This study will be conducted in Western Australia’s capitol city, Perth, which covers 5,384 km ². In 2018, 2.1% of the Perth population (40,951 people) identified as Aboriginal and Torres Strait Islander people. During 2018, 2,122 babies of Aboriginal and Torres Strait Islander descent were born in Western Australia and of these, approximately 848 (39.5%) babies reside in the Perth area.

We will use an interrupted time series (ITS) design to determine the effectiveness of the CQI program delivered to selected community health staff at two community health regions on predominant breastfeeding rates at two and four months. Child health care is predominately provided by Child and Adolescent Health Service through 19 community health regions across the Perth Metropolitan area. The intervention will be delivered at 2 community health regions and the remaining 17 community health regions will be the control group.

While a randomised trial (when properly implemented) provides the most valid method for measuring the effectiveness of an intervention, we determined that this design was not feasible in this context. ²⁷ The type of intervention (i.e. targeted at community health staff) would necessitate evaluation using a cluster-randomised trial, with community health care services representing clusters. Inclusion of all Perth metropolitan community health care services would be extremely costly and require a large number of research staff. In settings where a randomised trial is impracticable, interrupted time series experiments have been advocated. ²⁸ While these designs can be at a higher risk of bias compared with randomised trials, ITS designs do allow statistical investigation of potential biases such secular trends, seasonal trends, random fluctuations, and autocorrelation compared to other quasi-experimental designs such as controlled before and after studies. ²⁹ Furthermore, the ITS design can be strengthened with the inclusion of control regions, which allow for an assessment as to whether any observed impact of the CQI intervention is likely to be causal, or due to extraneous events (e.g. other intervention(s) occurring around the time the CQI intervention is rolled out).

The study will be conducted over 54 months and include three distinct periods: 24 month pre-intervention period (including ethical and governance approvals), 6 month intervention period and 24 month post-intervention period. There will be an additional period of 12 months for data analysis and report writing period ( Figure 1).

All aspects of the study will be governed by a Study Guidance Group established at the beginning of the project. This group will be majority Aboriginal and Aboriginal chaired. They will provide their expertise on survey and interview questions, interpretation of the data and guide translation.

Pre-intervention period (12 months)

During the pre-intervention and intervention period we will be collecting survey and interview data from mothers of Aboriginal and Torres Strait Islander infants about their perceptions, barriers and facilitators of breastfeeding. The Breastfeeding Data Collection Form used in the 2006-2007 north metropolitan Perth breastfeeding cohort will be revised and updated to collect data on breastfeeding patterns. ³⁰ Mothers of Aboriginal infants up to 18 months old living in the Perth metropolitan area, regardless of whether they breastfed their infant will be eligible to complete the survey anonymously. Mothers who complete the survey will have the opportunity to participate in interviews. We will aim to survey 200 mothers and interview 20 mothers. Mothers will receive a $25 voucher for completing the survey and another $25 voucher if they complete the interview.

We will collect survey and interview data from health service providers including Aboriginal Health Workers, midwives, community health nurses and any staff member who may have contact with mothers and discuss breastfeeding. The survey will seek participants’ views on breastfeeding, including barriers and enablers to breastfeeding, and current practices to develop an understanding of gaps in existing policies and guidelines, and the educational support needs of both hospital and community-based primary health care professionals to ensure they are well prepared to provide culturally competent information and support to Aboriginal mothers and families. Health service providers will also have the opportunity to participate in interviews to expand on the topics outlined above. We aim to survey 80 health service providers and interview 15 health service providers from all health service regions, regardless of whether they are in the control or intervention regions.

Data collected from mothers and health service providers will be provided to health service providers participating in the intervention arm of the study to provide them with background. In addition, we will develop the CQI training for health staff that will be provided during the intervention period.

Intervention period (6 months)

Tailored CQI training to staff (2 months)

Training will be conducted and delivered by two highly experienced external consultants who have extensive experience working In clinical practice, management, evaluation, quality improvement, and policy, community and workforce development. Training will be delivered to staff via a series of webinars and face to face sessions. These webinars will cover a number of core topics to familiarise participants with the basics of CQI. Topics included five main areas: 1.

introduction to CQI,

This method of delivery has previously been trialled with the STORK project. ¹⁹ Ongoing mentoring will be provided to staff who participate in the program. The goal of the CQI program will be to work with the data from the health service to develop an identified program of evidence-based CQI activity. This will be achieved through supporting the participation and empowerment of employees to make changes in the workplace through a culture of ownership, continuous learning and support.

Implement breastfeeding CQI cycles (4 months)

Teams and individuals will conduct two CQI cycles over two months. CQI cycles will be developed on topics chosen from the data from interviews or what health service providers deem important within their workplace. A staff member will be available to support teams to complete their CQI cycles and generate data to inform the team of their goals. This staff member will also be mentored by the CQI trainers.

Post-intervention period (24 months)

Interviews will be undertaken with health professionals to gather their perceptions of the intervention and the impact this has had in relation to their attitudes to, and experiences with, breastfeeding. We will also collect the data and information on individual PDSA cycles that teams and individual undertook. All people who participated in the training will be invited to participate in interviews to determine whether the CQI training and implementation was an acceptable and feasible method to improve breastfeeding for Aboriginal babies. These data will highlight successful strategies and identify areas for improvement in using CQI initiatives to drive positive change within healthcare teams.

To determine whether the CQI intervention had an effect on breastfeeding rates of Aboriginal babies we will collect administrative data on breastfeeding rates in addition to infant hospitalisations, emergency department presentations, and vaccinations.

Study population

Administrative data will be collected for all Aboriginal and Torres Strait Islander babies who have attended Community health facilities for the 24 months prior to the intervention study period and 24 months after the intervention.

Intervention and control regions

The CQI intervention will be delivered at two community health regions where there are a high proportion of Aboriginal infants attending the service. Location-based control regions will be represented by 17 community health centres. These centres will not receive the CQI training or complete PDSA cycles. These 17 regions will provide an appropriate control group for excluding extraneous events as the reason for any observed impact of the CQI intervention. This is due to all 19 regions located within the same healthcare service. Therefore, any co-interventions which may be rolled out are likely to equally affect the control and intervention regions. We believe that contamination between control and intervention regions is unlikely due to the nature of the intervention and the required effort needed from participating groups to complete the intervention. We also do not anticipate that any characteristics that might be related to the outcomes will change between the control and intervention regions. To ensure this is the case we will discuss with control and intervention regions whether there were any changes in practice that may affect the outcome. We will also collect information from each of the control regions as to whether there were any events or interventions that occurred around the time the intervention was rolled out in the intervention regions.

Data collection

Table 1 provides an overview of the data collected to determine the effectiveness of the intervention, the method of collection, assessment period and source of data. Non-identifiable breastfeeding data, emergency department presentations and immunisation data will be extracted for intervention and control regions from routine data capture systems managed by PeopleWA (health service utilisation data: https://www.wa.gov.au/organisation/department-of-the-premier-and-cabinet/office-of-digital-government/peoplewa) and the National Health Data Hub (immunisation data: https://www.aihw.gov.au/reports-data/nhdh). Ideally our primary outcome would be exclusive breastfeeding at 2 and 4 months; however, only whether the infant is predominantly breastfed or received any breastmilk is captured in the Child Development Information System Database. Predominantly breastfed is defined as an infant breastmilk and only water with nothing else. ³¹ ‘Any’ breastmilk is defined as the sum of predominant and complementary (introduced non-human milk or solids) breastfeeding.

Data will be collected over 54 months, with 24 months pre-intervention, 6 months during intervention delivery, and 24 months post intervention delivery. For the primary outcomes ‘predominant breastfeeding at 2 months’ and ‘predominant breastfeeding a 4 months’, across the two intervention regions, we anticipate there will be approximately 10 Aboriginal babies per month (unpublished data). Two-monthly proportions of babies predominantly breastfed will be the unit of analysis in the regression models (see ‘ITS Data Analysis’ section for details). Two-monthly units will therefore yield 27 data points for analysis (12 pre-intervention, 3 during intervention delivery, 12 post intervention delivery).

Primary outcomes

Individual-level data will be separately combined across intervention regions and the control regions. Combining data across the sites will create more stability in the time series, due to the relatively small number of Aboriginal and Torres Strait Islander babies born per region per year. We will aggregate the data within two-monthly intervals and calculate the proportion of babies who are being predominantly breastfed (separately at 2 months and 4 months). We will analyse the two-monthly proportions using segmented linear regression. We will fit separate models for the intervention and the control series using the following terms: time (in two-month units, treated as continuous, ranging from 1 to 27), time during intervention delivery (i.e. the transition period, ranging from 0 to 2), and time after the intervention delivery (ranging from 0 to 11). Visual inspection of the data and residuals against time will be used to signal potential concerns regarding non-linear trends, serial autocorrelation and seasonal patterns. If the series are non-linear, we will attempt to linearize by applying, for example, a logit transformation to the proportions. We will estimate the model parameters using restricted maximum likelihood, allowing for first-order serial autocorrelation. ³²

We will calculate the following effects of the intervention: immediate level-change post intervention delivery and level-change at six months post intervention delivery. These will be calculated by comparing the predicted outcome response from the full model with the extrapolated outcome response based only on the pre-intervention trend (i.e. the counterfactual) at the time points of interest. We will plot interrupted time series graphs using the recommendations of Turner et al. ³³ Analyses will be undertaken in the statistical packages Stata (mixed package) and R (glarma package).

Secondary outcomes

For the secondary outcomes, we will fit segmented regression models using the same model structure and approach as per the primary outcomes. For dichotomous outcomes (6 month immunisation coverage, 6 month immunisation timeliness), we will calculate proportions and analyse using linear regression. For count outcomes (all cause emergency department presentations, all cause hospitalisations), we will use generalised linear regression with a negative binomial distribution and a log link. We will adjust for first-order serial autocorrelation, and for the count outcomes, additionally adjust for potential seasonal effects using Fourier terms. ³⁴

Sensitivity analyses

We plan to undertake a range of sensitivity analyses of the primary outcomes to examine whether the intervention effect estimates are sensitive to our analysis decisions, including: •

Assuming a different model structure. Given the uncertainty in the assumed model structure, we will fit the following alternative structure: time (in two-month units, treated as continuous, ranging from 1 to 27), intervention (1: post-intervention delivery, 0: pre-intervention delivery), and time after the intervention delivery (ranging from 0 to 11). This model does not have a transition period, but allows for a level-change post-intervention delivery, and a slope change between the ‘pre and during intervention delivery’ period and the ‘post intervention’ delivery period.

Descriptive analysis will be completed for survey data. All interview data will be synthesized descriptively, key themes will be identified will be performed using thematic analysis and Nvivo software. We will use applied thematic analytic techniques to analyse our data. Our thematic analysis will be performed through the process of coding in six phases to create established, meaningful patterns.

These phases are: familiarization with data, generating initial codes, searching for themes among codes, reviewing themes, defining and naming themes, and producing the final report. We will pinpoint, examine, and record patterns within our data. Our themes are considered to be patterns across data sets that are important to the description of a phenomenon and are associated with our research questions. Our themes will then become our categories for analysis. We will perform cross tabulations and scatter plots to assess fittingness, credibility, and consistency. We will then interpret the codes by comparing theme frequencies, identifying theme co-occurrence, and graphically displaying relationships between different themes.

Study Guidance Group members and project partners will be invited to participate in report writing with members of the project team, and will be provided with draft reports for comment prior to their finalisation and distribution to participating Community health regions. Presentations will be made to the relevant organisational governance structures and key executives. The purpose will be to discuss policy and practice implications from the study results.

We have provided the SPIRIT guidelines in Figshare (DOI: 10.6084/m9.figshare.26508511) and amended as needed for this study design. ^{35 , 36}

This research will be conducted in accordance with the NHMRC National Statement on Ethical Conduct in Human Research and Guidelines for Ethical Conduct in Aboriginal and Torres Strait Islander Health Research. The protocol for the study has ethics approval from the Western Australian Aboriginal Health and Ethics Committee (995; approval date 8/9/2020), King Edward Memorial Hospital HREC (RGS0000004409; approval 27/4/2021), Edith Cowan University HREC (2020-01064-MCAULLAY; approval 2/3/2020) and the Child and Adolescent Health Service HREC (RGS0000004409; 27/4/2021). This study adheres to the Declaration of Helsinki.

Consent was received for all interviews and surveys conducted all participants. Interviews with families required written consent. Interviews with health service providers required verbal or written consent. Verbal consent was approved as it was difficult for health staff to complete face to face interviews, therefore online interviews and verbal consent for these was ethically approved. Surveys were completed online and required participants were asked to consent through the survey. Ethics has been approved for all consent processes described.

For administrative data on babies (outlined in Table 1), it is not feasible to obtain consent from all families of Aboriginal babies born during the study period due to high costs, the high proportion of mothers who leave the hospital within 6 hours after birth and accessibility of some families. It is necessary for us to obtain outcome data for all babies born during this period to achieve the objectives of the project and ensure the results are generalisable and valid. We will apply for a waiver of consent to obtain administrative data through the organisation outlined in Table 1. Ethical approval for administrative data has yet to occur as of the date of this publication.

Extended data: We will provide PICFs and interview guides on request and if approved by our governance group.