Defining electronic-prescribing and infusion-related medication errors in paediatric intensive care – a Delphi study

Recent decades have seen two distinct but related healthcare developments; the patient safety movement and the drive for increased use of health information technology (HIT) or eHealth [1–4]. The potential of HIT to reduce medication errors, the most common preventable cause of patient harm, has been identified in numerous safety agency and governmental reports [5–7]. Paediatric patients are at increased risk from medication error, with the paediatric and neonatal intensive care setting being of particular concern [8, 9]. The complexity of illness, the use of multiple high-risk medications and the vast range in weights of patients cared for are significant factors [10, 11]. As a result, there has been an influx of technology based interventions into this environment, aimed at mitigating these inherent risks [12, 13]. Electronic-prescribing, more commonly referred to as computerised provider order entry (CPOE) in the US, and the use of smart-pump technology are primary examples [14–16].

Paediatric-specific HIT systems do not however exist, and the current evidence for the effectiveness of HIT in preventing paediatric medication errors is limited [15, 17–19]. The specific functionalities required for the paediatric setting have been highlighted [20, 21]. The considerable financial cost and extensive process changes that HIT brings to healthcare warrant robust evaluation and rigorous research. Additionally, the introduction of new errors on implementation of HIT is a widely reported phenomenon [22–24]. Capturing the impact of these well intended initiatives is difficult [12]. One impediment is the absence of suitably specific error definitions [22, 25, 26]. The ‘niche’ nature of paediatrics, and the fact that many existing definitions pre-date HIT implementation, are particular barriers [27, 28]. Heterogeneity of medication use processes, with differing levels of HIT integration, and local customisation of both home-grown and commercially available systems, are further impediments [23, 29–31].

A germane example of both the ‘niche’ component and the need for clearer definitions is the current movement in both Europe and the US to standardise paediatric and neonatal infusions [32–34]. The use of traditional individualised weight-based infusions is recognised as being error-prone, yet remains common practice in many paper-based neonatal and paediatric settings across Europe [33, 35, 36]. Furthermore, widespread implementation of smart-pumps has yet to happen, with many units continuing to administer high-risk medications via traditional infusion pumps [33, 34, 36]. Current definitions fail to adequately catch the types of errors associated with these contrasting infusion processes.

The Delphi technique is a methodology used in health research to achieve consensus among groups of experts on particular issues where none exists. It consists of multiple iterative rounds, involving an expert panel, who are provided with controlled anonymous feedback [37, 38]. It has previously been used, in the pre-HIT era, to produce practitioner-based definitions of both paediatric and adult prescribing errors, with accompanying lists of included medication incidents [27, 39].

The aim of this study was to achieve consensus in defining a range of new technology-generated scenarios related to electronic prescribing, the use of smart-pump technology, and the interface between these two HIT interventions. To facilitate this work, it was also necessary to seek consensus on a range of previously unaddressed weight-based infusion scenarios still common in paediatric and neonatal intensive care settings. We aim to mirror the process used in earlier similar studies, thereby enabling the results of this study to be used in conjunction with them.

A secondary aim is to identify any significant differences between the specific healthcare professional (HCP) groups in the expert panel. The medication use process is complex, involving multiple stages from point of medication ordering to patient administration, and is reliant on a range of HCPs. Successful HIT implementation requires an understanding of these different roles and their points of interaction with each HIT system [23]. In an era of rapid change in the use of HIT we hope, that by systematically defining novel medication errors, this study will assist future research on the impact of HIT medication safety initiatives.

This study demonstrates that achieving consensus for different categories of HIT-related medication errors varies considerably depending on the particular associated HIT system. The complexity of electronic-prescribing systems is evident in the diversity of individual opinion on the seven included scenarios. All were included in Round 2, with two requiring further clarification in Round 3. This difficulty in obtaining consensus is indicative of the nuances in such complex HIT systems, particularly in the paediatric setting.

Scenarios 1–4 in this study are examples of the difficulty in customising a commercial system to facilitate the prescribing of a large range of medications to a diverse population. An example of Scenario 1 can be seen in Fig. 5.

Configuring robust dose range checking and clinical decision support in the paediatric setting is particularly challenging [42]; paediatric-specific electronic-prescribing or CPOE systems do not exist [18]. Scenarios 5–7, which involve failure to discontinue orders and duplication, are examples of errors commonly reported with implementation of CPOE [43, 44]. Although they can also occur with paper-based systems and similar scenarios are listed in the UK adult and paediatric practitioner-based studies, these do not capture the complexity of electronic-prescribing. When an electronic order is not discontinued, it stays ‘live’ for the patient’s full length of stay. This differs considerably from paper systems where the paper prescription and administration record becomes full after a set number of days and needs to be rewritten. Also, when an electronic order is altered on our system, a new order is automatically generated. In addition to the risks associated with certain fields inappropriately auto-filling as identified in scenario 2, the number of similar/duplicate orders is also increased.

The consensus position reached in this study was that failure to discontinue orders should not be considered a medication error. The participants considered that this did not pose a risk to the patient and in the dynamic intensive care setting may be clinically warranted. In contrast, it was determined that duplicate orders should be categorised as medication errors. This widely reported consequence of electronic-prescribing has the potential to lead to missed or extra doses due to a cluttered medication record. Ability to view medication orders from various screens compounds this risk. In addition, the existence of duplicate orders in our particular clinical information management system can increase the risk of incorrect manual assignment of infusion pumps to corresponding infusion orders. See Fig. 1.

The decision to provide a video demonstration of scenarios for Round 3 did not produce significantly higher levels of consensus. Despite this, feedback from participants indicated that the videos improved their understanding of the scenarios. We propose that this method may be useful in future studies.

Almost unanimous levels of agreement were found for the smart-pump related scenarios. This may be indicative of a widespread appreciation of both the high-risk nature of the commonly used infusions in PICU, and the risks associated with incorrect infusion pump programming. The programming errors identified in scenarios 8–10 are examples of new errors introduced by the use of smart-pump drug libraries. Scenarios 11–13, involving misalignment of standard concentration infusion (SCI) orders, syringes and pumps are also novel. The range of SCIs required to accommodate the complexity and diversity of the paediatric setting makes these particularly problematic. These errors highlight the need for progress towards closed-loop medication processes, however pre-prepared and barcoded infusion solutions remain rare in many European paediatric hospitals. With national standardisation of paediatric infusions underway in Ireland, and being actively pursued in the UK and the USA, clarity on defining these smart-pump related medication errors is important in assessing the limitations of current processes [32–34].

This study also produced consensus on the pre-electronic prescribing of weight-based paediatric infusion scenarios identified. Despite widespread recommendations against it, this practice is still common in Europe and continues to be recommended in paediatric reference books [32, 35, 45]. Scenarios 15–17, highlight common errors that occur when writing the ‘statement of rate’, a necessary component of the prescription to direct the setting of the pump. Although unlikely to cause harm they should not be overlooked, as the potential for harm is considerable should the statement of rate be misinterpreted by nursing staff unfamiliar with a particular infusion [46, 47]. With widespread discrepancies in medication error terminology, clarity on inclusion of these as prescribing errors may assist future studies and support the move away from this commonly used, yet error-prone method.

No single mitigation strategy is likely to avert the range of errors highlighted in this study. Future research is necessary to assess the impact of standard orders and order sets on electronic-prescribing errors. There are currently limited data on their impact on the efficiency or accuracy of the ordering process [48–50]. There are some data to indicate that they increase duplicate error rates [51]. This information is essential to ensure the on-going safe design of electronic ordering systems. The limitation of smart-pumps as a single intervention to prevent IV administration errors is well recognised. Until closed loop medication processes and bar-coded ready-to-use paediatric infusions are readily available in European hospitals, research into the effectiveness of more immediately available solutions, such as 2D barcode labelling, to reduce programming errors, is warranted.

Implementation of health technology systems introduces a range of medication errors not easily categorised using previously published medication error lists. The Delphi technique can successfully be used to define these novel medication errors and assist in measuring outcomes for HIT quality improvement and patient safety initiatives.

Electronic-prescribing errors may be the most difficult to clearly define, particularly as clinical information systems with higher levels of decision support become more widely employed. Further similar studies are likely to be required as the range of HIT generated medication errors expands across the entire medication use process. The use of video demonstrations may be a useful tool.

The results of this study, in conjunction with previously defined lists of medication errors will add to the quality of future medication error studies in both the paediatric and non-paediatric intensive care setting. This includes sites using electronic-prescribing systems, in particular the many adult and paediatric sites in Ireland and the UK using the Philips ICCA® system, sites using smart-pump technology and most neonatal and paediatric intensive care units.