Applicability of drug-related problem (DRP) classification system for classifying severe medication errors

Medication errors (MEs) are one of the major concerns in patient safety [1‐3]. A medication error is any preventable event that may cause or lead to inappropriate medication use or patient harm [4]. There is a need to learn effectively from errors that have already happened using error reporting systems and other available data sources to improve medication safety [1, 5‐8]. Especially the reoccurrence of MEs causing severe harm or even the death of the patient should be prevented, which is also promoted by global actions [2].

In ME reporting systems and studies on medication safety, data have been typically analyzed using error classification systems and taxonomies. Theoretical or conceptual frameworks have usually guided the development of error classification systems, or the process has been data-driven [9]. Depending on the aim of the classification, the system can be contextual (e.g., the place or the medicine involved), modal (how the error happened, e.g., omission), or psychological (why the error occurred, e.g., skill-based error) [10]. Classification systems and taxonomies need to relate to similar events, describe essential clinical and systemic factors, and support analysis purposes [11].

Although there are multiple taxonomies for MEs, they often share the limitation of being in too general level and describing only outcomes, not the causes of the errors [12]. This deficiency has been noted in many widely internationally used classification systems (e.g., World Health Organization, National Coordinating Council for Medication Error Reporting and Prevention) [13, 14], but the challenges to describing MEs accurately enough and with the causes and contributing factors are still partly unresolved. Without knowing the causes and contributing factors of MEs, essential information for error prevention and risk management are lacking [12, 15]. Adequate information on what happened and why is crucial, especially when preventing severe MEs [16‐18]. A classification system that gives enough information for medication error prevention and risk management, is desirable in the system-based approach to medication safety [11, 12]. Consequently, there is a need to develop ME classification systems further to support system-based medication safety.

Exploring the utility of existing classification systems, including also causes, can provide an option for acquiring more information about MEs to find novel and robust methods for describing ME data. Previous research has described the application of adverse drug event (ADE) classification systems to MEs, but the evidence is still limited [15]. According to our knowledge, none of the studies has evaluated how classification systems for drug-related problems (DRPs) could be applied to ME data, even though many of them also include causes for the problems [19].

Valvira received a total of 1654 complaints and statement requests from 2013–2017. Of these cases, 58 cases were medication-related and included in this study. The identified MEs had occurred in a wide range of social and healthcare settings, most commonly in assisted living facilities (n = 16, 25% of all settings, Table 3). There were 6 cases (10%) where more than one care setting was involved in the error process. In more than half of the cases (52%, n = 30), the ME caused the patient’s death or severe harm. Other characteristics of the data have been published elsewhere [17].

It was possible to classify all MEs according to Basger et al. [21]. In total, 100 MEs were identified from the cases (n = 58) (Table 2). In 53% (n = 31) of the cases, more than one ME was identified (mean number of MEs, 1.7 per case). Figure 3 presents an example of how MEs were identified and categorized from the case descriptions. A small proportion (8%, n = 8) of the MEs identified from the reports were classified in the “Other” category (a cause that cannot be classified into one of Basger’s et al.’s eight categories, Category 9, Table 2). The”Other” category included the following MEs (n = 8):

The identified MEs (n = 100) fell into all nine main categories of Basger et al.'s classification, 21/33 (64%) of the subcategories and 21/58 (36%) of the sub-subcategories (Table 2). Most of the MEs (n = 89, 89%) were categorized into the following six main categories: “Drug selection” (n = 22, 22%), “Drug use process” (n = 18, 18%), “Monitoring” (n = 14, 14%), “Dose selection” (n = 13, 13%), “Treatment duration” (n = 11, 11%) and “Logistics” (n = 11, 11%). Subcategory and sub-subcategories were used for all MEs if there were such defined in Basger et al.´s classification system.

The present study suggests that the aggregated DRP classification system by Basger et al. [21] can be applied to categorizing severe MEs. To our knowledge, this was the first international study to pilot test the applicability of a DRP classification system for this purpose. Further studies are needed to confirm our findings. Because of its authoritative nature, Valvira’s data had comprehensive qualitative and narrative documentation on the ME cases that enabled us to identify reliably multiple MEs and their causes in each case.

As Valvira’s data included both severe and non-severe MEs in different social and healthcare settings, our study indicated that the DRP classification might be applied to MEs with different levels of harm occurring in different care settings. Still, there were some challenges in using aggregated DRP classification system for all MEs included in our data. This concerns especially the potential limitation of not being able to categorize all MEs prevented before they reached the patient (i.e., near misses). With severe MEs, this situation is not usually an issue as they typically reach the patient and cause some problems or harm to the patient. However, this limitation of DRP classification can become a major challenge when classifying MEs that are potential (Fig. 1.). It would be problematic if the same classification system would not apply to near miss events as usually the ME data consists of all kinds of ME, ranging from near misses to most fatal ones. As there was only one near miss case in our study material, there is a need for further research with more extensive data acquired from other ME sources, such as ME reporting systems commonly used in healthcare organizations.

With Basger et al.’s aggregated DRP classification system, we were able to categorize both the problem (ME) and its cause. Subgroups and sub-subgroups of causes were seen to be useful in the categorization of the causes in a way that provides detailed information about the incident for ME prevention and risk management purposes from the systems approach, although further research is needed on ME categorization. Already existing ME classification systems could be further developed to include similar subgroups of error causes as the Basger et al.’s aggregated DRP classification system has. This kind of more detailed error cause categorization has been already experimented within e.g., a medical error study that used the WHO International Classification for Patient Safety [34]. Still, the strength of Basger et al.’s aggregated DRP classification system is that it is readily available and specifically designed for risk management in the medication use process. There is a need to further optimize ME classification so that it will provide enough information about error causes and contributing factors.

Having as much information as possible is a benefit when developing medication safety based on reported errors, especially severe ones [17, 18]. For example, prescribing error (as an outcome) can mean anything starting from the wrong drug or dose selection to duration of the medication treatment [35]. In our previous study with the same Valvira data as used in the present study, we were able to identify the most common ME types but were using an outcome-based ME taxonomy [17]. By this present analysis of the same data using Basger et al.’s aggregated DRP classification system, we were able to describe and understand the MEs and causes contributing to the incidents more in detail. Another strength of using Basger et al.’s aggregated DRP classification system is that it indicates whether the MEs were caused by healthcare professionals or by the patient.

Even though we found Basger et al.’s aggregated DRP classification system to be potential for categorizing severe and non-severe MEs, it could be further optimized for classifying MEs. Basger et al.’s aggregated DRP classification system had many subcategories that did not have any sub-subcategories to describe the cause for the problem (16/58). As we had rich qualitive data and were able to identify the cause(s) for the ME in most cases, we could have been able to add some sub-subcategories to the classification system (e.g., why too low medication dose was prescribed). Additional sub-subcategories could produce more information about the causes and contributing factors. On the other hand, they may increase the complexity of the classification system [21].

It became evident that we were not able to describe all contributing factors identified that caused the error by using the aggregated DRP classification system. For example, in a ME case presented in Fig. 3, we identified many more contributing factors than we were able to classify with the classification system (e.g., two different oxycodone concentrations in use, no double-checking procedures, an inexperienced physician who did not know the possibilities and resources for patient monitoring in assisted living facility unit). This result indicates that Basger et al.’s aggregated DRP classification system in its current form does not alone meet the need for comprehensively categorizing a wide range of factors contributing to severe MEs. Therefore, it is important to recognize the limitations of the DRP classification system and use other methods to supplement the understanding of the multiple causes and factors contributing to MEs.

Even though the aggregated DRP classification system helped to describe MEs and their causes in more detail, we are still lacking tools to describe complex medication errors. Severe errors often consist of multiple errors that are chained (e.g., Fig. 3) [16, 17, 31, 32, 36]. It is challenging to make these error chains visible using only structured classification, although there are some successful examples [16]. But in terms of learning in-depth from severe MEs and complex error cases, qualitative analysis (e.g., root cause analysis or causal tree analysis) can still be seen as more informative, because these tools enable the description of multiple error chains and the contributing factors in the medication use process [37, 38].

All kinds of classification systems need to have an “other” category for miscellaneous issues because classification systems can rarely be completely comprehensive. The prevalence of using this category has been even higher in the studies using different DRP classification systems for DRPs than in our study with MEs [19]. Our data contained documenting errors at the administration phase that we were not able to categorize using Basger et al.’s aggregated DRP classification system. Documenting errors have potential to cause DRPs and thus harm the patient; therefore, they should be added to the classification system [39, 40]. They could be categorized under the main category “Logistics.” At all, the main category “Logistics” could be renamed because some of its sub- and sub-subcategories relate to other than purely logistic issues, such transcribing errors (Category 6.2) and errors in drug selection (Category 6.3) (Table 2). We also found that the classification system was missing an explicit DRP category for problems in medication cessation and implementing phased dose changes. Both these subcategories are important from the DRP and ME perspective.

Furthermore, we noticed that the classification system should be developed from the pharmacists’ perspective to better consider risks related to the preparation and dispensing phases of the medication use process, as well as counseling and educating medicine users. All these situations are potential causes for errors and causes for DRPs [41, 42]. The classification system should distinguish between transcribing errors that occur in medication reconciliation and the transcribing phase. Basger et al. [21] themselves noticed that DRPs associated with care transfers may require a taxonomy of its own, because these problems may be challenging to describe using only the DRP classification system. Indeed, the medication discrepancy taxonomy (MedTax) has been developed few years after Basger et al. published their aggregated DRP classification system [43].

Our study suggests that the aggregated DRP classification system with some modifications could be used for analyzing and describing MEs and their causes, especially severe MEs. This finding aligns with the current definition framework in the study which defines that DRPs can be caused by MEs. Using a cause-based DRP classification system produces additional information which is essential for understanding why the MEs happened and how to prevent such MEs in future. This information is particularly valuable with severe MEs that are a priority to prevent. Because severe MEs are often complex processes including several errors and contributing factors, additional means to analyze the incidents are also needed. Further studies using different ME data are needed to validate our preliminary findings and learn more how DRP classification systems could be utilized in learning from errors and their contributing factors.