Drug-related problems associated with direct oral anticoagulants: an observational cross-sectional study of medical record review by pharmacists in a large teaching hospital

Background Prescribing DOACs presents with challenges in the elderly and patients with renal and hepatic impairment. To mitigate safety risks, pharmacists have a role in detection, prevention, and resolution of DOAC-associated drug-related problems (DRPs). Objectives To identify the types of DOAC-associated DRPs in patients on DOAC therapy and factors that predispose patients to DOAC-associated DRPs. Methods An observational cross-sectional study was conducted in SGH from January 1, 2017, to May 31, 2019, on patients prescribed with a DOAC (rivaroxaban, dabigatran, and apixaban). Data were electronically extracted for patient demographics, clinical characteristics, and details of DOAC-related DRPs identified by pharmacists. Matching of DRP group to non-DRP group at a ratio of 1:2 based on gender, race, and DOAC was performed. The DRP group included patients with detected DRPs while non-DRP group included patients without them. Descriptive analysis was used to summarize patient characteristics and types of DOAC-associated DRPs. In the matched population, conditional logistic regression was used to calculate unadjusted (UOR) and adjusted odds (AOR) ratio to detect association of DOAC-associated DRPs with age, renal function, ≥2 comorbidities, and DOAC indication (atrial fibrillation [AF] vs venous thromboembolism). Results A total of 8432 patients prescribed DOACs were analyzed, which consisted of 827 (9.8%) and 7602 (90.2%) patients with DRPs and no DRPs, respectively. The top DOAC-associated DRP was inappropriate drug regimen (n = 487, 60.1%). After matching, 2403 patients were analyzed, consisting of 801 patients from DRP group and 1602 from non-DRP group. Factors associated with DOAC-associated DRPs were statistically significant for renal function at creatinine clearance (CrCl) of >30 to 50 mL/min/1.73 m2 (AOR: 1.42; 95% CI: 1.14-1.76; P = .002), 15 to 30 mL/min/1.73 m2 (OR: 1.94; 95% CI: 1.42-2.66; P < .001), and <15 mL/min/1.73m2 (OR: 2.35; 95% CI: 1.13-4.88; P = .022), respectively, compared with a CrCl of >50 mL/min/1.73 m2 and DOAC indication for AF (AOR: 1.84; 95% CI: 1.47-2.30; P < .001) compared with venous thromboembolism. Conclusion Inappropriate drug regimen was the most common DOAC-associated DRP. Impaired renal function and patients with AF increased the likelihood of DOAC-associated DRPs.


| I N T R O D U C T I O N
Direct oral anticoagulants (DOACs) were favored by clinicians as the armamentarium for managing thrombotic conditions such as venous thromboembolism (VTE), atrial fibrillation (AF), venous thromboprophylaxis, and cancer-associated thrombosis.Compared with warfarin, DOACs require less frequent monitoring and have fewer drug-drug and drug-food interactions, fixed dose regimens, and favorable safety profile with lower rates of intracranial hemorrhage [1].Anticoagulants have been associated with preventable adverse drug events (ADEs) such as bleeding and thrombotic complications [2].The risk of ADEs from DOACs were higher in patients with renal or hepatic impairment, the elderly, or those taking concurrent interacting drugs [3].Pharmacists can play a collaborative role in preventing ADEs with DOACs by identifying and resolving DOAC-associated drugrelated problems (DRPs) during the medication delivery process [4].
The Pharmaceutical Care Network Europe (PCNE) defined DRP as an event or circumstance involving drug therapy that actually or potentially interferes with desired health outcomes [5].Prescribing DOACs presents with unique clinical challenges, which can lead to inadvertent prescribing errors and poor outcomes.DOACs require dose adjustment in patients with renal impairment and should be avoided in liver cirrhosis Child-Pugh classification B or C. Dose adjustments should be considered in the elderly (aged >80 years).There is safety and efficacy uncertainty for DOACs prescribed to patients with extremes in body weight (<60 kg or >120 kg).Concomitant use of DOACs with coexisting medications that are potent inhibitors or inducers of hepatic cytochrome P450 enzymes or p-glycoprotein transporters should be avoided as this can lead to supratherapeutic or subtherapeutic levels of DOACs in the body, potentially increasing risk of bleeding or thrombosis, respectively [6].

Many studies have highlighted the challenges in prescribing
DOACs, and DOAC-associated DRPs resulting from prescribing errors were commonly detected [7,8].DRPs are undesirable and can lead to increased morbidity, mortality, and healthcare costs [9].Implications of ADEs from DOAC-associated DRPs can be catastrophic leading to bleeding or thrombotic events [10].To mitigate safety risks of DOAC prescription, pharmacists have an essential role in detection, prevention, and resolution of DOAC-associated DRPs [11].
As the anticoagulation paradigm shifts toward DOACs with its desirable characteristics and increasing number of approved indications, DOAC prescriptions will inevitably increase.Insight into the causes of prescribing errors will enable development of strategies to reduce prescribing errors and DRPs and prevent ADEs.There is a dearth of studies looking at prescribing patterns of DOACs in the Asian population.This study aims to look at the types of DOACassociated DRPs from prescribing rivaroxaban, dabigatran, and apixaban and identify factors linked to DRPs in a large academic hospital in Asia.

| Study design and ethics review
This was a single-center cross-sectional study conducted in Singapore General Hospital in patients prescribed with DOACs from January 1, 2017, to May 31, 2019.The study was part of the quality improvement project in DOAC-prescribing habits.Nevertheless, the study was submitted to institution's ethics review board and exempted from review.

Essentials
• Pharmacists are crucial in detecting and resolving errors from prescribing direct oral anticoagulants (DOACs).
• All DOAC prescriptions were reviewed by pharmacists in a large hospital and prescribing errors were recorded.
• The common prescribing errors were dose-related, avoidance of adverse events, and drug interactions.
• Patients with renal dysfunction and on DOAC for atrial fibrillation were associated with prescription errors.

| Study objectives
To examine the distribution and types of DOAC-associated DRPs among adults prescribed DOAC therapy, and identify factors that predispose patients to DOAC-associated DRPs.Factors that predisposed patients to DRPs were age, renal function, the presence of comorbidities (dichotomized into <2 and ≥2 comorbidities), and DOAC indication (AF vs VTE).These factors were recognized by previous studies to be associated with DRPs [12].Renal function was calculated using the Cockcroft-Gault equation based on actual body weight.

| Pharmacist interventions and DRP classification
A pharmacist intervention is defined as "a recommendation initiated by the pharmacist in response to a DRP during any phase of the medication supply process" [5].All drug orders required mandatory review by pharmacists prior to supply or patient administration as part of the institution's safety policy.The DOAC orders were reviewed for compliance to licensed recommendations from Health Science Authority of Singapore (Supplementary Table S1).The pharmacists were alerted to deviations from labeled dose or indication as well as patients' clinical condition (eg.bleeding) that contradicted use of DOAC  S2.

| DRP group and non-DRP group matching
The DRP group and non-DRP groups were compared with look for factors predisposing patients to DOAC-associated DRPs.Matching DRP group to non-DRP group at a ratio of 1:2 was performed based on gender, race, and type of DOAC.This ensured even distribution of confounding factors such as gender, race, and the DOAC prescribed among the 2 groups.The DRP group included patients with detected DRPs who received interventions by pharmacists while non-DRP group included patients with no DRPs detected and no interventions performed.Patients from the non-DRP group were considered once for each matched patient from the DRP group.A 1:2 ratio was chosen because the power of the study to detect a difference in the proportion of patients in DRP group between any pair of DOACs seemed to taper off at this ratio.

| Sample size calculation
Sample size was calculated based on 95% CI, alpha error of 5%, and power of 80% to detect odds ratio of at least 1.5 for factors linked with DOAC-associated DRPs.Based on DRP group to non-DRP group matching ratio of 1:2, the minimum total sample size was calculated to be 876, with 292 for DRP group and 584 for non-DRP group [13].Similarly, a conditional logistic regression was used in the matched study population.Statistical significance was achieved when the 2sided P value was < .05.Statistical analysis was performed using Statistical Package for Social Sciences (SPSS, 28th edition).

| Handling of missing data
A total of 326 (3.9%) patients had missing data for this study, which was managed using the mean substitution method where in place of the missing data, the mean value of the same variable was used [14].
Missing data included the weight (n = 254) or the serum creatinine (n = 72).Mean substitution was used to prevent patients with missing data being omitted from analysis, which may affect the study power.
Figure 2 showed that rivaroxaban (92.9%) was the most implicated DOAC associated with overdosing by renal function followed by apixaban (6.4%).Rivaroxaban (63.2%) was the most common DOAC implicated for underdosing by renal function, followed by apixaban (34.8%).Apixaban was more commonly implicated for DRP due to underdosing by renal function compared with overdose.
After matching for confounders (gender, race, and DOAC), a total of 2403 patients on DOACs were analyzed.The DRP group consisted of 801 patients and the non-DRP group consisted of 1602 patients.The number of patients on rivaroxaban, dabigatran, and apixaban in the control group was reduced by 66.6% (from 3517 to  1176), 76% (from 342 to 82), and 90.8% (from 3746 to 344), respectively, after matching (Table 3).The average age for patients on all 3 DOACs was similar to that of the unmatched population (70 ± 13 years), and the majority were on DOACs for AF (64.3%) and VTE (43.2%) (Table 3).unmatched populations, which showed similar trend (Supplementary Table S6).

| D I S C U S S I O N
This study gave insight into the types of DOAC-associated DRPs identified in an academic medical center and the factors that predisposed patients to such DRPs.Our study identified that inappropriate drug regimen was the most common DOAC-associated DRP, followed by avoidance of adverse events and drug interaction.Among the 3 DOACs, patients on rivaroxaban had the highest proportion of interventions, while patients on apixaban had the least.Subanalysis of the matched populations showed that patients with poor renal function and AF were more likely linked to DOAC-associated DRPs.
Our study detected 9.8% prevalence of DRPs among patients on DOACS and observed that DRPs were more prevalent in patients on rivaroxaban, followed by dabigatran and apixaban.This was comparable to a study by Viprey et al., which found prevalence of 8.4% of DRPs from DOACs among hospitalized patients and identified that patients on rivaroxaban for AF had the highest prevalence of DRPs when compared with other DOACs [15].Furthermore, DOAC-associated DRPs were frequent in hospitalized patients, which was a similar finding to our study wherein 92% of DRPs were identified in the inpatient setting.
This study observed that the most common DOAC-associated DRP was related to dosing.This corroborated with other studies, which identified that inappropriate dosing of DOACs was the most prevalent DRP [15,16].The sub analysis of inappropriate dose found that underdosing due to renal function was identified as the most common cause of inappropriate drug regimen, followed by overdose based on renal function and wrong dose for indication.Similarly, a meta-analysis study of DOAC-related prescribing errors revealed that common prescribing errors were dose-related, contributed by lack of adjustment for renal function, wrong indication, advanced age, and altered weight [17].
Among the recommendations made by the pharmacists, 68% were accepted by the physicians.This was within the acceptance rates of 60% to 80% in most studies where pharmacists were incorporated disease or on dialysis was not approved by the Singapore licensing body, Health Sciences Authority (HSA).However, the use of apixaban in this population was approved by United States Food and Drug Administration (FDA) based on pharmacokinetic studies [27].Pharmacist recommendations were based on local guidelines for apixaban dosing.However, some prescribers follow FDA recommendations and continue to prescribe apixaban in patients with advanced renal disease or on dialysis, while other prescribers would switch to an alternative anticoagulant.
Our study showed that the most encountered interventions among the prescribers were from cardiology and internal medicine, who were also the main prescribers for DOACs for patients with AF and VTE.However, our study did not investigate the prescribing discipline in the non-DRP group, which could affect the distribution of interventions among the disciplines.Another study conducted showed that clinicians from internal medicine had the highest rate of DOAC prescription followed by cardiology [28].This also corroborated our findings, in which the same 2 disciplines contributed to the majority of DOACs prescribed among patients in the DRP group.
This study identified that impaired renal function and AF were factors that predisposed patients to DOAC-associated DRPs.However, Sell et al. found that number of concurrent medications and age were factors that predisposed a patient to DRPs.Unlike our study, this study focused on DRPs in general and not specific to DOAC therapy [29].Other studies showed that elderly patients on DOACs for AF were at higher risk of bleeding compared with younger patients [30], had multiple comorbidities and higher mortality risk, and required drug therapy optimization to prevent poor outcomes [31].Our study were associated with increased risk of bleed [32].
From our study, we found that patients on DOACs for AF were 1.8 times more likely to be associated with a DOAC-associated DRP compared with patients on DOACs for VTE.This could be due to  [38,39].A qualitative study showed that improving the safety of DOAC use should be a multidisciplinary and multifaceted approach involving doctor, pharmacist, and patient and system level improvement strategies [40].
Our study had several limitations.The prevalence of DRPs could be underestimated since only electronically documented interventions were extracted.We were unable to collect verbal or manually documented interventions.Also, we could not rule out undetected DRPs nor patients with DRPs resolved from earlier prescriptions.Handling of missing data via mean substitution, which may lead to inconsistent bias and data replacement, may not reflect the true value.We did not perform a sensitivity analysis excluding patients with missing data.
Being an observational study, patient adherence and concurrent medications were not examined.Although matching was performed to ensure that confounding factors were evenly distributed between the 2 groups, some residual confounders may remain.One of these factors was care setting because hospitalized patients tend to be more medically unstable compared with patients in outpatient settings and hence would encounter more DRPs.Our study but did not examine clinical outcomes such as bleeding and thrombosis or healthcare cost savings.

| C O N C L U S I O N
Common DOAC-associated DRPs were inappropriate drug regimen, avoidance of adverse events, and drug interactions.

RELATIONSHIP DISCLOSURE
There are no competing interests to disclose.
Data of adults aged 21 years with DOAC prescription were extracted from electronic medical records.Data included patient demographics (race, gender, and age) and clinical characteristics (body weight, chronic conditions, renal function, and DOAC prescriptions such as the type of DOAC and dose).The DOAC prescription and clinical characteristics prior to time of intervention were captured.
that could potentially lead to negative outcomes.The DRPs were addressed by the pharmacist by providing recommendations to the prescriber to change the drug order.Data of all DOAC-associated interventions were electronically extracted (DOAC, types of DRPs, recommendations, intervention outcomes, and prescriber disciplines).The DRPs were classified according to modified PCNE DRP classification.Any uncertainty in the classification of DRPs were discussed among the coinvestigators and consensus was reached based on the modified PCNE DRP definitions [5].The definitions for DRPs classifications and examples are displayed in Supplementary Table Descriptive analyses were used to summarize patient demographics, types, and distribution of DOAC-associated DRPs.Numbers (percentages) were presented for categorical data and mean values (SD) were presented for continuous data depending on the data distribution.To identify factors associated with DOAC-associated DRPs in the unmatched study population, a binomial logistic regression was used to calculate the unadjusted and adjusted odds ratio (adjusted for age, renal function, number of comorbidities, and DOAC indication).

Figure 3
Figure 3 showed that majority of interventions for DOACassociated DRPs originated from drug orders by cardiology (n = 284, 34%) and internal medicine (n = 193, 23%) disciplines.Table4showed the unadjusted and adjusted odds ratio for the factors associated with DRPs for the matched populations.The into the medical team to review medication orders[18][19][20].There were some reasons postulated in which 27% of recommendations were not accepted or partially accepted.We elucidated that majority of DRPs addressing suboptimal doses of DOACs were not accepted because physicians deliberately intended to reduce doses due to concern with increased bleeding risks especially in the setting of reinitiating DOACs postoperatively when hemostasis was not established, or in elderly, frail, or underweight patients.Clear communication on physicians' prescribing intent will prevent the pharmacist inadvertently detecting the order as a DRP and putting forth unnecessary recommendations.Dose adjustments for DOACs were complicated and multifactorial.Renal dose adjustment for DOACs was also dependent on the indication.For patients with AF, the renal threshold to adjust for rivaroxaban was creatinine clearance of <50 mL/min/1.73m 2 (dose is reduced from 20 mg/day to 15 mg/day) and not recommended for use when renal function falls to <15 mL/min/1.73m 2[21].Dabigatran could be used at full doses till creatinine clearance of <30 mL/min/ 1.73 m 2[22].Apixaban dose adjustments were multifaceted, which included considerations for serum creatinine, weight, and age[23].For VTE management, full therapeutic doses of all DOACs were recommended until creatinine clearance of <30 mL/min/1.73m 2 (for rivaroxaban and dabigatran) and <25 mL/min/1.73m 2 (for apixaban) in which DOACs should be avoided since most VTE studies excluded these patients[24][25][26].The differences in dose requirements for different indications accounted for 14% of dosing errors under "wrong dose for indication", which was also encountered in other studies[17].The study found disparity among prescribing patterns of apixaban in patients with advance renal disease (creatinine clearance < 15 mL/ min/1.73m 2 ) or on dialysis.The incongruence in prescribing practice could be due to differences in dose recommendations by various authorities.Therapeutic use of apixaban in patients with end-stage renal F I G U R E 2 Distribution of inappropriate drug regimen based on dose among the DOACs.
Factors that predispose patients to DOAC-associated DRPs were impaired renal function and AF.This study provides opportunities for implementation of focused system-level improvement strategies to reduce inappropriate DOAC prescription.FUNDING The publication cost was supported by SingHealth Group Allied Health.AUTHOR CONTRIBUTIONS M.Y.Y.C., S.K.K, and L.H.L. contributed to the conception of the study and methodology.M.Y.Y.C., S.K.K., X.Y.T., S.Y.H., and D.J.J.C. contributed to the data collection, analysis, and interpretation.M.M.Y.C. and D.J.J.C. contributed to the authorship of the manuscript.All investigators contributed to the revisions and gave final approval for publication.
Patient demographics and indications of unmatched population.
T A B L E 3 Patient demographics for matched population (matched 1:2).
population was heterogenous, with patients on DOACs for multiple indications, and had wider age distribution, and hence, we did not find an association between age and comorbidities with DOAC-associated DRPs.creatinine clearance of >50 mL/min/1.73m 2 .As DOACs rely on renal function for elimination, dose adjustment and increased monitoring are warranted to prevent accumulation of DOAC levels, which can lead to a bleeding event.Whittemore et al. showed that patients on DOACs with impaired renal function accompanied by other factors IT) enhancements to prompt renal dose adjustments, access to dose guides, and training are some strategies to improve prescribing habits.Studies showed that additional IT enhancements implemented to support the prescription of DOACs could lower the number of DRPs AF, atrial fibrillation; CrCl, creatinine clearance; DRP, drug-related problem; VTE, venous thromboembolism.F I G U R E 3 Distribution of interventions based on discipline.8 of 10 -