INFANT will be implemented in collaboration with eight practice and policy partners including key agencies involved in MCH service delivery, population health and Aboriginal health. Partners will integrate the promotion and endorsement of INFANT to all Victorian LGAs (n=79), leveraging routine communication channels (e.g., statewide newsletters) and via presentations from the research implementation team at existing meetings of service providers. Health services seeking to adopt INFANT will express interest via the INFANT website (infantprogram.org) and be guided through the sign-up process by the research implementation team. Cultural and linguistic adaptations to meet the needs of multi-cultural and Aboriginal and Torres Strait Islander families within Victoria will be explored with relevant peak representative organizations as a sub-project, to be reported elsewhere.

To be eligible for adoption, health services will be required to nominate at least one staff member to complete the online INFANT implementation training. Organisational staff eligible to facilitate INFANT include MCH nurses, dietitians, community health workers or nominated personnel. INFANT implementation training comprises eight hours of free online training, providing guidance on all aspects of INFANT implementation and facilitation, with a focus on how to deliver group sessions and integration within existing delivery systems to maximise adoption and sustainability. The training includes case studies of successful implementation, an online discussion forum to promote peer networking (moderated by research staff), and interaction with the research implementation team to develop an implementation plan (a blueprint for local implementation). Following implementation training, health services will be eligible to have their trained staff deliver INFANT with yearly renewal contingent on facilitators completing a short online annual refresher course. Trained facilitators will be provided with online access to all resources including a facilitator and implementation manual as well as parent handouts (downloaded as pdfs) and the My Baby Now app.

Implementation strategies have been selected from the Expert Recommendations for Implementing Change (ERIC) compilation (32) to address barriers identified from the RCT and related small scale implementation studies (15, 26, 28, 33). Key strategies to support local implementation include assessment of organisational readiness; development of an implementation blueprint; implementation support through site visits, resource sharing, and collaborative learning; support with accessing new funding; revising professional roles; implementation audit and feedback; and accessible, incentivised training. Supplementary File 4 presents the 16 INFANT implementation strategies, categorised according to the six key implementation processes (34): planning, educate, finance, restructure, quality management, and policy context. Implementation strategies are described using Proctor et al.’s (35) recommended ‘specifiying and reporting’ guidelines.

Implementation outcomes, as detailed below, will be explored to understand (i) the role of, and barriers and enablers to, the use of implementation strategies and perceived impacts on implementation and/or effectiveness, and (ii) the rationale for uptake and modification, and impacts on fidelity and/or effectiveness (30).

Descriptive statistics will be reported for quantitative survey items related to organizational readiness, acceptability, appropriateness, feasibility, adaptation and penetration. Qualitative interview data will be transcribed verbatim and coded deductively [informed by CFIR (37), PSAT (38) and FRAME (39)] and inductively to capture issues within and outside of these frameworks. This has been successfully used by the team in previous implementation research (25, 42). Findings will be compared with the choice of INFANT implementation strategies to determine future refinements. Cost data analysis will use a healthcare funder perspective and detailed pathway analysis will specify scale-up and implementation activities in order to measure costs of associated resource use (2023 reference year, costs measured in Australian dollars (AUD)). Total and incremental cost, cost per local government area, cost per participant child and cost per eligible child will be reported.

An embedded non-randomised controlled trial will assess the effectiveness and cost-effectiveness of INFANT when delivered at scale. The trial will involve comparing primary and secondary effectiveness outcomes between a group of INFANT participants (intervention arm) and a non-randomised comparison group recruited within the same communities prior to the implementation of INFANT (usual care arm). This approach enables comparison of outcomes between parents and infants exposed and not exposed to the intervention living in the same geographical areas and therefore sharing socio-demographic characteristics and having access to the same health services. The risk of bias will be reduced by replicating strategies for recruitment, follow-up and assessment in both trial arms and adjusting for unbalanced baseline characteristics. This trial design was deemed appropriate for assessing the effect of INFANT when implemented at scale in real-world conditions, enabling natural dissemination and uptake to be captured. A cluster randomized design with communities allocated to intervention or usual care was deemed not feasible because the prospect of being randomised to the control arm is likely to reduce participation. Within communities, the prospect of parent groups being randomised to a control arm is unethical (given the known efficacy of the intervention) and contamination between groups would be unavoidable. A stepped-wedge cluster randomised study design (43) was also considered but ruled out as randomisation of LGAs to an implementation time period may interfere with natural uptake under real world conditions.

Parents will be eligible if they are a first-time parent/primary caregiver of an infant aged 0-3 months, owning an internet enabled personal mobile phone and able to communicate in English. Infants born before 37 weeks gestation will be excluded.

Families will be recruited over a 6-month period for the comparison arm and over a two-year period for the intervention group (Table 1). The comparison (usual care) group will be recruited prior to the implementation of INFANT by health services who have expressed interest prior to undergoing INFANT training. If sample size is not achieved from these LGAs, recruitment will be extended to other LGAs ensuring a mix of metro, regional, rural LGAs and various levels of socioeconomic disadvantage. There will be a three-month gap between recruitment of the last comparison participant and the commencement of recruitment for INFANT implementation participants. This gap ensures comparison participant infants will be too old to enroll in INFANT sessions and thus reduce potential for contamination.

Eligible parents will be invited to participate via SMS (with up to 2 reminders) sent by participating organisations. The SMS will contain a link to an online screening and consent process with consenting parents invited to complete online surveys when infants are aged ≤ 3 months (T1), 12 months (T2) and 18 months (T3) (Table 1). Up to three reminders will be generated for non-responders to follow up survey invitations. Strategies will be employed to promote retention (e.g. personalised newsletters, reimbursement for time spent completing surveys) which have been used successfully in the original INFANT trial, which achieved 85% retention over two years across both study arms (13).

The primary outcome for the effectiveness trial will be child lifestyle pattern score ‘Discretionary food and screen time’ at 18 months. This is a composite measure of child diet, physical activity and screen time derived from a previous analysis (44) of the INFANT RCT (13) and INFANT Extend RCT (33) using principal component analysis. This has been selected as the primary outcome based on previous research (44) confirming the early clustering of these behaviours into unhealthy versus healthy lifestyle patterns in children as young as 18 months of age. Further, these lifestyle patterns track from early childhood into later childhood and adulthood, and are associated with overweight and obesity (45).

Secondary outcomes of the effectiveness trial will include (i) proportion of mothers with any breastfeeding at 12 months; (ii) energy-balance behaviours including child diet (fruit, vegetables, non-core drinks and snacks), child time physically active and sedentary, scores on the ‘Fruit, vegetables and outdoor time’ lifestyle pattern (12 & 18 months), and scores on the ‘Discretionary food and screen time’ lifestyle pattern (12 months) (44); (iii) infant BMIz at 12 months and 18 months, and (iv) child classification as overweight or obese according to WHO definitions (46).

The trial aims to recruit 1,500 primary carer/infant dyads in each arm (total 3,000). Assuming a conservative attrition rate of 33%, it is expected that complete data on 2,000 primary carer/infant dyads (1,000/arm) will be collected. The statistical software PASS version 14.0.9 (NCSS, LLC) was used for all power calculations, assuming that 190 first-time parent groups will run in each arm with an average number of 5.4 children with complete data per group and a coefficient of variability in group size of 0.32, as observed in the INFANT RCT (13) (α=0.05; two-sided tests). A target sample size of 1,000 infants per arm achieves 80% power to detect a mean difference between arms of at least 0.17 points in the lifestyle score ‘Discretionary food and screen time’ at 18 months [primary outcome; standard deviation within clusters (SD) =1.21, intracluster correlation coefficient (ICC) = 0.048]. In our former INFANT trial (44) the difference in “Discretionary food and screen time” pattern score was 0.29 (95% CI: 0.08, 0.49).

The target sample size (1,000 infants per arm) also achieves 80% power to detect the following differences in secondary outcomes:

Infant diets will be measured by a validated Food Frequency questionnaire (FFQ) with average percentage categorical agreement against three days of 24 hour recall of 62-64% (49). Infant physical activity will be measured using a single parent-reported question with good test-retest reliability (ICC 0.62) (50). Infant screen time will be measured using a single parent-reported question with excellent reliability (ICC 0.84) (51). Change in BMIz measured at 12 and 18 months will be derived from child height and weight which is routinely collected by MCH nurses and will be made available in de-identified form to the research team. These data are typically available for 80% and 71% of infants at 12 and 18 months respectively (52). Lifestyle pattern scores will be calculated as linear combinations of the standardized dietary (fruit, vegetables, water, discretionary sweet foods, discretionary savory foods), outdoor and television viewing items, using the principal component analysis factor loadings previously published from the combined INFANT and INFANT Extend datasets (44).

Covariates will include self-reported parent socio-demographic characteristics (including marital status, education level, age, sex, Indigenous status, language spoken at home, country of birth, years lived in Australia, health care card holder) parent weight and height, pre-pregnancy weight, self-rated coping, sleep and health, infant sex, birth weight and temperament. Baseline characteristics will be compared between arms using linear mixed models (LMM) or generalized estimating equations (GEE) with link and distribution selected based on the type of variable. The effect of the intervention on the primary outcome (‘Discretionary food and screen time’ lifestyle pattern) will be estimated using a LMM including arm, time (12 and 18 months), interaction arm-time and any baseline variable showing clear imbalance between arms (covariates) as fixed effects. The same LMM model will be used for numerical secondary outcomes. All mixed models will include MCH group as a random effect to account for clustering. The effect of the intervention on proportion of mothers with any breastfeeding at 12 months will be estimated using a GEE model with logit link and binary distribution including arm, time (12 and 18 months), interaction arm-time and baseline confounders. If appropriate, multiple imputation will be used to handle missing data.

A cost-effectiveness analysis will determine if the implementation of INFANT represents ‘value for money’ when measured incrementally against usual practice. A ‘trial-based evaluation’ (i.e., costs/outcomes exactly as per the trial) will be conducted using primary and secondary outcomes. A ‘modelled economic evaluation’ will also be conducted, using the difference in BMIz between the intervention and control arms, and extending the target population, time horizon and decision context. A cohort multi-state lifetable model (53) will be used to translate changes in infants’ BMIz to long-term health outcomes (healthy adjusted life years). Simulation-modelling using the @RISK software package will calculate 95% uncertainty intervals around epidemiological inputs and cost estimates.