Medication incidents and pharmacist interventions in relation to directly acting oral anticoagulants in hospital settings: evaluation using Reason’s Accident Causation Theory

Medication incident data from the incident reporting system (48-months period) and pharmacist interventions data from the prescribing system (26-month period) were extracted. Reason’s Accident Causation Model was used to identify potential causality of the incidents. Pharmacists’ intervention data was thematically analysed.


Introduction
Thromboembolic events present major clinical concern. Consequences can be serious, resulting in morbidity or mortality [1]. It is estimated that one in ve people die due to causes involving clots [2]. Anticoagulants are rst-line therapy for thromboembolic events. They are indicated for prophylaxis and treatment of venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE). Additionally, they are used to reduce the risk of secondary complications such as stroke in patients with atrial brillation (AF) [3,4]. In recent years, the traditionally used vitamin K antagonist (VKA), warfarin has been gradually replaced by direct oral anticoagulants (DOACs), previously known as novel oral anticoagulants (NOACs).
Currently, there are four DOACs licensed in the United Kingdom (UK) including: apixaban, rivaroxaban, dabigatran and edoxaban [3]. The approval of this drug class has revolutionised oral anticoagulation pharmacotherapy and considerably expanded clinical use [5]. DOACs display a preferred safety pro le; they have fewer problematic interactions, a xed-dose regimen and do not require routine international normalised ratio (INR) monitoring, unlike VKAs. Moreover, DOACs have a faster onset effect and a relatively short half-life compared to VKAs. Therefore, anticoagulation outcomes are achieved quicker whilst still reaching the desired effectiveness [6,7]. These advantages have encouraged a shift in favour of DOACs in treatment guidelines, consequently increasing national prescribing rates [8].
Despite wide use, research studying DOAC-related medication incidents is lacking. Though reports of adverse events relating to DOACs and the wider anticoagulant class are available [5,[9][10][11], analysis of error cause is limited. As they continue to be integrated into clinical practice, a better understanding of the DOAC-related incident types and what entails occurrence is required. Determination of causes will help identify risk reduction strategies and allow further guidance from regulatory bodies to ultimately reduce medication incidents associated with DOACs.
In this study, a medication incident is de ned as "a medication related incident or event which actually resulted in or had the potential for a detrimental result consequence to a patient" [12]. It is important to understand that the outcome of a medication incident is not always harmful, but rather risks patient safety. Incidents can occur at any stage of the medication process: prescribing, transcribing, dispensing, administering and monitoring [13]. Previous studies have detected and quanti ed error types according to the medication process stages [14][15][16]. Inappropriate prescribing due to incorrect dosing has been highlighted in literature as a major issue in relation to DOAC prescribing [17,18]. Patient height and weight, baseline activated partial prothrombin time, INR, haemoglobin, urea, electrolytes, liver function tests and creatinine clearance (CrCl) are imperative assessments before DOAC initiation [19].
Reason's Accident Causation Model is a widely used theoretical framework. This model can be applied to identify potential causality of errors [20]. Fig. 1 displays a simpli ed diagram of the model categories. Errors can be classi ed into active and latent failures. Active failures are de ned as unsafe acts carried out by individuals in direct contact with the patient or system. These can be sub-classi ed into slips (action-related execution error), lapses (memory-related execution error), mistakes (planning error) and violations (rule-breaking error). Latent failures are system failures that arise from high level organisation decisions [21]. Application of this model and subsequent identi cation of incident causes will stimulate the basis for future interventions in minimising medication incidents.

Aim of the study
The aims of this study were to assess medication incidents associated with DOACs in the hospital setting using Reason's Accident Causation Model and to evaluate the nature of pharmacists' interventions in minimising DOACs medication incidents.

Ethics approval
This study was approved by the University of Birmingham School of Pharmacy Research Ethics Panel in October 2020 (UoB/SoP/2020-03) (see Appendix 1). The NHS Foundation Trust approved this study as an audit (CARMS-16618) and no further NHS ethical approval was required.

Method
A two-part data analysis study was conducted. Firstly, medication incidents reported to DATIX by healthcare professionals was analysed over a 48-month period (September 2016 -September 2020). DATIX is a widely used, web-based, voluntary incident reporting and risk management system. This database collates occurrence of all events that have resulted in or have the potential to result in patient safety violation [23]. Next, pharmacist interventions submitted to the Prescribing Information and Communications System (PICS) were reviewed over a 26-month period (August 2018 to September 2020). PICS is an electronic clinical decisions support system, speci c to the study setting. The system is designed to minimise medication related errors via various automatic rule-based prescribing checks. PICS is also a communication platform; it allows healthcare professionals to voluntarily log occurrence of events/interventions [24].

Setting
Both databases, DATIX and PICS were obtained from one of the largest tertiary teaching hospitals in England. The patient population across the hospital sites is estimated to be 2.2 million [25].

Data processing and analysis
Search terms (DOAC, NOAC, apixaban, rivaroxaban, dabigatran, edoxaban and anticoagulant) were used to extract relevant data from both databases in the preceding ve years up to 30/09/2020. The acquired data was processed on Microsoft Excel in an anonymous form. Then, data was ltered according to the inclusion criteria: adult patients (≥18 years old) who were prescribed a DOAC. Data cleaning was used to remove duplicate records, incomplete and unclear information. Quantitative analysis was used to investigate the identi ed medication incidents from DATIX. Categorisation according to incident type was conducted primarily by one author followed by independent checks by two authors (VP and ZJ).
Classi cation of categories was determined by identifying the common reoccurring events. Descriptive statistics including frequency and percentages were used to analyse this data.
Reason's Accident Causation Model was used to determine the contributory factors associated with medication incidents and to ultimately establish potential causality. The free text data from the DATIX database were examined to classify cause of medication incident according to the model categories.
Organisation into sub-categories dependent on the most common themes was conducted to enable further investigation. Quantitative analysis via descriptive statistics was performed to determine the major cause of medication incidents.
Data from the PICS database in relation to pharmacist interventions and associated rationale was classi ed as per nature of clinical interventions. The classi cation system used was adapted from a previously reported study [26]. Two additional categories: 'documentation' and 'other' were also added.
Sub-categories were included as appropriate.

Evaluation of DOAC incidents
A total of 419 incidents were identi ed over a 48-month period from the initial DATIX system search. 241 reported incidents were excluded. Reasons include not DOAC-related (i.e., regarding warfarin, enoxaparin, tinzaparin), duplicate records and incomplete information (i.e., DOAC unspeci ed, use of unclear abbreviations). Of the remaining 178 DOAC-related incidents, a further 54 cases were excluded as they were not deemed as medication incidents. For instance, these were concerning access, transfer and cancellation of procedures. Hence, 124 reports were included in this study following inclusion and exclusion ltering.
A number of factors resulted in medication incidents as shown in Fig. 2. The majority of the incidents occurred during the prescribing and administration stage of the medication process. The most common errors resulting in an incident were missing drug/omission (26.6%, n=33), wrong drug (16.1%, n=20) and wrong dose/strength (11.3%, n=14). Table 1 shows the contributory factors that resulted in medication incidents in line with Reason's Accident Causation Model. Almost all (89.5%, n=111) medication incidents were classi ed as active failures. The active failures comprised of lapses (29.8%, n=37), slips (24.2%, n=30), mistakes (22.6%, n=28) and violations (12.9%, n=16). The rest of the incidents were classi ed as latent failures (10.5%, n=13). These categories were sub-categorised, as summarised in Table 1.

Missing drug/omission
Various scenarios resulted in drug dose omission, each with differing error causes as de ned by Reason's Accident Causation Model (see Table 1). The majority of drug omission incidents were due to lapses including lack of plan adherence (48.6%, n=18). A reoccurring theme was patient discharge from hospital without anticoagulation supply. Failure to restart DOAC post procedure/scan was also a common cause resulting in drug omission (10.8%, n=4). This indicates lack of staff awareness of the importance of missed DOAC doses and effect on anticoagulation. A few cases of drug omission due to violation concerned to take out (TTO) prescriptions which had inadvertently not been updated by the prescriber prior to patient discharge (18.8%, n=3). Latent failures resulting in drug dose omission involved insu cient team communication/handover (7.7%, n=1).

Wrong drug
Medication incidents due to wrong drug supply comprised a high percentage of incidents. Causes of error were largely due to slips and mistakes (see Table 1). Slips involved dispensing errors such as selecting the wrong drug due to incorrect system/clerking documentation (16.7%, n=5). There were two reported cases where the look-alike, sound-alike drug rosuvastatin was dispensed instead of rivaroxaban (6.7%, n=2). A large proportion of slips involved drug supply to the incorrect patient (26.7%, n=8).

Wrong dose/strength
The most common dose/strength related medication incident was the prescribing of wrong dose for indication (28.6%, n=8). This error is classi ed as a mistake (see Table 1). For example, a patient diagnosed with left leg DVT was commenced on rivaroxaban 15mg once daily. However, the patient should have been prescribed 15mg twice daily for the rst 21 days as per national guidance [3]. Latent failures resulting in wrong dose/strength supply involved the double dose administration of DOAC to overcome the effect of missed doses (15.4%, n=2).

Evaluation of pharmacist interventions
Following the initial PICS database search, 1024 pharmacist interventions were identi ed over a 26month period. 336 intervention cases were excluded from the study due to the same reasons for exclusion of reported incidents as aforementioned (i.e., not DOAC related, unclear, incomplete text). The remaining 688 submitted interventions speci c to DOACs formed the data sample included in this study.
Changes in pharmacological strategy comprised the highest proportion (38.1%, n=262). Interventions related to quantity of drug followed (26.5%, n=182) and then those related to patient education (14.5%, n=100) (see Fig. 3). Start/restart DOAC accounted for more than half of the pharmacological strategy interventions (51.5%, n=135) (see Table 2). Drug change was the second most common pharmacological strategy intervention (21.0%, n=55). Almost all of the quantity of drug interventions were associated with DOAC dose changes (91.2%, n=166). The rationale for the interventions varied, as shown in Table 2.

Dose change
Pharmacist-led interventions owing to inappropriate dose prescribing contributed to the largest overall percentage of recorded interventions (see Table 2). In many circumstances, multifactorial rationale including age, weight and renal function were assessed to establish suitable doses. Renal function was the most common reason for dose adjustment (29.4%, n=67). The majority of these cases involved renally impaired patients requiring dose reduction and a few increased doses as renal function improved. Age and weight were also highlighted as important considerations when determining the appropriate patient dose (16.7%, n=38 and 18.0%, n=41 respectively). 13.2% (n=30) of dose modi cation interventions were related to indication and/or treatment guidelines such as the switch from initiation to maintenance doses or changing between prophylactic and therapeutic doses.

Start/restart medication
Key rationale for this intervention included the initiation or re-initiation of DOAC therapy on discharge (22.2%, n=30). A common scenario involved inpatient low molecular weight heparin therapy and reinitiation of DOAC on discharge, in line with the hospital Trust policy guidelines [27]. New diagnosis of thromboembolic indications, such as AF and PE resulted in the initiation of appropriate DOAC therapy (7.4%, n=10). Anticoagulation is contraindicated during state of active bleeding. Restarting anticoagulation post-procedure or post-scan comprised of 6.7% (n=9) and 3.0% (n=4) respectively (see Table 2).

Drug change
Foundation for changes in anticoagulation therapy involved drug-drug interactions (12.7%, n=7).

Patient education
General counselling formed the majority of patient education interventions (79%, n=79). Also, 19% (n=19) were related to patients newly initiated on a DOAC. The remaining 2% (n=2) concentrated on enhancing patient compliance (see Table 2).

Key ndings
This study shows that the majority of the DOAC-related incidents in a secondary care setting occurred in the prescribing and administration stages of the medication process. This is in line with previous studies which reported a high degree of anticoagulant incidents due to inappropriate prescribing and administration [10,28]. This demonstrates a need for further education surrounding DOACs and other anticoagulants for all staff groups. By contrast, a Danish study showed that most of the incidents occurred largely in the prescribing phase, suggesting a problem speci cally correlating the incidents to physicians and prescribers rather than the nursing staff [29].
Key medication incidents associated with DOACs were missing drug/omission, wrong drug and wrong dose/strength. Omissions accounted for a considerably high percentage, more than a quarter of the incidents in this study. Past research reviewing all medication incidents reported to the National Reporting and Learning System (NRLS) in England over a 6-year period (2005-2010) found that the most frequent incident type was those relating to drug dose omissions [30]. Targeted implementation of preventative strategies is required to reduce this incident type.
The application of Reason's Accident Causation Model in this study showed that most of the errors were due to active failures (slips, lapses, mistakes and violations). Lapses were the most common of the active failure category, closely followed by slips and mistakes. Causes of the medication incidents were largely due to the performance of the healthcare professionals, rather than faults in system organisation. Hence, it is vital that staff adhere to guidelines.
Our ndings show that pharmacists play an integral role in minimising medication incidents. Some key interventions include dose and drug alterations, stopping and starting treatment, documentation and patient counselling. Overall, dose changes contributed to the highest percentage of recorded interventions. This is consistent with several published studies investigating pharmacist interventions in other therapeutic areas [31][32][33][34]. Hence, this demonstrates lack of prescriber familiarity with dosing regimens and necessitates for further prescriber education.
A closer look at dose change rationale reveals renal function as the top cause. The National Patient Safety Agency (NPSA) in the UK has warranted a safety alert with regard to inappropriate anticoagulant dose prescribing, particularly concerning renal function [4]. Dose adjustment according to renal function is highly important to ensure optimal thromboembolic therapy whilst reducing the associated bleeding risks. The Medicines and Healthcare products Regulatory Agency (MHRA) advises calculation of CrCl prior to making dosing decisions [35]. These incorporated measures stress the frequency of inappropriate renal dosing and suggest the need for further implementation to reduce related incidents.

Recommendations for practice
With regards to prescribing errors, there are two distinct ways in lowering them: reducing errors during prescribing and reducing errors after prescribing [36]. The latter focuses on the pharmacist's role. As shown in this study and previous studies, pharmacist interventions are signi cant in identifying and minimising medication incidents. It is imperative that pharmacists continue reviewing drug charts and conducting patient medicine reconciliations, but on a larger scale. The involvement of more clinical pharmacists throughout the medication process increases the chance of error identi cation; therefore, implementation is recommended. Only a small percentage of incidents were due to latent system failures; the main cause of error consisted of active failures. This demonstrates that the causes are mainly due to errors in task execution and planning by healthcare professionals. Pharmacist participation in the training process will assist education of other multidisciplinary team members such as prescribers and nurses. Annual assessments in relation to the correct prescribing and administration process of DOACs speci c to staff groups are needed. Furthermore, we recommend mandating renal function information on prescriptions as priority [14]. This will allow ease of checking by the prescriber and the pharmacist.
In this study, medication incidents due to wrong/omitted verbal patient directions were identi ed as a signi cant issue. Thus, it is necessary to understand de ciencies in current counselling and cooperate with patients to implement effective counselling practices.

Strengths and Limitations
Large, comprehensive data samples were extracted over a substantial timeframe using sophisticated incident and intervention reporting databases. The commonly applied framework, Reason's Accident Causation Model was used providing indication of error causality allowing identi cation of areas of improvement for patient safety. However, both reporting systems operate voluntary. Underreporting, selective and incomplete reporting are recognised; our results are likely to be underestimated compared to the true values. In addition, data was obtained from only one large hospital Trust in the UK limiting generalisability. This study used a theoretical model to allow analysis and interpretation of the data in a structured way, which may enable other researchers to classify DOACs incidents and interventions accordingly.

Recommendations for research
Future observational research can be conducted to overcome bias in the voluntary reporting system. Qualitative studies consisting of semi-structured interviews of patients, nurses, prescribers and pharmacists to determine medication incidents and actioned interventions are needed. Additional research should aim to extend the scope of this study to incident severity and its impact on patient health outcomes. Development and evaluations of interventions to minimise errors are needed. Research should be extended to non-hospital settings.

Conclusion
DOACs related medication incidents commonly occur in hospital settings due to varying factors. It is important to stress on the healthcare professionals around the importance of guideline adherence, in particular renal function assessment to determine appropriate dosing schedules. Mandating renal function information on prescriptions is recommended to allow ease of checking. Whilst pharmacists play a crucial role in minimising incidents at present, additional strategies such as strengthening clinical governance, pharmacist involvement in the on-going training of staff and annual staff assessments are required to improve patient safety in relation to DOACs. [35] GOV.UK. Prescribing medicines in renal impairment: using the appropriate estimate of renal function to avoid the risk of adverse drug reactions. 2019; Available at: https://www.gov.uk/drug-safetyupdate/prescribing-medicines-in-renal-impairment-using-the-appropriate-estimate-of-renal-function-toavoid-the-risk-of-adverse-drug-reactions. Accessed 8 December, 2020.