Process evaluation
A number of strategies are planned to monitor intervention uptake, assess adherence to protocols and describe experiences with trial implementation.
Measures of intervention exposure
SILC-MCHN home visits: number of women identified as eligible; number of women who decline a visit; number of visits; timing of visits; and time spent during visits.
Drop-in centres: number of women who visit drop-in centres, reason for the visit and total number of visits to each drop-in centre.
Adherence to protocols
SILC-MCHNs will keep a log of what is provided to women during visits. Breastfeeding concerns women raise or issues identified by the providers will be documented on pre-coded data sheets at home visits and drop-in centres.
SILC-MCHNs will attend a total of four workshops during the trial period: one before implementation of the interventions and three while the interventions are in place to discuss adherence to SILC protocols, review processes and discuss any concerns SILC-MCHNs may have. Research team members will visit LGAs at regular intervals (estimated to be 3–5 times) during the intervention period, to monitor protocol adherence and discuss concerns about intervention implementation with SILC-MCHNs and MCH co-ordinators.
Intervention implementation evaluation by SILC-MCHNs and MCH co-ordinators
An evaluation of how the interventions have been carried out will be undertaken using surveys and focus groups with SILC-MCHNs and MCH co-ordinators. SILC-MCHNs attending the final SILC workshop at the conclusion of the intervention will be invited to participate in focus groups and complete a structured questionnaire including open-ended questions. The questionnaire will explore their views and experiences of being a SILC-MCHN. They will be able to report on any highlights, any problems or barriers in implementation of the interventions, and to make recommendations on the improvement of intervention implementation. There will be two focus groups led by the research team – one group for SILC-MCHNs from Trial arm 1 and the other group for SILC-MCHNs from Trial arm 2. Focus groups will give SILC-MCHNs an opportunity to reflect on and discuss their experiences of participating in the SILC interventions.
Face-to-face interviews will be conducted with MCH co-ordinators and managers from intervention LGAs. These interviews will provide an opportunity for participants to share their experiences and to provide information regarding problems or barriers to the introduction and embedding of the interventions, which will be important in terms of sustainability if the interventions are shown to be effective.
Sample size
Infant feeding outcomes are routinely collected by the MCHNs at KAS visits, and include infant feeding at hospital discharge, two weeks, three months (asked at the four months KAS visit) and six months postpartum (asked at eight months KAS visit). Given the non-contemporaneous nature of ascertainment of the infant feeding outcomes and that breastfeeding at four months is a key national indicator of children’s health, the four month KAS visit was determined the most appropriate time to collect infant feeding outcome data. The standard MCH questions related to infant feeding have the potential for misclassification of infant feeding out-comes, therefore a new question on infant feeding will
be added to the data collected at the four month KAS visit that provides a cross-sectional snapshot of the percentage of infants receiving any breast milk (‘in the last 24 hours’). Our primary outcome measure is thus the percentage of infants receiving any breast milk at four months.
There are no local representative data on infant feeding outcomes at four months, so we estimated the four month figures from three and six-month state-wide infant feeding data. We plan to include only LGAs with breastfeeding rates below the state average at hospital discharge and with more than 450 births per annum, so we used the most recent data available (2008/2009) to identify potentially eligible LGAs and calculate their breastfeeding rates [34]. The state average for any breastfeeding at hospital discharge was 87%, and for the LGAs whose rate was less than the state average, the average rate of any breastfeeding at hospital discharge was 81%. We then used the rate of any breastfeeding at three months in this group of LGAs (53%) and any breastfeeding at six months (39%) to estimate a figure for four months. We consider the drop is likely to be continuous, so estimate that 48% are giving any breast milk at four months.
While the unit of randomisation is the LGA, the MCH centres form the clusters. We aim to have the power to detect an increase in the overall breastfeeding rates at four months from 48% to 58%. With this change and assuming a standard error for the average breastfeeding rate in each LGA of 5%, sample size estimates show that we need approximately 4 LGAs in each trial arm. To achieve this standard error we need to have 374 women in each arm (without allowing for clustering). Given that we will be obtaining breastfeeding data on all women in each LGA, with an estimated average of about 1000 births per LGA per annum, we can obtain the required sample size allowing for an inflation factor of approximately 2.7 (1000/374), which if the geometric mean number of MCH centres is 86 (calculated on data from MCH centres in the potentially eligible LGAs), allows for an intra-cluster correlation (ICC, ρ) of 0.02 (2.70= 1 + 85*ρ). We could not identify relevant literature on which to base our estimate for the ICC in breastfeeding trials, therefore we used the conservative estimate of 0.02 which is larger than ICCs found in cluster trials for smoking cessation in pregnancy [42]. This was also consistent with the variation we found in statewide MCH data.
Calculating the number of women who will be eligible to potentially receive the intervention is based on the fact that the proposed intervention will only have the ability to affect women who are breastfeeding at discharge, so to gain this 10% increase overall (i.e. 48% to58%) we will need to increase the rate of breastfeeding from 59% (or 48%/81%*100) to 72% (58%/81%*100) among those who are breastfeeding at hospital discharge. For a simple random sample this would require 224 women per each intervention trial arm (with alpha 0.05 and 80% power), i.e. 448 in total. Taking into consideration the potential effect of clustering we have inflated the sample size assuming an intra-cluster correlation (rho) of 0.02 (as discussed above), so we require approximately 400–500 women in each intervention arm.
Data collection
The data items for each component of the study are summarised in Table 1. The interventions will run in
LGAs for a nine to twelve month period depending on birth numbers available in participating LGAs. The first two months of the interventions will be a pilot phase/run-in period. This will give SILC-MCHNs the opportunity to provide feedback on issues they may encounter and the research team the opportunity to refine protocols. Data collected during the run-in period will not be used in the final analysis.
Baseline data
Baseline breastfeeding rates will be collected to allow a before and after analysis of breastfeeding rates in each LGA, and to assess comparability of trial arms at trial commencement. Baseline breastfeeding outcomes will be collected for a period of three months before infants exposed to the interventions will have their routine MCH appointments (when breastfeeding outcomes are collected). These data will be obtained from routinely collected MCH centre data (three, four and six month infant feeding data) and from a new data item to collect infant feeding ‘in the last 24 hours’ (in all participating LGAs). Inclusion of this new data item will give higher quality cross-sectional data, enhancing the capacity to discern changes that may arise from the intervention. It will be completed by MCHNs with mothers at their routine four month KAS visit.
Outcome data
There will be two components to outcome data collection. Breastfeeding outcomes will be obtained from routinely collected MCH centre data and from the new data item collecting infant feeding ‘in the last 24 hours’ at four months.
In addition, information will also be obtained directly from women via a postal survey. All women who give birth during the intervention time-frame in the participating LGAs will be sent a questionnaire when their infants are six months of age, with the exception of those who meet the exclusion criteria described previously. Data will include socio-demographic details and infant feeding outcomes (any breastfeeding at survey completion, and duration of any and only breastfeeding). Women will be asked if they have had difficulties with breastfeeding, and if and where they obtained support or help with breastfeeding. Information will also be collected about home visiting (as part of the intervention or
other home visiting services) and breastfeeding services (including the intervention drop-in centres). Women will be given the opportunity to report on their views and experiences of the available services.
Outcome variables
The primary outcome of the study is any breastfeeding at four months. A single data item asking about infant feeding ‘in the last 24 hours’ will be used to ascertain this outcome at the routine four month KAS visit. The four month KAS participation rate from the last state-wide report was 93.6% [43]. Three and six month infant feeding outcomes will be obtained from routinely collected MCH centre infant feeding data collected at the four month and eight month KAS visits respectively. The eight month KAS participation rate from the last state-wide report was 85.6% [43].
Other variables of interest (or potential confounders for the primary outcome) which will be abstracted from routinely collected MCH data are: