I write in response to the article published in your journal “Passing the acid test? Evaluating the impact of national education initiatives to reduce proton pump inhibitor use in Australia” (Bruno C et al., BMJ Qual Saf 2019). I am writing as NPS MedicineWise evaluation found a significant impact on general practitioner prescribing of proton pump inhibitors (PPIs) following their educational programs in 2009 and 2015. I acknowledge that the article is well written, the methods are well described, and the approach includes a number of sensitivity analyses. However, I would like to highlight some key points on the analysis methods used that differ from the approach taken by NPS MedicineWise and about which I have some concerns.

The 2015 NPS MedicineWise educational program on PPIs was part of a larger educational strategy over a decade to support best-practice prescribing of PPIs by general practitioners (GPs) and included Pharmaceutical Benefits Scheme (PBS) feedback to GPs in Australia. This consists of a letter that includes information of the individual GP’s prescribing of PPIs compared to their peers. Programs launched in 2009 and 2018 on PPI prescribing for GPs also included face-to-face educational visits to GP practices. The 2015 campaign, which did not include these face-to-face visits, aimed to reinforce the impact of the earlier 2009 campaign.

NPS MedicineWise has conducted an evaluation of the 2009 and 2015 education programs on GP prescribing of PPIs and found a significant decrease in the volume of high and standard PPIs prescribed to concessional patients.

There are a number of ways in which our analysis differs to that described in the Bruno et al. article:
• Our analysis found a reduction in high/standard PPI dispensed volumes from GP prescriptions for concessional patients over four times that found in the Bruno et al. article. Our analysis was focussed on GP prescribing, as they were the target prescribers for our education programs. In contrast, the Bruno et al. article includes all dispensed prescriptions including those by specialists. Specialists may be prescribing PPIs more frequently for people with more complicated conditions for which long-term PPI use is appropriate.
• The authors of the Bruno et al. article have used a novel approach by looking at a sample of 10% of individual longitudinal PBS patient data and used novel data analysis methods. In contrast, we use the complete PBS data for concessional patients and analyse dispensing volumes using best practice time series analysis.

• We analyse trends over multiple years, whereas the Bruno et al. analysis used data from one year before and after the 2015 NPS MedicineWise program, although they did adjust for seasonality. Benefits of a multiple year analysis include taking into account additional factors such as the impacts of previous programs, consumers stockpiling medicines once they reach the PBS safety net and other environmental factors.

• In the Bruno et al. article they only assessed a change in the dispensing levels after the interventions although there is a clear upward trend in PPI prescriptions over the years prior to the intervention. They did not assess a change in slope (i.e. change in trend) which NPS MedicineWise considers in evaluation work using time series analysis. One possible reason for an upward trend is a continual increase in the size of the population with gastro-oesophageal reflux disease (GORD) due to population growth, and this is particularly strong in older populations. I would recommend assessing total dispensing over time as an age-standardised or age-specific population rate to control for change in the size of the population at risk of being prescribed PPIs.

Since NPS MedicineWise launched the Choosing Wisely Australia initiative in April 2015, three recommendations from medical colleges and societies about the use of PPIs have been published – from the Royal Australian College of General Practitioners (RACGP) in April 2015, the Gastroenterological Society of Australia (GESA) in October 2016 and the Royal Australasian College of Physicians’ Paediatrics & Child Health Division in September 2017. The 2018 NPS MedicineWise educational program included recommendations on PPIs from RACGP and GESA.  One of the program’s aims was to grow GP awareness of the RACGP recommendation and GPs were encouraged to regularly review patients with GORD on PPIs, with the aim of reducing or ceasing PPIs altogether. 

A GP survey conducted in February 2019 asked GPs to rate their agreement with the RACGP statement before and after participating in the NPS MedicineWise program.  The result was a significant 13% increase in GP awareness of the RACGP recommendation after participating, demonstrating that educational activities do have an important role in behaviour change.

There is a need for further research into the impact of reducing PPI use on GORD symptom control. Further evaluation is planned and the MedicineInsight dataset provides a unique opportunity to explore the reasons general practitioners prescribe PPIs to patients. MedicineInsight, managed by NPS MedicineWise with support funding from the Australian Government Department of Health, is a large-scale national general practice dataset that extracts and collates longitudinal, de-identified patient health records from general practice clinical information systems.

We appreciate the concerns raised and agree that accurately presenting the findings of our study regarding patient-reported possible PICC-related complications is important. We took several steps to ensure transparency in how we presented our data. First, we were vigilant about consistently defining our outcomes as “possible complications” in all key areas of the paper including the main outcomes and conclusions section of the abstract, study measures section in the methods, main findings in the results section and in the discussion. The fact that the term “possible” or “potential” was not always used or some terms appeared to be used interchangeably was to improve readability of the article. In general, however, we were conscientious about clearly noting that we asked patients about signs or symptoms of a potential complication or adverse effect. Second, we purposefully chose to not present the data as complication rates (implying that these were actual events), but as the percentage of patients reporting a given event – thus we faithfully represented what we were told by patients during follow-up assessments. Third, we took these approaches because our primary objective was to accurately present the data collected from our patients; if something mattered enough to a patient to tell us about it, then it should matter to us as healthcare professionals regardless of internal standards that we may use to define serious medical complications. While that’s not to say that future work to better understand how patient reported outcomes relate to ascertainable or medically defined outcomes isn’t needed, our reporting was accurate and achieved the primary objective of this study - which was to highlight PICC-associated patient concerns and opportunities for care improvement.

O’Reilly-Shah et al. present a novel approach to quality improvement in anaesthesia by attempting to elicit change in practice using ‘nudge theory’ derived from the field of behavioural economics [1]. Translating new research evidence into common clinical practice is an important quality issue and cheap and effective strategies to achieve this are of interest. O’Reilly-Shah et al. hypothesised that using ‘nudge-type’ interventions, an audit-feedback dashboard as well as changes to mechanical ventilator default settings, might improve anaesthesia provider compliance with this ‘lung protective’ ventilation strategy in the general operating theatre environment.
I disagree with the conclusion that the authors have drawn, that these interventions might improve clinical and financial outcomes. My disagreement stems from the clinical rationale of the intervention, which overlaps onto the assumptions built into the ‘nudge theory’ of behavioural economics itself.
Sunstein & Thaler, who have influenced large scale policy making in several countries, describe nudge theory as a form of choice architecture that “alters people’s behaviour in a predictable way without forbidding any options or significantly changing their economic incentives” [2, 3]. Their philosophy is described as “libertarian paternalism”, as it influences choices that make people better off as judged by themselves, while preserving their freedom to choose otherwise. This is presumably what the interventions chosen by O’Reilly-Shah et al. were based on, that changing ventilator defaults did not force anaesthetists to select a certain tidal volume, and audit-feedback dashboards might foster change through an increase in awareness of peer group behaviour.
Assessing preferences and costs may not be straightforward, as described in a critique of nudge theory by Berg & Davidson [3]. They argue that many choice decisions implicitly have a trade-off, and it is difficult to predict in advance which direction the balance tips. Furthermore, selecting a particular tidal volume (a continuous variable) for a particular patient during an operation might be a complex decision and not a binary option, which makes the cost or indeed the opportunity cost, of a particular choice difficult to ascertain, in economic parlance.
A study much vaunted by the New York Times Magazine demonstrating improvement in hand hygiene rates amongst clinicians after the introduction of screen savers depicting bacterial cultures from physicians’ hands, failed to be replicated in another study using a similar methodology, which suggests that audit-feedback dashboard techniques may not be robust [4, 5]. O’Reilly-Shah et al. state that most practitioners accepted the target ventilation strategy without debate but they also admit that there is equipoise in the anaesthetic literature regarding the benefit of a low-tidal volume strategy for elective surgical patients [1]. As such, the increased compliance rate might not have reflected the provider’s ‘better preference’ per se, but a trend towards the default lower tidal volume as it may not have been an important factor either way.
A better example using the authors’ approach might be the default setting of ‘pressure-controlled’ ventilation as opposed to ‘volume-controlled’ ventilation in paediatric patients undergoing general anaesthesia, to avoid the danger of delivering overtly dangerous large tidal volumes. This is sensible practice in paediatric, but in hospitals with mixed adult and paediatric populations, this may not always be appropriate. It would then represent a binary decision and the benefit would be more obvious.

References
1. O'Reilly-Shah VN, Easton GS, Jabaley CS, Lynde GC. Variable effectiveness of stepwise implementation of nudge-type interventions to improveprovider compliance with intraoperative low tidal volume ventilation. BMJ Qual Saf. 2018 May 18
2. Thaler RH, Sunstein CR. Nudge: Improving Decisions About Health, Wealth, And Happiness. New Haven : Yale University Press, 2008, p5-6.
3. Berg C, Davidson S. Nudging, calculation, and utopia, Journal of Behavioral Economics for Policy, Vol. 1, Special Issue, 49-52, 2017
4. Dubner SJ, Levitt SD. Selling soap. The New York Times Magazine, September 24, 2006.
5. Charters M, Cheng AL, Esaki RK, Kuo AS, Neal J, Thawani JP, Chenoweth C. The impact of a screensaver in promoting hand hygiene. Am J Infect Control. 2009 Dec;37(10):867-8