Examining prescribing behaviours in Australian primary care: three case studies using the MedicineInsight electronic health record dataset

Background Research using primary care electronic health records (EHRs) provides opportunities but is not without challenges. This thesis presents three projects using applied biostatistics and epidemiological methods to research medicine prescribing with the Australian primary care, EHR dataset MedicineInsight. Methods Data extracts from Australian general practice EHR database MedicineInsight between 2006 and 2020 were analysed. Project one investigated influenza vaccination and subsequent antibiotic prescribing in 1–4-year-olds attending general practices in two influenza seasons using case-control methods. Project two evaluated a prescriber education program targeting inappropriate antibiotic prescribing using two approaches: interrupted versus Bayesian structural time series. Project three analysed adults initiating antihypertensive therapy to ascertain alignment between treatment guidelines and subsequent prescribing patterns using descriptive statistics and count models. Results Children receiving influenza vaccination were less likely to be prescribed antibiotics in 2018 (aOR = 0.65; 95% CI: 0.62–0.69) and 2019 (aOR = 0.78; 95% CI: 0.73–0.82). I estimated if all Australian 1–4-year-olds were vaccinated, up to 100,000 antibiotic prescriptions could be avoided annually. Antibiotic prescribing reduced contemporaneously with the prescriber education program by 22.5% (95% CI: 17.9–27.1) using interrupted and 17.4% (95% CrI: 10.7–24.2) using Bayesian structural time series methods. In adults initiating antihypertensive therapy, 42% received the same regimen only, 33% had their regimen changed, and 25% received no further prescriptions. Of those whose regimen changed, 50% had a dose increase while only 28% had a new medicine added, which is guideline preferred. Conclusions Influenza vaccination and prescriber education were associated with reduced antibiotic prescribing. Antihypertensive prescribing and guideline alignment could improve, and use of fixed-dose drug combinations may increase alignment with treatment guidelines. General practice EHR data can be used to investigate important public health questions about prescribing in Australian primary care, but satisfaction of the strong assumptions underpinning the time series analyses was questionable. Tailoring research questions to the strengths of the EHR dataset and conducting robust validation of analyses increases the likelihood of accurate, defensible and implementable findings.