Philosophical approach
The quantitative component of this study involved an analysis of Australian GP practice data (i.e., entries from medical records of GPs) in the MedicineInsight dataset. This component fits the positivist paradigm with a logical deductive approach. Although this reflects an empiricist epistemology, we recognise that there are multiple flaws in these assumptions (e.g., have GPs diagnosed mastitis correctly? Did patients purchase the antibiotics and take them?). In contrast, the qualitative component is based on a more interpretivist approach, involving in-depth interviews with GPs. We based the interview guide around the COM-B (‘capability, opportunity, motivation-behaviour’) system as a framework for understanding the barriers/enablers to GPs’ use of guidelines [
15]. We used an inductive approach to coding and analysis of the interviews. We recognise that our attitudes influence the way we collect and analyse the data, and our prior knowledge of the topic, findings from the dataset study, and other factors were all be brought into our research conclusions. This component has a relativist ontology (meanings are constructed subjectively) and subjectivist epistemology (researchers are part of the investigation) [
16]. In this paper, we bring the two components together with a pragmatic approach: there are multiple perspectives of reality or worldviews.
Study design
We used a convergent mixed methods design, where quantitative and qualitative data are collected and analysed separately over a similar period, and results are merged and compared [
17]. In this study, quantitative analysis of a large dataset and qualitative data collection and analysis occurred over the same timeframe, with regular interaction between the researchers working on each component. The interview guide about GPs’ use of guidelines
built on the preliminary findings from the dataset, and findings from both components were
merged for analysis [
17]. The findings are described in a
weaving approach by presenting quantitative and qualitative findings topic by topic [
17]. The data are brought together in a joint display [
18], the Pillar Integration Process (described below) [
19]. We followed the Good Reporting of A Mixed Methods Study (GRAMMS) framework for writing up the study [
20].
Quantitative component
MedicineInsight is a large-scale primary care dataset of longitudinal de-identified electronic health records (EHRs) in Australia [
21]. The MedicineInsight program collates routinely collected EHR data from clinical information systems from consenting general practices; currently over 500 practices with over 3,000 GPs involved. It includes information from 9% of all Australian GPs and 13% of all Australian patients who saw a GP at least once during the financial year (2018–2019) [
22].
The independent MedicineInsight Data Governance Committee approved the quantitative component (protocol 2019–003) and the Human Research Ethics Committee of the University of Adelaide and La Trobe University exempted it from ethical review due to the use of non-identifiable data.
Using data from 2021–2022, we restricted our analysis to females of reproductive age (18–44 years inclusive) with one or more documented clinical encounters related to mastitis and documentation relating to a pregnancy within the previous 12-months of the encounter. Mastitis encounters were identified by searching the ‘Encounter reason’ free text field for the term ‘mastitis’. We also searched the ‘Test reason’ and ‘Prescription reason’ free text field for the term ‘mastitis’. We excluded the free text term ‘granulomatous mastitis’ as this was considered unlikely to be related to lactational mastitis. Clinical encounters for mastitis occurring within 14 days of a previous mastitis encounter were defined as belonging to the same treatment episode. Only the first episode per individual was included in the analysis. Documented pregnancies were identified using the separate ‘pregnancy’ dataset which included data on date of last menstrual period and estimated date of confinement. We also searched the ‘Encounter reason’ free text field using terms related to pregnancy (i.e., ‘Antenatal’, ‘Pregnancy’, ‘Hyperemesis gravidarum’, ‘Morning sickness’), postpartum (‘postnatal’, ‘postpartum’, ‘baby check’, ‘6 week check’), or breast feeding (i.e., ‘breast feeding’, ‘breastfeeding’, ‘lactation’) to identify women with a recent pregnancy. This was undertaken to increase the likelihood of the clinical encounter being related to lactational mastitis. Notably, the MedicineInsight program uses the terms sex and gender interchangeably and presents sex/gender information as a single binary variable (i.e., female/male). Pensioner concession status is an indication of low income and was extracted as yes/no.
We report the proportion of women prescribed oral antibiotics on the same date as a mastitis encounter. Prescribed antibiotics were identified from the corresponding ‘Prescriptions’ dataset. Secondary outcomes included the proportion of women ordered clinical investigations for mastitis including breast ultrasound, breast milk culture, nipple swab culture, blood test (i.e., C-reactive protein [CRP], Erythrocyte Sedimentation Rate [ESR], Full Blood Examination [FBE]), and breast aspirate. These were identified by searching the ‘Requested tests’ free text field for the previously listed terms. Additional secondary outcomes included the proportion of women prescribed other medications, including topical or intravenous antibiotics, antifungals, lactation suppressants (i.e., cabergoline, bromocriptine), or lactation stimulants (i.e., domperidone). We are assuming “ultrasound” applies to a diagnostic ultrasound, but may also refer to therapeutic ultrasound, therefore we recognise the estimate for ultrasound is a likely to be an overestimate of number of actual diagnostic ultrasounds ordered. The dataset only includes biochemistry pathology results, so we were unable to analyse bacteriology or radiology data. Stata MP 17 (Stata, College Station, Texas) was used for analysis of the MedicineInsight dataset.
Qualitative component
The qualitative component used semi-structured interviews to explore GPs’ perspectives of the issues they faced when managing mastitis, making decisions about prescribing medications, and how they used guidelines, such as the Therapeutic Guidelines. The qualitative component received approval from La Trobe University Human Research Ethics Committee (HREC Ethics Application Number: HEC21054). The study followed all relevant guidelines and regulations for conducting ethical research.
Recruitment and procedure
An invitation to participate in the study was posted on the Facebook group GPDU (GPs Down Under) with approval from the group administrator. The group has over 9,000 GP members from around Australia. The invitation briefly explained the purpose of the research and what participation involved, with a stock image (female doctor with a female patient) and a link to a short survey in REDCap [
23,
24]. Interested GPs provided basic information to assess their eligibility (i.e. had seen breastfeeding woman in previous year; location; gender; age) and their contact details. The invitation was posted on 17 May 2021 and 27 October 2021. We received between 5 and 10 expressions of interest after each post. Several GPs were recruited using snowballing from initial participants. Eligible participants were contacted via email and sent the Participant Information and Informed Consent Form and an interview was arranged at a convenient time. They were asked to return the signed Informed Consent Form (via post or electronically) prior to the interview.
Interviews were conducted by MC and SBC between June 2021 and March 2022. The interviews were conducted online, using the Zoom platform, or via telephone. Each interview lasted between 30 to 45 min. The interviews were audio recorded, with permission, and the audio-recording was transcribed verbatim by a professional transcribing service and anonymised before analysis. Transcripts of interviews were emailed to participants to allow for member checking and verification prior to analysis.
Directly after each interview, the researcher made field notes of general impressions and reflections from the interview. After each interview, participants were sent an AUD$100 gift voucher to acknowledge their time commitment.
Interview schedule
A semi-structured interview schedule was used to guide the interviews. We based the Interview schedule on the Capability, Opportunity, Motivation-Behaviour (COM-B) framework which is structured to understand clinicians’ behaviour and likely barriers to perform according to best practice [
15]. The schedule covered the following topics:
Capability includes knowledge about prescribing during lactation,
Opportunity includes social norms about management of breastfeeding women, and
Motivation includes reflective aspects (beliefs around use of guidelines) and automatic aspects (established habits in prescribing for women with mastitis). The COM-B theory/framework has been useful in exploring barriers and enablers of Australian GPs’ management of children’s check-ups [
25]. Basic participant demographic data were also included in the interview schedule, to help describe the sample. For example, participant gender, location, years’ experience as a GP, where they conducted their GP training and the number of children they had, were collected at the beginning of the interview. The schedule was adapted in an iterative manner and the final version is provided as a supplementary document (Additional file 1. Interview guide).
Research team and reflexivity
LHA is an expert in mastitis and breastfeeding medicine research and led the study. LEG is an expert in pharmacoepidemiology in pregnancy and lactation and serves as an expert adviser to the Therapeutic Guidelines. He led the analysis of the MedicineInsight dataset for the quantitative study component. SBC has a background in health promotion and is an experienced mixed-methods public health researcher. She conducted interviews and led the analysis of the qualitative component of the study. MC has a background in microbiology and over ten years’ experience in breastfeeding/early parenting research. She arranged and conducted interviews. Prior to the interviews, SBC and MC conducted practice interviews with LHA who role-played different GPs to familiarise the interviewers with the topic, and allowed for minor changes to the interview schedule. Regular meetings helped to provide different perspectives on the findings during data collection and preliminary analysis phases, with LHA providing an insider view as a medical practitioner and the other team members reflecting on their experience as parents of young children, and experience from other research projects.