Adverse drug events (ADEs) and avoidable ADE hospital admissions are common, costing health systems billions of dollars every year [1-7]. Internationally, approximately 7% of hospital admissions result from drug-related problems, of which 59% are considered avoidable through safer prescribing [3,4,8]. Most drug-related admissions are caused by commonly prescribed drugs, notably nonsteroidal anti-inflammatory drugs (NSAIDs), antiplatelet medications, and anticoagulants, which together account for one-third of ADE admissions [3,4,9].

High-risk prescribing is prescribing that places patients at increased risk of ADEs. The single greatest predictor of ADEs and high-risk prescribing is the number of medications a person is taking [10]. With demographic ageing, there are increasing numbers of older people prescribed multiple medications for multiple co-existing medical conditions [11]. In New Zealand, approximately 10% of people aged 65 years and older are taking ten or more regular medications, and high-risk prescribing is common, often involving NSAIDs [12-14]. The individual circumstances of a patient may justify high-risk prescribing, but to minimize harm it is necessary that medications are regularly reviewed and stopped or started as appropriate [15].

Most on-going medications are prescribed in primary care. Despite strong evidence to guide safe prescribing, a gap remains between existing evidence and current prescribing practice. Translating research evidence into practice is difficult. There are many barriers to regular medication review in everyday practice [16]. The large variation in prescribing between practices and regions in New Zealand suggests room for improvement [13,17,18].

In New Zealand primary care, most quality improvement processes are delivered through Primary Health Organisations (PHOs), professional groupings of practices for administrative and quality improvement purposes [19]. The most effective, cost-effective, and practical approach to safer prescribing in everyday practice is not yet known [20,21]. There is evidence to suggest education programmes can improve prescribing but education alone is not enough to induce lasting change [13,22]. Complex interventions as part of ongoing quality improvement programs show the most promise, in particular interventions combining audit and feedback, education, incentive for participation, and patient engagement [20,23-26]. The Australian Veterans’ Medicines Advice and Therapeutics Education Service (MATES) quality improvement program in primary care has shown promising results in the Australian Veterans population, especially when delivering a focused message targeting single medications and less so when delivering a combination of messages targeting general topics such as interactions and potentially inappropriate medications in older people with polypharmacy [24]. The MATES programme is based on sound theoretical underpinnings and delivers 4 interventions per year. The MATES intervention uses practice prescribing audits, patient-specific feedback, education to doctors, and a practice mail-out to selected at-risk patients to encourage their engagement.

Adapted from the Australian Veterans’ MATES programme, we developed the Safer Prescribing and Care for the Elderly (SPACE) intervention to prompt medication reviews and support safer prescribing in the New Zealand primary care context. We recently piloted the SPACE intervention in two New Zealand primary care practices in preparation for this proposed randomized trial, focusing on the topic of NSAIDs and antiplatelet prescribing. This topic was chosen because these drugs are commonly prescribed and are associated with serious ADEs including bleeding and renal impairment. The SPACE intervention was found to be feasible to implement using existing primary care structures and both acceptable and useful to patients, doctors, and the PHO clinical advisory pharmacists [27]. In the pilot study, we developed practice audit queries to identify patients with high-risk prescribing of NSAIDs and/or antiplatelet medications; integrated the SPACE intervention into practice management software; developed processes to collect, encrypt, and link study data; and derived information for calculating sample sizes for the randomized trial.

We will assess whether the SPACE intervention can reduce the rate of high-risk prescribing of NSAIDs and/or antiplatelet medications and related ADE hospitalizations over 12 months using existing PHO infrastructure and systems in New Zealand primary care.

We will conduct a cluster randomized control trial. The clusters will be primary care practices. Data collection and analysis will be at the level of patient.

The study will be conducted in primary care practices in Auckland and Northland, New Zealand.

Primary care practices will be eligible to participate if:

The study will target all physicians working in participating practices since patients can receive a prescription from any physician in a practice.

Patient inclusion criteria:

All data collected on patients is anonymized prior to leaving the practice using a unique identifier (National Health Index number) to enable linking of clinical data over time and linking to hospitalization data.

The SPACE intervention is designed to prompt medication reviews and support safer prescribing decisions in primary care practice. The intervention comprises a practice audit to identify for each doctor a list of their patients with high-risk prescribing for the chosen topic; an outreach visit from a clinical advisory pharmacist to physicians to provide education about the prescribing topic and to go through with each physician their list of patients identified as having high-risk prescribing; a tick-box for physicians to indicate the action they will take in response to the feedback for each patient (“review medications + patient mail-out,” “review medications + no mail-out,” or “no action”); and a mail-out from the practice to patients selected by physicians with information about their medications and a letter encouraging them to discuss their medications when they are at the practice next seeing their physician [27]. All prescribing decisions are made as usual by the doctor in discussion with the patient. The individual circumstances of the patient may justify high-risk prescribing and, after review, the prescribing may or may not be changed. The prescribing topic for the trial is the prescribing of NSAIDs and antiplatelet medications.

Control practices will deliver care as usual. If the SPACE intervention is shown to be effective, we aim to deliver the intervention to control practices after 12 months.

All practices and doctors will participate as usual in PHO quality improvement initiatives and medical education activities. See Figure 1 for the flow of practices through the study.

Assessment time-points will be baseline, 6 months, and 12 months. The outcomes of interest are the difference between intervention and control groups at 6 months controlling for baseline, and the difference between intervention and control groups at 12 months.

The primary outcome measure is:

The difference in proportion of the participants (those vulnerable at baseline) receiving high-risk prescribing of NSAID and/or antiplatelet medications between the control and intervention groups at 6 months. That is, the proportion of “vulnerable-at-baseline patients” (with gastro-intestinal, renal, or cardiac risk factors) receiving high-risk prescribing of NSAID and/or antiplatelet medications at 6 months according to the definitions listed in Table 1. Whether difference in proportion between the two groups is sustained at 12 months will also be examined. Participants will be considered to have high-risk prescribing at each time-point if they fulfil any of the high-risk prescribing criteria set out in Table 1 in the 14 weeks leading up to each time-point.

Secondary outcome measures include:

Whether difference in proportion between the 2 groups is sustained at 12 months will also be examined for secondary outcomes 1, 2, 3, and 5 months.

These outcomes have been used in similar trials previously [23]. Data will also be collected to enable a subsequent cost-effectiveness evaluation of the intervention from a societal and health funder perspective. The cost-effectiveness of the intervention will be measured as the cost per reduction in high-risk prescribing and cost per reduction in hospitalizations from the health funder (District Health Board) perspective. The data collected will include the cost of delivering the intervention (including pharmacist and doctor time for the feedback outreach session, travel time and costs for the outreach visit, audit time and cost); cost of medications; cost of hospitalizations.

The sample size calculation for this study is based on previous trials demonstrating a clinically relevant 25-45% relative risk reduction in the proportion of high-risk prescribing, [23,25] one trial of which also demonstrated how such reduction (3.7% to 2.2%) in high-risk prescribing can translate to significant reductions in hospitalizations due to bleeding complications [23]. Based on the local pilot data, we estimated an average of 200 “vulnerable patients” per practice and an 8% high-risk prescribing rate [28]. We estimated an intracluster correlation coefficient of 4.68 x 10^-7 for the primary outcome based on a cluster randomized trial examining similar outcome of NSAIDs prescribed to patients with a history of peptic ulcer and not prescribed gastro-protection [25]. Assuming approximately 12% of patients would be lost to follow-up over the 12 month study period, data from 8000 patients from 40 practices (20 practices in each group) with an average of 200 vulnerable patients per practice would be required to detect a statistically significant difference of 6% in the intervention group and 8% in the control group of high-risk prescribing at 12 months (P=.90, alpha=.05).

Practices will be purposively sampled and recruited, aiming to include both medium sized (3000-7999 enrolled patients) and smaller practices (0-2999 enrolled patients) in the Auckland and Northland regions. Practices and physicians will be invited by a colleague using a comprehensive list of practices. Consent for participation will be at the practice and physician level. Written informed consent will be obtained from all participating physicians. Participants, ‘vulnerable patients’, will be identified using a standard query applied to the practice enrolled population. Consent will not be sought from individual patients because outcomes data are collected in routine patient care and will be anonymised prior to extraction for analysis and linking.

Practices will be randomized 1:1 to intervention or control. Randomization will be stratified by practice location (Auckland vs Northland) and practice size (medium [3000-7999 enrolled patients] and smaller practices [0-2999 enrolled patients]). Block randomization will be carried out using randomly varying block sizes of 2, 4, and 6. Random sequence generation and allocation of randomization will be undertaken by a statistician not involved in recruitment or baseline data extraction.

Analyses will be performed according to the intention-to-treat principle, with the use of mixed-effect models to account for clustering in the data. The primary and secondary outcomes will be analyzed using generalized linear mixed effect model, GLIMMIX, with the individual as the unit of analysis and the practice included as the random effect to control for the effects of clustering. GLIMMIX with Group x Time interaction will be used to assess the overall difference between intervention and control. The model will adjust for the stratification factors including practice location (Northland or Auckland) and practice size. Baseline covariates including (age, sex, and baseline number of long-term medications) will be adjusted if appropriate.

Study approved by the University of Auckland Human Participants Ethics Committee: Ref 020092, expires 9 Oct 2020.

The SPACE intervention is based on sound theoretical underpinnings, is acceptable and useful in the New Zealand primary care context, and identifies and reaches patients with high-risk prescribing who are at increased risk of ADEs. The intervention builds on existing primary care infrastructure and uses existing primary care staff to deliver a safety improvement intervention. The intervention combines audit and feedback with mail-out to motivate patient engagement. It is amenable to repeat use, and could be used in an on-going quality improvement program to target different high-risk prescribing topics.

If shown to be effective and cost-effective, the SPACE intervention could be rolled out nationally and used regularly by PHOs to support safer prescribing in practices and minimize avoidable ADE hospital admissions in the short-to-medium term. Since the SPACE intervention is designed to support behavior change, it could be applied to other evidence-based topics, including other prescribing topics and test ordering and monitoring.

Interventions that have been shown to improve practice have been published [23,25]. However, most previous trials of primary care interventions in this area have involved time-series or noncontrolled trials [20,22,24]. The SPACE trial uses a cluster randomized controlled trial design to provide robust evidence to assess whether such an intervention can change prescribing and improve clinical outcomes.

Since the SPACE intervention is designed to prompt medication reviews and support safer prescribing, the primary outcome measure is designed to reflect a change in prescribing behavior (a reduction in the rate of high-risk prescribing in patients vulnerable at baseline). Since the ultimate aim is to change prescribing behavior overall and improve patient outcomes, we will also measure the rate of high-risk prescribing overall and related ADE hospital admissions.

There is a risk of contamination between intervention and control practices and between intervention practices and those waiting for the intervention; doctors might change their prescribing behavior if they are alerted to the study prescribing topic. However, rolling delivery of the intervention is the only practical and feasible way to progress this study, given the educational outreach visit component of the intervention and the limitations of our study team. We are also limited to using small to medium sized practices (fewer than 8000 enrolled patients), since all doctors in participating practices must receive one-on-one feedback from the clinical advisory pharmacist at an outreach visit.

This study addresses an important translational gap, testing an economically sustainable intervention designed to support safer prescribing in routine practice. The new knowledge generated will help to address the most important threat to patient safety in primary care: high-risk prescribing and adverse drug events.