Prescribing Data in General Practice Demonstration (PDGPD) project - a cluster randomised controlled trial of a quality improvement intervention to achieve better prescribing for chronic heart failure and hypertension

Pharmacological management of hypertension (HT) reduces cardiac events, stroke, hospitalisations, health care costs and improves quality of life for hypertensive patients (HT) [1, 2]. Appropriate treatment of chronic heart failure (CHF) with ACE inhibitors, angiotensin II receptor antagonists, beta-blockers and diuretics have also shown benefits in terms of survival and averted hospitalisations, irrespective of the underlying cause [3, 4].

Australian [5, 6] and overseas studies report that treatment of HT [7‐15] and heart failure [16‐19] are not well aligned with evidence-based guidelines. For example, in Australia, fewer than 50 percent of heart failure patients admitted to any of three hospitals in Tasmania were being treated with target doses of the recommended drugs [18]. Among Australian patients attending general practice, under-prescribing for heart failure was found both in terms of the number receiving the recommended drugs and the dosage levels [20, 21]. A national Australian survey reported the prevalence of untreated HT at 15.2% [13] and four consecutive GP audits of self-reported prescribing practices concluded that there was room for improvement in the management of hypertensive patients with co-morbidities [14].

The consequences of suboptimal care for these conditions include increased hospitalisation, higher mortality, [22, 23] greater symptom severity [23] and increased costs to the health care system [22, 24‐26]. It is clear that best practice guidelines alone cannot secure improvements in practice [5, 27‐29].

Educational interventions and quality improvement (QI) activities can improve the quality of prescribing in general practice settings [5, 30]. Multi-faceted interventions, particularly those involving interactive educational sessions for healthcare providers, and/or patient education are reported to be more effective than passive interventions [10, 31‐37]. Multi-faceted interventions have shown modest improvements in patient outcomes such as the proportion of patients meeting blood pressure targets [8, 23] and also changes in prescribing patterns such as the proportion CHF patients receiving a beta blocker or taking target doses of ACE inhibitors [38].

In evaluating these types of complex QI interventions, pragmatic randomised controlled trials (RCTs) are emerging as a way of bridging the gap between traditional RCTs which have a good internal validity and observational studies, which have good external validity [39‐41].

The Prescribing Data in General Practice Demonstration (PDGPD) project uses a multifaceted complex intervention aimed to improve GP prescribing behaviour for patients with HT and CHF in alignment with clinical practice guidelines.

This paper provides details of the intervention components and protocol, and methods and measurements used to evaluate the impact of the quality improvement initiative.

The project is a partnership between National Prescribing Service (also known as NPS MEDICINEWISE), an independent, evidence-based organisation delivering continuing education for health professionals to enable better decisions about medicines and medical tests [42], and the Australian General Practice Network (AGPN), the body representing a network of 111 local general practice networks (henceforth referred to as Networks) which covers 90% of registered GPs in the country [43].

The project governance included several external and internal groups that oversee various aspects of the project including the evaluation and the implementation. (see Additional file 1: Appendix 1).

The challenges of implementing and evaluating real-world interventions to improve health and professional behaviours are well documented [27, 30, 58‐60]. Participation of doctors in educational interventions and their willingness to accept evaluation audits rely on practice-based and doctor-based factors not often reported in the literature.[53, 54] A doctor’s prescribing behaviour also depends on personal characteristics, years of experience, motivation, marketing strategies used by pharmaceutical industries, demands from society and patients, knowledge of guidelines, confidence and regulatory actions [34, 60‐63].

There are methodological difficulties in designing sound evaluations to identify the impact that context can have on success of prescribing interventions [54, 57, 64]. For instance, GPs prescribe in apparent conflict with guidelines for reasons that are complex, and can vary from previous experience with medical misadventures, to patient-factors such as patient age, education, ethnicity, social class, likely compliance and lifestyle considerations. [62, 64] Uncertainty of intervention effectiveness is further compounded by the bias in reporting of studies and the inability to attribute success to individual intervention components [23, 65].

The evaluation of PDGPD project uses a pragmatic cluster RCT design, where one of two interventions or control status is randomly assigned to individual practices. A pragmatic cluster-randomised trial is considered an appropriate research design in public health and educational interventions as it allows more variability in the entry criteria, reduces the impact of contamination within groups, and provides administrative convenience for implementation [39, 40, 57, 66]. PDGPD will have access to several data points to investigate immediate and mid-term impact of the quality improvement initiative. However, several potential weaknesses in the design of PDGPD are acknowledged: first, the voluntary nature of GP participation in the cluster design where a practice [cluster] agreed to participate but not all GPs in the practice received the intervention. This may lead to dilution of the intervention effect when data from all patients seen by all GPs in the practice are analysed. The intervention is designed to encourage GPs to communicate key messages with other non-participating GPs in the practice and to welcome non-enrolled GPs to attend clinical discussion meetings within the same intervention group; further, the impact of participation rate by practice can be incorporated in the regression analysis model. Second, the cross-over design (Figure 1) where the same facilitator delivers one intervention after another in the same practice and one topic in one intervention practice and the competing topic in another practice carries the risk of cross-contamination; allocation of clinical topic order was done at random and facilitators were trained at focusing on discrete messages of one topic at a time. Thirdly, the wait-control group provides pre-intervention comparisons only for the first six months of the trial; for practical reasons in non-research settings (i.e. need for an intervention in exchange for participation) the control group received the intervention six months later.

Finally, the PDGPD project is expected to bring the following benefits:

MW conceived and developed the design of the study, developed ethics application and obtained ethics approval, conducted literature searches, undertook and supervised data quality assurance, led the first draft and revised this manuscript and commented on subsequent versions. MCM conducted literature searches, contributed to the definitions of prescribing indicators to measure impact, produced data specifications and analysis plan for the impact analysis plan and led subsequent drafts of the manuscript to completion. JDE contributed to indicator development, project design, project implementation, data coordination, and revision of the manuscript versions. JM provided input into project design, indicator development, clinical considerations and day-to-day strategic decisions on project direction. NS, JG, JR and JE made intellectual contributions into trial design and project implementation issues and commented on drafts of the manuscript. All authors read and approved the final manuscript.