility, stability, and completeness prior to preparation. The pharmacist discussed each deviation from standard with the responsible physician and eliminated errors before preparing PN.

The MEs were defined according to the US National Coordinating Council for Medication Error Reporting and Prevention (NCC MERP) as “any preventable event that may cause or lead to inappropriate medication use or patient harm while the medication is in the control of the health care professional, patient, or consumer. Such events may be related to professional practice, health-care products, procedures, and systems, including prescribing, order communication, product labeling, packaging, and nomenclature, compounding, dispensing, distribution, administration, education, monitoring, and use” (25). Therefore, a prescribing error in our study was determined as deviation from a “defined standard,” not intended from the physician that may result in patient harm. “Defined standard” meant orders that were not in accordance with the manufacturer’s summary of product characteristics, local PN-guidelines, as well as national (26) or international consensus guidelines (8, 18). Our hospital guidelines are based on these consensus guidelines and were only used as reference where no recommendation was given in consensus guidelines. This refers especially to calcium concentrations in PN prescriptions for peripheral vein administered infusions for which no standard is given in the guidelines mentioned above.

In a first step after study inclusion, the pharmacist categorized the MEs into seven different types of error: patient data, drug choice, dosage, indication, compatibility, concentration range, and osmolarity (Table 1). In the categories dosage and concentration range, only deviations, which exceeded a 5% tolerance range from the defined standard were counted as ME. We selected a tolerance of 5% because it is a conventional pharmaceutical practice in the preparation of pharmaceuticals. We calculated the overall error rate as the percentage of errors relative to total drug orders and the error rate of each category relative to drug orders according to the category with 95% confidence intervals (CIs). The calculation of CI was performed with Microsoft^® Excel^®, Version 14.3.0, Microsoft Corporation, Redmont, Washington, USA.

In a second step, the potential severity of each ME was assessed independently by three experts from three different tertiary level NICUs under the hypothesis that the ME had reached the patient. The experts were senior neonatologists not affiliated with our institution and not involved in the development of our institutional guidelines. None of the experts was involved in any therapy regime, and all were blinded for patients and their outcome. Severity classification was based on the NCC MERP index (27) (Figure 1). For the expert classification, every prescribing error was described anonymously including patient data (age and weight), medication, and recent laboratory results. Expert training was carried out via a standard operating procedure that includes instructions how to use the NCC MERP index while predicting severity. The experts had the choice between the NCC MERP index categories C, D, E, F, G, H, and I. Categories A and B were excluded according to the hypothesis under which the assessment was done. A consensus was reached if all experts voted for the same level. If a minimum of two experts came to the same result, a “majority decision” was achieved. If there was no consensus or majority decision, errors dropped out of analysis but are discussed individually in the results section and discussion.

The pharmacist classified the errors into the seven error-type categories patient data, drug choice, dosage, indication, compatibility, concentration range, and osmolarity (Figure 2).

Most errors were assigned to the category “concentration range” (n = 77 in 1,277 orders, 6.0% error rate; 95% CI 4.7–7.3). The error rate was calculated relative to the drug orders with calcium gluconate and for peripheral vein administration. All of these were calcium concentrations above 0.4% in solutions intended for peripheral vein infusion. The concentrations of calcium gluconate in these PN solutions ranged from 0.42 to 1.71% (see Table S1 in Supplementary Material). There were 14 dosing errors in 3,012 orders, giving an error rate of 0.46%; 95% CI 0.22–0.71. Most “dosage” errors (n = 12) were related to wrong dosages of water-soluble vitamins and trace elements in PN (Table 2), most of which would have led to errors of less than 10% increase from the accepted range. One error would have led to an overdose of potassium chloride of 43%, another one would have caused an overdose of sodium chloride of 26% (Table 2). The third most frequent error-type category was “indication” errors (n = 12 in 3,012 orders, 0.40% error rate; 95% CI 0.17–0.62). All errors would have led to an omission of a PN component even though this component was indicated. Six (50%) errors resulted from the omission of water-soluble vitamins, five (42%) errors from omission of trace elements, and one (8%) from omission of calcium gluconate. Six errors in 3,012 orders (0.20%; 95% CI 0.04–0.36) were assigned to the “drug choice” error category. In five of these cases, the wrong concentration of glucose was prescribed. The standard for preparing PN is a 50% glucose solution. The physician prescribed the correct dose of glucose but chose the wrong concentration (5% instead of 50%). Inadvertent order of 5.85% sodium chloride instead of 20% sodium chloride solution, the standard in pharmacy PN compounding, occurred once. Five errors in 1,898 orders (0.26%; 95% CI 0.03–0.49) were assigned to the error-type category “compatibility.” The error rate was calculated relative to the drug orders with simultaneously prescribing of calcium gluconate and sodium glycerophosphate or trace elements and water-soluble vitamins. These formulations would have contained no amino acids and, therefore, would have been chemically unstable. In three instances, this was due to the simultaneous order of sodium glycerophosphate and calcium gluconate and twice due to the simultaneous order of water-soluble vitamins and trace elements. The category “patient data” contained three events (0.1%). In each instance, a wrong patient age was conveyed to the pharmacy. The single event (0.06%) in the error-type category “osmolarity” was an excess osmolarity (933.7 mmol/l) intended for a peripheral vein administration. Due to the small number of errors in the categories patient data and osmolarity, the 95% CIs were not calculated.

The three experts classified the 118 errors independently according to the NCC MERP index (Figure 3). In eight instances, errors were not classified unintentionally by two experts and, therefore, they are not presented in the chart.

One expert did not classify one error of the category drug choice (the wrong concentration of glucose was prescribed) and one error of the category dosage (underdose of trace elements were prescribed, 0.71 ml/kg/day instead of 1 ml/kg/day).

The other expert did not classify one error of the category drug choice (the wrong concentration of glucose was prescribed), three errors of the category concentration range (calcium gluconate 0.5, 0.65, and 0.46% instead of the maximum 0.4%), and two errors of the category dosage (underdose of trace elements were prescribed, 0.71 ml/kg/day instead of 1 ml/kg/day, same error as the other expert and underdose of trace elements 0.92 ml/kg/day instead of 1 ml/kg/day) (see Table S2 in Supplementary Material).

Majority decisions were found for 85 (72%) errors. In 23 (20%) of these errors, the three experts found consensus, all concerning category C. Most of events (38%, n = 45) were categorized into category C, 22% (n = 26) into category D, and 12% (n = 14) in category E. In 27 (23%) errors, the assessment by the three experts was spread between three NCC MERP categories. Five errors were classified only by two experts into two different categories; one error was classified only by one expert. The variability between the experts while assessing the prescribing errors is shown in Figure 4.

In the category “concentration range,” an agreement of all three experts was reached in four errors, all concerning category C. For the remaining 73 errors, the assessment of the experts varied between C and D (34 errors), C, D, and E (14 errors), C and E (5 errors), D and E (11 errors), D, E, and F (8 errors), and E and F (one error). In the category “dosage,” ten errors, all relating to a wrong dosage of water-soluble vitamins or trace elements, were assigned to category C by all experts. The error related to an overdose of sodium chloride was categorized into C and E and the one related to an overdose of potassium chloride was assigned to C, D, and F. Six errors in the category “indication,” all related to the omission of water-soluble vitamins, were assigned to category C by all experts. The six remaining errors fell in category C and D (omission of calcium gluconate), C and E or C and F (both omission of trace elements). The errors in the category “drug choice” were assessed differently. While one expert classified nearly all errors into category E, the two other experts assigned most of the six errors to category C or D. For three errors, a majority decision for category C was reached. In the category “compatibility,” the two errors “simultaneous prescribing of water soluble vitamins and trace elements in the absence of amino acids” (40%, n = 2), were classified by all three experts into category C. For the three remaining errors, “simultaneous prescribing of sodium glycerophosphate and calcium gluconate in the absence of amino acids” (60%, n = 3), the assessment of the experts varied between C, D, and E. In the category “patient data,” an agreement of all three experts was found for one error (category C) while for two errors the assessment differed between the categories C and E. The error categorized in C, D, and E resulted from an order with a patient age of 28 years although the patient was 1-day old. The error categorized into C and E was an order with a patient age of 11 weeks although the patient was 2 years and 11 weeks old. The error in the category “osmolarity” was assigned for majority decision to category D and from one expert to category E.

One limitation of this study is that we only studied the situation of prescribing PN in one department of our hospital. The results are difficult to generalize because CPOEs, standards, guidelines, and methods to categorize MEs may differ depending on institutional policies. Another limitation is that only one pharmacist conducted the analysis, which may cause observation bias. Therefore, the classification of error severity was delegated to three independent experts who are unaffiliated with our institution.

While MEs can occur at all stages of the medication process, we focused on prescribing errors. To enhance the medication safety in the whole PN process, further studies are needed to analyze errors in the transcription, preparation, and administration of PN.

Additionally, we did not address errors that might have been missed by the pharmacist. But as we observed no clinically overt complications or metabolic derangements related with TPN in the study period, we assume that there was at least no missed error of clinical significance.

We used the NCC MERP index to categorize our prescribing errors. The index was developed to classify MEs on the basis of patient outcome. However, the classification in our study was based on the estimated potential of harm to the patient, not on actual outcome. Additionally, the experts assessed the errors independently without interaction among each other. An error assessment of an expert group could have resulted in a different outcome. Expert classification of the relevance of errors in this study has shown a larger variability than expected. All experts were provided the same standard operating procedure for the judgment of the clinical relevance of errors. It was requested to categorize the errors on the assumption that the “worst possible event” had occurred. We believe that this is most likely the explanation for the high variability. So, despite the identical specification for all experts, it is to some degree left to the individual imagination of each expert, which would be the most extreme but still realistic combination of unfavorable circumstances. We chose to assume the worst case, because it is most significant for the individual patient security and should most certainly be avoided. However, judgment of the clinical relevance, which is to be expected most commonly based on the clinical experience of the experts might have resulted in a more homogenous classification of events.

Parenteral nutrition order in pediatric inpatients is highly individualized and includes a large number of calculations and characteristics to respect. It is, therefore, particularly prone to errors. However, only a small percentage of errors carries the risk to result in patient harm. If physicians and pharmacists are aware of these risks, measures can be developed to avoid errors. Above and beyond an alert system in the CPOE, a clinical pharmacist is proven to be very effective to improve medication safety in a medical team and to prevent adverse drug events.