Description of the Interventions
The April 2009 and April 2015 NPS MedicineWise PPI programs sought to provide and reinforce evidence-based recommendations to guide GPs in the appropriate primary care management of GORD, and to promote dialogue between GPs and patients about the relative benefits, risks, harms and costs of treatments. The programs included a review of recent safety updates from the Australian Therapeutic Goods Administration and tools for GPs to facilitate step-down PPI therapy in patients whose reflux symptoms were well-controlled (19).
GPs were further provided with feedback detailing their prescribing behaviours during the calendar years preceding each intervention (2008 and 2014). This feedback was provided in the form of an individualised report sent to all registered and practicing GPs in Australia. The report used national medicines dispensing records from the Pharmaceutical Benefits Scheme (PBS) to summarise all PPI medicines prescribed by each GP, and how their prescribing of these distinct medicines compared to that of all other GPs in Australia, for each medicine. The report also highlighted how their prescribing aligned with best practice recommendations. This feedback was provided to approximately 20,000 GPs in 2009 and 24,000 GPs in 2015 and included the number of PPI prescriptions dispensed each month, the strengths of the prescribed PPIs, the cost of the prescribed PPIs, and the number of their patients estimated to be receiving long-term PPI treatment (defined as patients with six or more PPI dispensings).
In 2009 and 2015, these feedback reports were followed by a clinical audit and a case study for a sample of GPs receiving the reports. See Table 2 for the details and the reach of each of these aspects of the interventions.
The key recommendations to GPs in the NPS MedicineWise programs in 2009 and 2015 were:
Review all patients currently being treated with PPIs
Confirm that the indication for treatment remains; evaluate whether or not the strength and frequency of PPI dosing can be reduced; and evaluate whether or not PPI therapy can be discontinued
Encourage lifestyle modifications and review the concomitant use of medicines that may exacerbate symptoms
Decrease PPI treatment to low strength or intermittent, symptom-driven therapy (pro re nata) once symptoms are controlled
Always discuss the expected duration of treatment and have a plan for stepping down or discontinuing treatment when PPI treatment is initiated.
Setting and Data
Australia maintains a universal healthcare system entitling all citizens and permanent residents to subsidised medicines through the PBS and subsidised outpatient medical services through the Medicare Benefits Schedule. The Australian Government Department of Human Services (DHS)—administering body for the PBS and MBS—supplied summary data of monthly dispensing records of PPI medicines subsidised through the PBS, prescribed by a GP, for each GP; and monthly records of every medical service (21) billed to the government by each GP in Australia from January 2006 through June 2016. GPs comprised registered general practitioners, trainees and non-vocationally recognised doctors. The DHS generated a unique identifier code for each GP in the data and this code allowed us to link dispensing and services data.
PBS data capture prescription medicine dispensing that has resulted in a subsidy paid by the PBS; the data do not capture dispensing of medicines priced below the PBS co-payment threshold or medicines dispensed privately. The price of many PPI medicines is below the general PBS co-payment threshold (range: $29.50 - $38.30 between 2006 and 2016) but above the concessional PBS co-payment threshold (range: $4.70 - $6.20 between 2006 and 2016). To ensure complete ascertainment of PPIs dispensed during the study period, we restricted our analyses to concessional beneficiaries (22).
Measures
As actual GP prescribing records are not available in our dispensing data, we examined the impact of the NPS MedicineWise interventions on the rate of dispensed PPI prescriptions issued by GPs as a surrogate measure of GP prescribing. Our outcome measures were constructed as the monthly number of standard and low strength PPIs (Table 1) dispensed through the PBS (numerators) per 1,000 reimbursable GP consultations (denominator) (23, 24). In the case of esomeprazole, the only PBS-subsidised strengths available in Australia are classified as “high” and “standard” strength. For the purpose of our analyses, we considered dispensings of the highest available strength esomeprazole (“high”) as a standard strength PPI, and the lowest available strength esomeprazole (“standard”) as a low strength PPI.
Statistical analysis
We used time series intervention models with an autoregressive residual process to analyse the dispensing rate of standard and low strength PPIs separately. We adjusted our data seasonally to account for the well-known “stockpiling” phenomenon that results in increased dispensing of many medicines subsidised through the PBS towards the end of each calendar year, and subsequent reduced dispensing during the following January and February (22). We conducted our analyses using the seasonally adjusted data but added the unadjusted data series to the final graphic presentations.
We used two separate change-in-trend variables to represent the NPS MedicineWise interventions in 2009 and 2015. We hypothesised that the impact of the 2009 intervention would diminish with time and included a decay term for the 2009 program in our models to test this hypothesis. If the estimate of a decay effect was not significant, we set the decay parameter to zero and re-estimated the model. We did not include a decay term for the 2015 intervention as fewer time points were available between April 2015 and the end of our data series (June 2016). We used a linear underlying trend to fit the dispensings rate for standard strength PPIs and a square root of linear trend to represent the underlying trend in the dispensing rate for low strength PPIs.
We used the statistical package, mgcv, to perform seasonal adjustment of the data series (25) and estimated the intervention models using generalised non-linear least squares with the package nlme (26). We performed all analyses in R v3.3.3 (27) and used a p-value of less than 0.05 to indicate statistical significance.