Implementation science to design, test and scale up effective Kangaroo Mother Care in Oromia region, Ethiopia

To develop a locally tested and optimised Kangaroo Mother Care (KMC) scale‐up model to achieve high population‐based effective coverage of KMC in Oromia region.


| INTRODUC TI ON
Ethiopia has made remarkable progress towards reducing under-five mortality; however, the neonatal mortality rate remains stagnant at 33/1000 live births and neonatal mortality accounts for 56% of under-five deaths. 1 Addressing neonatal mortality is a critical challenge to the health system in its efforts to improve overall child health. KMC has been demonstrated to promote physiologic stability, facilitate early breastfeeding, provide a thermally supportive environment and reduce the risk of serious infections. Compared to conventional care, KMC has been shown to reduce mortality of clinically stable low-birthweight infants by approximately 40%, nosocomial infection/ sepsis by 55% and hypothermia by 66%. 2 KMC also promotes bonding between infants and their mothers during the first hours and days of life. [2][3][4] Despite the growing body of evidence for KMC as a costeffective intervention in improving outcomes of low-birthweight infants, implementation has been limited and global coverage has remained low. [5][6][7] In Ethiopia, KMC coverage is also low, even though it was first introduced in 1996 and the target set by the Ministry of Health was to reach 80% of preterm babies by the year 2020.
There is a great need to scale up KMC implementation in areas where it has been initiated, and to extend implementation further.
The initiation and scale-up efforts should be implemented in a manner that will institutionalise activities within the health system. This will require a locally contextualised framework, based on implementation research with a view that considers the context and engages all stakeholders. 5,8 By studying the barriers and potential facilitating factors to achieving high effective KMC coverage, using data in repeated rapid cycles of model developmentimplementation-evaluation-refinement, and partnering with the government health system to pilot-test the models, we aimed to develop a promising model to achieve 80% effective KMC coverage in Oromia, Ethiopia.

| Site selection and population
Oromia is the largest regional state in Ethiopia, covering an area of 284,538 square kilometres and an estimated population of 38 million. 9 Study sites were selected jointly with the regional health bureau using criteria including accessibility (within 300 km of Addis Ababa), the availability and number of neonatal care providers, the availability of neonatal services, the number of facility births in 2014, the estimated number of newborns weighing <2500 g and <2000 g, and the reported number of neonatal deaths. A total of five public hospitals and 39 health centres catering an estimated population of 1.1 million were selected (Table 1) to give a minimum of 310 < 2000 g newborns that will allow us to estimate 80% effective KMC coverage with 5% precision. The details of the main multisite study protocol are discussed elsewhere. 7

| Study Design
The study used a mixed-methods design: formative research and continuous programme learning to understand contextual factors and identify barriers. This informed the design of scalable models for delivering KMC across the facility-community continuum.
These models were implemented in routine care settings and evaluated with an eye towards wider national or state-level scale-up.
Newborns with birthweight of <2000 g, who were born in or who presented to KMC study facilities, were enrolled into a < 2000 g co-  Figure 1).
To run the study, three teams with different roles were organised: implementation support team to provide support to the government health system to effectively implement the models of KMC service delivery; programme learning team to conduct formative assessment and continued programme learning through qualitative data collection, analysis and use to refine the models; and independent outcome measurement team to continuously measure the model outcome indicators.
In-depth interviews and focus group discussions conducted as part of the formative research and continuous programme learning Community engagement was crucial to improving effective coverage, institutionalising KMC within a health system and developing KMC culture within communities 12 (see Figure 2). Second, we described the evolutions of the KMC implementation models during the study period. Finally, we presented the effectiveness of the final KMC implementation model assessed using primary model effectiveness outcome measures.

| Outcome
The primary outcome was population-based effective coverage of KMC: the proportion of infants with birthweight <2000 g who re-

| RE SULTS
Highlighting the contextual uniqueness of the KMC Oromia site, we present results in three main areas: health system building blocks, implementation models, and effective KMC coverage. (iv) Health management information: Data from the Health management information system were aggregated and often incomplete.

| Health system building blocks
With the local governments and the Ministry of Health, we created simple registries and clinical logs to track the care provided and health outcomes among small babies. Health facility staff were able to monitor the data and see improvements.
(v) Health infrastructure: There was limited space for newborns, especially for mothers to stay with their newborns. The available space tended to be dark, dirty, and often with poor ventilation.
Latrines were far away, showers were broken, and there were no cooking facilities or food. Throughout the study, the implementation team adopted a zero-separation policy of mother and baby, and worked with facilities to renovate space, fix latrines and washing places and keep them tidy and clean.
(vi) Service delivery: KMC services were limited and initially were not systematically provided at the study facilities. Throughout the study, we introduced and standardised the KMC service provision.
Details are provided in the section on implementation process.

| Community engagement
Health extension workers and health development army leaders are the elements that link the formal health system with the community and community-based organisations. However, during the initial phase of project implementation, follow-up and support between the formal health system and health development army leaders seemed to be a bit loose and was affected by the prevailing unrest within the region at the time. Family support to KMC-implementing mothers was lacking, even though there was very high readiness among mothers to do whatever they could to save their babies. In addition, there was a perception within the community that small babies do not survive, weakening the efforts to provide proper care.
As implementation was scaled up, community engagement was facilitated through: the introduction of KMC family discussion sessions on KMC rooms; the presence of champion mothers during these discussions; the use of KMC reminder charts and recording the babies' weight progress in the chart; and post-discharge follow-up from specifically identified newborn health focal persons from catchment health centres.  Figure 3 and Table S1.

| Initial KMC implementation model
The first KMC implementation model was informed by reviewing available evidence from literature, using experiences of the study team members, and conducting a comprehensive formative assessment. The formative assessment helped to understand the context- The formative assessment data were used in an interactive workshop with health workers and regional health bureaus. The deidentified raw data extracted from transcription of audio-recorded interviews were analysed using a framework developed by the research team. The process allowed participants to review and reflect on the data extracted from interviews of community members, health extension workers, health development army leaders and mothers, and to suggest solutions to address identified challenges.
Five key challenges to KMC implementation were identified   monitoring and evaluation quality improvement tool because they duplicated the balanced scorecard that the health system was using for the same purpose. We added promoting and monitoring hand hygiene (hand washing or hand rubbing) and the designation of a newborn care focal person and champion for each facility. Furthermore, guided by the programme learning data, we added the use of KMC champion mothers to teach community members.
Implementation of the second model provided additional lessons that led to further refinement to the KMC facility component.
In the third model, we dropped integration of performance review and mentoring into the hospital review meetings because through the programme learning, we learned that the hospital internal review meeting lacked focus and discussed administrative issues. Following learning gained from the implementation of the second model, we also added interventions. In one of the KMC facilities, we redesigned newborn care ward to introduce family-integrated newborn care, and in all hospitals, we introduced KMC in the neonatal intensive care units and labour and delivery wards, and added orientation of support staff (janitors, porters) on KMC to promote KMC practice, and use of senior mothers who practised KMC for longer time to teach mothers who were newly admitted to the unit.

| The third (final) KMC implementation model
The final KMC implementation model was the result of continuous improvements made to the initial models. Figure 3 and Table S1 show the changes made. The final model ( Figure 4) devoted more attention to interventions at the KMC facility directed at immediately initiating KMC to all eligible babies, continuing quality KMC, and improving experiences of mothers and health workers involved in the care. Figure 4 and (54%), and that at 7th day post-discharge was 108/286 (38%). In the 24 h period before discharge, the coverage of any skin-to-skin care was 67%, and that of exclusive breastfeeding was 70%. Coverage before discharge and on the 7th day post-discharge was 54% and 60%, respectively ( Figure 5). Exclusive breasƞeeding, (%) EffecƟve KMC (>8 hours of skin-to-skin care and exclusive breasƞeeding), (%) KMC in the 24-hour period before discharge from facility (n=307) KMC at home on 7th day post discharge (n=286) the facility and community government structures, KMC champion health workers, reorganising the health ecosystem to support the power of mothers, data-driven continuous model improvement, and strengthening of the linkages between communities and the health system.
The study has a few limitations. First, there were political unrest and protests in Oromia region during the implementation period. As a result, some Health Extension Workers and Health Development Army members were unable to effectively carry out their routine activities including identification and referral of home births and postdischarge follow-up of mothers. Second, the study did not assess the effect of increased effective KMC coverage on health outcomes such as morbidity, mortality and development. Third, although the iteration of the models resulted in dropping and adding ineffective and effective interventions, our study did not assess the contribution of individual interventions to effective coverage of KMC.
Further studies evaluating the impact of increased population-based effective coverage of KMC on health outcomes and studies that measure the effects of individual components within a package of interventions on outcomes will be important.

| CON CLUS ION
Our study demonstrated that an implementation model that achieves a 54% population-based effective KMC coverage can be