Context and setting
In Australia, the General Practitioner (GP) is the cornerstone of primary care coordination. About 84% of Australians see a GP every year, and 77% of patients have a preferred GP [12]. The goal of the proposed intervention was to help GPs identify their vulnerable patients and promote follow up appointments during the period of restrictions. Some particularities of the Australian health system determined the technology choice:
- Geographical location - Veterans are distributed across Australian states and territories. While there are a few GPs specialised in veteran care, most have less than 4 veterans under their care. Patients are free to choose their GP irrespective of geographical location. While this may increase patient satisfaction and access, the lack of patient registration makes it harder for practices to define their population, potentially reducing continuity of care [13].
- Technological readiness - Australian GPs have had near universal use of electronic health records for more than 10 years [14] and a large penetration of secure messaging infrastructure for receiving laboratory test results.
- Public funded, privately operated model – Due to the health model in Australia, any intervention focused on GPs must be highly collaborative and involve practitioners from the start. There is a high degree of agency during GP appointments, and payers have limited influence on practice.
The Veterans’ MATES program
The Veterans’ MATES program is a multifaceted intervention, composed of an educational component and an audit and feedback component delivered to general practitioners (GPs), supported by educational components delivered to pharmacists, other relevant healthcare professionals and veterans. Interventions are created in three sequential steps.
The first step is an epidemiological inquiry to identify trends and potential issues in healthcare access and use. Examples include: long term prescription of medicines recommended for acute issues; doses above guideline recommendations; and lack of screening tests for an eligible population. The program has access to the DVA health claims database, updated monthly. This database includes all dispensed medicines requiring prescription, claimed healthcare services and laboratory services, home care and aged care.
The second step is the development of educational material and audit and feedback documents. This is a collaborative process with heavy stakeholder involvement, including multiple health professionals and behaviour change specialists.
The final step is the identification and delivery of the intervention to veterans and their main healthcare provider. This step requires use of patient-level information contained in the database to print and post via regular mail personalised audit and feedback documents at scale, reaching tens of thousands of veterans and GPs per intervention.
The program has been extensively described elsewhere [15]. It has been shown to be effective for changing professional behaviour in different domains [15], including promoting medicine review [16], osteoporosis screening [17], uptake of health services [18], reducing inappropriate proton pump inhibitor use [19] and hypnotic use for insomnia [20].
The Veterans’ MATES program ran the first intervention in 2004 and, since then, has delivered 4 interventions each year. In 2019, a digital precision public health initiative was started to make the best use of digital technology infrastructure and replace all printed materials. The proposed solution was primarily aimed at improving intervention efficacy by increasing GP engagement and reducing the lag between the detection of an issue and its notification to the GP.
Study design and sample
To evaluate the effectiveness of a digital precision public health intervention in promoting continuity of care during national emergencies, we performed a non-randomised experimental study. As part of the Veterans’ MATES program, we developed a digital intervention to promote care provision for patients vulnerable to poor outcomes following COVID-19 infection compared to usual care (paper based intervention, sent by post).
Patient allocation to postal or digital group was done in two sequential steps. First, eligible patients were identified based on information contained in the administrative health claims database. Identification of vulnerable individuals was based on the best information at the time, and focused on the population at highest risk of poor outcomes from COVID-19, which was persons aged over 70 years with the following comorbidities, hypertension [21-28], chronic heart disease [21-28], diabetes [21-29], chronic airways disease [22-28], cerebrovascular disease [21, 22, 25, 30] , chronic liver disease [27] , chronic renal failure [22, 24, 26-28] , malignancy [21, 22, 25, 26, 28, 31] or being immunocompromised [27]. Identification algorithms were composed of clinical rules with varying levels of complexity, looking for past diagnostic codes (ICD-10) during hospitalisations, use of medicines indicating treatment for one of the target comorbidities (e.g., chronic carvedilol use in heart failure), and combinations of services and medicines used.
After patient eligibility checking, the primary GP was identified using a proprietary algorithm based on the frequency and recency of appointments. General practitioners with at least one eligible patient were eligible for the intervention. All GPs who had the capability to receive the digital intervention (access to electronic health record and secure message delivery) were included in the digital group. The remainder was included in the post group.
Intervention development
The main goal of this intervention was to promote continuity of care during lockdown and social restrictions. The underlying theory was that the provision of personalised recommendations in the form of an audit and feedback document delivered directly to a GP’s clinical software would influence patient recall and trigger an early appointment. The initiative was conducted using a collaborative, pragmatic approach, influenced by Greenhalgh et al diffusion of innovations model [32], in order to develop a solution that could be implemented at national scale. The model summarises a collection of theoretical and empirical findings, and highlights the interplay between an innovation, the adopter, the context in which the innovation takes place, the implementation and the diffusion process. The model suggests innovation developers to consider 9 dimensions during intervention creation: 1) Innovation; 2) Adopter; 3) Assimilation; 4) Communication and Influence; 5) System Antecedents for Innovation; 6) System Readiness for Innovation; 7) Outer Context; 8) Implementation Process; 9) Linkage.
The processes involved in intervention development complements the 3-steps used by Veterans’ MATES interventions, suited for rapid care coordination (see Figure 1). The development of all content and interventions is based on the best evidence available at the time and supported by repeated reviews by healthcare professional panels. The audit and feedback document was developed and submitted to a stakeholder review group, including health professionals (pharmacists, GPs, among others). The behaviour change techniques used included goal setting (e.g., “Schedule appointments to ensure vulnerable patients are still receiving necessary care”), prompts (e.g., medicine dispensing suggesting respiratory disease), and information about health consequences (e.g., rationale for recommendations). A sample can be seen in Figure 2.
All veterans living in the community (68,872 individuals) received educational material delivered by post about COVID-19 infection prevention prompting them to maintain regular contact with care providers and to continue to adhere to health plans, as well as how to access care during the pandemic given the expansion of video and telehealth appointments and free medicine delivery services for eligible persons. The online version of the printed materials can be seen in [33].
Outcomes and statistical analysis
The main outcome of the intervention was the time to first appointment with the primary GP. Australia had in place restrictions on visitors to aged care, thus, we excluded veterans living in aged care from the time-to-event analysis. To compare the digital and the postal groups, we performed a time-to-event analysis, considering the first visit to the primary GP since the intervention delivery date as the target event. Determinants of an early GP appointment were analysed using Cox regression. The analysed variables included age at the time of intervention, gender and number of appointments in the previous year (April/2019 to March/2020).
Given the large sample and the purposeful sampling, we considered a 99% confidence interval (p <= 0.01) for all hypothesis tests. All analyses were performed in Python 3.7. The main statistical library used for time-to-event was lifelines 0.25 [34].