A cluster randomised controlled trial of the community effectiveness of two interventions in rural Malawi to improve health care and to reduce maternal, newborn and infant mortality

Trials, Volume 11, Year 2010

Ukuhunyushwa

Background: The UN Millennium Development Goals call for substantial reductions in maternal and child mortality, to be achieved through reductions in morbidity and mortality during pregnancy, delivery, postpartum and early childhood. The MaiMwana Project aims to test community-based interventions that tackle maternal and child health problems through increasing awareness and local action.Methods/Design: This study uses a two-by-two factorial cluster-randomised controlled trial design to test the impact of two interventions. The impact of a community mobilisation intervention run through women’s groups, on home care, health care-seeking behaviours and maternal and infant mortality, will be tested. The impact of a volunteer-led infant feeding and care support intervention, on rates of exclusive breastfeeding, uptake of HIV-prevention services and infant mortality, will also be tested. The women’s group intervention will employ local female facilitators to guide women’s groups through a four-phase cycle of problem identification and prioritisation, strategy identification, implementation and evaluation. Meetings will be held monthly at village level. The infant feeding intervention will select local volunteers to provide advice and support for breastfeeding, birth preparedness, newborn care and immunisation. They will visit pregnant and new mothers in their homes five times during and after pregnancy.The unit of intervention allocation will be clusters of rural villages of 2500-4000 population. 48 clusters have been defined and randomly allocated to either women’s groups only, infant feeding support only, both interventions, or no intervention. Study villages are surrounded by ‘buffer areas’ of non-study villages to reduce contamination between intervention and control areas. Outcome indicators will be measured through a demographic surveillance system. Primary outcomes will be maternal, infant, neonatal and perinatal mortality for the women’s group intervention, and exclusive breastfeeding rates and infant mortality for the infant feeding intervention.Structured interviews will be conducted with mothers one-month and six-months after birth to collect detailed quantitative data on care practices and health-care-seeking. Further qualitative, quantitative and economic data will be collected for process and economic evaluations.Trial registration: ISRCTN06477126. © 2010 Lewycka et al; licensee BioMed Central Ltd. The intervention will be evaluated in a cluster-randomised controlled trial. A cluster design has been chosen because the allocation and loci of delivery of the interventions (community clusters) are groups rather than individuals. 48 rural clusters have been identified. 24 have been allocated to receive the women’s group intervention and 24 will act as controls. Using a factorial design, each of these arms of the women’s group trial was randomized a second time, 12 clusters to receive the volunteer counseling intervention and 12 to act as controls. An independent surveillance system for outcomes – births, morbidity, mortality, and care seeking – has been designed and implemented in all 48 clusters, covering a population of 144,000 (Figure (Figure11). Two-by-two factorial design. Community-based women’s group mobilisation intervention and home-based infant feeding and care counselling, with basic health service strengthening for both intervention and control areas Will community mobilisation through women’s groups reduce perinatal, neonatal, infant and maternal mortality rates through changes in care practices and health-seeking behaviour? Will volunteer infant feeding and care counselling for pregnant and breastfeeding mothers in their homes reduce infant mortality through changes in knowledge and practices regarding exclusive breastfeeding, family planning and other care practices and health-seeking behaviours? Women’s group activities will lead to: reductions in maternal, perinatal, neonatal and infant mortality, reductions in maternal and infant morbidity, and increases in recognition of high-risk symptoms, increased health-care seeking behaviour and changes in care-taker practices. Infant feeding and care counselling sessions will lead to: reductions in infant mortality, reductions in maternal and infant morbidity, increases in exclusive breastfeeding rates in the first six months, increases in health-care seeking behaviour, and changes in care-taker practices. The study endpoints are shown in Table Table1.1. The interventions will run for two years from the date when they are hypothesised to start having an effect, nine months after the start of the interventions, allowing for new pregnancies to have the maximum benefit of exposure to interventions. If a valuable effect is shown at the end of this period, the intervention models – refined on the basis of experience – will be implemented in the control areas. Study outcomes A cluster-randomised controlled trial, with a two-by-two factorial design, will be used to test the impact of two interventions on specific outcome indicators. 48 clusters or ‘zones’ were defined from the Enumeration Areas generated during the 1998 census (Figure (Figure2a).2a). Each zone contained a population of approximately 8000 in total. A population of around 3000 was selected from the centre of each zone, leaving a ‘buffer area’ around it to reduce contamination between neighbouring intervention and control areas (Figure (Figure33). a – Aggregation of census enumeration areas into 48 study ‘zones’, b – Random allocation of zones to four different combinations of intervention. Study villages and buffer areas in three clusters near Mchinji Boma. Mchinji district lies to the west of Lilongwe, in the Central Region of Malawi. It has international borders with Zambia and Mozambique, and a population of about 375,000 in 2004 [33], of whom about 80% live in rural areas and make a living through subsistence farming. The main ethnic group are the Chewa (90%), and the predominant religion is Christianity (97%). Comparison of some key socioeconomic and health indicators for Mchinji and Malawi are shown in Table Table22. Socioeconomic and health indicators for Mchinji and Malawi Sources: Malawi Population and Housing Census, 1998, Malawi DHS, 2004, World Bank 2006, State of the World’s Newborns report 2001, UNDP 2009 In 2006 it was reported that 99% of women in Mchinji attended antenatal care at least once during their pregnancy, but only 58% delivered at a health facility [34]. Maternal and perinatal health care is provided by personnel from one district hospital (a first referral and secondary health facility), four rural community hospitals (one government and three mission hospitals), one maternity unit, six health centres that provide maternity care, two dispensaries and two private clinics that offer antenatal care. Quality is compromised by the severe shortage of personnel, low morale of the health providers, and irregular drug supplies. Traditional Birth Attendants are available in all localities. The trial is being implemented by MaiMwana project, a Malawian trust established in 2003, as a collaboration between the Department of Paediatrics, Kamuzu Central Hospital, the Mchinji District Hospital and the UCL Centre for International Health and Development. The target population will be rural communities with the least access to health services, who might benefit most from community-based interventions to improve maternal and child health. The district administrative centre was excluded because it is more urbanised than the rest of the district and therefore not comparable to other clusters. The target group for participation in both of the interventions will be women of childbearing age (WCBA), between the ages of 15 and 49 years, and particularly pregnant mothers. Girls aged between 10 and 15 years will also be monitored and encouraged to participate in order to identify and support early teenage mothers. Older women will also be encouraged to attend as they influence decision-making around pregnancy, childbirth and child care and have valuable experiences to share. All women aged 10 to 49 years (inclusive) who agree to take part, will be enrolled into the study, regardless of whether they are married or not. Women who have no possibility of conceiving during the study period will be enumerated but excluded from the final sample (for example women who have had hysterectomies or terminal family planning procedures). Participation in intervention activities will be voluntary, and women’s groups will be free to establish their own membership criteria. A cohort of 44,000 women aged between 10 and 49 years was defined during the baseline phase of the study, and each is visited monthly by study personnel. From the beginning of the study period we identified all pregnancies, births and deaths occurring within the cohort of WCBA, with follow-up until at least one year after delivery. The cohort members are listed in a master document to which new names can be added: the cohort is open and new participants may be enrolled during the study period if they move into the study area, or if they reach 15 years of age. These are summarised in Table Table33. Elements of the women’s group and infant feeding and care counselling interventions The women’s group intervention seeks to build the capacities of communities to take control of the mother and child health issues that affect them. The intervention is community-based in that it defines the community as the agent of change [35]. To achieve this, 24 local female facilitators, one per cluster, were recruited and trained. The facilitators formed between six and 12 groups in their clusters and invited all women of childbearing age to attend. They guide the groups of women through a four-phase community mobilisation action cycle developed to be appropriate, accessible and feasible for the Malawian context from similar models in Bolivia and Nepal (Figure (Figure4)4) [36,37]. In the first phase, consisting of eight meetings, the groups identify and prioritise the mother and child health problems they feel are most important and discuss what may contribute to these problems and how they might be prevented and managed. In the last meeting of this phase they share their discussions with men in the community. In the second phase, consisting of four meetings, the groups plan locally feasible strategies to address the problems they have prioritised. In the last meeting of this phase they share their discussions with the whole community. In the third phase, consisting of four meetings, the membership of the groups expands to include all community members, including men, working together to implement the strategies that have been identified. In the fourth phase, four meetings, the groups evaluate what they have done, and plan for the future. The facilitators received minimal health training but use participatory rural appraisal tools and picture cards to facilitate discussions and enable groups to access their collective knowledge and capacities. With these capacities the groups take increasing control of the intervention over the course of the cycle. They do not receive any resources from MaiMwana Project, except the guidance of the facilitator, employed by MaiMwana project, and supported by four trained supervisors and a senior supervisor. Women’s group cycle. The volunteer counselling intervention seeks to change the behaviour of individuals in relation to care and care-seeking for mothers and children. The intervention is community-based in that it defines the community as the target of change [35]. In particular, the intervention seeks to provide health education to raise the awareness, change the attitudes and build the self-efficacy of mothers in relation to exclusive breastfeeding. To achieve this, 72 female volunteers, between two and four per cluster, were recruited and trained to identify pregnant women in their area and visit them at home to provide counselling and support to exclusively breastfeed their infants and perform other key infant care practices. The volunteer counsellors visit these women at five key times in pregnancy and after birth (Table (Table4).4). The first visit is conducted in the third trimester, and the mother receives information regarding the importance of early and exclusive breastfeeding, PMTCT, birth preparedness and family planning. The second visit is conducted in the first week after birth (and where possible within the first three days), when the mother receives information on the importance of exclusive breastfeeding, vaccinations, warmth and hygiene and family planning. The mother also receives information about danger signs for herself and her child, advice about breast problems and support for good attachment and positioning. The third visit is conducted at one month after birth and covers the same issues as the previous visit, with the addition of information about weaning. The fourth visit occurs at three months after birth and covers the same issues as the third visit. The fifth visit is conducted at five months after birth and covers the same issues as the third and fourth visits with the addition of information about weaning foods. The volunteer counsellors received minimal health training but use a picture book to facilitate learning. The volunteer counsellors are supervised by 24 government Health Surveillance Assistants, and the intervention is coordinated by one supervisor employed by MaiMwana project. Volunteer infant feeding and care counselling intervention meeting guide The 48 study zones were randomly allocated to one of four groups; 12 zones receiving infant feeding counselling and facilitation, 12 zones receiving facilitation only, 12 zones receiving infant feeding counselling and no facilitation, and 12 zones receiving neither (Figure (Figure11 and Figure Figure2b).2b). All zones will benefit from the district-wide training and support for health services from the project’s initiation. Random number generation was done in STATA 7.0, and each of the 48 clusters was allocated to one of four possible combinations of interventions. In this way, each intervention was stratified according to the presence or absence of the other one, in order to balance any effects of one intervention on outcomes of interest in the other. SL and DO generated the random number sequence and allocated the clusters to intervention groups. All women aged between 10 and 49 years residing within the clusters were eligible for inclusion in the interventions and follow-up of maternal and child health outcomes. Records of a woman’s migrations within and outside of the study area are recorded in order to allow for both ‘intention to treat’ and ‘per protocol’ analysis. Due to the nature of the interventions, blinding of study participants to their allocation was not possible, though analysts and trial monitors will be blinded to the study allocation until the definitive analysis is performed. Data will be collected independently from intervention implementation, and no results will be fed back to inform the interventions. The sample size for the cluster randomised controlled trial was arrived at by comparing statistical power for different estimates of population parameters. Parameters estimated included baseline mortality rates, the projected size of the reduction in maternal, neonatal and infant mortality and increase in exclusive breastfeeding rates due to the interventions, the number of births in each cluster over the trial period, the number of clusters in the intervention and control groups, the statistical power of the study, and the inter-cluster coefficient of variation. Realistic values of some of these parameters were difficult to estimate, as few population-level mortality data were available, especially at district and sub-district level. The sources of data and values used for these estimates and sample size calculations are summarised in Table Table5.5. We aimed to achieve the smallest sample size that would allow adequate statistical power to detect an impact within a reasonable time-frame and would be logistically feasible to implement. Parameters used to estimate sample size, and the estimated size of reductions or increases that would be detectable * Malawi DHS data used is the national estimate, as data were not disaggregated for Mchinji District Initial sample size estimates were made using national estimates of crude birth rates and mortality rates from the Malawi DHS for 2000 [38], subsequently revised after the results for the 2004 DHS survey were released [2]. Estimates were made with 80% power, a two-sided 5% significance level and an inter-cluster coefficient of variation (k) of between 0.15 and 0.30, using the methodology laid out by Hayes et al [39]. An estimate of k = 0.25 came from work on interventions to reduce HIV incidence rates [40], and later from work in Makwanpur, Nepal [27]. We assumed that variability in neonatal, infant and maternal mortality between clusters would be less in rural Malawi, due to fairly uniform exposure to risk factors such as poor hygiene and poor quality or absent delivery care and malaria. In addition, we reduced potential heterogeneity by excluding the main district administrative centre, as it was felt to be socio-economically and demographically different from rural villages. In recognition of the lack of certainty for this estimate, sample sizes were calculated for a range of values of k. Through the two-by-two design, it will be possible to assess the interaction between the two interventions, though the study is not powered to evaluate the combined impact of the interventions on mortality compared to single interventions alone, as this would have required an unfeasibly large sample. We will explore qualitatively and quantitatively whether or not there is a synergistic relationship between the two interventions, resulting in an effect greater in magnitude than either intervention alone. We might expect that women in an enabling environment (provided by women’s groups) would find it easier to discuss health issues and make health-seeking decisions for themselves and their babies than women in control areas where social barriers to care-seeking have not been addressed. More specifically, women in areas receiving both interventions might be more likely to use the Volunteer MaiMwana Counsellor and recognise the importance of her advice. More of these women may have decided to use PMTCT services than those not receiving facilitation, and will therefore be in a better position to make informed choices about infant feeding and family planning based on knowledge of their HIV status. Conversely, individual visits made by volunteers to women in their homes may serve to reinforce messages and issues arising from women’s group discussions. Data will be collected in two phases; baseline data during intervention development, and prospective data during intervention implementation. The baseline phase was completed in 2004. Formative qualitative data were collected through semi-structured interviews and focus groups. These data were collected with the specific purpose of: a) exploring current care practices to help to develop the structured questionnaires and interviews; and b) exploring the aims, setting, target population, methods and resources of the interventions to assist in their development. During the baseline data collection phase, the 692 study villages were mapped, each household enumerated, and a census was done in order to collect basic socioeconomic and demographic data about household members. A list of all female residents aged between 10 and 49 years was produced. Prospective data relevant to the evaluation of the two interventions will be collected in three ways – impact, process and economic. To assess the impact of the interventions on mortality, all WCBA will be visited monthly to identify pregnancies and births, as well as neonatal, infant and maternal deaths. WCBA will be visited by trained enumerators once per month, and events recorded in a register holding an up-to-date list of all women in the cluster (generated from the baseline survey, plus new residents). One enumerator will visit all WCBA in one cluster. All deaths of infants and women will be followed up by a supervisor who will verify the death and conduct a verbal autopsy interview between two and six weeks after the death. This interview will help to elicit the causes and contributing factors of the deaths [41]. There are five supervisors, each based at a nodal office. To assess the impact of the interventions on morbidity, care practices and behaviours, women who are identified as pregnant will be followed up until 6 months after birth by trained interviewers. Interviewers will administer one-month and six-month post-partum interviews to collect detailed information about demographic characteristics, maternity history, health-seeking behaviours, care behaviours and maternal and infant morbidity. To assess the impact of the interventions on growth and nutrition, women and infants will be visited by Health Surveillance Assistants (HSAs) at one-month and seven-months to collect anthropometric and biodata. Infant weight and height and mother’s mid-upper arm circumference will be measured, and blood spots taken from infant heel pricks will be tested for haemoglobin levels using a Haemocue machine. Six HSAs have been recruited and have received extensive training according to Malawi government guidelines in theoretical and practical aspects of blood spot collection by the Community Health Sciences Unit (CHSU) technicians. The HSAs will also be supervised by CHSU technicians quarterly and attend full week-long six-monthly refresher trainings. An integrated process evaluation will collect data on the key factors that may mediate the impact of the interventions in order to understand why they work or fail to work, and how different contextual factors influence their success. This will include collecting information on the context in which the interventions are being implemented, the mechanism through which the interventions are working, and the proximal outcomes of the interventions. On the basis of this information, hypotheses will be developed and, where possible, tested to explain and better understand the impact of the interventions. The data will be collected, using predominantly qualitative methods such as key informant interviews, focus group discussions and observations. A range of different respondents will be consulted, including members of intervention and control communities, members of women’s groups and women who have been counselled, key community, district and national informants, MaiMwana staff, and health facility staff. All inputs will be audited and cost-effectiveness analysis will be carried out in order to assess the replicability and scalability of the interventions, and the potential for them to be adopted as larger scale public health interventions in Malawi. Full costs will include start-up and running costs for both interventions, and will be collected through the project accounting systems. In addition to financial costs, economic costs will be estimated, valuing resources at their “opportunity cost”, or value in their best alternative use (including time, resource use and donated items) [42]. We will estimate the cost per maternal, infant, neonatal and perinatal death averted and the cost per newborn infant exclusively breastfed. We will collect information on the costs associated with monitoring and evaluation, although research costs will not be included in the analysis. Appropriate sensitivity analyses will be used to explore the implication of uncertainty of assumptions. The cost of scaling up the interventions at national level will be estimated in order to explore cost-saving opportunities and to investigate issues of generalisability beyond the trial context and beyond domestic boundaries. All quantitative data collected will be delivered to the main office for data entry in a relational database management system in Microsoft Access run on a dedicated server and workstations. Each WCBA will be given a unique ID number generated from the cluster, village and household she comes from. All quantitative data from the mortality surveillance, morbidity, care practice and behaviour questionnaires and process evaluation will be linked to the WCBAs through this unique ID. After checking and entry, all questionnaires will be archived in a locked room for future reference. Qualitative information will be audio recorded after receiving verbal consent from respondents. Data in Chichewa will be translated into English by a bilingual speaker. To ensure conceptual, grammatical, and syntactical equivalence, translations will be reviewed collaboratively by the researcher who obtained the data and the translator (both bilingual Chichewa-English speakers), and the lead researcher (English speaker). This team will make decisions about the best terms to use. All data will subsequently be transcribed and imported into a database in MAXqda 2 (VERBI Software version 2). One enumerator per cluster will identify births and deaths, and each event will be cross-checked by one interviewer. Supervisors will make regular field visits to check the quality of work done by enumerators and interviewers and observe some interviews. Each supervisor is responsible for between six and ten clusters. Interviewers will meet with enumerators weekly in order to check on their work and receive updates on births and deaths in their area. Supervisors will meet with interviewers and enumerators fortnightly to check on their work, discuss problems and provide quality control feedback. A sample of 200 one-month and six-month interviews has been selected to be independently re-done by the supervisor, in order to be able to estimate recall and interviewer error rates. Quantitative data will be checked in three stages. The first check will be performed after completion of the questionnaire, by the MEO in full and a nodal data checker (NDC) based at one of the five nodal offices. The second check will be done by a team of two data checkers (DC) based at the main office. The last stage of data checking will be done at the point of data entry by the four data entry clerks (DEC). Further checks will be carried out internally within the electronic data-handling environment. Data entered into the study databases will be regularly reviewed for inconsistencies and missing information. Lists of women interviewed and key fields to be verified will be produced, such as ID numbers, dates of birth, and reported pregnancies, births or deaths that have received no further follow-up. Minimising loss to follow-up is an important aspect of trial conduct. Certain features of this location and population dynamic need special attention for outcome tracking: a) Residents of Mchinji move seasonally to maximise their access to fertile land during the farming season, both within the district and across international borders into Zambia and Mozambique; b) Residents of Mchinji may go home for delivery and some time after birth (to other villages within Mchinji or to other districts); c) Non-residents may come into Mchinji from other districts for delivery and some time after birth; d) Families may move after a woman’s death, making it difficult to find respondents who know the details of what happened; e) High population turnover in trading centres and commercial farm estates; f) Women being busy working in their gardens, or at community gatherings such as funerals and chieftainship ceremonies, making them unavailable for interview; g) Weather conditions making roads impassable and conducting interviews difficult. For residents or respondents who are temporarily unavailable, the main strategy is to keep following up until an outcome is ascertained. All women who have ever lived in the study areas are maintained in the database, and appear every month in the register. Any events (such as pregnancy or birth) that are reported but no further details are known, are selected and lists produced to remind field-workers of the need for follow-up. In most cases this causes delays in getting complete data, though basic data for estimating mortality rates are still available. In recognition of the fact that certain data collected after a long delay will no longer be valid, sections in the questionnaire (such as infant feeding recall) are skipped. For residents or respondents who are permanently unavailable, basic information about dates and timings of events is sought from other community members such as friends or neighbours. Contamination may occur when people from one cluster have contact with people from another. In the rural villages of Mchinji, there are many opportunities for social mixing. Friends, relatives or neighbours may mix socially, or contact may be made through travel or migration between intervention and control clusters. There might be direct participation of residents from control areas in WG or VMC activities, or more likely, informal discussion of ideas from WG or VMC activities – control area residents may gain some benefit from hearing health messages received by intervention participants. The usual effect of this kind of contamination would be ‘dilution’ of the differences between treatment arms [43]. In order to reduce the possibility of contamination, we opted to use clusters of villages rather than individual villages as the unit of randomisation, thus reducing rates of travel across cluster boundaries [43]. Furthermore, each zone had a defined ‘buffer area’ around the perimeter (figure (figure2).2). A population of 3000 in each zone was required to achieve the desired sample size, but rather than selecting villages at random from each zone, only villages at the centre of the zonal area were eligible for inclusion in surveillance and intervention activities. This reduced the possibility of communication between neighbouring study villages in intervention and control areas. Women’s group facilitators and volunteer counsellors are residents of the zone in which they work. This reduces the possibility that they might transfer intervention benefits to neighbouring communities. For the WG trial, whilst health messages are discussed in group meetings, it is unlikely that neighbouring control communities would spontaneously mobilise themselves without the presence of a facilitator. So the purported benefits of community empowerment and social capital are unlikely to spread beyond intervention areas. An independent Data Safety and Monitoring Board (DSMB) will meet several times according to the DAMOCLES statement, [44] to review progress and advise on the conduct of the trial. They will assess compliance with the protocol, data quality and completeness, recruitment figures, sample size assumptions and ethical considerations. The meetings will not include analysis of outcome data by intervention allocation until completion of the trial. Baseline data will be reviewed to evaluate how well balanced the clusters were after randomisation and suggest any adjustments that may need to be made. We do not foresee any adverse effects of community mobilisation or peer counselling, so we do not intend to apply stopping rules. Data from the trial will be analysed by intention to treat at both cluster and participant levels. Participants will be assigned to the cluster they were resident in at the time the trial began, even if they move to another cluster or out of the study area during the trial period. Primary and secondary outcomes will be compared between intervention and control groups with random effects logistic regression models, taking account of clustering. Potential confounding factors will be included in the model, and measures of baseline mortality will be included in order to adjust for any imbalances in randomisation at the outset. Cluster-level data will be compared using modified t-tests. All estimates will be presented with 95% confidence intervals. Data collected from the project both during the baseline phase and the intervention phase mean that we will be able to look at overall changes in outcome indicators over time. Factors other than the study interventions may influence health indicators seen, such as other mother and child health programs in the area, or positive reporting and the Hawthorne effect [45]. If mother and child health programs are only active within some of the cluster areas then these can be examined separately if necessary. Analysis of qualitative data will involve iteratively developing a system of codes and memos jointly by the lead researcher and the researcher who collected the data. This method will seek to develop an analytical framework based on the data by coding and memoing pertinent excerpts that illustrate emerging themes. Subsequently, in an iterative process, the researchers will refine their analysis ensuring that the themes that are built up are cross-checked with other data firstly within and then between transcripts so that the validity of emerging explanations is tested and improved. The qualitative data will also be used to interpret and contextualise the quantitative data collected. MAXqda 2 software will be used to facilitate this analysis. Ethical permission for this study was granted by the Malawi National Health Sciences Research Committee in January 2003 (Ref: MED/4/36/I/167) and the ethics committee of the UCL Institute of Child Health and Great Ormond Street Hospital. It is registered with ISRCTN06477126. Verbal and written consent was received from community leaders after full consultation and discussions. The regional, district and village leaders, and local health and development professionals will have ongoing access to the research programme and will be the first to be briefed on study findings and outcomes through written and verbal reports. Before each instance of data collection, the process and advantages and disadvantages of taking part will be explained to all participants. Verbal consent will be obtained, and participants informed that they can stop taking part at any time. Participation in intervention activities is voluntary, and women may choose to start or stop as they wish. The study is designed to test the community effectiveness of two community-level initiatives to reduce maternal, infant, neonatal and perinatal mortality in rural Africa. Encouraging community action for maternal and newborn care alone will not grant success. For health promotion interventions to work, the supply side of health care services must reach a minimum standard. The study team considers it unethical to strengthen services only in intervention and not control areas. Control communities will benefit from low-cost improvements in equipment, supplies and training at all primary level facilities in the district in intervention and control areas. When the study workers identify minor or chronic illness in mothers or infants in either intervention or control areas they will encourage referral to the appropriate health facility. All information will remain confidential. Access to information will be limited to interviewers and their supervisors at sites of collection, to auditors and data feeders at the collation point and thence to the senior data management officers and principal investigators. No analyses or reports will include the names of participants. A key objective is to take lessons learned quickly to scale in other districts, through our partners. The Ministry of Health in Malawi will be involved throughout the implementation of this project. At a local level we will work closely with the district development committees, local community representatives and locally active NGOs. The Mchinji District Health Office will play a key collaborative role. We will also work closely with UNICEF, Malawi and the National AIDS Commission, to implement PMTCT activities. Community-based team members have been recruited locally and carry out their activities in their home areas. We will attempt to maintain intervention costs to a minimum so that they can be adopted by the Ministry of Health. Control areas will be the first beneficiaries of scale-up if either of the interventions prove to have a positive impact.

I-AI Digest

Study Justification:

– The study aims to address the UN Millennium Development Goals of reducing maternal and child mortality in rural Malawi.
– The interventions being tested are community-based and aim to increase awareness and local action to improve maternal and child health.
– The study will provide evidence on the effectiveness of these interventions in reducing mortality rates and improving health care-seeking behaviors.

Highlights:

– The study uses a cluster-randomized controlled trial design to test the impact of two interventions: a community mobilization intervention through women’s groups and a volunteer-led infant feeding and care support intervention.
– The interventions will be implemented in 48 rural clusters, with 24 clusters receiving each intervention and 24 acting as controls.
– Outcome indicators will be measured through a demographic surveillance system, including maternal, infant, neonatal, and perinatal mortality rates, as well as exclusive breastfeeding rates.
– The study will also collect qualitative, quantitative, and economic data for process and economic evaluations.

Recommendations:

– The study recommends implementing the interventions in control areas if a positive effect is shown at the end of the two-year study period.
– The study also suggests exploring the potential synergistic relationship between the two interventions and evaluating their combined impact on mortality rates.

Key Role Players:

– Local female facilitators for the women’s group intervention
– Local volunteers for the infant feeding and care support intervention
– Supervisors and senior supervisors to support the facilitators and volunteers
– Health Surveillance Assistants to supervise the volunteer counsellors
– MaiMwana project, a Malawian trust established in 2003, collaborating with the Department of Paediatrics, Kamuzu Central Hospital, the Mchinji District Hospital, and the UCL Centre for International Health and Development
– Ministry of Health in Malawi, district development committees, local community representatives, and locally active NGOs
– UNICEF, Malawi, and the National AIDS Commission for implementing PMTCT activities

Cost Items:

– Start-up and running costs for both interventions
– Training and support for health services
– Monitoring and evaluation costs
– Cost per maternal, infant, neonatal, and perinatal death averted
– Cost per newborn infant exclusively breastfed
– Cost of scaling up the interventions at the national level
Please note that the above information is a summary of the study and may not include all details.

Amandla Obufakazi

The strength of evidence for this abstract is 7 out of 10.
The evidence in the abstract is strong because it describes a cluster-randomised controlled trial design to test the impact of two interventions on specific outcome indicators. The study uses a two-by-two factorial design and has a large sample size of 48 clusters with a population of 144,000. The interventions aim to improve maternal and child health through community mobilisation and infant feeding and care counselling. The study will collect both quantitative and qualitative data for process and economic evaluations. To improve the evidence, the abstract could provide more details on the specific outcome indicators being measured and the statistical power of the study.

Abstract

The intervention will be evaluated in a cluster-randomised controlled trial. A cluster design has been chosen because the allocation and loci of delivery of the interventions (community clusters) are groups rather than individuals. 48 rural clusters have been identified. 24 have been allocated to receive the women’s group intervention and 24 will act as controls. Using a factorial design, each of these arms of the women’s group trial was randomized a second time, 12 clusters to receive the volunteer counseling intervention and 12 to act as controls. An independent surveillance system for outcomes – births, morbidity, mortality, and care seeking – has been designed and implemented in all 48 clusters, covering a population of 144,000 (Figure (Figure11). Two-by-two factorial design. Community-based women’s group mobilisation intervention and home-based infant feeding and care counselling, with basic health service strengthening for both intervention and control areas Will community mobilisation through women’s groups reduce perinatal, neonatal, infant and maternal mortality rates through changes in care practices and health-seeking behaviour? Will volunteer infant feeding and care counselling for pregnant and breastfeeding mothers in their homes reduce infant mortality through changes in knowledge and practices regarding exclusive breastfeeding, family planning and other care practices and health-seeking behaviours? Women’s group activities will lead to: reductions in maternal, perinatal, neonatal and infant mortality, reductions in maternal and infant morbidity, and increases in recognition of high-risk symptoms, increased health-care seeking behaviour and changes in care-taker practices. Infant feeding and care counselling sessions will lead to: reductions in infant mortality, reductions in maternal and infant morbidity, increases in exclusive breastfeeding rates in the first six months, increases in health-care seeking behaviour, and changes in care-taker practices. The study endpoints are shown in Table Table1.1. The interventions will run for two years from the date when they are hypothesised to start having an effect, nine months after the start of the interventions, allowing for new pregnancies to have the maximum benefit of exposure to interventions. If a valuable effect is shown at the end of this period, the intervention models – refined on the basis of experience – will be implemented in the control areas. Study outcomes A cluster-randomised controlled trial, with a two-by-two factorial design, will be used to test the impact of two interventions on specific outcome indicators. 48 clusters or ‘zones’ were defined from the Enumeration Areas generated during the 1998 census (Figure (Figure2a).2a). Each zone contained a population of approximately 8000 in total. A population of around 3000 was selected from the centre of each zone, leaving a ‘buffer area’ around it to reduce contamination between neighbouring intervention and control areas (Figure (Figure33). a – Aggregation of census enumeration areas into 48 study ‘zones’, b – Random allocation of zones to four different combinations of intervention. Study villages and buffer areas in three clusters near Mchinji Boma. Mchinji district lies to the west of Lilongwe, in the Central Region of Malawi. It has international borders with Zambia and Mozambique, and a population of about 375,000 in 2004 [33], of whom about 80% live in rural areas and make a living through subsistence farming. The main ethnic group are the Chewa (90%), and the predominant religion is Christianity (97%). Comparison of some key socioeconomic and health indicators for Mchinji and Malawi are shown in Table Table22. Socioeconomic and health indicators for Mchinji and Malawi Sources: Malawi Population and Housing Census, 1998, Malawi DHS, 2004, World Bank 2006, State of the World’s Newborns report 2001, UNDP 2009 In 2006 it was reported that 99% of women in Mchinji attended antenatal care at least once during their pregnancy, but only 58% delivered at a health facility [34]. Maternal and perinatal health care is provided by personnel from one district hospital (a first referral and secondary health facility), four rural community hospitals (one government and three mission hospitals), one maternity unit, six health centres that provide maternity care, two dispensaries and two private clinics that offer antenatal care. Quality is compromised by the severe shortage of personnel, low morale of the health providers, and irregular drug supplies. Traditional Birth Attendants are available in all localities. The trial is being implemented by MaiMwana project, a Malawian trust established in 2003, as a collaboration between the Department of Paediatrics, Kamuzu Central Hospital, the Mchinji District Hospital and the UCL Centre for International Health and Development. The target population will be rural communities with the least access to health services, who might benefit most from community-based interventions to improve maternal and child health. The district administrative centre was excluded because it is more urbanised than the rest of the district and therefore not comparable to other clusters. The target group for participation in both of the interventions will be women of childbearing age (WCBA), between the ages of 15 and 49 years, and particularly pregnant mothers. Girls aged between 10 and 15 years will also be monitored and encouraged to participate in order to identify and support early teenage mothers. Older women will also be encouraged to attend as they influence decision-making around pregnancy, childbirth and child care and have valuable experiences to share. All women aged 10 to 49 years (inclusive) who agree to take part, will be enrolled into the study, regardless of whether they are married or not. Women who have no possibility of conceiving during the study period will be enumerated but excluded from the final sample (for example women who have had hysterectomies or terminal family planning procedures). Participation in intervention activities will be voluntary, and women’s groups will be free to establish their own membership criteria. A cohort of 44,000 women aged between 10 and 49 years was defined during the baseline phase of the study, and each is visited monthly by study personnel. From the beginning of the study period we identified all pregnancies, births and deaths occurring within the cohort of WCBA, with follow-up until at least one year after delivery. The cohort members are listed in a master document to which new names can be added: the cohort is open and new participants may be enrolled during the study period if they move into the study area, or if they reach 15 years of age. These are summarised in Table Table33. Elements of the women’s group and infant feeding and care counselling interventions The women’s group intervention seeks to build the capacities of communities to take control of the mother and child health issues that affect them. The intervention is community-based in that it defines the community as the agent of change [35]. To achieve this, 24 local female facilitators, one per cluster, were recruited and trained. The facilitators formed between six and 12 groups in their clusters and invited all women of childbearing age to attend. They guide the groups of women through a four-phase community mobilisation action cycle developed to be appropriate, accessible and feasible for the Malawian context from similar models in Bolivia and Nepal (Figure (Figure4)4) [36,37]. In the first phase, consisting of eight meetings, the groups identify and prioritise the mother and child health problems they feel are most important and discuss what may contribute to these problems and how they might be prevented and managed. In the last meeting of this phase they share their discussions with men in the community. In the second phase, consisting of four meetings, the groups plan locally feasible strategies to address the problems they have prioritised. In the last meeting of this phase they share their discussions with the whole community. In the third phase, consisting of four meetings, the membership of the groups expands to include all community members, including men, working together to implement the strategies that have been identified. In the fourth phase, four meetings, the groups evaluate what they have done, and plan for the future. The facilitators received minimal health training but use participatory rural appraisal tools and picture cards to facilitate discussions and enable groups to access their collective knowledge and capacities. With these capacities the groups take increasing control of the intervention over the course of the cycle. They do not receive any resources from MaiMwana Project, except the guidance of the facilitator, employed by MaiMwana project, and supported by four trained supervisors and a senior supervisor. Women’s group cycle. The volunteer counselling intervention seeks to change the behaviour of individuals in relation to care and care-seeking for mothers and children. The intervention is community-based in that it defines the community as the target of change [35]. In particular, the intervention seeks to provide health education to raise the awareness, change the attitudes and build the self-efficacy of mothers in relation to exclusive breastfeeding. To achieve this, 72 female volunteers, between two and four per cluster, were recruited and trained to identify pregnant women in their area and visit them at home to provide counselling and support to exclusively breastfeed their infants and perform other key infant care practices. The volunteer counsellors visit these women at five key times in pregnancy and after birth (Table (Table4).4). The first visit is conducted in the third trimester, and the mother receives information regarding the importance of early and exclusive breastfeeding, PMTCT, birth preparedness and family planning. The second visit is conducted in the first week after birth (and where possible within the first three days), when the mother receives information on the importance of exclusive breastfeeding, vaccinations, warmth and hygiene and family planning. The mother also receives information about danger signs for herself and her child, advice about breast problems and support for good attachment and positioning. The third visit is conducted at one month after birth and covers the same issues as the previous visit, with the addition of information about weaning. The fourth visit occurs at three months after birth and covers the same issues as the third visit. The fifth visit is conducted at five months after birth and covers the same issues as the third and fourth visits with the addition of information about weaning foods. The volunteer counsellors received minimal health training but use a picture book to facilitate learning. The volunteer counsellors are supervised by 24 government Health Surveillance Assistants, and the intervention is coordinated by one supervisor employed by MaiMwana project. Volunteer infant feeding and care counselling intervention meeting guide The 48 study zones were randomly allocated to one of four groups; 12 zones receiving infant feeding counselling and facilitation, 12 zones receiving facilitation only, 12 zones receiving infant feeding counselling and no facilitation, and 12 zones receiving neither (Figure (Figure11 and Figure Figure2b).2b). All zones will benefit from the district-wide training and support for health services from the project’s initiation. Random number generation was done in STATA 7.0, and each of the 48 clusters was allocated to one of four possible combinations of interventions. In this way, each intervention was stratified according to the presence or absence of the other one, in order to balance any effects of one intervention on outcomes of interest in the other. SL and DO generated the random number sequence and allocated the clusters to intervention groups. All women aged between 10 and 49 years residing within the clusters were eligible for inclusion in the interventions and follow-up of maternal and child health outcomes. Records of a woman’s migrations within and outside of the study area are recorded in order to allow for both ‘intention to treat’ and ‘per protocol’ analysis. Due to the nature of the interventions, blinding of study participants to their allocation was not possible, though analysts and trial monitors will be blinded to the study allocation until the definitive analysis is performed. Data will be collected independently from intervention implementation, and no results will be fed back to inform the interventions. The sample size for the cluster randomised controlled trial was arrived at by comparing statistical power for different estimates of population parameters. Parameters estimated included baseline mortality rates, the projected size of the reduction in maternal, neonatal and infant mortality and increase in exclusive breastfeeding rates due to the interventions, the number of births in each cluster over the trial period, the number of clusters in the intervention and control groups, the statistical power of the study, and the inter-cluster coefficient of variation. Realistic values of some of these parameters were difficult to estimate, as few population-level mortality data were available, especially at district and sub-district level. The sources of data and values used for these estimates and sample size calculations are summarised in Table Table5.5. We aimed to achieve the smallest sample size that would allow adequate statistical power to detect an impact within a reasonable time-frame and would be logistically feasible to implement. Parameters used to estimate sample size, and the estimated size of reductions or increases that would be detectable * Malawi DHS data used is the national estimate, as data were not disaggregated for Mchinji District Initial sample size estimates were made using national estimates of crude birth rates and mortality rates from the Malawi DHS for 2000 [38], subsequently revised after the results for the 2004 DHS survey were released [2]. Estimates were made with 80% power, a two-sided 5% significance level and an inter-cluster coefficient of variation (k) of between 0.15 and 0.30, using the methodology laid out by Hayes et al [39]. An estimate of k = 0.25 came from work on interventions to reduce HIV incidence rates [40], and later from work in Makwanpur, Nepal [27]. We assumed that variability in neonatal, infant and maternal mortality between clusters would be less in rural Malawi, due to fairly uniform exposure to risk factors such as poor hygiene and poor quality or absent delivery care and malaria. In addition, we reduced potential heterogeneity by excluding the main district administrative centre, as it was felt to be socio-economically and demographically different from rural villages. In recognition of the lack of certainty for this estimate, sample sizes were calculated for a range of values of k. Through the two-by-two design, it will be possible to assess the interaction between the two interventions, though the study is not powered to evaluate the combined impact of the interventions on mortality compared to single interventions alone, as this would have required an unfeasibly large sample. We will explore qualitatively and quantitatively whether or not there is a synergistic relationship between the two interventions, resulting in an effect greater in magnitude than either intervention alone. We might expect that women in an enabling environment (provided by women’s groups) would find it easier to discuss health issues and make health-seeking decisions for themselves and their babies than women in control areas where social barriers to care-seeking have not been addressed. More specifically, women in areas receiving both interventions might be more likely to use the Volunteer MaiMwana Counsellor and recognise the importance of her advice. More of these women may have decided to use PMTCT services than those not receiving facilitation, and will therefore be in a better position to make informed choices about infant feeding and family planning based on knowledge of their HIV status. Conversely, individual visits made by volunteers to women in their homes may serve to reinforce messages and issues arising from women’s group discussions. Data will be collected in two phases; baseline data during intervention development, and prospective data during intervention implementation. The baseline phase was completed in 2004. Formative qualitative data were collected through semi-structured interviews and focus groups. These data were collected with the specific purpose of: a) exploring current care practices to help to develop the structured questionnaires and interviews; and b) exploring the aims, setting, target population, methods and resources of the interventions to assist in their development. During the baseline data collection phase, the 692 study villages were mapped, each household enumerated, and a census was done in order to collect basic socioeconomic and demographic data about household members. A list of all female residents aged between 10 and 49 years was produced. Prospective data relevant to the evaluation of the two interventions will be collected in three ways – impact, process and economic. To assess the impact of the interventions on mortality, all WCBA will be visited monthly to identify pregnancies and births, as well as neonatal, infant and maternal deaths. WCBA will be visited by trained enumerators once per month, and events recorded in a register holding an up-to-date list of all women in the cluster (generated from the baseline survey, plus new residents). One enumerator will visit all WCBA in one cluster. All deaths of infants and women will be followed up by a supervisor who will verify the death and conduct a verbal autopsy interview between two and six weeks after the death. This interview will help to elicit the causes and contributing factors of the deaths [41]. There are five supervisors, each based at a nodal office. To assess the impact of the interventions on morbidity, care practices and behaviours, women who are identified as pregnant will be followed up until 6 months after birth by trained interviewers. Interviewers will administer one-month and six-month post-partum interviews to collect detailed information about demographic characteristics, maternity history, health-seeking behaviours, care behaviours and maternal and infant morbidity. To assess the impact of the interventions on growth and nutrition, women and infants will be visited by Health Surveillance Assistants (HSAs) at one-month and seven-months to collect anthropometric and biodata. Infant weight and height and mother’s mid-upper arm circumference will be measured, and blood spots taken from infant heel pricks will be tested for haemoglobin levels using a Haemocue machine. Six HSAs have been recruited and have received extensive training according to Malawi government guidelines in theoretical and practical aspects of blood spot collection by the Community Health Sciences Unit (CHSU) technicians. The HSAs will also be supervised by CHSU technicians quarterly and attend full week-long six-monthly refresher trainings. An integrated process evaluation will collect data on the key factors that may mediate the impact of the interventions in order to understand why they work or fail to work, and how different contextual factors influence their success. This will include collecting information on the context in which the interventions are being implemented, the mechanism through which the interventions are working, and the proximal outcomes of the interventions. On the basis of this information, hypotheses will be developed and, where possible, tested to explain and better understand the impact of the interventions. The data will be collected, using predominantly qualitative methods such as key informant interviews, focus group discussions and observations. A range of different respondents will be consulted, including members of intervention and control communities, members of women’s groups and women who have been counselled, key community, district and national informants, MaiMwana staff, and health facility staff. All inputs will be audited and cost-effectiveness analysis will be carried out in order to assess the replicability and scalability of the interventions, and the potential for them to be adopted as larger scale public health interventions in Malawi. Full costs will include start-up and running costs for both interventions, and will be collected through the project accounting systems. In addition to financial costs, economic costs will be estimated, valuing resources at their “opportunity cost”, or value in their best alternative use (including time, resource use and donated items) [42]. We will estimate the cost per maternal, infant, neonatal and perinatal death averted and the cost per newborn infant exclusively breastfed. We will collect information on the costs associated with monitoring and evaluation, although research costs will not be included in the analysis. Appropriate sensitivity analyses will be used to explore the implication of uncertainty of assumptions. The cost of scaling up the interventions at national level will be estimated in order to explore cost-saving opportunities and to investigate issues of generalisability beyond the trial context and beyond domestic boundaries. All quantitative data collected will be delivered to the main office for data entry in a relational database management system in Microsoft Access run on a dedicated server and workstations. Each WCBA will be given a unique ID number generated from the cluster, village and household she comes from. All quantitative data from the mortality surveillance, morbidity, care practice and behaviour questionnaires and process evaluation will be linked to the WCBAs through this unique ID. After checking and entry, all questionnaires will be archived in a locked room for future reference. Qualitative information will be audio recorded after receiving verbal consent from respondents. Data in Chichewa will be translated into English by a bilingual speaker. To ensure conceptual, grammatical, and syntactical equivalence, translations will be reviewed collaboratively by the researcher who obtained the data and the translator (both bilingual Chichewa-English speakers), and the lead researcher (English speaker). This team will make decisions about the best terms to use. All data will subsequently be transcribed and imported into a database in MAXqda 2 (VERBI Software version 2). One enumerator per cluster will identify births and deaths, and each event will be cross-checked by one interviewer. Supervisors will make regular field visits to check the quality of work done by enumerators and interviewers and observe some interviews. Each supervisor is responsible for between six and ten clusters. Interviewers will meet with enumerators weekly in order to check on their work and receive updates on births and deaths in their area. Supervisors will meet with interviewers and enumerators fortnightly to check on their work, discuss problems and provide quality control feedback. A sample of 200 one-month and six-month interviews has been selected to be independently re-done by the supervisor, in order to be able to estimate recall and interviewer error rates. Quantitative data will be checked in three stages. The first check will be performed after completion of the questionnaire, by the MEO in full and a nodal data checker (NDC) based at one of the five nodal offices. The second check will be done by a team of two data checkers (DC) based at the main office. The last stage of data checking will be done at the point of data entry by the four data entry clerks (DEC). Further checks will be carried out internally within the electronic data-handling environment. Data entered into the study databases will be regularly reviewed for inconsistencies and missing information. Lists of women interviewed and key fields to be verified will be produced, such as ID numbers, dates of birth, and reported pregnancies, births or deaths that have received no further follow-up. Minimising loss to follow-up is an important aspect of trial conduct. Certain features of this location and population dynamic need special attention for outcome tracking: a) Residents of Mchinji move seasonally to maximise their access to fertile land during the farming season, both within the district and across international borders into Zambia and Mozambique; b) Residents of Mchinji may go home for delivery and some time after birth (to other villages within Mchinji or to other districts); c) Non-residents may come into Mchinji from other districts for delivery and some time after birth; d) Families may move after a woman’s death, making it difficult to find respondents who know the details of what happened; e) High population turnover in trading centres and commercial farm estates; f) Women being busy working in their gardens, or at community gatherings such as funerals and chieftainship ceremonies, making them unavailable for interview; g) Weather conditions making roads impassable and conducting interviews difficult. For residents or respondents who are temporarily unavailable, the main strategy is to keep following up until an outcome is ascertained. All women who have ever lived in the study areas are maintained in the database, and appear every month in the register. Any events (such as pregnancy or birth) that are reported but no further details are known, are selected and lists produced to remind field-workers of the need for follow-up. In most cases this causes delays in getting complete data, though basic data for estimating mortality rates are still available. In recognition of the fact that certain data collected after a long delay will no longer be valid, sections in the questionnaire (such as infant feeding recall) are skipped. For residents or respondents who are permanently unavailable, basic information about dates and timings of events is sought from other community members such as friends or neighbours. Contamination may occur when people from one cluster have contact with people from another. In the rural villages of Mchinji, there are many opportunities for social mixing. Friends, relatives or neighbours may mix socially, or contact may be made through travel or migration between intervention and control clusters. There might be direct participation of residents from control areas in WG or VMC activities, or more likely, informal discussion of ideas from WG or VMC activities – control area residents may gain some benefit from hearing health messages received by intervention participants. The usual effect of this kind of contamination would be ‘dilution’ of the differences between treatment arms [43]. In order to reduce the possibility of contamination, we opted to use clusters of villages rather than individual villages as the unit of randomisation, thus reducing rates of travel across cluster boundaries [43]. Furthermore, each zone had a defined ‘buffer area’ around the perimeter (figure (figure2).2). A population of 3000 in each zone was required to achieve the desired sample size, but rather than selecting villages at random from each zone, only villages at the centre of the zonal area were eligible for inclusion in surveillance and intervention activities. This reduced the possibility of communication between neighbouring study villages in intervention and control areas. Women’s group facilitators and volunteer counsellors are residents of the zone in which they work. This reduces the possibility that they might transfer intervention benefits to neighbouring communities. For the WG trial, whilst health messages are discussed in group meetings, it is unlikely that neighbouring control communities would spontaneously mobilise themselves without the presence of a facilitator. So the purported benefits of community empowerment and social capital are unlikely to spread beyond intervention areas. An independent Data Safety and Monitoring Board (DSMB) will meet several times according to the DAMOCLES statement, [44] to review progress and advise on the conduct of the trial. They will assess compliance with the protocol, data quality and completeness, recruitment figures, sample size assumptions and ethical considerations. The meetings will not include analysis of outcome data by intervention allocation until completion of the trial. Baseline data will be reviewed to evaluate how well balanced the clusters were after randomisation and suggest any adjustments that may need to be made. We do not foresee any adverse effects of community mobilisation or peer counselling, so we do not intend to apply stopping rules. Data from the trial will be analysed by intention to treat at both cluster and participant levels. Participants will be assigned to the cluster they were resident in at the time the trial began, even if they move to another cluster or out of the study area during the trial period. Primary and secondary outcomes will be compared between intervention and control groups with random effects logistic regression models, taking account of clustering. Potential confounding factors will be included in the model, and measures of baseline mortality will be included in order to adjust for any imbalances in randomisation at the outset. Cluster-level data will be compared using modified t-tests. All estimates will be presented with 95% confidence intervals. Data collected from the project both during the baseline phase and the intervention phase mean that we will be able to look at overall changes in outcome indicators over time. Factors other than the study interventions may influence health indicators seen, such as other mother and child health programs in the area, or positive reporting and the Hawthorne effect [45]. If mother and child health programs are only active within some of the cluster areas then these can be examined separately if necessary. Analysis of qualitative data will involve iteratively developing a system of codes and memos jointly by the lead researcher and the researcher who collected the data. This method will seek to develop an analytical framework based on the data by coding and memoing pertinent excerpts that illustrate emerging themes. Subsequently, in an iterative process, the researchers will refine their analysis ensuring that the themes that are built up are cross-checked with other data firstly within and then between transcripts so that the validity of emerging explanations is tested and improved. The qualitative data will also be used to interpret and contextualise the quantitative data collected. MAXqda 2 software will be used to facilitate this analysis. Ethical permission for this study was granted by the Malawi National Health Sciences Research Committee in January 2003 (Ref: MED/4/36/I/167) and the ethics committee of the UCL Institute of Child Health and Great Ormond Street Hospital. It is registered with ISRCTN06477126. Verbal and written consent was received from community leaders after full consultation and discussions. The regional, district and village leaders, and local health and development professionals will have ongoing access to the research programme and will be the first to be briefed on study findings and outcomes through written and verbal reports. Before each instance of data collection, the process and advantages and disadvantages of taking part will be explained to all participants. Verbal consent will be obtained, and participants informed that they can stop taking part at any time. Participation in intervention activities is voluntary, and women may choose to start or stop as they wish. The study is designed to test the community effectiveness of two community-level initiatives to reduce maternal, infant, neonatal and perinatal mortality in rural Africa. Encouraging community action for maternal and newborn care alone will not grant success. For health promotion interventions to work, the supply side of health care services must reach a minimum standard. The study team considers it unethical to strengthen services only in intervention and not control areas. Control communities will benefit from low-cost improvements in equipment, supplies and training at all primary level facilities in the district in intervention and control areas. When the study workers identify minor or chronic illness in mothers or infants in either intervention or control areas they will encourage referral to the appropriate health facility. All information will remain confidential. Access to information will be limited to interviewers and their supervisors at sites of collection, to auditors and data feeders at the collation point and thence to the senior data management officers and principal investigators. No analyses or reports will include the names of participants. A key objective is to take lessons learned quickly to scale in other districts, through our partners. The Ministry of Health in Malawi will be involved throughout the implementation of this project. At a local level we will work closely with the district development committees, local community representatives and locally active NGOs. The Mchinji District Health Office will play a key collaborative role. We will also work closely with UNICEF, Malawi and the National AIDS Commission, to implement PMTCT activities. Community-based team members have been recruited locally and carry out their activities in their home areas. We will attempt to maintain intervention costs to a minimum so that they can be adopted by the Ministry of Health. Control areas will be the first beneficiaries of scale-up if either of the interventions prove to have a positive impact.

Izinto zokwakha

N/A

Emisha

Based on the information provided, the following innovations could be potential recommendations to improve access to maternal health:

1. Community mobilization through women’s groups: This intervention involves forming women’s groups in rural communities to increase awareness and take action on maternal and child health issues. The groups go through a four-phase cycle of problem identification, strategy development, implementation, and evaluation. This approach aims to empower communities to address their own health challenges and improve care practices and health-seeking behaviors.

2. Volunteer-led infant feeding and care support: This intervention involves recruiting and training local volunteers to provide counseling and support to pregnant and breastfeeding mothers in their homes. The volunteers offer advice on exclusive breastfeeding, birth preparedness, newborn care, and immunization. This approach aims to improve knowledge and practices related to infant feeding and care, leading to a reduction in infant mortality.

3. Cluster-randomized controlled trial design: The study design used in this project allows for the evaluation of the effectiveness of the interventions. Clusters of rural villages were randomly allocated to receive either the women’s group intervention, the infant feeding support intervention, both interventions, or no intervention. This design helps to assess the impact of the interventions on specific outcome indicators, such as maternal and infant mortality rates and exclusive breastfeeding rates.

4. Demographic surveillance system: An independent surveillance system was designed and implemented in all 48 clusters to collect data on births, morbidity, mortality, and care-seeking behaviors. This system provides valuable information for measuring the impact of the interventions and monitoring changes in health indicators over time.

5. Integration of basic health service strengthening: Both intervention and control areas received support for strengthening basic health services. This integration ensures that the interventions are implemented within a supportive health system, addressing issues such as personnel shortages, drug supplies, and quality of care. Strengthening health services is crucial for improving access to maternal health care.

These innovations aim to address the UN Millennium Development Goals of reducing maternal and child mortality by improving care practices, health-seeking behaviors, and community awareness. The cluster-randomized controlled trial design allows for rigorous evaluation of the interventions, providing evidence for their effectiveness and potential scalability.

AI Innovations Description

The recommendation for improving access to maternal health is to implement a cluster-randomised controlled trial of community-based interventions in rural Malawi. The trial will test the impact of two interventions: a community mobilisation intervention run through women’s groups, and a volunteer-led infant feeding and care support intervention.

The community mobilisation intervention will involve local female facilitators guiding women’s groups through a four-phase cycle of problem identification, strategy identification, implementation, and evaluation. The groups will meet monthly at the village level. The goal of this intervention is to increase awareness and local action related to maternal and child health problems.

The infant feeding and care support intervention will select local volunteers to provide advice and support for breastfeeding, birth preparedness, newborn care, and immunisation. These volunteers will visit pregnant and new mothers in their homes five times during and after pregnancy.

The interventions will be evaluated using a two-by-two factorial cluster-randomised controlled trial design. 48 rural clusters have been randomly allocated to receive either the women’s group intervention, the infant feeding support intervention, both interventions, or no intervention. The impact of the interventions will be measured through a demographic surveillance system, with primary outcomes including maternal, infant, neonatal, and perinatal mortality rates, as well as exclusive breastfeeding rates.

The trial aims to assess whether community mobilisation through women’s groups and volunteer-led infant feeding and care counselling can reduce mortality rates and improve care practices and health-seeking behaviors. The interventions will run for two years, and if successful, the models will be implemented in control areas.

The trial is being implemented by the MaiMwana project, a collaboration between the Department of Paediatrics, Kamuzu Central Hospital, the Mchinji District Hospital, and the UCL Centre for International Health and Development. The target population for the interventions is rural communities with limited access to health services. The interventions will focus on women of childbearing age, particularly pregnant mothers, but will also involve girls aged 10-15 and older women who influence decision-making around pregnancy and child care.

Data will be collected through structured interviews, qualitative methods, and economic evaluations. The trial will be monitored by an independent Data Safety and Monitoring Board, and ethical permission has been obtained. The results of the trial will be shared with community leaders, health professionals, and other stakeholders, and lessons learned will be used to scale up the interventions in other districts.

AI Innovations Methodology

The study described is a cluster-randomized controlled trial that aims to test the impact of two interventions on improving maternal and child health in rural Malawi. The interventions being tested are a community mobilization intervention run through women’s groups and a volunteer-led infant feeding and care support intervention.

The community mobilization intervention involves local female facilitators guiding women’s groups through a four-phase cycle of problem identification, strategy identification, implementation, and evaluation. The women’s groups will meet monthly at the village level to address maternal and child health issues.

The infant feeding and care support intervention involves selecting local volunteers to provide advice and support for breastfeeding, birth preparedness, newborn care, and immunization. These volunteers will visit pregnant and new mothers in their homes five times during and after pregnancy.

To evaluate the impact of these interventions, a cluster-randomized controlled trial design is being used. 48 rural clusters have been identified, with 24 clusters allocated to receive the women’s group intervention and 24 clusters acting as controls. Within each arm of the women’s group trial, 12 clusters were randomized to receive the volunteer counseling intervention and 12 clusters acted as controls.

The impact of the interventions will be measured through a demographic surveillance system, which will collect data on births, morbidity, mortality, and care-seeking behaviors. Primary outcomes include maternal, infant, neonatal, and perinatal mortality for the women’s group intervention, and exclusive breastfeeding rates and infant mortality for the infant feeding intervention.

Data will be collected through structured interviews with mothers one month and six months after birth, as well as through qualitative, quantitative, and economic data collection for process and economic evaluations.

The methodology for simulating the impact of these recommendations on improving access to maternal health involves comparing outcome indicators between the intervention and control groups using statistical analysis. Random effects logistic regression models will be used to compare primary and secondary outcomes, taking into account clustering. Potential confounding factors will be included in the analysis, and measures of baseline mortality will be used to adjust for any imbalances in randomization.

In addition to analyzing quantitative data, qualitative data will be collected and analyzed to understand the factors that mediate the impact of the interventions and to provide contextual information.

Ethical permission has been obtained for the study, and consent will be obtained from participants before data collection. Confidentiality of information will be maintained, and access to information will be limited to authorized personnel.

The study aims to take lessons learned from the interventions and scale them up to other districts in collaboration with the Ministry of Health in Malawi and other partners. Control areas will also benefit from improvements in equipment, supplies, and training at primary level facilities.

Umbhali

Dima Health

Statistics:

Citations: 65

Identifiers:

DOI: 10.1186/1745-6215-11-88

ISSN: 17456215

Research Areas:

Community Interventions, Food Security, Genetics and Genomics, Health System and Policy, Infectious Diseases, Maternal Access, Maternal and Child Health, Quality of Care, Sexual and Reproductive Health, Social Determinants

Study Countries:

Malawi, Mozambique, Multi-Countries, Zambia

Study Design:

Cohort Study, Cross Sectional Study, Exploratory Study, Grounded Theory, randomised-control-trial

Study Approach:

Mixed-methods, Qualitative, Quantitative

Participants Gender:

Male & Female

Yabelana ngalokhu: