Understanding prescribing errors for system optimisation: the technology-related error mechanism classification

Discussion

The unintended consequences of health IT, including TREs, have been recognised for over two decades.12 13 Addressing TREs is a key component of CPOE optimisation.11 18 The TREM classification reported in this paper provides a systematic means to understand how technology-related prescribing errors occur, allowing targeted CPOE modifications to be made. The classification was initially developed through a review of 1164 prescribing errors at two adult hospitals using two different CPOE systems.16 It has now been expanded with a review of prescribing errors at a tertiary paediatric hospital.

Updating the classification with the prescribing error data from a paediatric hospital resulted in three new mechanism categories and the re-organisation of subcategories. One new subcategory was specific to paediatric CPOE functionality, ‘Editing errors that occur when using the dose calculator’, with dose calculators used for the vast majority of paediatric prescribing. The other updates to the classification could equally apply to the adult setting; however, we have also added examples from paediatrics. For instance, we encountered construction errors where the weight of the child was entered incorrectly when constructing an order. Similarly, we found that off-label prescribing, that is, use of medications for indications for which they are not licensed, was more frequent among the paediatric population requiring editing of preprogrammed order sentences or constructing orders. However, the core elements of the classification, first applied to CPOE systems in two adult hospitals, remained applicable to the paediatric CPOE when applied 8 years later, providing an indication of its utility and relevance despite design changes to CPOE systems during that time. Our updated classification is more comprehensive as a result of including a contemporary dataset of prescribing errors from another setting and can be applied to CPOE for both paediatric and adult populations.

The TREM classification focuses on ‘how’ TREs occur, which allows for targeted CPOE modifications. We have demonstrated the feasibility of informing CPOE modifications using the TREM classification. Recommendations arising from our work have already been implemented within the study hospital’s CPOE; for example, changing the drop-down menu options for the intravenous route and new design options to avoid inadvertent selection of controlled-release opioids. Multiple optimisation strategies informed by the TRE data have been formulated and published online in the Health Innovation Series to allow greater dissemination of specific system optimisation recommendations and user tips.29–32 Data on TRE mechanisms would also be a powerful addition to human factors approaches for evaluating and analysing CPOE systems and in the design of solutions to address TREs.33

Research shows that health IT managers are flooded with requests to modify health IT systems but often have limited resources to respond to these requests.9–11 34 Thus, prioritisation of CPOE modifications based on factors such as the safety, efficiency and frequency of issues identified is required. It is also recognised that data to support the identification, prioritisation and evaluation of CPOE optimisation to address TREs are needed. To date, the majority of data on TREs have originated from voluntary incident reports from hospitals or regulatory bodies, such as the Food and Drug Administration.18 While incident reports can provide valuable insights into TREs, they cannot provide information on the relative frequency with which TREs are occurring or allow for meaningful comparisons between organisations.23 35 36 Incident reporting systems are also less likely to capture TREs with minimal or no patient impact. However, these less serious TREs may significantly affect health IT usability and thus should be addressed to improve user experience by removing obstacles to task completion and improving workflows.

In order to effectively prioritise CPOE optimisation activities, representative data on multiple dimensions of TREs are required. In figure 1, we showed two key dimensions of TREs. Figure 2 expands on this concept with a third dimension of TREs, that is, the outcomes of the error, and also shows other examples of the error manifestation. In work spanning the last two decades, our team has developed classifications that can be used to describe TRE dimensions, and examples of these classifications are provided in figure 2.2–6 16 26 36–38

Armed with data on multiple dimensions of TREs, CPOE managers could prioritise optimisation activities according to their goals or known areas of risk. For example, they could target the most frequent clinical error type by examining the underlying mechanism. Similarly, to reduce TREs with high-risk medication orders, such as high-strength potassium fluids or opioids, the underlying mechanisms of these errors could inform CPOE optimisation to support prescribers when ordering. Table 3 shows further examples of how CPOE optimisation goals can be mapped to the TRE dimensions shown in figure 2.

To apply the TREM classification, data on prescribing errors within a CPOE are required to identify and classify the multiple dimensions of TREs. It is important, however, to recognise that this may be a labour-intensive process and thus, potential users of the classification should consider how best to capture error data within existing medication safety processes for example, pharmacist review of medication orders. Future work could explore automation of TRE detection to reduce record review workload for potential users. Alternately, the TREM classification could also be used to proactively improve ordering for high-risk medications and prescribing scenarios. For example, the risk of selection errors could be examined for high-risk medications and changes made to limit options in drop-down menus.31

Back to top