- What is the problem to be addressed?
- Adverse drug events to outpatient medications
Medications have transformed the lives of Canadians suffering from debilitating medical and mental health conditions. As a result, medication use has increased dramatically, and with it the incidence of adverse drug events, harm caused by the use or misuse of medications.1 Adverse drug events result in 2 million emergency department visits and 650,000 admissions in Canada each year, and rank between the 4th and 6th leading cause of death.2,3 Optimizing the health benefits of medications while limiting their potential for harm is an urgent public health priority across patient populations, health settings, and medical disciplines.4,5
- Information discontinuity about adverse drug events
Patients who experience adverse drug events to outpatient medications commonly seek care in emergency departments because most events are unexpected, serious, and require urgent medical attention.3,6-10 After assessment and treatment, patients are usually discharged back into their communities where community-based providers assume their care. Yet, information about the adverse drug event(s) often does not follow the patient (Figure 1).11 While discharge summaries and consultation notes are intended to communicate information to family physicians, adverse drug event information is commonly omitted, incomplete, or documented using ambiguous terminology and non-standard abbreviations. This creates the potential for miscommunication and errors. Other care providers, such as community-based specialists may not receive these documents. Even if all care providers were to receive all relevant information, it is often filed away in thick paper charts or buried in outpatient electronic medical records (EMRs), where it cannot effectively remind providers of medication contraindications at the time of prescribing. In a prior study, 22% of adverse drug events diagnosed in-hospital were integrated into patients’ outpatient records, indicating a high degree of information discontinuity.12
- Repeat adverse drug events
Information discontinuity about adverse drug events leads to inappropriate medication re-exposures.13-15 In a Dutch study of 227 seniors admitted to hospital with an adverse drug event, 27% were restarted on the culprit drug within 180 days, most within 90.12 An Ontario study estimated that 55% of seniors admitted to hospital for hypoglycemia while on glyburide (which is contraindicated in seniors), or for a fall while on atypical neuroleptics or benzodiazepines (which increase falls risk), were re-exposed within 180 days.14 For example,
“I saw a diabetic who was discharged from hospital, where he was admitted for hypoglycemia (low blood sugar) due to glyburide (a blood sugar lowering drug). The physician asked him to stop glyburide, and gave him a prescription for gliclazide, a drug with a lower risk of causing hypoglycemia. But he presented here with low blood sugar. When I looked at the patient’s blister pack I discovered that both glyburide and gliclazide had been dispensed. The patient was given a discharge prescription for gliclazide, but no note was made to discontinue glyburide, and neither the family physician nor pharmacist were aware of what had happened.”
Hospital pharmacist, 2012
We completed a multi-center study of 1,296 patients presenting to 3 BC emergency departments with an adverse drug event.16 Of these, 32.5% presented with a repeat adverse drug event, of which 75.3% were avoidable had information about prior adverse drug events been shared between healthcare providers (Table 1). The most common types of repeat events were adverse drug reactions (i.e., harm caused by the use of a drug at normal doses),1 and harmful events due to non-adherence.16,17
We have designed, piloted, and refined a software application called ActionADE that allows healthcare providers to document standardized adverse drug event information in a user-friendly and standardized manner at the patient’s bedside.18,19 After its integration with other health systems, it will use reported data to create patient-level safety alerts when providers attempt to re-prescribe or re-dispense the same or similar medications as ones that previously caused harm (Figure 2). In 2018, we released the beta (or pre-release) version of ActionADE, allowed early adopters to pilot it in one emergency department, and refined the software based on their feedback. We aligned partnerships with stakeholder groups, including patient partners, the BC Ministry of Health (Ministry) and health authorities. We developed a governance structure, negotiated an outcomes-based agreement, and developed an implementation plan (see Section 2.3.3).
ActionADE’s implementation and integration with other health information systems has the potential to reduce adverse drug events, and thereby prevent avoidable morbidity, enhance the patient and provider experience, and improve health outcomes and health system sustainability across BC.
- What are the research objectives?
We will conduct two parallel trials to evaluate the effect of ActionADE on patient and health system outcomes. The main trial will focus on evaluating ActionADE’s effect on re-exposures to medications that previously caused adverse drug reactions. We will leverage the trial’s infrastructure to conduct a secondary trial to collect preliminary information about ActionADE’s effect on non-adherence.
- Main Trial
Primary Objective: The primary objective of our main trial is to evaluate the effect of providing information continuity about adverse drug reactions between care providers and across health sectors using ActionADE on re-exposures to culprit medications over 12-months compared to standard care.
Secondary Objectives: Secondary objectives are to evaluate the effect of providing information continuity about adverse drug reactions on health services use, mortality, and healthcare costs over 12-months compared to standard care. We will compare outpatient and emergency department visits, admissions, hospital-days, mortality, and healthcare costs.
- Secondary Trial
Primary Objective: The primary objective of our secondary trial is to collect preliminary information about the effect of providing information continuity about non-adherence between care providers and across health sectors using ActionADE on adherence to the same medication over 12-months compared to standard care.
Secondary Objectives: Secondary objectives are to evaluate the effect of using ActionADE to provide information continuity about non-adherence on health service use, mortality, and costs over 12-months versus standard care.
- Why is a trial needed now?
- Adverse drug event prevention is an international public health priority
Medication use is rising due to an aging population and expanding treatment indications for chronic disease prevention and treatment.20 This has led to a steadily increasing burden of adverse drug events.5,21,22 As a result, the World Health Organization has identified adverse drug event prevention as an urgent public health priority internationally,5,23 with repeat events identified as an ideal target for intervention.14,16
- Implementing ActionADE meets identified provincial priorities
ActionADE meets three of four identified BC Health Priorities:24 ActionADE will improve primary and community care coordination for seniors with complex medical conditions who are at high-risk of repeat adverse drug events (Priority 1), and has the potential to reduce adverse drug event-related admissions for this group.15,25 ActionADE will help clinicians provide care to patients with mental health conditions and/or substance misuse who are prone to non-adherence by providing clinicians the ability to share information about non-adherence and drug-seeking behavior (Priority 1). ActionADE is a health technology innovation that will facilitate patient-centered, team-based care by facilitating information sharing across health sectors (Priority 4), and to rural/remote areas (Priority 3).
- Vanessa’s Law
In 2014, Vanessa’s Law (Protecting Canadians from Unsafe Drugs Act) came into effect.26 It mandates reporting of adverse drug reactions by health institutions. Currently, no software exists that enables frontline clinicians to report adverse drug reactions in a user-friendly manner (Figure 3).27 As a result, clinicians currently do not report adverse drug reactions,28,29 making it impossible for health institutions to comply with the new legislation.30 By ensuring optimal integration of ActionADE into clinical workflow and enabling reporting to positively impact patient safety, we anticipate that ActionADE will enable health institutions to comply with Vanessa’s Law.
- Timeliness of ActionADE’s integration with other health systems
Our plan to evaluate ActionADE is timely: Vancouver Coastal Health Authority (VCH) is engaged in a large-scale Cerner EMR implementation, which will provide us with an opportunity to scale ActionADE to other acute care hospitals across VCH, if ActionADE proves beneficial. Vancouver Island Health Authority (VIHA) has implemented the Cerner EMR, and thus, will be able to leverage work completed in VCH to integrate ActionADE with VIHA’s Cerner implementation in the future.
- Limitations of existing solutions
Few information technologies have been designed to prevent repeat adverse drug events, but have failed due to design limitations or lack of integration with other systems.17 In many prior systems, data entry was poorly designed and not adapted to clinical workflow resulting in underreporting of adverse drug events.17 Without documentation of events, systems could not generate automated alerts to prevent re-exposures.7,9,31,32 Most EMR systems have focused on allergy documentation and flagging, which make up less than 1% of clinically significant adverse drug events. One prior system enabled documentation of a broader range of events, but was limited to one hospital and did not address information discontinuity across health sectors.33
- Limitations of prior health information technology evaluations
Most prior medication safety technology evaluations demonstrated reductions in errors and potential adverse drug events, but only a few demonstrated tangible benefits on patient-oriented outcomes (e.g., actual adverse drug events; Table 2).34,35 We will address this by using patient-oriented outcomes selected with our patient partners. In addition, most prior health information technology evaluations used pre-post designs and did not control for important confounders or capture unintended effects.17,34-40 Therefore, evidence about their effectiveness is weak, and should not be used to guide evidence-based decision-making.17,35-42
- Why a randomized controlled trial is needed to evaluate ActionADE
Interrupted time series (ITS) designs that employ routinely collected data have emerged as robust quasi-experimental designs to evaluate the effect of system-level interventions on health outcomes. They can address some of the limitations of pre-post designs, and are less expensive than randomized trials.43 In addition, their results may have greater external validity than that of randomized trials due to the absence of volunteer bias.
Various limitations preclude our use of an ITS design to evaluate ActionADE: The most important is a lack of high-quality longitudinal data on adverse drug events.44-47 Administrative data and other retrospective adverse drug event identification methods suffer from poor sensitivity for adverse drug events, precluding their use to establish pre-intervention trends required for ITS designs. In addition, multiple VCH and provincial polypharmacy reduction interventions are currently underway which have the potential to influence rates of adverse drug events independently of our intervention, and would threaten to bias our results (history bias).48,49 Therefore, we will use a randomized controlled trial design to evaluate ActionADE, allowing us to evaluate causality between the use of ActionADE and health outcomes while minimizing potential sources of bias.50
- How will the results of this trial be used?
Our results will guide evidence-based decision-making about ActionADE’s scaling across VCH and BC, while informing its adaptation to and implementation in other settings, including physician offices and residential care. Our implementation evaluation (see Section 3.2) will allow us to capture lessons from early implementations that we can apply to later ones. Our results have the potential to inform decisions about how clinical information can be included in drug information systems in other provinces and countries.23 We are poised to scale ActionADE, as it is modular and can be incorporated into EMRs. We anticipate that this seminal trial will advance knowledge about information continuity and medication safety, lead to further research questions, and provide insights about how health information technology can be evaluated using experimental designs.
- Are there any risks to the safety of participants involved in the trial?
All participants whose medications are withdrawn due to an adverse drug reaction may experience negative physiological effects from withdrawal, and adverse events from newly started medications. 51,52 These are unavoidable risks of discontinuing and replacing medications which will be similar in the intervention and control groups, and occur as part of routine medical care.
In our main trial, communicating adverse drug reactions through ActionADE may lead to fewer re-exposures to culprit medications in the intervention group. While this may lead to fewer repeat adverse drug reactions, withdrawal of first-line therapies may increase the risk of poor disease-related outcomes. In the control group, adverse drug reactions will not be communicated to PharmaNet and community pharmacy computer information systems (CIS) using ActionADE. This group may be at higher risk of experiencing repeat events, but may also be more likely to remain on first-line therapies, and thus could experience better disease-related outcomes. Both of these risks are experienced as part of routine medical care. There is clinical equipoise as to which of these risks is greater.
Past studies have shown that implementation of health information technologies in high-pressure environments can be associated with unanticipated data entry errors leading to erroneous prescriptions reaching patients.40,51-56 To avert this risk, we have field-tested ActionADE extensively in an emergency department, allowing providers to become proficient with data entry before their entries impact clinical care. Others have reported risks including increased time to deliver care and alert-fatigue leading to inappropriate alert overrides.34,35,40,41 We have minimized these risks through software refinement, careful integration into clinical workflow, end-user training and engagement, and high-specificity patient-level (in contrast to generic) alerts.
As adverse drug events will be transmitted from ActionADE to PharmaNet and community pharmacy CIS in the intervention group, these data will be visible to a larger number of care providers. This poses a greater risk of loss of confidentiality of adverse drug event-related health information in the intervention group. However, PharmaNet access is strictly legislated, and only available to providers in the patient’s circle of care.57 Therefore, we believe this risk is small. In focus groups with patients with vulnerable conditions (e.g., HIV), patients were comfortable with this information being communicated given the access controls in place (unpublished data).
- PRELIMINARY WORK TO PREPARE FOR THE PLANNED TRIAL
- Opportunity for Intervention
In BC, community pharmacists enter dispensed medications into PharmaNet for billing purposes. PharmaNet is accessible to care providers in all health sectors, including hospitals, long-term care, and community-based offices.58 It contains a 96-character ‘Adverse Reactions’ field intended to capture information about adverse drug events. However, only community pharmacists can enter data into this field. No software was ever developed to enable hospital-based providers to enter information into this field. Currently, when community pharmacists use the ‘Adverse Reactions’ fields, they rely on anecdotal information reported by patients, and enter information in free-text format, leading to the use of non-standard terminology and abbreviations, creating the potential for errors. As a result, the ‘Adverse Reactions’ field is rarely used: Among 2,284 patients presenting to hospital with an adverse drug event, none were recorded in PharmaNet (unpublished data).31
- Designing ActionADE
In 2013, the Canadian Institutes of Health Research (CIHR no. 293546) funded our team to design ActionADE. The software captures information about adverse drug events and medication contraindications and has been designed to communicate this information to PharmaNet and other health systems. It is compatible with HL-7, a set of international standards for transfer of clinical and administrative data between software applications.
We developed ActionADE by completing ethnographic workplace observations and participatory workshops to understand work practices and workflow of diagnosing, treating and documenting adverse drug events, and understand barriers to reporting.18 We then outlined ActionADE’s high-level functional requirements:30,59-61
- Documentation must be rapid, user-friendly, and integrated into clinicians’ workflow;
- Adverse drug event reports must be standardized, and use unambiguous non-abbreviated terminology;
- ActionADE must be co-designed with end-users, and subsequently field-tested and refined;
- Users must be able to update adverse drug event reports, as events are diagnosed over time;
- Clinicians must be able to refute events if an alternative diagnosis is made;
- Adverse drug event reports must be patient-specific and impact clinical decision-making;
- ActionADE must be interoperable with other systems;
- Its integration strategy must enable communication to other clinicians in the patient’s circle of care;
- New medication prescriptions and dispensations must be checked against previous reports; and,
- Patient-specific medication-level alerts must have high specificity to avoid alert fatigue.62
To date, no other health information technology exists that meets these requirements.17,30
We developed ActionADE using an iterative process (Figure 4).18 We completed a systematic review and identified 107 adverse drug event reporting systems.63 We extracted information on reporting concepts and data fields, and used visual thinking software to remove duplicates and sort fields to understand differences and interdependencies. We completed 350 hours of ethnographic workplace observations to understand work practices and workflow.60 This allowed us to develop a preliminary data entry form and anticipate the need for end-users to update reports as new information became available.64 End-users iteratively refined paper-based drafts in workshops, prioritizing data entry elements while omitting redundant or irrelevant ones.61,64,65 End-users then piloted a paper-based version in clinical settings.19
- Preliminary work during the Business Case Development period
- Programming, testing and refining ActionADE
During the Business Case Development period (CIHR no. 398187; Score 4.23; Rank 1), we released the beta (pre-release) version of ActionADE in partnership with Connected Displays at Vancouver General Hospital (VGH) as a stand-alone application (Figures 5 & 6). Clinicians provided iterative feedback about the software in a test environment, and subsequently in the emergency department. This allowed us to test feasibility, refine ActionADE, and capture important lessons for our planned implementation and health system integration. We held participatory workshops with end-users to obtain feedback, and anticipate how, under what conditions, and when reported events should be displayed as alerts to other care providers, allowing us to calibrate the timing, location, specificity, and frequency of alerts to avoid alert fatigue.
- Aligning partnerships
On June 6th, 2018, the Ministry’s Executive Planning Board approved ActionADE as a new provincial standard for documenting adverse drug event information and approved its integration planning. We formed a Steering Committee with patient partners, and members from the Ministry, VCH, VIHA, the Lower Mainland Pharmacy Services, the BC College of Pharmacists, and the BC SUPPORT Unit to oversee ActionADE’s implementation planning, evaluation for the randomized trial. We developed subcommittees to oversee Clinical Workflow, refine Outcomes and Metrics, develop a Solution Architecture, and negotiate an Outcomes-Based Contract which established timing and value of performance-based reward payments and an independent audit partner. Our patient partners have been and will continue to be active participants at all levels of governance of this project, and in developing and executing the proposed trial. UBC Legal Counsel performed a Privacy Impact Assessment of ActionADE as a stand-alone application, while our team developed business requirements, drafted change management processes, conformance standards for community pharmacy integration, and policy implications in collaboration with Ministry. ActionADE’s preliminary solution architecture describing its relationships with other information systems is HL7 compatible (Figure 7). Given the complexity of ActionADE’s integration with other systems, we developed a phased implementation plan.
- Phased implementation plan
In Phase 1, we will integrate ActionADE with PharmaNet, BC’s medication dispensing database, and community pharmacy CIS using a Minimum Viable Product (MVP). The ActionADE MVP will leverage existing PharmaNet data fields to communicate adverse drug event information from participating hospitals to community pharmacies via PharmaNet. By using existing PharmaNet functionality, the MVP will be cheaper to build than if we added new data fields to PharmaNet, while allowing us to iteratively roll out the intervention and retain flexibility during the research phase, if refinements are needed. The ActionADE MVP will support our research needs, while allowing us to plan ActionADE’s broader health systems integration in BC.
In Phase 2, Ministry will fund the redesign of PharmaNet to develop the new data fields required for ADE documentation, integration with the acute care EMRs being implemented in BC (Cerner and Meditech), community EMRs, and community pharmacy CIS, if ActionADE proves beneficial. This will ensure access to adverse drug event information for providers across the continuum of care. Phase 2 will allow scaling of ActionADE in BC, including to 6 health authorities, and other health sectors (e.g., general practitioners).17,59
Due to the complexity of the health system integration required for Phase 2, the implementation of iActionADE will extend beyond the Rewarding Success grant. This proposal focuses on Phase 1, the research phase.
- PROCESS OF IMPLEMENTATION EVALUATION
- ActionADE integration with PharmaNet and community pharmacy CIS
We will start Phase 1 by integrating ActionADE with PharmaNet and community pharmacy CIS through the ActionADE MVP. This will enable hospital providers to use the ActionADE MVP to communicate information from hospitals to community pharmacies. We will start a qualitative process evaluation after validating the end-to-end ActionADE MVP build in a test environment, and after completing change management activities at each participating site. We have drafted high-level end-user engagement and change management strategies with Ministry. We will iteratively refine these in year 1 with our patient partners and key clinicians at each site to ensure that we take clinicians’ information needs, and the culture and organizational structure of each site into account. We believe that this will optimize our chances of successful implementation.
- Process of implementation evaluation
While randomized trials are regarded as the gold standard for establishing the clinical effectiveness of interventions, their results may not provide information about why an intervention was successful or not, or how a beneficial intervention may be replicated in other contexts.66,67 Process evaluation can help optimize implementations while providing insight to help distinguish between interventions that are inherently faulty (i.e., intervention failure) from interventions that were poorly delivered (i.e., implementation failure).
The primary objective of our process evaluation is to identify facilitators and barriers to the ActionADE MVP implementation. Secondary objectives include identifying contextual factors (such as differences in staffing mix or local variations in scope of practice), and the extent to which we were successful in altering adverse drug event documentation and communication at participating hospitals.66
We will conduct a mixed-methods process evaluation alongside the planned randomized trial. We will implement the ActionADE MVP in the three acute care hospitals participating in the randomized trial and will start process evaluation as we begin onboarding users in clinical settings. Trained research assistants will collect qualitative data by completing in-situ observations, interviews, and focus groups with ActionADE MVP users at each site to understand facilitators and barriers to ActionADE MVP’s use and identify contextual factors. We will consider six domains in which change must occur in relation to ActionADE MVP: technical, professional, organizational, economic, ethical, and legal.68 We will use an embedded researcher model where the researcher is an integral member of the intervention team implementing the ActionADE MVP.69 This will allow us to anticipate and observe unintended consequences.69 We will track insights from our qualitative process evaluation and how they influenced implementation processes using a study log. We will consider these in relation to our quantitative findings in order to develop hypotheses about how resolution of process issues may influence the adoption and use of the ActionADE MVP. We will analyze qualitative data after each implementation cycle to identify factors that can improve the next implementation. We will code qualitative data using NVivo. We will use quantitative methods to evaluate the extent to which we were successful in altering adverse drug event documentation and communication. Outcomes include the number and types of adverse drug events reported, and the number and type of reports transmitted to PharmaNet and visualized in community pharmacies. We will produce summary statistics appropriate to the data distribution and observe trends over time and by site. We will capture the number and type of alert overrides by querying the PharmaNet database, and analyzing the documented reasons for overrides.
Process evaluation will allow our team to learn from earlier implementations to optimize later implementations and enhance ActionADE MVP’s adoption, and ultimately inform iActionADE’s implementation strategy. Our qualitative data is likely to assist us in interpreting our quantitative results (e.g., by identifying possible explanations for differential adoption), and may assist in interpreting our trial’s overall findings. Identified contextual factors will assist us in understanding the trial’s generalizability to other settings and develop recommendations for future implementations. Finally, our process evaluation may allow us to identify unanticipated effects that we will need to track closely during the trial.
- THE PROPOSED RANDOMIZED TRIAL
- What is the proposed trial design?
We propose a quadruple-blinded randomized control trial among adults presenting to participating hospitals with an adverse drug event. Care providers, study participants, outcomes assessors, and statisticians will be blinded to treatment allocation. Clinical pharmacists will record diagnosed adverse drug events in the ActionADE MVP in emergency departments and on inpatient wards of participating hospitals. We will enroll patients presenting with an adverse drug reaction into our main trial, and patients presenting with an adverse event due to non-adherence into our secondary trial.
- What is the planned trial intervention?
- Experimental intervention: Information continuity
Patients in the intervention group will have adverse drug events and medication contraindications communicated to other providers using the ActionADE MVP. Their information will be transmitted and stored in PharmaNet and be made accessible to community pharmacists. In community pharmacies, adverse drug event reports will generate pop-up alerts when attempts are made to re-prescribe or re-dispense a medication to the same patient in which an adverse drug event or medication contraindication was reported. Community pharmacists will be able to manually override adverse drug event and contraindication alerts (while documenting the reason for the override) if they wish to re-dispense a culprit medication, or a same-class medication despite the pop-up alert.
- Control group: Standard care
Patients in the control group will have their adverse drug events and medication contraindications recorded in the ActionADE MVP, but their records will be retained locally, as is the current standard of care. Practically, this means that their information will be retained on the ActionADE MVP server without transmission to PharmaNet. Their information will not be visible to other providers via PharmaNet and will not generate automated alerts.
All patients will continue to have information about adverse drug events and medication contraindications incorporated into written emergency department, consultation, and medical progress notes, as well as into discharge summaries, as is the current standard. Care providers can print out reports they generate in ActionADE MVP or copy any documentation from the chart that they would normally provide to their patients, as per their current practice. All patients will be able to request copies of their information at any time. All participants will receive standard medical care during the index emergency department visit and hospitalization.
- What are the practical arrangements for allocating participants to trial groups?
We will stratify randomization by age category (<80, >80), and the hospital where the adverse drug event was recorded (based on login information). 70,71 A statistician otherwise uninvolved in the study will generate a list of treatment assignments using permuted blocks of varying size for each stratum. A randomization list for each stratum will be stored in the ActionADE MVP and will be concealed. Only ActionADE MVP’s developer will have access to the randomization sequence.
Clinical pharmacists will document adverse drug events (see letter from Lower Mainland Pharmacy Services) into the ActionADE MVP as part of clinical care. We will program an algorithm into the ActionADE MVP software that will automatically determine each patient’s eligibility for participation in the trial using the patient’s recorded demographic information (e.g., age) determined through ActionADE MVP’s linkage with PharmaNet, and information entered by pharmacists at the point-of-care (e.g., adverse drug event type). After the software has determined the patient’s eligibility and stratum the software will allocate the patient to the next unused treatment assignment of the appropriate stratum. If the allocation is to the intervention arm, adverse drug event information will be pushed to PharmaNet. If the allocation is to the control arm, adverse drug event information will be retained on the local ActionADE MVP server.
- What are the proposed methods for protecting against sources of bias?
We will minimize treatment allocation bias by generating unpredictable sequences, concealing the randomization lists from all study personnel, and using an integrated randomization module (currently under development). We will mitigate ascertainment and withdrawal bias by using administrative data (see Section 4.8) to determine outcomes. Only patients who move out of BC in the year after their index adverse drug event will be lost to follow-up. Contamination bias is possible; clinicians may become aware of the ongoing trial as they will be able to visualize standardized information about adverse drug events in the intervention group via PharmaNet, if they re-access those same patients’ PharmaNet profiles after data transmission from the ActionADE MVP to PharmaNet has occurred. Healthcare providers may also use other means to better communicate adverse drug events (e.g., by documenting adverse drug events more thoroughly in discharge summaries). To understand whether contamination occurred, we will randomly sample patients before the trial implementation, and compare the proportion of adverse drug events documented in their discharge summaries to those documented among control patients after implementation. We will conduct the primary analysis using intention-to-treat principles, in which patients with protocol violations will be analyzed in the group to which they were allocated.
- Research Ethics
Our ethics board has approved this protocol (H018-1332), and provided a waiver for obtaining informed consent as this trial meets the Tri-Council Policy Statement (TCPS) minimal risk criteria (Figure 8).72
- Rationale for requesting the waiver of informed consent
We seek to evaluate the effectiveness of a potentially sustainable public health intervention that is being introduced by the Ministry as a new part of routine care in BC. However, the process of gaining informed consent would likely have influenced patient and provider behaviour, and could have negatively impacted the internal validity of the trial. Patients would likely have been hypersensitized to the issue of adverse drug events, which could have led to increased reporting of this information to subsequent providers. This would have led to a dilution of the intervention effect, resulting in a downwardly biased estimate, while not reflecting patients’ usual behaviours. Similarly, patients diagnosed with non-adherence may have made greater efforts to take their medications, or could have been more relunctant to report non-adherence in the future. This would put them at greater risk of overly aggressive medical management due to subsequent providers uptitrating their medications based on false reports of adherence. Thus, it is imperative, that when evaluating ActionADE MVP’s impact on health outcomes, participant behaviours mirror their behaviours in real-life.73
- Minimal risk criteria
The documentation, transfer of information, and care of individuals in the control group is unchanged from the current standard. We believe that patients in the intervention arm may be more likely to have their medications permanently withdrawn than patients in the control group. However, both removal and re-exposure to culprit medications can lead to adverse outcomes, both of which are unavoidable risks that occur in routine medical practice. The probability and magnitude of the possible risks implied by participation in the research are no greater than those that patients would encounter outside of the trial. The automated communication of medication safety information is low-risk (TCPS2, Section 3.7a).72,74
We anticipate no negative impact on the physical, mental or spiritual health of participants, nor on their physical, economic, or social circumstances by participating in the trial. We will keep personally identifying information (e.g., provincial health number (PHN)) on the secure ActionADE MVP server with strict access controls that are subject to a Privacy Impact Assessment (already completed for the stand-alone ActionADE application; to be updated for ActionADE MVP in year 1). We link study data to administrative data through PopData BC, a multi-university, data and education resource facilitating interdisciplinary research. PopData BC will link individual-level data from ActionADE MVP’s server with adminstrative data using PHNs. Once linked , PopData BC will strip the records of identifiers (e.g., PHN; TCPS2, Section 3.7b). Our trial requires no additional contact with study subjects beyond what is required to provide standard care (TCSP2, Section 3.7e).
We will inform community pharmacists in the catchment areas of participating hospitals of the ongoing study to ensure that they are aware that their patients may or may not have new adverse drug event reports in their PharmaNet profiles, and that this information may be incomplete (TCPS2, Section 3.7d).
Finally, no robust alternative design is available to evaluate this public health intervention without bias (see Sections 1.3.7 and 4.5.1; TCPS2, Section 3.7c). It is for this reason, that we decided in collaboration with our patient partners, to not include quality-of-life or patient experiences as outcome measures for this trial, as this would have required obtaining informed consent and would have threatened the internal validity of our trial.
- Data safety monitoring board (DSMB)
A DSMB will monitor the study for harmful outcomes attributed to adverse drug events and brought to the attention of the research team. As we will ascertain outcomes using linked administrative data, which take time to accumulate and be released, we will be unable to conduct interim analyses for efficacy or safety. We will convene the DSMB within two weeks of being notified of any adverse event that may be attributable to the trial.
- What are the planned inclusion/exclusion criteria?
- Inclusion Criteria
We will enrol consecutive adults (>19 years) who are diagnosed with an adverse drug event in the emergency department or on a hospital ward of a participating institution, and whose event is reported in the ActionADE MVP. Patients with reported adverse drug reactions will be enrolled into the primary trial, while patients with reported adverse events due to non-adherence will be enrolled into the secondary trial.
- Exclusion Criteria
We will exclude patients without Medical Service Plan coverage (e.g., out-of-province patients), as this will prevent linkage with the administrative data necessary for outcomes ascertainment. We will also exclude patients who suffer adverse drug events to culprit medications that are not recorded in PharmaNet (e.g., over-the-counter medications, HIV/AIDs medications), as we will not be able to ascertain re-dispensing of these medications, our primary outcome, because these medications are not captured in PharmaNet.
- What is the proposed duration of follow-up?
We will link patient-level ActionADE MVP data, collected as part of routine care, to health outcome data in PharmaNet and other administrative health databases after 12 months of follow-up.
- What are the proposed primary and secondary outcome measures?
- Main trial
Primary outcome: The primary outcome will be culprit medication re-dispensing during follow-up. The summary measure will be the proportion of patients with a culprit medication re-dispensation within 12 months. The effect measure will be the between-group difference in the proportion with a 12-month re-dispensation.
Secondary outcomes: Secondary outcomes will be the number of outpatient and emergency visits, admissions, hospital-days, and mortality during the 12 months following the index adverse drug event. We will ascertain these using administrative data. As adverse drug events are not reliably coded within BC’s administrative data, we will be unable to reliably capture downstream adverse drug events, or attribute health services use to these adverse drug events, and will instead capture all-cause health services use.44,45,47
- Secondary trial
Primary outcome: The primary outcome will be adherence to culprit medications over the follow-up period. We will ascertain adherence using culprit medication re-dispensing from the PharmaNet database as a proxy for adherence. We will define adherence as the dispensation of >80% of prescribed medication doses within 12-months, based on the recorded medication dose and frequency, and the volume and dates of re-dispensations.
Secondary outcome: The secondary outcomes will be the number of outpatient and emergency visits, admissions, hospital bed-days, and mortality over the follow-up period. We will ascertain secondary outcomes using administrative data.
- How will outcomes be measured?
Community pharmacists must enter all dispensed medications into PharmaNet for billing purposes. Therefore, dispensing data are considered accurate and complete. We will capture the generic names of patients’ dispensed medications, prescribed doses, routes and frequencies, and the dates and volumes of dispensations through patient-level linkage of the ActionADE MVP data with PharmaNet data using PHNs. This will allow us to ascertain both re-dispensation outcomes for our main trial, and adherence for our secondary trial. We will capture inpatient and outpatient health services utilization, mortality, and cost by linking the ActionADE MVP data with the Discharge Abstract Database, Medical Service Plan, Vital Statistics, and Client Registry after the last recruit completes their 12-month follow-up period.
- What is the proposed sample size and justification for assumptions underlying the power calculations?
- Sample size for the main trial
We have estimated the required sample size based on the main trial. It is a superiority protocol with two independent groups randomized 1:1, with the primary outcome being the proportion of patients who experience a culprit medication re-dispensation within 12 months of the index adverse drug event. We believe that an absolute difference of 5% would be the minimal clinically important difference to change practice.
From our prior work, we anticipate that over one third of our patients will be seniors (i.e., >80 years of age).16 A previous study found that 26.6% (95% CI=17.3% to 38.5%) of seniors were re-prescribed a culprit medication within 6 months after an adverse drug event-related hospitalization.12 An Ontario database study estimated that 54% of elderly were re-exposed to a high-risk medication within 6-months of a presentation related to an adverse drug event.14 Using a conservative estimate of 20% of patients being re-exposed to culprit medications over a 12-month period in the control arm, we will require 2072 patients (1036 per group) to detect a 5% absolute reduction (from 20% to 15%) in culprit medication re-dispensing using an alpha of 5%, a beta of 15%, and the two-sided Z-test approach.
In a prior study, less than 1% loss to follow-up occurred within 12 months due to unresolved linkages between Medical Service Plan and PharmaNet data. This happens only when an individual leaves the province or loses Medical Service Plan coverage (e.g., non-payment), both of which are rare. To ensure adequate precision at 12 months anticipating 5% mortality, we will conservatively adjust for a combined 10% loss, therefore targeting a minimum sample of 2302 patients (1151 per group, based upon n/(1-0.1) adjustment) for our primary trial.
- Sample size for the secondary trial
We anticipate approximately 1450 patients with an index event of non-adherence to accumulate over that time for the secondary trial. We will have at least 80% power to detect risk ratios of 0.7 to 0.85 over a range of non-adherence proportions (0.2 to 0.5).
- Will health service research issues be addressed?
A robust analysis of ActionADE MVP’s economic impact is crucial to informing resource allocation decisions related to iActionADE’s planned provincial implementation by the BC Ministry of Health and the health authorities in Phase 2. We will conduct an economic analysis to examine the incremental costs and benefits of implementing the ActionADE MVP compared to the standard of care based on our primary trial. Conducting the analysis from the outset of our trial will allow us to complete a comprehensive micro-costing and cost-effectiveness analysis.
Our objectives are to evaluate the differences in costs, benefits, and the cost benefit ratio of communicating adverse drug event information using the ActionADE MVP compared to current standard of care.
We will begin the project with a micro-costing analysis to examine the inpatient and outpatient costs of patients who have experienced an adverse drug event, as well as the costs involved in developing and administering the ActionADE MVP software. We will use a full absorption costing approach following the identification, measurement, and valuation protocol to ascertain patient costs.75 Given the heterogeneity of patients with adverse drug events, we will use both top-down gross costing and bottom-up activity-based micro-costing to estimate fixed utilization costs.76-79 We have drafted a process map reflecting varied permutations of patient journeys (Figure 9), and a universal costing algorithm to empower other researchers estimating costs associated with adverse drug events in their own jurisdictions.80 We will ascertain the cost of building, implementing, and maintaining the ActionADE MVP through consultation with its developers, change management leaders, and health care managers and decision-makers, and will include costs of developing, integrating, implementing and maintaining the software. Figure 10 depicts the mathematical equations.
Given the primacy of Ministry and health authority stakeholders to this project, we will adopt health system and governmental perspectives in our analysis. We will use a Trial-Based Economic Analysis approach to conduct a cost-consequence and cost-effectiveness analysis, since this research will be taking place alongside a randomized trial.81,82 Using the approach developed by Glick et al., the targeted sample size for the trial will deliver an adequate level of power to ascertain the cost-effectiveness of the intervention (Figure 11).83 We will begin the economic evaluation with a cost-consequence analysis where the costs of the trial, including acute care utilization and intervention expenditures will be presented in tabular format along with its consequences (i.e., number of re-dispensations, emergency department visits, hospital length of stay, etc.) to provide maximum transparency for decision makers (Figure 12).84
Within the cost-effectiveness analysis, we will divide incremental costs by incremental differences in units of effectiveness to obtain a ratio of cost-effectiveness.81,82 The unit of effectiveness will be the number of acute care-free days that patients experience after their initial visit for an adverse drug event. An equation of this cost-effectiveness ratio is presented in Figure 13. We will analyze any missing data to identify appropriate methods to address gaps, possibly including multiple imputation.85 We will account for sampling or stochastic uncertainty in the incremental cost effectiveness ratio by using nonparametric bootstrapping methods to resample cost-effect pairs from the trial data, and plot those points on the cost-effectiveness plane.86 We will address decision uncertainty by plotting a cost-effectiveness acceptability curve.87 In comparing costs to benefits in this way, we will discern the relative value of the ActionADE MVP in comparison to standard care, and in so doing will provide a quantifiable measure of cost-effectiveness to help inform future policy decisions in relation to longer-term use, and implementation of iActionADE for scaling across the province and to other health settings.
- What is the planned recruitment rate?
We plan to recruit patients from the emergency departments of VGH, Saint Paul’s (SPH), and Richmond Hospitals (RH). These hospitals currently do not have the Cerner EMR implemented. Because the Cerner implementation strategy has recently changed, and the Cerner implementation sequence has not yet been finalized, we have strategically decided to evaluate ActionADE through a MVP that is independent of the Cerner implementation strategy or sequence. This strategy ensures the feasibility of our trial regardless of the Cerner implementation, and provides us with the flexibility of recruiting other VCH hospitals, if needed.
The planned VCH hospitals (VGH, SPH and RH) have a combined emergency department census of 197,000 adult visits per year. Each hospital has a minimum of one full-time dedicated clinical pharmacist in the emergency department (VGH has 3 full-time equivalents distributed over 7 days/week). The emergency department pharmacists will provide over 170 of combined pharmacy hours of coverage per week and will document adverse drug events in the ActionADE MVP (see letter from the Lower Mainland Pharmacy Services) in addition to ward pharmacists and physicians. Emergency department pharmacists are typically scheduled during the highest volume times of the day and week, allowing us to maximize recruitment during those times.
- Recruitment rate for the main trial
Previous studies have found that 4.3-6.3% of adults present to the emergency department with an adverse drug reaction (Table 3).3,7,8 We have estimated the recruitment rate for the main trial using a conservative estimate of 4% of adults presenting with an adverse drug reaction (7800/year), 70% being evaluated by a pharmacist (i.e. documented in the ActionADE MVP) enabling us to enroll from 15 adverse drug reactions per day across three sites. Even if only 40% of adverse drug reactions meet the trial’s inclusion criteria, we expect to be able to recruit 180 patients for the main trial monthly and will attain the required sample size within 12-14 months (Table 4). We will track the numbers of enrolled patients on a monthly basis for each participating site and will use our planned process evaluation (see Section 3.2) to troubleshoot lower than anticipated rates of enrolment. We will continue enrolment until we have attained our required sample size for the main trial.
- Recruitment rate for the secondary trial
Prior studies have found, that 2.9-4.1% of adults present with an adverse event due to non-adherence(Table 3).3,7,8 We have estimated recruitment for the secondary trial using a conservative estimate of 2.5% of adults presenting with adverse events due to non-adherence (4900/year), 70% being evaluated by a pharmacist (i.e. documented in ActionADE MVP) enabling us to enroll from 10 patients per day across three sites. If only 40% of adverse events due to non-adherence meet the secondary trial’s inclusion criteria and are documented in the ActionADE MVP, we expect to recruit ~1450 patients into the secondary trial in 12 months (Table 4).
In earlier research, we successfully enrolled 10,807 patients at high-risk for an adverse drug event at our planned study sites over a 10-month study period from the same three participating emergency departments.31
- Are there likely to be any problems with adherence?
As patient enrolment and delivery of the intervention will be fully automated, we do not anticipate any problems related to adherence.
- What is the likely rate of loss to follow up?
Follow-up losses will be minimal as we will capture data using administrative data. Loss to follow-up occurs for patients who have moved out of BC, or have duplicate PHNs. We expect losses to follow-up to be negligible, as only 0.009% of BC’s population emigrated in 2016.88 Deaths before completion of follow-up (12-month) will decrease the power/precision of the trial. We have increased our enrolment by 10% to account for this.
- How many centers will be involved?
We have selected two tertiary care hospitals (VGH, SPH) and one urban community hospital (RH) in the Lower Mainland. We have previously collaborated successfully with these sites, and have excellent engagement with their pharmacy departments.7,8,31 We plan to implement the ActionADE MVP consecutively at VGH, then SPH, and finally at RH, and can add a fourth site, if required, through our use of an MVP approach rather than a Cerner-integrated product. This minimizes risk of not enrolling sufficient patients for our trial.
- What are the proposed types of analyses?
We will summarize baseline data by treatment group using frequency and percent (for categorical variables) and mean and standard deviation, or quartiles (for continuous variables).
- Main Trial
We will perform intention-to-treat analyses for both primary and secondary outcomes. We will use cumulative incidence functions to estimate the cumulative incidence of culprit drug re-dispensation at 12 months, treating death prior to culprit drug re-dispensation as a competing risk and emigration from the province as a censoring event. We will stratify all analyses by age category and hospital. We will determine point and interval estimates for the between-group differences. We will use Gray’s test to test for a between-group difference. We will analyze secondary outcomes using zero-inflated Poisson regression (number of outpatient and emergency department visits, deaths) and the Kruskal-Wallis test (days in hospital weighted by length of follow-up). We will supplement our intention-to-treat analyses with a sensitivity analysis that excludes patients who after randomization received a final diagnosis that the index event was not an adverse drug event.90
- Secondary Trial
We will evaluate the primary outcome, adherence to the culprit medication, on an intention-to-treat basis. We will define adherence as the dispensation of >80% of prescribed medication doses during follow-up, based on the recorded medication dose and frequency, and the volume and dates of re-dispensations. We will determine the proportion of adherent patients in each group, determine point and interval estimates for the between-group differences, and compare them using a Mantel-Haenszel test, stratifying by age group and hospital.
- What is the proposed frequency of analyses?
We will complete only one analysis, after all patients have completed a minimum follow-up period of 12 months. There will be no interim analysis.
- Are there any planned subgroup analyses?
There is a potential for effect modification by both age, sex, and gender. We will ascertain gender-based variables using the BC Medical Registry to ascertain the number of dependents in each subject’s household, and income band using an approximation based on the first 3-digits of the subject’s postal code. All required data are available through PopData BC and provide a uniform source of events that are reliably captured in BC. To explore the possibility of effect modification we will add age group-by-treatment and sex-by-treatment terms to the models. Additionally, we will stratify by sex, and conduct gender-specific analyses using selected social determinants such as family composition and socioeconomic status. We will conduct these sub-analyses to determine if the differences by sex can be explained by gender-related variables, which reflect socially constructed roles, expectations, relationships, and other traits that society has ascribed to women, men, and people of diverse gender identities. We will ascertain family composition using the dependent number for each study participant. We will also conduct sensitivity analyses based on the certainly of the index adverse drug event.
- TRIAL MANAGEMENT
- What are the Arrangements for Day-to-Day Management of the Trial?
Our team of researchers and patient partners is based out of the Centre for Clinical Epidemiology and Evaluation (C2E2) in Vancouver and will oversee the day-to-day coordination and management of the trial. Three research coordinators will be responsible for the overall study conduct. Our qualitative research coordinator has extensive qualitative research experience and will oversee qualitative research assistants (graduate students) carrying out the planned process evaluation. A quantitative research coordinator will be responsible for the main and secondary trials, and will track recruitment of study subjects, plan and coordinate data linkage, and clean the data. A statistical analyst will analyze the data under the supervision of the study statistician. A health economist research coordinator will complete all activities required for the economic evaluation and will complete the required data analysis. The research team will provide bimonthly updates to the provincial Steering Committee who will provide oversight to the planned research. The research team will provide immediate updates to our DSMB with any patient or healthcare provider concerns about negative outcomes possibly attributable to the ongoing trial. The DSMB will meet in person or via teleconference within two weeks of any complaint.
- What will be the role of each principal applicant and co-applicants proposed?
Dr. Corinne Hohl (C2E2, UBC) is an emergency physician and health services researcher with extensive research experience in the field of medication safety. As nominated principal investigator, Dr. Hohl will oversee all aspects of the project and will commit 20h per week to the trial. Dr. Ellen Balka (C2E2, Simon Fraser University) is a Professor in SFU’s School of Communications and qualitative researcher whose work focuses on health information technology design and implementation. She will oversee qualitative research and issues related to health information technology integration and implementation and will dedicate 10h per week to the trial. Dr. Craig Mitton (C2E2, UBC) is a Professor in UBC’s School of Population and Public Health, and health economist who will oversee the economic analysis and dedicate 5h per week to this work. Dr. Penelope Brasher, a senior scientist at C2E2 will oversee data monitoring for quality during the trial, devise a detailed statistical analysis plan in collaboration with the nominated principal investigator, the quantitative research coordinator, and oversee the statistical analyses. She will dedicate 8h per week to the trial. Our three patient partners (Johanna Trimble, Laurel Radley, and Kevin Barry) are integral members of our research team and will continue to commit time (estimated at 3h/week each) towards the research, and the committees governing the proposed research.
- Describe the trial steering committee and data safety and monitoring committee
- ActionADE provincial Steering Committee:
In 2018, we formed a Provincial Steering Committee with researchers (Hohl & Balka), patient partners (Trimble & Radley) and representation from Ministry (Pharmaceutical Services & Project Management branches), VCH and VIHA (Executive-level), BC College of Pharmacists, Lower Mainland Pharmacy Services and the BC SUPPORT Unit. The Steering Committee has overseen ActionADE’s pilot implementation at VGH and will oversee the development and implementation of the ActionADE MVP. The Steering Committee will oversee the planned research in Phase 1 and provide guidance on Phase 2 planning. Our monthly reports to this committee will ensure that decision-makers are kept abreast of the ongoing research process, any management or safety issues that arise, and that they are aware of new knowledge that is being generated as a result of this work. Our Steering Committee has reviewed the Outcomes-Based Contract that has been negotiated with our stakeholder organizations, will oversee the independent auditor function, and ensure execution of the performance-based reward payments. All parties have agreed to the proposed governance structure and the proposed performance management plan, which will allow any party to escalate any disagreements through BC’s Arbitration Act.
- Data Safety Monitoring Board
The DSMB is an independent committee comprised of a senior emergency physician who is otherwise uninvolved in this work, a health services researcher, a statistician, and a patient partner. The DSMB will be responsible for assessing any reported instances of adverse events possibly attributable to the trial and will review study recruitment and data quality. The DSMB can provide a recommendation for trial suspension if there is a suggestion of harm, but the final decision will rest with the trial’s Steering Committee. We will develop a detailed DSMB Charter prior to initiating the trial.
- POTENTIAL IMPACT
Repeat adverse drug events cause 1,500 deaths and substantial health resource utilization in BC annually.16 By preventing unintentional re-exposures to harmful medications, ActionADE has the potential to create value by improving healthcare quality and preventing avoidable morbidity. It has the potential to avoid substantial costs by reducing adverse drug event-related outpatient and emergency department visits, and hospital admissions thereby enhancing health system sustainability. Modeling worst and best-case scenarios, the potential cost-avoidance is substantial due to the high incidence of repeat adverse drug events in BC (Figure 14). Through our collaboration enabled by CIHR, we expect to advance the quadruple aims of enhancing patient care, improving population health, reducing healthcare costs, and improving the work of care providers, while providing an example of successful collaboration, innovation, and implementation of new health information technology to transform medication safety for all.