Using an Electronic Decision Support Tool to Reduce Inappropriate Polypharmacy and Optimize Medicines: Rationale and Methods

Background Polypharmacy and inappropriate continuation of medicines can lead to a significant risk of adverse drug events and drug interactions with patient harm and escalating health care costs as a result. Thorough review of patients’ medications focusing on the need for each drug can reduce the potential for harm. Limitations in performing effective medicine reviews in practice include consultation time constraints and funding for pharmacy services. We will aim to overcome these problems by designing an automatic electronic decision support tool (the medicines optimization/review and evaluation (MORE) module) that is embedded in general practice electronic records systems. The tool will focus on medicines optimization and reducing polypharmacy to aid prescribers in reviewing medicines and improve patient outcomes. Objective The objectives of this study are: (1) to develop an electronic decision support tool to assist prescribers in performing clinical medication reviews with a particular focus on patients experiencing multimorbidity and polypharmacy, and (2) evaluate and assess the use of the electronic decision support tool, providing pilot data on its usefulness in supporting prescribers during consultations with patients. Methods The first three study phases involve development of clinical rules outlining clinical interventions and the creation and validation of the MORE decision support tool. Phase four is a community-based, single-blind, prospective, 6-month controlled trial involving two interventions and two control general practices, matched for practice demographics. We will be measuring the number of times prescribers engage with the tool, total number of interventions suggested by the tool, and total number of times prescribers change medicines in response to recommendations. There will also be prospective follow-up of patients in the intervention group to examine whether changes to medications are upheld, and to determine the number of hospitalizations or emergency department visits within 6 months of a medicine intervention. Comparisons between control and intervention practices will measure the changes in proportions of patients with polypharmacy and inappropriately prescribed medicines before and after the introduction of the electronic decision support tool, proportions of patients receiving appropriate treatment in each practice, and changed, maintained, or improved health status, hospitalizations, and deaths in the study year. Initiation rates of inappropriately prescribed medicines will be measured as a secondary outcome. As well as external assessment of the extent of use and application of the tool, prescribers will receive monthly practice progress reports detailing the proportion of their patients experiencing polypharmacy and taking inappropriately prescribed medicines identified for review. Results Phase one has now been completed and the decision support tool is under development. Final data analysis is expected to be available in December 2016. Conclusions This study will establish whether the MORE decision support tool stands up to real world conditions and promotes changes in prescribing practice.


DESIGN AND METHODS
The project plans a phased approach, using literature review and clinical consultation to develop the algorithms to be implemented. The research team work with an established partner organisation experienced in developing clinical software to develop the tool. Face and content validity are established using historical data to ensure feasibility, with the testing phase being undertaken in two general practices across the country and assessed against a comparison group of two practices. The study would be strengthened with more control practices. The study is appropriately powered for changes in poly-pharmacy, the endpoint intended to be assessed. The study is not powered for assessing changes in health outcomes, this should not be seen as a limitation as the study was not designed for this. If the study is successful, larger studies that assess the impact on health outcomes would be required.

TEAM CAPABILITY -RESEARCH OUTCOMES
The lead investigator is an early career researcher, with a significant track record relative to opportunity spanning Australia, the UK and New Zealand. The research team is an impressive mix of clinicians and researchers spanning the disciplines of pharmacy, pharmacoepidemiology, general practice, biostatistics, and clinical pharmacology and is in partnership with BPAC, an organisation with significant track record in the area. The more experienced researchers should be able to provide appropriate mentorship for Dr Smith and the team's track record demonstrates their capacity to manage grants of this nature.

TEAM CAPABILITY -RESEARCH UPTAKE
If this research is successful, it will be highly usable. The planned phased method maximises the chance of development of a successful product and the significant number of researchers from a general practice background will ensure the likelihood of a tool with appropriate application to general practice

IMPACT ON NEW ZEALAND HEALTH DELIVERY
If the tool is successful it will be able to be rolled out to general practices across New Zealand where it is likely to have a significant impact quickly. Medication use is the most common intervention in health care and consequent with that medication error and adverse drug events are most common cause of harm. Reducing errors and inappropriate prescribing will contribute to reduced harm, improved outcomes and reduced costs.

OVERALL/GENERAL COMMENTS (To Applicant, HRC and Committee)
I think overall this is an impressive grant. The project is well thought through and the research team and partner organisation appropriate to the task. The only limitation of the research is that phase four is a small study (4 practices) which means health outcomes cannot be assessed. This arises solely due to the maximum funding limit as the inclusion of more practices would add considerably to costs.

SCIENTIFIC MERIT
The extent of inappropriate medicine use, especially in those taking multiple medicines represents a significant health challenge for all health care professionals involved, patients and carers and puts a significant burden on health care systems. Hence interventions designed to address this are of high scientific, professional and practical importance. Hence the importance of this proposal. If successful, this proposal has the potential to significantly advance practice. The first two objectives of this study involve significant development and the third objective describes a feasibility study. Notwithstanding the feasibility study the investigators have included a null hypothesis with sample size calculation in this proposal. The approach is original and builds on the existing work of bpac.

DESIGN AND METHODS
The proposal describes a justification for the different aspects of the research design and methods. The research team and collaborators is more than capable of conducting this project. However there is a significant risk for this proposal in that no or limited preliminary data has been presented and that the successful implementation of the proposal includes all aspects of the development, evaluation and assessment of the EDS module. Specifically the project design comprises four phases which are conducted in sequence. Delay or deficiencies in any of the phases will therefore have an impact on subsequent phases of the project. Also, presumably some aspects of the project, such as the conduct of literature reviews and compilation of prescribing resources, should/could have already been undertaken. The proposal differentiates between polypharmacy and hyperpolypharmacy. A justification for the use of a higher cutoff of 10 medicines per patient has not been detailed. The use of a hyperpolypharmacy cutoff may present challenges in the identification of suitable patients to meet the sample size within the two general practice intervention sites, given an estimated 11% rate of hyperpolypharmacy. The reference to support the influence of EDS on prescribing is the bpac website. Could an alternative or additional peer reviewed reference have been used? The proposal provides little detail of how or what specific individual data will be used in the MORE module to the extent to which this data will allow for a meaningful individualized recommendation versus a "generic" recommendation. Multimorbidity is mentioned throughout the proposal including the aim of this study, however how multimorbidity will be considered in the MORE module has not been detailed. The example clinical rules listed in the proposal seem to have an individual medicine or drug class as the unit of analysis and do not discuss conditions or the handling of multiple conditions. Will the MORE module consider factors such as patient preference or a shared decision making approach? The proposal implies that the MORE module will minimize the need for a manual review of medicines and be time saving for GPs. Could it be the case that presenting medicine recommendations to GPs in complex (hperpolypharmacy) patients with multimorbidity, may require more GP time to consider the recommendations? (given that GPs are unlikely to blindly implement the recommendations made by the module). The MAI does not identify specific medicines that are often prescribed inappropriately. Unlike Beers criteria and STOPP/START it is an implicit tool rather than an explicit tool. The second part of objective 1 is unclear. Is this a feasibility or pilot study? Although paediatric and adolescent patients may be selected, the hyperpolypharmacy criterion may rule out many who may benefit from EDS due to complexity in the use of medicines. Regarding the (boxed) example of the MORE module -would the recommendation regarding the use of omeprazole in a child come up if the PPI was being prescribed for the first time? Based on the prompt provided it should. What would happen for the same patient if they did not have (GI) symptoms or ADRs and they were currently taking the PPI? Would the alert only come up if the patient was taking 10 or more medicines?

TEAM CAPABILITY -RESEARCH OUTCOMES
The research team combines an excellent balance of academics, practitioners, and decision support/IT implementation. The combined skill set of this group is more than capable of effectively conducting this proposal. It is noteworthy that the PI, an early career researcher, has an experienced academic mentor to support her in all aspects of the research.

TEAM CAPABILITY -RESEARCH UPTAKE
The publication track record of the research team is significant and spans the related areas of pharmacy, clinical pharmacology and general practice, all of which are of direct relevance to this proposal. The professional standing of the research team is well established and evident through numerous current or previous leadership positions of key organisations and committees at both the local and national levels. There is evidence that some members of the research team have worked together previously, in the form of co-authored publications. More broadly the core research team will be supported by a list of national and international collaborators. It is noted that most of this support will be offered in-kind. Notwithstanding the significant records of these external collaborators, the proposal does not articulate in detail how the core research team will practically facilitate the involvement of the external collaborators or the expected time commitment of them. Furthermore, given that the proposal has been described as a feasibility study, it may be argued that additional local NZ experts may arguably be in a better position to provide input into content and feasibility issues. Nevertheless, if successful this proposal, as the investigators note, has the potential to be rolled out due to the collaboration with bpac. However, even if this proposal was successful, it would seem prudent to repeat and upscale the project before national implementation.

IMPACT ON NEW ZEALAND HEALTH DELIVERY
The extent of inappropriate medicine use, especially in those taking multiple medicines represents a significant health challenge for all health care professionals involved, patients and carers and puts a significant burden on health care systems. Hence interventions designed to address this are of high scientific, professional and practical importance. Hence the importance of this proposal. If successful, this proposal has the potential to significantly advance practice.

SCIENTIFIC MERIT
I was excited to read this application which deals with an important health issue, and is proposing a pragmatic response.
In relation to scientific merit the primary outcome is vague and appears to comprise four (five if you count differences in both group 1 &2) quantitative co-primary outcomes and one qualitative outcome. Although the mixed methods approach is potentially valuable, usually a single pre-specified primary outcome is preferred (or, at most, a co-primary outcome).

DESIGN AND METHODS
I was concerned about the design of the study. Although the BPAC team sound excellent, I was concerned that Phases II and III do not explicitly include testing by GPs. Presumably GPs should be involved in the development of the module. You could prejudice success in Ph IV if GPs are not adequately involved in Phase II (you acknowledge all of the limitations in the literature regarding alerts being overridden by busy Drs -I thus feel it's incumbent on the team to pilot test the module with GPs in Phase II, to determine that GPs feel it meets their needs).
I was likewise not at all convinced of the rational for limiting validation to the technical aspects (ie validation using fictitious patients, as opposed to determining acceptability etc by GPs piloting the tool).
I was also concerned about the statistical methods. Although intervention sites will be randomly selected, they will then be matched with other practices. So the design is quasi-experimental. The power calculation does not account for clustering at practice level, which would appear to be relevant. I note that a statistician is involved but I feel these points need to be explicitly addressed in the rebuttal as I do not have confidence in the analytic approach as presented currently.
It's not clear why the proposal target hyperpolypharmacy (rather than polypharmacy), and will only comprise 10 clinical rules (these aspects seem arbitrary -what if the 11th clinical rule appears to be important?).

TEAM CAPABILITY -RESEARCH OUTCOMES
The team appears strong and capable with a good mix of GP, pharmacy, pharmacology, technical and statistics input. I was surprised that leading NZ figures in the field of de-prescribing (eg Mangin has published with Garfinkel) were not even listed in the collaboration section.

TEAM CAPABILITY -RESEARCH UPTAKE
I was initially uncertain (thinking BPAC was a commercial provider) but after reading the BPAC website was convinced of this aspect of the application. Thus, although there is little actual concrete detail in the application, BPAC appears to be able to engage end users successfully.

IMPACT ON NEW ZEALAND HEALTH DELIVERY
Potential impact is limited by the scientific and design concerns listed. If the intervention is not shown to be successful it could be because the development was not rigorous, or because the efficacy study was poorly designed or analysed. Otherwise the application is strong in terms of the partnership with BPAC and potential for impact through wide uptake of a successful module.

OVERALL/GENERAL COMMENTS (To Applicant, HRC and Committee)
nil additional

SCIENTIFIC MERIT
This project targets the problem of polypharmacy and reducing the risk of adverse drug reactions. It is difficult to assess how effective the intervention is likely to be given that the rules to be included in the MORE module have not been defined. Some initial work on the rule development would have greatly strengthened the application and assisted reviewers in assessing the likely impact of the intervention. It would be important to establish that interventions in general practice based on these rules will result in reduced prescribing in people with hyper-polypharmacy. This was not clear in the application.

DESIGN AND METHODS
I felt that the timeline for this application was not feasible. For example, the intervention is to run for 6 months until Sept 2015, with patients being followed for up to 6 months to determine outcomes. It is not clear to me from the timeline that this would be possible given the project ends in March 2016. Given the rules have not been established, it is not clear how many patients the investigators will need to follow up. Conceivably, with 10 rules, this could included a large number of patients. Will the budget be sufficient for these follow up calls given the time involved? How many patients do the investigators expect to follow up?

TEAM CAPABILITY -RESEARCH OUTCOMES
Collectively, the team has the experience necessary to conduct the proposed research. However, I question whether the project can be completed within the suggested timeframe.

TEAM CAPABILITY -RESEARCH UPTAKE
The application demonstrates meaningful engagement with end users and is likely to result in good participation.

IMPACT ON NEW ZEALAND HEALTH DELIVERY
This is not clear from the application. As previously stated, this would depend on the rules chosen, and a number of other factors, including uptake by GPs, potential to reduce hyper-polypharmacy, and the degree of patient benefit derived from application of the rule. I feel the investigators need to provide more evidence on each of these links for their chosen rules before this is implemented/funded. This application would be greatly strengthened by more clarity on the rules to be chosen. This phase was a part of the project, but the lack of clarity in the application makes it difficult to assess the feasibility of the project. The timeline, especially for follow-up of patients is also a major concern.
APPLICANT RESPONSE TO REFEREE REPORTS -(ALL HRC ROUNDS)

Title of Research
Integrating patient data to optimise medicines and reduce polypharmacy Page 1 We would like to thank the referees for their positive comments including: 1. The research team-suitability, capability and experience of the team spanning multiple disciplines (referee# 67, 10, 33 & 27) 2. The impact on New Zealand health delivery and international significance of the proposed project which will contribute to reducing harm and improving outcomes and reducing costs (#10) 3. The strength of the partnership between bpac nz and the University of Otago bringing together experts in IT, clinicians and researchers (#27, 10, 33) 4. The pragmatic design of the project and that the developed tool would be highly useable (#10) with wide uptake (#33) and meaningful engagement with end users (#67) The referees had some comments and questions around the following topics: 1.
Referees # 67 and #33 requested further information around phase 1 -the development of the clinical rules in the MORE module. Limited information was given here due to space constraints. From our initial work, we anticipate that 10 rules will be chosen using literature searches, expert clinical advice and internationally validated prescribing quality tools such as STOPP/START to enhance the quality use of medicines (#27). This phase has not been completed as funding is needed to for time and expertise to evaluate the literature and validated tools. If it is deemed that a rule will benefit a large target group, then the recommendation will appear even if the patient does not have hyper-polypharmacy (e.g. proton pump inhibitors (PPIS) in children) as our intention is to improve quality use of medicines overall (#27).

2.
Individualised data used to make recommendations (#27). Individual patient data such as age, sex, co-morbidities, current and previously prescribed medicines, test results and other factors such as CVD risk assessment and renal impairment will automatically be included by the MORE module. This will allow an individualised recommendation to be made to the prescriber (e.g. this patient has reduced renal function -consider reducing the dose to xx or changing this medicine to xx) rather than a generic, irrelevant or broad recommendation (e.g. prescribe with caution in renal impairment) that is likely to be ignored.

3.
Timelines were questioned by referees #67 and #27. The timelines are compliant with the full 18 months available for the research, with phases adhering to strict timelines. BPAC and the researchers are used to complying with tight deadlines and have proven this with the successful completion of other grants and commercial projects. Some phases of the project are already underway e.g. ethics applications and Maori consultation, which form the first two months of the project timeline, to allow some contingency to the proposed method.

4.
Sample size #67 and #33. It is very difficult at this stage to undertake a power calculation that accounts for clustering (#33) as we will not know the patient size of each cluster until after recruitment; therefore, we cannot determine the intra-cluster correlation coefficient. Referee #10 agreed that we have adequately powered the study to measure changes in polypharmacy. Although this is a small sample size of 4 practices (2 intervention, 2 control), we believe that patient cohort of between 20,000-40,000 individuals in total will provide enough information needed to determine a sample size for a large nation-wide randomised control trial, if the proposed study is successful. A small sample will also allow us to gain in-depth knowledge from participating prescribers around the usability of the tool, the barriers, facilitators and future recommendations.

5.
Experiment design and outcome measures. This study is not quasi-experimental as suggested by #33. The intervention practices will be randomly assigned, then matched to control practices. The outcome measures for this study are robust quantitative data that address APPLICANT RESPONSE TO REFEREE REPORTS -(ALL HRC ROUNDS)

Title of Research
Integrating patient data to optimise medicines and reduce polypharmacy Page 2 the primary outcome of determining the usability of the MORE module. We consider the design and outcome measures appropriate for this stage of developing an innovative and novel tool.

6.
Polypharmacy vs hyper-polypharmacy (#27 and #33). The ultimate outcome from this project is to improve the quality use of medicines in New Zealand. Improvement measures may include reduction in the number of patients with hyper-polypharmacy (10 or more concurrent) medicines and/or polypharmacy (5 or more concurrent medicines). We specifically chose hyperpolypharmacy as the primary outcome measure for this study as these patients are most likely to benefit from a medicines review, yet often most difficult to review due to the high number of medicines. Hyper-polypharmacy patients are at increased risk of hospitalisation due to drugdisease, drug-drug or drug-food interactions. ≥5 medicines would capture a huge number of patients, many who will be taking 5-9 medicines considered essential for their condition/s e.g. usually ≥5 medicines are required as standard treatment post heart attack.

7.
Multimorbidity was raised by referee #27. Multimorbidity is not a measured outcome in this study but it contributes indirectly. Multimorbidity (multiple diseases) is often considered a consequence of aging, and is broadly thought to be a cause of polypharmacy due to improved treatment options for diseases.

8.
The MORE module will not account for patient preferences and shared care (referee #27), but it will allow prescribers more time in their consultations to consider these factors and will prompt prescribers to do this. The MORE module will automatically review the patient's medicines, taking this time consuming task away from the prescriber and freeing up more time for discussion with the patient.

9.
Further reference (#27). Electronic decision support development in New Zealand is relatively new, but an HRC grant to measure health outcomes and changed general practice behaviour regarding TIA/stroke diagnosis and management showed that EDS use resulted in fewer deaths and decreased health costs. Ranta

10.
Referee #27 commented on the strong connections and international collaborations with leading experts for this project. Referee #33 commented about the lack of national collaborators, however all members of the project team currently are leaders in this field and also work with and have established networks with national researchers in primary care and quality use of medicines; in particular, Dr Tordoff collaborates with D Mangin (as mentioned by referee #33). This national expertise can be accessed as needed. It is expected that those in the clinical advisory group would participate in 2 x ½ day meetings during the study project and an additional ½ day, accessed as required to review documents or provide feedback (#27).

11.
Referee #33 asked about the involvement of GPs in the development of the tool. Dr Lloyd, a practicing GP, will oversee the development of the MORE module as the divisional head of information services at BPAC. Prof Tilyard, a practising GP, will also assist with the development. As outline in phase 3, the MORE module will be validated using a historic database; GPs will also be invited to pilot and comment on the module.
We would like to thank the referees for their overall comments that this is an impressive grant (#10) which has the potential to impact NZ health delivery through wide uptake and participation (#33, #67) and to significantly advance practice (27).

Health Research Council 110 Stanley St Auckland, 1010
Re: 14/730 Integrating patient data to optimise medicines and reduce polypharmacy

Dear Mr Davidson
Thank you for your letter regarding our application for a Research Partnership for New Zealand Health Delivery. Please find below responses to the questions raised in this letter.

1, 2. Importance and Research team
Thank you for acknowledging that targeting the issue of polypharmacy at a general practice level is a useful approach. Thank you also for recognising the strong research team behind this grant, our positive track record and the good links for successful dissemination though BPAC.

3a. Managing the MODULE recommendations in a 15 minute consultation
The number of available clinical decision support tools has significantly increased over the last 5 years; therefore GPs are very familiar with using and managing these tools within a 15 minute consultation. To-date clinical decision support tools developed by BPAC have been used over two million times during general practice consultations. The proposed research has also incorporated features that will hopefully assist GPs in using the MORE module, including: -Saving time: The MORE module will automatically review each patient's medicines for the ten clinical rules, taking this time consuming task away from the prescriber and freeing up more time for discussing the information with the patient. -Education: Intervention practices will also receive written education materials and face-to-face academic detailing on the different clinical rules recommended by the MORE module, meaning less GP time spent digesting the information during the consultation thus more time to implement the changes and consider patient preferences. -Prescriber resources: We will provide a printed hand-book/guide for GPs outlining the key messages for each of the clinical rules. They can then have this on-hand to refer to during a consultation, if needed. -Patient information: Links to validated patient information will be available through the MORE module, which prescribers can print for patients or refer them to the listed websites/information. Examples of potential patient resources may include information sheets from Health Navigator NZ 1 or a modified version of the Personal Decision Guide for Medicines available from the national prescribing service in Australia. 2 -Evaluation: As part of the qualitative evaluation of the MORE module (through focus groups and/or interviews), we will obtain feedback on how GPs integrated the module into their consultation and its impact on time management.
3b. Impact on improving quality service delivery