FormalPara Key Points

ActionADE is a novel adverse drug event reporting platform that is integrated with a central provincial drug dispensing database to enable communication with community pharmacists.

Reporting of known ADEs to older medications (e.g., hydrochlorothiazide) was common in ActionADE.

Among patients experiencing prescription drug-related harm, every tenth adverse drug event reported in ActionADE and shared with community pharmacies resulted in one avoided drug re-exposure.

1 Introduction

Adverse drug events (ADEs) are unintended and harmful events associated with medications, and represent a major burden on patients and the health care system [1, 2]. They are a leading cause of unplanned emergency department (ED) visits and hospital admissions [3,4,5]. Between 8.2 and 12.0% of ED visits to Canadian acute care hospitals are ADE-related, 11.9 to 36.9% of which require admission; 32% of ADE-related ED visits are repeat events, of which 75% are preventable [3,4,5,6,7].

Preventable repeat ADEs occur because ADE information is rarely shared effectively within a patient’s circle of care [8]. Electronic medical record systems generally do not transmit ADE information between systems due to lack of uniform data standards and interoperability [9, 10]. As a result, serious ADEs leading to ED visits and hospital admissions are not automatically communicated to the medication’s prescriber or the dispensing community pharmacy. In a previous study, only 22% of ADE notifications were integrated into patient charts with re-exposures to culprit medications occurring in 27% of patients within six months of an ADE-related admission [8].

Voluntary reporting systems or websites such as Health Canada’s adverse drug reaction reporting form [11] were designed to enable documentation of ADEs for post-marketing surveillance. These systems are poorly suited to capture the complex nature of ADEs and are not integrated into electronic medical records or medication dispensing databases [12]. Adverse drug event reporting into stand-alone electronic systems or websites is time consuming for clinicians and does not result in a perceived benefit to the patient [12]. Despite ADE reporting being mandated in Canada through the Protecting Canadians from Unsafe Drugs Act (i.e., Vanessa’s Law) [13], clinicians often only document ADEs in hospital charts [14, 15]. As a result, existing systems like British Columbia Patient Safety and Learning System (BCPSLS), a web-based, free-text entry ADE reporting form that communicates events directly to Health Canada [16], are not protecting patients from re-exposures due to lack of interoperability or clinical utility.

We designed, piloted, and refined a web-based software application called ActionADE to address this gap in reporting. ActionADE allows health care providers to document ADEs in a rapid, user-friendly, and standardised manner at the bedside, and subsequently communicates the most relevant information to other systems [17,18,19]. ActionADE interoperates with PharmaNet, British Columbia’s (BC’s) central medication dispensation database, where ADE data are stored and automatically pushed to community pharmacy systems. When community pharmacists attempt to re-fill a prescription for a patient with an existing ADE record for the same or a same-class medication (subsequently called a culprit drug), a preventative alert appears. Pharmacists must respond to the alert in order to continue dispensing the medication.

2 Objectives

Our primary objective was to quantify the time needed to report and describe ADEs reported in ActionADE during the first two and a half years of its implementation. Secondary objectives were to explore patient sex- and age-based associations among reported ADEs, and to assess the rate of re-dispensations of culprit medication after sharing standardised ADE information via PharmaNet with community pharmacies.

We hypothesised that ADE reporting in ActionADE would require little time commitment, adverse drug reactions would comprise a large proportion of reports in ActionADE, patient sex and age would be associated with ADE types and outcomes, and that ADE reports would be used to inform community pharmacists’ re-dispensation decisions.

3 Methods

3.1 Design

This was a prospective observational cohort study in four hospitals in the Lower Mainland of BC, Canada between April 1, 2020 and October 31, 2022.

3.2 Setting

ActionADE is a web-based clinical ADE reporting application that was pilot-tested as a paper form at Vancouver General Hospital (VGH) in 2018, prior to its beta release in 2019 [17, 18]. We achieved integration of ActionADE with PharmaNet on December 7, 2020, after which completed ADE reports could be transmitted to PharmaNet [9, 20, 21]. We expanded ActionADE’s implementation to four hospitals (Vancouver General, Saint Paul’s, Lions Gate, and Richmond Hospitals), with implementation support focused on pharmacists working primarily in EDs and on medical wards, as they would be the most likely to see incoming patients with ADEs to outpatient medications. Our in-hospital implementation methods are described in detail elsewhere [22]. Community pharmacists received email communications about the intervention from their pharmacy software vendor, were able to access information on a website designed by the BC Ministry of Health, and had access to a continuing education webinar hosted by the BC Pharmacy Association [23, 24]. On May 3, 2021, we began a randomised controlled trial in which we transmitted adverse drug reaction and allergy reports (a subset of ADEs) to PharmaNet in the intervention group. At this time, we were able to determine the number of ActionADE alerts and rates of culprit medication re-dispensation in community pharmacies only for patients enrolled in the intervention group for this trial [20].

3.3 Participants

We included consecutive patients for whom practicing clinicians (i.e., hospital-based clinical pharmacists, physicians, or nurse practitioners) documented an ADE in ActionADE. There were no patient-level exclusions. We considered an ADE report complete if the clinician entered information on the culprit drug, ADE type, symptom/diagnosis, certainty, and outcome into ActionADE. We excluded duplicate or incomplete ADE reports and those marked as erroneous or where a clinician changed the ADE certainty to “Refute” after the initial report was made (e.g., after test results became available).

3.4 Data Collection and Variables

Clinical pharmacists, nurse practitioners and physicians reported ADEs into ActionADE during the process of clinical care provision. ActionADE automatically imported the patient’s name, age, biological sex at birth, and their 14-month PharmaNet medication dispensation history. Clinicians used these data to obtain a best possible medication history [25]. Clinicians then selected the suspected culprit drug(s) and entered ADE type, symptoms/diagnosis, certainty, and outcome using dropdown menus and/or predictive entry from standardised pick lists [9]. Clinicians were able to visualise definitions for the dropdown options when hovering over the term in the entry form (Appendix Table 1). Clinician diagnosed and entered ADE types including adverse drug reactions (ADRs), allergies, incorrect drugs, subtherapeutic dose, supratherapeutic dose, treatment failure, drug withdrawal, drug interaction, drug misuse, non-adherence, and other (e.g., drug-disease interactions, drug-food interactions, treatment without indication). Clinicians entered up to three symptoms and diagnoses relevant to the event from the MedDRA®Footnote 1 Preferred Terms list, indicated their certainty as to whether the event was caused by the drug (Certain, Probable, Possible, Unlikely, Refute) and the most serious outcome of the event (listed here from least to most severe: Unknown, Other, Worsen Pre-existing Condition, Emergency Visit, Hospitalisation Extended, Hospitalisation, Permanent Disability, Life Threatening, Fetal Defect, Death). We defined serious events as those associated with fetal defect, permanent disability, hospitalisation or extended hospitalisation, life-threatening events, and/or those resulting in death.

When patients with ADE reports in ActionADE and PharmaNet went to a community pharmacy, the community pharmacist received an alert if the patient tried to fill a prescription for a culprit or same-class drug to which an ADE had been recorded. Community pharmacists could only continue with the dispensation by responding to the alert using pre-specified response codes. We accessed the community pharmacists’ response codes through Population Data BC, which provides researchers access to publicly funded community pharmacist interventions and medication dispensations [20, 26]. We considered all alerts that were overridden, as dispensed medications, and non-overridden alerts as non-dispensations.

3.5 Statistical Analyses

Where available, we used datetime stamps from the system from the beginning and end of data entry to assess the time it took clinicians to complete a report. We presented summary statistics on the top culprit drugs, ADE type, symptoms/diagnoses, certainty, and outcome by ADE seriousness. We highlighted common ADE types and symptoms/diagnoses for the most commonly reported culprit drugs by ADE seriousness. We calculated descriptive statistics for age and sex characteristics of patients in the cohort (e.g., means/standard deviations, median/interquartile range, or frequency & percentages) and examined associations between age and/or sex and ADE characteristics using regression models adjusted for clustering at the patient and hospital level. We entered age as a continuous variable into regression models, and multiplied effect sizes by 10 to create effect sizes per 10-year groupings. We calculated rates of attempted re-dispensations and avoided re-dispensations among patients with ADRs or allergies enrolled in the intervention group of a randomised control trial [20]. We considered a p value less than 0.05 as statistically significant. We suppressed cells with fewer than five patients or reports to ensure privacy and confidentiality.

4 Results

During the study period, 3591 ADE reports were initiated by clinicians, 347 of which were subsequently refuted, 70 marked as erroneous, 70 incomplete, and 32 duplicates, and were thus excluded (Fig. 1). The median time required for clinicians to complete an ADE report was 119 seconds (IQR 61–235 s) (data not shown).

We included 3174 ADE reports which occurred in 2730 patients (Fig. 1). The average age of patients was 67 years and half were male. Most patients had only one report (87%), with a maximum of six reports (Table 1). Serious ADEs occurred in 30.6% of patients (837/2730). Male patients were more likely to experience a serious ADE than females (53.8% vs 46.2%; p = 0.029) (Table 1). Among the 3174 included ADE reports, the most common ADEs were ADRs (73.7%), non-adherence (11.3%), and allergies (7.8%; Table 2). The most commonly implicated culprit drugs included hydrochlorothiazide (3.8%), ramipril (3.6%), and furosemide (3.1%). The most commonly reported ADE manifestations were hyponatremia (7.0%), acute kidney injury (5.5%), and rashes (4.2%). Clinicians assigned a certainty level of probable or certain in almost three-quarters of the reported ADEs (72.2%; Table 2).

Fig. 1
figure 1

Flow diagram of adverse drug event (ADE) patients and their events through the study

Full size image
Table 1 Descriptive statistics of included patients, by severity
Full size table
Table 2 Descriptive statistics of included ADE reports, by severity
Full size table

Certain ADEs (i.e., common culprit drugs, types and symptoms) were reported more frequently than others (Tables 3, 4). Adverse drug reactions to hydrochlorothiazide resulting in hyponatremia represented 50.0% of all reports for hydrochlorothiazide (60/120) and 1.9% of reports overall (60/3174; Table 3). These reports were also the most common among the 906 serious ADEs, where they represented 55.6% of all serious ADE reports to hydrochlorothiazide (25/45) and 2.8% of serious ADEs overall (25/906; Table 4). Adverse drug reactions to empagliflozin were commonly reported and resulted in euglycemic diabetic ketoacidosis in 50.0% of empagliflozin reports (34/68), and 1.1% of the 3174 overall reports (34/3174; Table 3). These ADEs were common among serious ADEs, with 60.0% of serious ADEs to empagliflozin resulting in euglycemic diabetic ketoacidosis (2.0% of all serious ADEs; Table 4). Adverse drug reactions to chlorthalidone were less common (N = 43), but 64.7% (11/17) of all serious reports for chlorthalidone were ADRs resulting in hyponatremia, making up 1.2% of all serious ADEs (11/906; Table 4).

Table 3 Characteristics of the most commonly reported adverse drug events (n ≥ 5) among the top 10 most commonly reported culprit drugs
Full size table
Table 4 The most commonly reported ADEs (n ≥ 5) among 837 patients with 906 serious ADEs
Full size table

The types of ADEs reported varied by patient age, with drug misuse, allergies, and non-adherence being more commonly reported among younger patients, whereas dose-related ADEs, use of incorrect drugs, and ADRs were more common among older patients (Fig. 2). Younger patients were more likely to experience a worsening of a pre-existing condition or a change in medication (i.e., “other”), whereas older patients were more likely to be hospitalised due to their ADE (Fig. 3). Males were more likely to report ADEs related to non-adherence, whereas females were more likely to have an allergy reported (Fig. 2). Males were more likely than females to have their hospitalisation extended due to an ADE (Fig. 3). Males experienced life-threatening ADEs at a younger age than females (mean age of 54 vs 73; p = 0.001).

Fig. 2
figure 2

Adjusted effects of patient age and sex on type of adverse drug event (ADE) reported (N =3174). *We adjusted each estimate for the other variable in the regression (i.e., age for sex and sex for age) and for clustering by patient and presenting hospital

Full size image
Fig. 3
figure 3

Adjusted effects of patient age and sex on the outcome of adverse drug events (ADEs) reported (N = 3174). *We adjusted each estimate for the other variable in the regression (i.e., age for sex and sex for age) and for clustering by patient and presenting hospital

Full size image

During the study period, data from 639 adverse drug reactions and allergies that occurred in 592 patients were transmitted to PharmaNet. Of 592 patients with transmitted reports, 200 (33.8%, 95% CI 30.0–37.6%) attempted to have the culprit drug re-dispensed. Of these, 44.5% (89/200; 95% CI 37.6–51.4%) occurred within two weeks of the ADE report. Community pharmacists did not re-dispense the culprit drug in 33.0% (66/200; 95% CI 26.5–39.5%) of cases when presented with information about the prior ADE. In 33 attempted re-dispensations (33/200; 16.5%, 95%CI 11.4–21.6%), community pharmacists contacted prescribers for advice resulting in 15 (15/200; 7.5%, 95% CI 3.9–11.2%) re-dispensations and 18 (18/200; 9.0%, 95% CI 5.0–13.0%) dispensation reversals.

5 Discussion

ActionADE is one of the first interoperable electronic ADE reporting platforms integrated into the clinical workflow of hospital clinicians and community pharmacists that transmits relevant ADE information across health sectors to inform clinical care. ActionADE captured a high volume of reports in 18 months from four participating hospitals. We attribute the high volume of reporting to the user-friendliness of ActionADE and the short amount of time required to generate reports (<2 min/report). Adverse drug event reporting requirements through Vanessa’s Law were met as a by-product of data generated for clinical care provision [27]. We believe these features and others described below motivated clinicians to report ADEs [12].

To characterise ADEs resulting in hospitalisation, prior studies regularly collected data retrospectively (i.e., from administrative data or through medical record review) [28,29,30,31,32]. In our prospective cohort, we reported fewer allergic reactions (8 % vs 13–23%) and more ADRs (74% vs 0.7–29%) than these retrospective studies. Rates of reported ADRs vary significantly between studies using retrospective versus prospective case identification [28, 33,34,35,36]. Administrative data are poorly sensitive for ADEs, often capturing fewer than 10% of events depending on the algorithms used [35, 37, 38]. As a result, retrospective data collection will systematically underestimate ADRs like hyponatraemia and acute kidney injury, which are difficult to diagnose with certainty. ActionADE allowed clinicians to document ADEs even when they may not have been certain about their diagnosis by allowing them to rate the certainty of the ADE diagnosis (e.g., suspected vs probable), and providing the option to subsequently edit or even refute events. We excluded refuted events in our analyses, 95% of which were allergy reports. Had they been included the proportion of ActionADE allergy reports would be equal to those previously been reported (17% vs 13–23%) [28, 30]. ActionADE’s design may increase reporting of select events by addressing some of the limitations encountered with other methods commonly used to generate ADE data for surveillance, quality improvement and research, and may help redirect prevention efforts to more prevalent and serious ADEs for greater patient and system-level impact.

The most commonly reported ADEs in ActionADE were ADRs to diuretics, ACE inhibitors, and anticoagulants. These findings align with those of prior studies in which the most commonly reported ADRs were from the cardiovascular medication class [28, 29]. Thiazide diuretics, including hydrochlorothiazide and chlorthalidone, were responsible for over one-fifth of reported events, most manifesting as hyponatremia, and commonly requiring hospitalisation. Other antihypertensive agents (e.g., ramipril, amlodipine) rarely caused ADEs that clinicians reported. While thiazide diuretics do not feature among the top ten drugs prescribed in BC, ramipril and amlodipine do [39]. This indicates that thiazide diuretics may cause a disproportionate number of serious ADEs in comparison to alternative blood pressure-lowering agents. Spontaneous reporting into stand-alone systems remains the basis of many international post-marketing drug surveillance systems. Thiazide diuretics have been on the market for some time, and hyponatraemia is a known side effect [40]. It is likely that these reactions are not being spontaneously reported to post-market surveillance systems, where reporting is driven by drug novelty, reaction severity, and the ease with which the ADE can be diagnosed [15]. Thus, they have not generated safety signals to trigger further investigation, even though they are among the most common events seen in clinical care. While ActionADE was designed primarily as a clinical communication tool, not a surveillance system, its data provide an opportunity to better understand which ADEs are causing the greatest burden to both patients and the health care system while supporting more patient-centred and individualised care.

We found that younger ED patients were more likely to be diagnosed with allergies, non-adherence and drug misuse whereas older patients were more likely to experience ADRs and dosing issues. These findings are in line with studies of similar phenomena in other settings [41,42,43,44,45]. Male patients were more likely to be diagnosed with non-adherence compared to females who were diagnosed more frequently with allergies. It is unclear whether these differences were due to biological sex- or gender-related factors [46]. We also saw differences in reported rates of life-threatening and fatal ADEs through interactions between sex and age, but were unable to draw conclusions about why those differences occurred without further understanding of the patient pharmacokinetics, behaviour and access, as well as prescriber and clinician reporting behaviour.

ActionADE has the potential to improve patient safety. One in three patients diagnosed with an ADE to an outpatient medication that was reported in ActionADE and shared via PharmaNet with community pharmacies sought a re-dispensation of the culprit drug. Provision of patient-specific medication-class level information about prior ADEs was associated with avoided re-dispensations of same or same class medications. This potentially prevented repeat ADEs in one in ten patients with an ADE report shared with PharmaNet.

This study is not without limitations. First, while we aimed to address the limitations of many other ADE reporting systems through ActionADE’s user-friendly design and integration [10], reporting remained an additional task for care providers. Thus, underreporting remains, the degree of which we were unable to estimate. Providers may have been less likely to report ADEs when ongoing medication use was indicated (e.g., hypoglycaemia from insulin in diabetic patients). This could have accounted for the lower proportion of events to warfarin or insulin reported in ActionADE compared to other studies, as both drugs are deemed essential medications that would have needed to be continued in almost all cases [30]. Second, reports in ActionADE are not representative of all ADEs to outpatient medications, as reports were generated in acute care hospitals where ActionADE is implemented. Our results are thus not representative of events treated in other health care locations (e.g., walk-in clinics, general practitioners’ offices, long-term care). Third, the community pharmacists responded to ADE alerts in various ways including overriding them, consulting with a physician, changing the drug, or changing the dose. We did not have access to dispensing data, or any instructions provided to the patient outside of PharmaNet (e.g., to decrease the dose) and thus were unable to ascertain dosing or drug changes. The outcome of avoided re-dispensation is therefore a conservative measure that does not encompass all the actions a pharmacist may have taken in response to the ADE information presented. These data could be generalisable to other regions if they were to implement ActionADE or a similar system. ActionADE is web-based and could be used as a stand-alone system or integrated with existing legacy systems in other regions with development of workflows and systems architecture. Finally, we cannot infer causality between the display of ActionADE information and avoided re-dispensations, as data from the control group were unavailable for analysis. We will report these results once enrolment into the trial has been completed.

6 Conclusions

ActionADE is a novel ADE reporting platform that addresses many limitations of stand-alone systems. Reporting of known ADEs to older medications was common and may reflect their burden on patients and the health system. Reporting and communicating ADEs to community pharmacies using ActionADE was associated with avoided re-exposures to culprit or same-class medications.