Implementing a Scalable, personalised, behaviour Change digitAL hEalth programme in primary care for type 2 diabetes treatment: the SCALE cluster-randomised study protocol

Abstract

Introduction

Type 2 diabetes mellitus (T2DM) is a fast-growing chronic disease, with at least 1.3 million people diagnosed in Australia. In the Western Sydney Local Health District (WSLHD), an estimated 13.1% of all adults have T2DM. The condition significantly contributes to cardiovascular, heart and kidney diseases and causes a large disease burden. Lifestyle modifications, such as improved nutrition, increased physical activity and stress reduction, are recommended as first-line treatments for T2DM management. However, the current primary care system cannot meet the growing demands for diabetes care, necessitating the development of innovative, scalable, cost-effective solutions. Digital health technologies present a promising approach for promoting self-management in individuals with T2DM.

Methods and analysis

This cluster-randomised controlled trial aims to evaluate the feasibility and effectiveness of Gro-AUS, a localised version of the Gro Health app in Australia, to support T2DM management in Australian primary care settings. The trial will be conducted across multiple general practice clinics within the WSLHD, an area with a high prevalence of T2DM and significant cultural diversity in patient populations. Participants will be randomly assigned by clinic to either the intervention group (digital health programme) or control group (standard care). Primary outcomes include improvements in glycaemic control, cardiovascular risk factors and diabetes remission, with secondary outcomes such as weight loss, physical activity and mental well-being. Data will be collected using electronic and paper methods, with secure storage and de-identification ensuring participant privacy. The study’s mixed-method approach ensures inclusivity for patients with varying levels of digital literacy. Data will be securely stored, de-identified and used to assess the effectiveness of the intervention. Findings are expected to inform future models of diabetes care in Australia, providing evidence for the scalability of digital health technologies in chronic disease management.

Strengths and limitations

This trial is by nature unblinded. The recruitment style for a stepped-wedge trial may also bias participant engagement. However, it has direct implications for clinical practice as an effectiveness implementation trial. The design also allows for a much larger sample and more statistical power to examine outcomes.

Ethics and dissemination

This trial has been prospectively registered with the Australian New Zealand Clinical Trials Registry. Ethical approval has been granted by the WSLHD Human Research Ethics Committee prior to data collection. Results will be disseminated through publication in a peer-reviewed medical journal and shared via the Agency for Clinical Innovation, the Primary Care Health Network and through community engagement initiatives.

Trial registration

ANZCTR388639.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The SCALE cluster-randomised trial is testing a novel intervention for diabetes in a unique pragmatic design that combines implementation and translation with a rigorous assessment of outcomes.

• The multiethnic context allows for broad generalisability outside of western Sydney.

• All primary outcomes are objectively measured.

• Clusters in this trial cannot be blinded, and therefore, some sources of bias such as contamination bias and selection/attrition bias cannot be eliminated despite the randomized nature of this design.

Introduction

Type 2 diabetes mellitus (T2DM) is a major global public health concern, with its increasing prevalence driven by lifestyle changes and population ageing.¹ In 2021, the International Diabetes Federation reported over 537 million adults worldwide were living with diabetes, a number expected to reach 783 million by 2045.² In Australia, T2DM is the fastest-growing chronic condition, with over 1.3 million people diagnosed,³ although this figure is an acknowledged underestimate.⁴ Western Sydney Local Health District (WSLHD) covers a large geographic region in the city’s west with about 1 million inhabitants. The area is a diabetes hotspot, with 13.1% of the adult population diagnosed with T2DM, exceeding the state average of 12%.⁵

T2DM is a major contributor to both global and Australian morbidity and mortality, causing a range of long-term complications.^{6 7} Cardiovascular disease remains the leading cause of death among individuals with T2DM.^{8 9} T2DM substantially increases the risk of chronic kidney disease, often leading to end-stage renal failure,¹⁰ and heightens the risk of stroke, neuropathy, retinopathy and peripheral vascular disease.¹¹ This negatively impacts quality of life and drives up healthcare costs. Emerging research also links T2DM to comorbidities such as cognitive decline, obstructive sleep apnoea, liver disease and greater susceptibility to infections.¹² In some countries, cancer and dementia have also become leading causes of death among people with diabetes.

Traditional models of diabetes care were focused on specialist endocrine management but have become unsustainable with the increasing prevalence of diabetes.¹³ In New South Wales, an Australian state of 8.1 million people, the most recent estimates suggest that there are fewer than 300 endocrinologists.¹⁴ The most recent estimates suggest ~700 000 adults live with diabetes across New South Wales (NSW). It is not possible for the limited endocrine services to manage all people who have diabetes. Therefore, primary and community healthcare services are fast becoming the central point for diabetes management.¹⁵ Despite the availability of evidence-based guidelines and positive trends in care delivery, many patients with diabetes are still unable to achieve recommended targets for glycaemic control, cholesterol levels and blood pressure.¹⁶ Barriers to effective management in Australian primary care settings include limited access to healthcare services,¹⁷ a lack of culturally competent care,^{18 19} challenges with medication adherence,²⁰ the financial burden of out-of-pocket costs due to a shortage of bulk-billing general practitioners^{21 22} and a lack of knowledge, confidence and motivation regarding lifestyle modifications.^{23 24}

Research consistently shows that lifestyle interventions are essential for managing T2DM, particularly through dietary changes, increased physical activity and behavioural modifications.^{25 26} Studies demonstrate that even modest weight loss (5–10% of initial body weight) can significantly improve blood glucose levels, cardiometabolic profiles and overall health outcomes.²⁷ However, long-term adherence to these interventions remains a challenge for many patients, often due to socio-economic barriers, lack of support or insufficient self-management skills.²⁷

To address these challenges, digital health technologies have emerged as promising tools for enhancing self-management and patient education in chronic conditions including T2DM.^{28 29} A meta-analysis of 25 randomised controlled trials (RCTs) involving 3360 participants with T2DM, type 1 diabetes and pre-diabetes revealed that mobile health app interventions can significantly improve glycaemic control, with HbA1c levels decreasing by an average of 0.90% for T2DM in 95% of the studies.³⁰

One such intervention is the Gro Health Structured Education and Self-Management Programme, which offers artificial intelligence-enhanced tools for behaviour change, weight management and T2DM care. Co-designed with over 20 000 patients and multidisciplinary clinical teams, including diabetes specialists and endocrinologists from the National Health Service, Gro Health has demonstrable benefits for patients and clinicians. The programme adopts a holistic approach, incorporating anti-discrimination and anti-stigma principles throughout its design and implementation.31,33

Gro Health: a personalised digital health platform for well-being

Gro Health is a comprehensive digital health platform that empowers individuals to self-manage their physical and mental well-being. Developed by Diabetes Digital Media (Coventry, UK), Gro Health offers a personalised approach to health management, focusing on mental well-being, sleep, activity and nutrition. Through evidence-based structured education, guided behavioural change activities, virtual meetups, community support, health tracking and data-driven insights, Gro Health facilitates precision digital health.³⁴ The Gro Health platform is designed to support personalised health management by tailoring interventions to individual health goals, ethnicity, gender, dietary preferences and activity levels. It uses the COM-B model of behaviour change to address key determinants of sustained behaviour modification: capability, opportunity and motivation. The platform provides evidence-based structured education, dietary and physical activity guidance, psychological support and goal-setting tools. It also integrates with wearable devices to enable real-time progress tracking and feedback.^{33 34}

To enhance accessibility and usability, Gro Health is available across multiple digital platforms, including web, iOS, Android, smartwatches, smart TVs and digital assistants. The platform supports 18 languages, including English, Chinese, Arabic, Hindi and Urdu, and is accessible 24/7, allowing for self-paced engagement and individualised support.

Previous research demonstrates Gro Health achieves high user retention and engagement, including culturally/linguistically diverse communities, and is effective in achieving clinically relevant reductions in weight and blood glucose, reducing and/or eliminating diabetes medication and diabetes remission.3132 34,37 The programme has recently been reviewed and localised for use in Australia (Gro-AUS) including clinical safety, data storage, encryption and security and information governance accreditation.³⁸

This article presents a protocol for a cluster-randomised trial examining the feasibility and efficacy of Gro Health as a scalable, cost-effective intervention to assist adults living with T2DM in Western Sydney, delivered through Western Sydney Diabetes (WSD), an integrated care initiative established in 2014 to combat diabetes in the region.

Aims and objectives

The aim of this study is to evaluate the effectiveness of implementing the Gro-AUS digital health intervention in an Australian primary care outpatient setting to improve the management of adults with T2DM. The primary outcome will be changes in glycaemic control, specifically HbA1c levels, from baseline to 3, 6 and 9 months post-baseline.

To meet these aims, the study objectives are as follows:

1. Assess changes in HbA1c, diabetes medication use and cardiovascular risk factors (lipid profile, blood pressure, body weight, body mass index (BMI)).

2. Evaluate kidney function (estimated glomerular filtration rate (eGFR)) and determine the proportion of participants achieving diabetes remission.

3. Measure changes in health-related quality of life (EQ-5D-5L) and diabetes-related emotional distress (Problem Areas In Diabetes (PPAID) scale).

4. Examine the feasibility of implementing Gro-AUS (referral uptake, completion rates, engagement).

5. Hospitalisation outcomes related to diabetes complications using Health Information Exchange data.

Methods and analysis

Study design

This research project will use an open-label cluster randomised stepped-wedge design, a best-practice approach that allows all clusters to access the intervention at different time points while serving as controls during their waiting period.^{39 40} The stepped-wedge design is illustrated in figure 1. The stepped-wedge design was chosen to mirror the usual process of rolling out an intervention such as this, which improves both general practice and patient buy-in. We anticipate better recruitment and retention with this design than with a standard cluster-randomised trial comparing to a waitlist or placebo control.

Figure 1. Stepped-wedge schematic for the study. Programmatic roll-out will be randomised by general practitioner practices (cluster) with sequential implementation of the intervention resulting in each study group contributing time as unexposed (grey) and exposed (blue) to the Gro-AUS intervention.

The trial is registered with the Australia New Zealand Clinical Trials Registry (388639) and has been approved by the WSLHD Human Research Ethics Committee, approval number: 2024/PID02400 - 2024/ETH02078.

The study will recruit 12 general practices (clusters), which will be randomised to either receive the intervention at commencement of the trial or to wait until their designated time period. Each GP practice aims to recruit 20 patients, with ‘wedges’ occurring at 3-month intervals over a total period of 12 months, accommodating four practices per interval. Prior to patient recruitment, practices will undergo a 3-month training phase. Overall, the trial will span 15 months, with recruitment facilitated by WSD, WSLHD staff and Wentwest (WSPHN).

Study settings

Participant recruitment will take place at GP practices throughout the region. To formalise their involvement, participating GPs and/or practice managers will sign a Clinical Trial Research Agreement (Collaborative or Cooperative Research Group Studies Form). WSD will conduct information sessions to engage GPs, providing comprehensive details about the project and clarifying their roles. The study coordinator will train GP staff on recruitment procedures, patient consent and data collection activities, while the WSLHD Research Support Team will offer training in Good Clinical Practices (GCP) to uphold high research standards.

Participating GP practices will be eligible for cost recovery, paid in three instalments over the trial’s 15-month duration, to compensate for time and resources devoted to patient recruitment and data collection.

Participant recruitment

Recruitment will target 240 individuals with T2DM. The study team will collaborate closely with participating GP practices to facilitate this process.

GPs and GP nurses will play a key role in introducing the study to patients, identifying eligible individuals during routine consultations or through practice records. They will provide an overview of the study with information materials and consent forms, to help patients understand participation requirements.

Eligible patients will be invited to provide informed consent at the GP clinic, with the clinical trial coordinator available for assistance. The full consent form is available as a supplementary material. Following consent, an eligibility assessment will be conducted based on the study’s criteria. Once eligibility is confirmed, participants will complete baseline data collection, including pathology results and survey responses, either directly with GP staff or the clinical trial coordinator, based on their preference and site logistics.

This approach integrates patient recruitment into routine clinical practice, leveraging the trusted relationship between patients and their GPs while maintaining flexibility and support from the study team. The study funding provides for reimbursement for both individual patients and GP practices to assist with recruitment.

Consent

Informed consent will be obtained from all patient participants prior to enrolment. This will be carried out by trained GP staff or clinical trial coordinators proficient in GCP guidelines. Potential participants will receive detailed information about the study, including its objectives, procedures, potential benefits, risks and their right to withdraw at any time without consequence. Participants will have sufficient opportunity to ask questions and clarify any uncertainties before providing written informed consent. For details on the consent process and how participant data are shared, see the supplementary appendix for the participant information and consent forms.

Consent can be obtained in person or via eConsent, which will be sent through messaging services endorsed by the hospital. No identifiable information will be used in these communications to ensure privacy and confidentiality. This approach allows participants to choose their preferred consent method while upholding rigorous ethical standards.

Recruitment criteria

Inclusion criteria

• People with T2DM attending GP practices in WSLHD.

• Owning a smartphone.

• Able to fluently speak/read one of the languages included in Gro-AUS (Hindi, Chinese, Arabic, English).

• HbA1c >7.5% within the last 6 months.

• Recent diabetes diagnosis (within the last 10 years).

Exclusion criteria

• Diabetes other than T2DM.

• Current pregnancy or breastfeeding.

• History of severe mental health diagnoses, in particular eating disorders.

• No access to a smartphone capable of accessing the Gro-AUS app.

• Severe kidney disease (chronic kidney disease stage 2, eGFR <30).

• Previous bariatric surgery.

• Porphyria.

• Carnitine deficiency (primary), carnitine palmitoyl transferase I or II deficiency, carnitine translocase deficiency, beta-oxidation defects, medium-chain acyl dehydrogenase deficiency, long-chain acyl dehydrogenase deficiency, short-chain acyl dehydrogenase deficiency, long-chain 3-hydroxyacyl-CoA deficiency, medium-chain 3-hydroxyacyl-CoA deficiency, pyruvate carboxylase deficiency.

• Severe complications/comorbidities that may compromise their ability to follow study protocol.

Intervention

Participants will be provided with the Gro-AUS application for a 12-month period that will begin when their GP clinic is allocated to the intervention. The intervention will also include follow-up by their local GP at 3-month intervals to assess progress. The trial will introduce the app to three groups of GP clinics at staggered 3-month intervals. On activation for each cluster, the clinical trial coordinator will assist all enrolled patients in downloading and activating the application. Follow-up assessments will occur at 3, 6 and 9 months post-recruitment for all participants, regardless of their intervention or intervention commencement status.

Participants will be provided free access to the app during the trial and will receive an access code on downloading it, granting them access for 12 months. The app will automatically log usage metrics.

Recruitment and retention

Patients will be recruited to the study by their GP, with assistance from the study team where necessary. The funding for the trial provides reimbursement for both patients and GP clinics to assist in recruitment, with full reimbursement coming at the end of the trial period for both groups. This is aimed at improving retention, as there are often issues with dropout in app-based interventions. Participant usage of the application will be monitored through process metrics such as app opens and use of materials.

Study outline

The Gro-AUS trial will consist of three phases: a pre-trial phase, a patient participant recruitment and baseline data collection phase, and an intervention delivery and follow-up data collection phase.

Pre-trial phase

Training of GP practices staff

Provide comprehensive training to study personnel on the study protocol, data collection procedures and ethical guidelines.

Patient participant recruitment and baseline data collection phase

• Screening and eligibility assessment: identify and assess potential patient participants to ensure study eligibility criteria are met.

• Baseline data collection: collect baseline data on patient participants’ demographic information, health status and other relevant variables.

• Randomisation: randomly assign GP clinics to one of the three study groups.

Intervention delivery and follow-up data collection phase

• Intervention delivery: deliver the intervention to each cluster over a 7 to 15-month period, ensuring consistent implementation and adherence to the study protocol.

• Follow-up assessments: conduct three follow-up assessments at 3, 6 and 9 months after the baseline assessment to collect data on patient participant outcomes and adherence to the intervention.

Randomisation and masking

GP clinics will be randomised into three groups at the proposed 3-month intervals. Patient participants will be recruited prior to allocation to minimise potential systematic variation between study groups. The randomisation schedule will be created by the study statistician in Stata v15 or v16 using a custom script, with clinics block-randomised together once all clinics have been recruited, based on random samples drawn from a normal distribution.

As with all pragmatic stepped-wedge trials, randomisation poses logistical challenges. GP clinics must maintain flexibility to align with the randomisation schedule. To facilitate this, the study coordinator will ensure each clinic has a central contact person for communication. All clinics will receive a 1-month notification before entering the activation phase of the trial. The stepped-wedge design minimises disruption and alleviates many ethical concerns; it allows patients involved in the trial to receive the treatment without being disadvantaged by the control period.

Sample size

Power calculations were conducted using the power cluster command in Stata 15. Given the complex relationship between stepped-wedge trials and presumed power, several variables will impact the study’s likely power. We expect clusters to be relatively heterogeneous due to the diversity of GP clinics in the region. Thus, the study should have slightly higher precision than a standard cluster-randomised trial. However, we conservatively assume precision similar to that of a simple cluster-randomised parallel-assignment study.

A previous study has reported a reduction from 53.3 mmol/mol to 50 mmol/mol (8% to 7.7%) in HbA1c using the Gro Health programme compared with a 0.5 mmol/mol reduction in wait-list controls. Using a SD of 4 mmol/mol and a lower retention rate of 71% from clinical studies of Gro Health’s effectiveness, a sample size of 12 practices with 20 patients each will provide sufficient power to detect a clinically relevant HbA1c reduction of 0.3% between the treatment and control groups, consistent with FDA pharmacotherapy treatment approval standards. This sample size is considered feasible to recruit within the timeframe and region of the study. The sample size is predicated on a single previous piece of research which likely will not perfectly match the patient population in this study as it is very different. We believe that this sample size reflects the best balance of practicality and usefulness and should be sufficient to detect clinically important effects.

Outcome measures

The primary outcome measure collected at baseline and at 3, 6 and 9 months will be:

Within-subjects difference in HbA_1c from baseline

HbA1c is a blood test that is used to diagnose T2DM and monitor blood glucose control in people living with diabetes. HbA1c reflects average glycaemia over the preceding 6–8 weeks. The test is subsidised by Medicare up to four times in a 12-month period. In some patients, HbA1c may be measured more frequently than 3 monthly to closely monitor glycaemic control.^{41 42}

In addition, at the same time-points, we will collect the following secondary outcome measures:

Health-related quality of life (assessed using EQ-5D)

EQ-5D-5L is a standardised measure of health-related quality of life comprising five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. A visual analogue scale complements the descriptive system, allowing patients to rate their overall health. The EQ-5D-5L is a copyrighted instrument and will be used under agreement with EuroQol.⁴³

Diabetes distress (measured using Problem Areas in Diabetes scale)

The PAID is a self-report questionnaire that contains 20 items that describe negative emotions related to diabetes (eg, fear, anger, frustration) commonly experienced by patients with diabetes. Completion takes approximately 5 min.^{42 44}

Other diabetes and cardiometabolic markers (eGFR, lipids, fasting glucose, BMI, blood pressure)

eGFR is a crucial biomarker of kidney function and per current diabetes management guidelines and is a fundamental component of routine care in diabetes.^{45 46}

Lipid profiles, including total cholesterol, triglycerides, HDL and LDL cholesterol, will be assessed as part of routine care for patients with T2DM.^{45 47} Addressing lipid abnormalities is critical in T2DM because of their high prevalence and significant contribution to cardiovascular disease risk. Dyslipidaemia, characterised by a mixed profile—elevated triglycerides, reduced HDL-C and atherogenic LDL-C particles—affects over 75% of individuals with T2DM and is closely linked to insulin resistance. As a key modifiable CVD risk factor, managing dyslipidaemia is essential to reducing cardiovascular complications in these patients.⁴⁸ Blood pressure and weight will be measured as is standard at GP appointments.

Diabetes remission will be defined as HbA1c <6.5% and prescribed no medications aside from metformin. This will be an exploratory secondary endpoint.

Gro-AUS app process measure

The percentage of people referred to the intervention who take it up, completion rate of patient participants and rate of engagement as measured through participant logging into the app.

Hospitalisations related to diabetes

Hospitalisations, including hospitalisation for diabetes-related complications identified using ICD-10 coding through linkage to NSW hospital data systems.

Changes in diabetes-controlling medication

Glucose-lowering medication use for T2DM management will be collected to evaluate the impact of the Gro-AUS intervention on diabetes medication changes.

Mean HbA1c, lipid profile, eGFR and fasting glucose levels will be obtained from pathology results ordered by GPs as part of routine care for patients with T2DM and entered into the WSLHD REDCap server. The EQ-5D and PAID questionnaires will be administered electronically through a REDCap link sent to patient participants. Alternatively, for participants unable or unwilling to complete the questionnaires electronically, the trial coordinator will conduct telephone interviews to gather the required data, which will then be entered into REDCap.

Data collection and retention

As per local legislation, data will be retained for 15 years post-publication before being destroyed by method of file deletion. Study personnel will attempt to obtain routine clinical information for patients who are lost to follow-up unless they discontinue entirely from the trial, including accessing GP and hospital records.

Due to the complex nature of recruitment, with individual GP clinics operating as sole traders in Australia, it will not be possible to publish data from this clinical trial. Data forms, including instruments used, can be requested from the study team.

Data will be managed by the clinical trial coordinator and chief investigator. This will include regular checks for possible data errors in the database. Security will be maintained as the data sits on secure servers, which are regularly audited and require authentication to access.

This study does not have an independent data monitoring committee, as the risk to participants was deemed low in ethical review. However, there will be a data steering committee which will meet quarterly and includes an independent statistical advisor to oversee the trial progress. There is no plan for interim analyses.

Adverse events

The clinical trial coordinator will be primarily responsible for collecting adverse event data. We have recruited an independent medical expert to adjudicate the severity of adverse events and whether they are related to the trial. Given the extremely low risk of the clinical trial, we do not anticipate a large number of reported adverse events.

All clinical trials in Australia may be audited by government bodies; however, there is no current auditing plan in place for this trial.

Statistical analysis

The analysis will compare primary and secondary outcomes between the baseline and intervention periods adjusted by the allocated intervention period, GP practice and covariates. We will follow the intention-to-treat principle, with both a full intention-to-treat (ITT) analysis including all randomised participants and a modified ITT principle including all randomised participants who logged into the Gro-Aus application at least once.

The primary statistical analysis will take the form of a within-subjects generalised linear mixed model design with random effects for cluster (GP practice) and fixed effects for the tranche (period) that the GP clinic was randomised in (study group), as well as controlling for the intra-cluster and intra-period correlation. We plan to also include key possible confounders that are likely to differ between practices in a systematic way, including age, gender, suburb of residence (SEIFA) and time since diabetes diagnosis in the full linear model predicting within-subject change in HbA1c.

For other secondary outcomes, we plan to use similar logistic and binomial models for binary and discrete data, respectively. We will also report unadjusted estimates for all outcomes to allow for better interpretation of the regression outputs. We will plot the difference over time for all outcomes. In addition, we plan to conduct subgroup analyses by clinic where possible and may use the data for secondary investigations in the future.

We plan to cost the provision of this programme using the Gro-AUS general costs. Hospital and GP outcome measures such as hospitalisations will be used to generate an estimate of the benefits for this project, which will allow for the calculation of incremental cost-effectiveness ratios comparing intervention and control.

We plan to use multiple imputation to address missing data. We will not set a threshold for missingness to use multiple imputation.⁴⁹

All statistical analyses will be conducted by independent blinded study statisticians.

Authorship guidelines

We do not plan to use professional writers. Authors for studies published from this clinical trial will have had to work on the publication as an author.

Discussion

There is currently a dearth of programmes that can be used in community health for patient self-management of diabetes that are supported by gold-standard evidence. One of the primary barriers to collecting this evidence is the complex nature of intervening in GP clinics. This study has been funded to provide high-quality evidence to assess the benefits of Gro-AUS in this setting in Australia.

The stepped-wedge design has been used effectively in recent years to evaluate complex diabetes interventions in primary care. A 2023 trial reported the effectiveness of a primary care enhancement for chronic disease in Australian general practices, with the trial setup assisting in the rollout of the intervention as designed.⁵⁰ A 2020 trial examining a virtual clinical pharmacy in a hospital setting also showed the utility and benefits of this trial design.⁵¹

This type of trial design does have limitations that must be acknowledged. While there are some benefits for statistical power from using a stepped-wedge design, it remains a complex implementation-focused trial method. This requires collaboration with several local services and GP clinics, which has increased the burden on the research staff.

The cluster-randomised design will likely result in some measure of contamination across GP clinics. This may come about as patients are able to move between clinics, and clinicians discuss their work with their fellows. However, the large geographic spread of our target area and the relatively small number of patients per clinic will reduce the potential for this to be a significant limitation for this research.

It is similarly challenging to ensure that patients have had no prior exposure to applications of this nature. However, this is also a strength of the study, as there are relatively few patients in clinical settings who have had no exposure at all to similar mHealth interventions, and the number is declining over time as smartphones become ever more ubiquitous.

In addition, there is no possibility of blinding. Participants will not be blinded to when their clinic will have the intervention available, and clinicians will know from the start of the study when they are being activated. This introduces unavoidable biases into the eventual results. While allocation is concealed by the randomisation process—clinics will all be randomised at the same time—the lack of blinding for all but the statistical staff means that the study will nevertheless have some biases associated with it.

Recruitment also introduces the possibility of selection biases. Individual clinicians at practices have the opportunity to direct their recruitment that introduces the possibility of substantial differences between practices who are participating in this study. This is an inherent issue with cluster randomisation and will be partially controlled by the statistical methodology but does not eliminate the possibility of biased results.

The study and research design also has some significant strengths. It is almost impossible to conduct traditional RCTs on easily available interventions without some measure of innovative design. Patients and GPs are highly capable of identifying their own diabetes self-management applications, and it is increasingly difficult to establish a true ‘control’ group in the current diabetes management environment. This issue is largely ameliorated through the provision of the intervention to all GP clinics in the study in a way that mimics traditional rollout of interventions.

Additionally, the study allows us to test the real-world effectiveness of the intervention as well as provide evidence around rollout at the same time. Rather than the tightly controlled scenario of more traditional RCTs, this study will allow us to directly infer about whether the intervention is beneficial for clinical practice in the targeted area immediately. This largely eliminates the significant burden of research translation, as the study is by definition translated once completed.

Western Sydney is one of the most multicultural areas of Australia, with a large portion of inhabitants born overseas as well as large urban communities of Aboriginal people living in the area, as well as significant refugee populations and other diverse groups. We aim to recruit a culturally diverse sample in this research through our network of GPs. This has implications for generalisability—the results will likely translate well to other high-income locations with highly diverse communities, but may not be applicable to low-income regions or areas with less diversity. In addition, Australia’s healthcare system is somewhat unique, meaning that international generalisability may be limited in situations where primary care is substantially different such as in China.

The views expressed in this article are those of the author(s) and do not necessarily represent the official position of their employers.