Effect of Integrated Community Case Management on Access and Utilization of Maternal, Newborn and Child Health and Immunization Services in Hard-to-Reach Communities in Migori County, Kenya: A Quasi-Experimental Study

Background Integrated community case management (iCCM) improves access to management of leading causes of under 5 (U5) mortality. Evidence of iCCM on maternal and newborn health and immunization services is scanty. The objective of this study was to determine the additional effect of iCCM on antenatal, skilled birth attendance (SBA) and immunization coverage in hard-to-reach communities. Methods A quasi-experimental (nonequivalent control group pretest – posttest) design for iCCM in Migori county. The intervention was iCCM training, mentorship/coaching and supportive supervision of 20 community health volunteers (CHVs). Twelve months pre-post intervention Kenya Health Information System (KHIS) data between July 2017-Sept 2019 reviewed. Differences in proportions for MNCH indicators pre – post-training were tested through test of proportions and considered statistically signicant at P ≤ 0.05 values. CHVs geographical coverage and their multi-purpose ability in curative, preventive and promotive activities demonstrating the limits of this cadre (19). This underscores the conceptualized enabling environment that promotes utilization of quality iCCM services including CHV deployment (availability and geographic distribution), quality of services (CHV training, supervision and adherence to protocols), enabling policy (supported by ministry of health, CHV motivation through monetary and non-monetary incentives, volunteer, treatment algorithms), demand creation (social mobilization, care seeking behaviour and availability of other health services) and availability of CHV supplies (medicines and diagnostics) for community–based treatments of childhood illnesses (12, 13). community health data (referrals) on maternal – antenatal birth health – postnatal and immunization a two-sample test of post-iCCM

promotes adoption of community newborn care practices as shown in the Malawi, Nepal, Bangladesh, and Uganda pilot that are key to reducing preventable neonatal mortalities (24).
Kenya is not on track for the attainment of U5 mortality rate to at least as low as the global target of 25 per 1000 live births by 2030 despite improvements in child mortality rates (25). Kenya's U5 mortality sharply declined from 115 to 52 deaths per 1,000 live births in 2014 (26). However, disparities persist in the management of the top 3 killers of U5 children in Kenya: pneumonia, diarrhoea and malaria and many children continue to die unnecessarily due to poor access to recommended treatment particularly pneumonia and diarrhoea. In 2013, Kenya developed the iCCM implementation framework anchored on the Ministry of Health (MOH) Community Health Strategy (MOH, 2007) and Child Survival and Development Strategy (MOPH, 2010) as well as the 'Policy Guidelines on Control and Management of Diarrhoeal Diseases in Children below Five Years' (27). Consequently MOH Malaria Control Unit, through funding from Global Fund round 10 and 11, facilitated trainings of CHVs in malaria case management in malaria endemic areas, including Migori County (28). The policy on use of antibiotics (Amoxicillin dispersible tablets for suspected pneumonia) by CHVs at the community level is under formulation (29). The iCCM strategy presented a platform for accelerating the control and management of childhood diarrhoea, malaria, pneumonia, neonatal causes and malnutrition at the community level and strengthen the health system, building upon the facility-level integrated management of newborn and childhood illness (IMNCI) activities thus contributing to the attainment of the SDG 3.2 by signi cantly reducing mortality attributed to the ve conditions. These community case management frameworks provide a roadmap for CHVs implementation and monitoring at household and community levels.
Despite national improvements, Migori county compared poorly (vs national) in most maternal, newborn and child health indicators -key being diarrhoea management, nutrition, immunization and skilled birth attendance as reported in the 5-year (2010-2014) KDHS ndings (26,30). For instance, 59% vs 74% of U5 with diarrhoea received oral rehydration therapy, 3% vs 8% U5 with diarrhoea received ORS and Zinc, 69% vs 78% children received birth polio immunization, 47% vs 71% children were fully immunized at 12 months, 56% vs 58% women attended at least four antenatal care visits during pregnancy and 53% vs 61% of births were delivered in a health facility. In 2016 through donor support, it became the rst county in Kenya to develop the county iCCM implementation plan (2016-2020) that focused on child survival envisioned "zero tolerance for preventable child deaths" (28). This was an innovative approach to accelerate reductions in incidence of the top childhood diseases in the county. Trainers of trainers in iCCM were developed in the county and trained CHVs in three hard-to-reach community units -Karapolo, Got Orango and Mobachi (implementation progress reports not available). Through support from the USAID funded AFYA Halisi project, the county trained CHVs from three facility-linked community units in hard-to-reach areas to implement community-based identi cation, management/referral of common childhood illnesses in under-5 s. This was an additional role to their other health preventive and promotive roles at the community and household level. These efforts aimed at improving the coverage of care in communities with poor access to health facilities for care.
Evidence around iCCM in Kenya has largely focused on the experiences of CHVs as agents of behavior change in the community, knowledge and skills assessment after training, coaching and after implementation (31,32). This study examined the contribution of the iCCM strategy on the provision of a spectrum of quality maternal, newborn and child health services in hard-to-reach communities. Speci cally, this study sought to determine the effect of iCCM implementation on diarrhea, pneumonia and malnutrition case identi cation and management; additive effect of iCCM implementation on maternal and newborn health (antenatal care coverage, skilled birth attendance and newborn/postnatal care) and immunization coverage in in the hard-to-reach communities of Migori county.

Study design
This was a nonequivalent control group pretest -posttest design, a quasi-experimental design that involves an intervention in an experimental/study group not randomly assigned and a control group, observed before and after its implementation (33). Data (from the Kenya Health Information System (KHIS), formerly the District Health Information System (DHIS2)) on community and health facility indicators for diarrhea (identi ed and treated) and pneumonia (identi ed and referred) at the community and antenatal, skilled birth attendance, newborn and immunization were reviewed. This was data for two intervention CUs (Komenya and Sagegi) and their link health facilities (Ongito and Kombe dispensaries respectively) and two control CUs (Misiwi and Gunga) -that only received basic CU training -and their link health facilities (Luciel and Nyamanga dispensaries respectively) of Migori county. The intervention was iCCM training conducted in June 2018 followed by other packages as summarized in Table 2 below. The pre-intervention period was between July 2017 -June 2018 and post-intervention period from Oct 2018 -Sept 2019. A 3-month 'preparation' interval between iCCM training and implementation was observed between the pre and post periods (July, August and September 2018). This period allowed preparation of the CHVs through issuance of recording/reporting tools, equipment (mid-upper arm circumference (MUAC) tapes, thermometers, respiratory timers) and the medications (Zinc and ORS) for iCCM implementation.
A "community health volunteer/worker" (CHV) was de ned as a health worker delivering health care in the community, trained in some way in the context of the intervention, and having no formal health professional or paraprofessional certi cate or tertiary education degree; regardless of whether or not they receive monetary payment (16). Importantly, the recruitment and management of CHVs is carried out by village and facility health committees. National demographic and health survey (2014) statistics showed that Migori county ranked top in diarrhoea (28%), second in fever/malaria (48%) and fth (12.7%) in children with pneumonia prevalence in the country (26). Due to intersubcounty differences in the prevalence rates of the top leading causes of childhood mortality in the county, two hard-toreach community units (CUs) linked to health facilities with poor maternal, newborn and child health indicators were selected for iCCM training and implementation. These were Sagegi -attached to Kombe dispensary (Kuria West subcounty) and Komenya -attached to Ongito dispensary (Uriri sub-county). The population in the villages in the selected CUs were more than the WHO recommended 5 km away from their respective link health facilities. The third CU trained on iCCM (Kanying'ombe) was an urban community unit and close to the link facility and therefore not included in this study as it did not meet the criteria for a hard-to-reach community. 'Hard-to-reach' referred to populations who have little regular contact with skilled pregnancy and childbirth services including people living in areas 'too far' from the health services. 'Too far' not only refers to the physical distance but also logistics and human resource capacity as used in the 'reaching every district' strategy in immunization (34).
Two control CUs in the hard-to-reach areas attached to health facilities where the project did not support and/or facilitate enhanced defaulter tracing for immunization during the national immunization rapid response initiative between July and September 2019 were selected. The CUs did not receive any additional support from the project or other partner support in community health services (apart from the mandatory CU training). Besides, the CUs must have been reporting every month for selection. The selected control CUs (Misiwi -attached to Luciel dispensary in Kanyasa ward and Gunga -attached to Nyamanga dispensary in Kachieng ward), were both from the expansive Nyatike sub-county of Migori county. All the study/experimental and control CUs had less than 40 percent coverage for 4th ANC and skilled birth attendance before the intervention (Fig. 1).

Community Health Volunteers And Community Units In Kenya
In Kenya, all CHVs are trained to diagnose malaria with a malaria rapid diagnostic test (mRDT) and treat it with artemisinin-based combination therapy (ACT); treat under-5 children with diarrhoea using oral rehydration salts (ORS) and zinc; refer suspected pneumonia, mild to moderate malnutrition, and sick newborns to a health facility (27). Before being certi ed as CHVs, all candidates receive training in basic community health modules as de ned by MOH which focuses on health and development, health promotion, and the Kenya community-based essential health package (  Post iCCM training, the CHVs carried out the following package of interventions (Table 3). Household visits -assessment/screening and classi cation of the child's condition(s) using a simpli ed IMNCIadapted algorithm (suspected pneumonia/fast breathing using respiratory timers); diarrhea; fever/malaria using thermometer & rapid diagnostic tests); malnutrition using MUAC tapes) 2 Referral of cases with general danger signs and other complicated cases (fast breathing/pneumonia, chronic diarrhea, bloody diarrhea, neonatal danger signs)* 3 Provision of treatment for the following conditions: • non-severe diarrhoea with oral rehydration salts and zinc; • non-severe malaria with artemisinin-based combination therapy Standard MOH reporting tools for community health services by CHVs were used for collecting routine service utilization data.

Data analysis
Data was extracted from the DHIS2 in Ms Excel, cleaned and exported to STATA version 13 for data analysis. Means and/or proportions of community and health facilities' maternal, newborn and child health and immunization indicators of interest were computed. Child health cases computed were diarrhoea (identi ed and treated), pneumonia (identi ed and referred), malaria (identi ed and treated) and undernutrition (identi ed and referred). The community health data elements of interest (referrals) on maternal health -antenatal care, skilled birth attendance; newborn health -postnatal visits and neonatal danger signs identi cation/management; immunization and nutrition from the two time periods were compared using a two-sample test of proportions for differences post-iCCM training. The overall effect and differences of the community health data elements on health facilities' maternal (antenatal and skilled birth attendance coverage), newborn and child health and immunization coverage indicators were compared between the two periods through the two-sample t-tests. It is important to note that for the pre-intervention period, data was analysed for 8 months from Nov

Community Health Volunteers Reporting
At least 86% of the households were visited by CHVs and over 95% of the CHVs reported during the two periods in the intervention sites with no differences in the proportion of households visited and the CHVs reporting in the pre and post-ICCM training in the two CUs. However, there was a signi cant difference in the proportion of households visited in the control sites with no differences in CHVs reporting between the two periods (Table 4).   (Table 5).

Maternal health
There was a double increase in the proportion of pregnant women referred from the community to the health facility for antenatal care (25.4% vs 45.8%, P < 0.0001) and pregnant women defaulters referred for antenatal care (9.8% vs 14.9%, P = 0.0327) in the intervention CUs. Consequently, there was a signi cant increase in the proportion of pregnant women referred from the community for antenatal care in the control CUs. However, there were no differences in the referrals from the community to health facilities for skilled births in both intervention and control CUs (Table 6).

Newborn health
There was a ve-times increase (1.4% vs 7.3%, P = 0.0337) in the newborns identi ed with danger signs and referred to the health facility for appropriate care in the intervention CUs. Even though there was a signi cant increase (43.9 vs 68.5%, P < 0.0001) in the proportion of newborns visited at home within 48 hours of birth in the control CUs, there were no newborns identi ed with danger signs referred for hospital care (Table 6).

Immunization
There were signi cant increases in proportions of children 0-11 months referred for immunization (4% vs 7.5%, P < 0.0001) and defaulters referred for immunization (2.2% vs 3%, P = 0.0366) in the intervention CUs. Conversely, there was a double reduction (10% vs 5%, P < 0.0001) in the children 0-11 months referred for immunization with no change in the proportion of identi ed defaulters referred for immunization in the control CUs (Table 6).

Malaria case management
Interestingly, a signi cant 16% decrease (57.6% vs 41.6%, P < 0.0001) in the number of fever cases who tested positive for malaria (RDT) in the control CUs with no change in the intervention CUs was observed. A 16% increase in the intervention CUs (32.0% vs 47.8%, P = 0.0001) and 7% increase in the control CUs (41.3% vs 33.9%, P = 0.0409) were observed in the cases of fever who tested positive for malaria and were treated with ACT as recommended (Table 6).   (Table 7).

Child health
There were no signi cant changes in the proportions of children U5 treated for diarrhoea and pneumonia at the health facilities in both the intervention and control sites (P ≥ 0.05) ( Table 7). Malaria case management has not been included since there were marked disproportionate errors in data reporting quality in the DHIS2 at the facility level for both intervention and control sites.

Immunization
There were signi cant increases in all the key antigens as per the Kenya Expanded Program of Immunization (KEPI) in the intervention sites. The antigens below the recommended averages of at least 90% coverage in the pre-intervention phase all improved to above 90% coverage for BCG (62.7% vs 102.6%, P = 0.0097), birth polio (58.7% vs 92.8%, P = 0.01) and Pentavalent 1 (82.4% vs 140.5%, P = 0.0008) in the intervention sites. The immunization coverages (pre and post) for all antigens in the control sites were lower than the recommended 90% for all antigens with no differences reported between the pre and post intervention periods (P > 0.05). The proportion of children under one year who are fully immunized (described as children who received BCG, three doses of Pentavalent, Polio and Pneumococcal vaccine each, Measles/Rubella 1 with a dose of Vitamin A) by their rst birthday signi cantly increased in the intervention sites (96.2% vs 130.8%, P = 0.0371) with no change observed in the control sites (53.7% vs 60.4%, P > 0.05), which had lower than the minimum expected 90% coverage. Even though the Measles/Rubella 2 coverage administered to children under 5 years at 18 months or at 24 months signi cantly increased from 55-87.5% (P = 0.0471) in the intervention sites, this was still lower than the expected 90% coverage in high-risk populations. The coverage for Measles/Rubella 2 in the control sites was lower than the expected 90% (21.7% vs 30.1%) with no differences between the two periods (P > 0.05) ( Table 7).

Discussion
This study set to demonstrate that apart from increased case detection and treatment and/or referral for diarrhoea and suspected pneumonia, iCCM had the potential to contribute to additional improvements in antenatal care, skilled births and immunization coverage and nutritional screening for undernutrition in the hard-to-reach communities. Through this community health platform, previously low service uptake and utilization in antenatal care, skilled births, community case (diarrhoea, malaria, suspected pneumonia and undernutrition) identi cation and management and immunization services improved in the catchment hard-to-reach communities in Migori. This nding is consistent with other studies in low and middle resourced countries in SSA where iCCM has been implemented (8, 10,19,32,40,41).
Household visitation by CHVs is an integral part of their responsibilities. The signi cantly low households visited in the control sites could be attributed to weak community-facility linkages, low supervision, erosion of skills/ knowledge, and lack of review of performance and targeted strategies to improve community health services in rural hard-to-reach areas. As community "gatekeepers," CHVs can be proactive in household visitations when supported with the basic community health diagnostic equipment (e.g. MUAC tapes, thermometers and respiratory timers) and basic drugs like ORS and Zinc that may help build trust between the community and the CHVs who are seen as community 'doctors.' Evidence in Uganda among CHVs showed that putting health-related knowledge into community action was among the important motivators for CHVs to conduct household visitations (42). However, there is also evidence from a peri-urban setting in Kenya among CHVs that lack of material resources necessary for CHVs' work and lack of motivation are demotivating factors contributing to low household visitations (31).
This study demonstrated that CHVs diagnosed uncomplicated diarrhoea cases and correctly treated 3 of every 4 cases identi ed post-training with the nationally recommended ORS and Zinc (23). Commodity stock-outs in the link health facilities that re-supply CHVs contribute to some cases missing treatment as per the recommended guidelines. Supply chain management is part of the health system bottleneck yet a key cornerstone for planning and sustainability of iCCM services that is likely to affect overall performance of CHVs trained on iCCM in poor and rural hard-to-reach areas (10,(43)(44)(45).
This nding is similar to a cross-sectional study conducted in rural Ethiopia and a multi-country study in SSA on implementation strength and quality of care provided by community health workers (46, 47). Our ndings also demonstrated improvement in the identi cation of childhood illnesses through simple diagnostic equipment. The improvement observed in the cases of suspected pneumonia detected by use of respiratory timers in taking respiratory rates and undernutrition cases by use of MUAC tapes is testament that through provision of diagnostic tools, high quality treatment for sick children can be achieved thereby contributing to quality interventions and health outcomes (40). Quarterly supportive supervision and mentorship conducted on regular basis ensured that knowledge and skills are maintained to promote quality community health care service utilization (32). These ndings support evidence from other studies that when well trained and mentored, CHVs can correctly manage multiple childhood illnesses (8, 9,15,19,32,41,46,48,49).
Community malaria case management improved at the intervention sites compared to the controls. First, the fever cases diagnosed as positive for malaria through RDT improved while there was a signi cant reduction in the same at the controls. Secondly, there was a marked increase in the proportion of cases correctly diagnosed and treated with the recommended ACT by CHVs both at the intervention and controls. These ndings are consistent with another study by Oresanya et al in Nigeria and Zalisk et al in Malawi (10,41). This can be explained by the fact that Migori is a malaria endemic zone where through the national malaria program, community units received prior training in malaria testing using RDTs and treatment using ACTs. This study underscores the fact that iCCM training and coaching/mentorship received reinforced the CHVs' knowledge and skills in managing common childhood illnesses by following the iCCM prescribed algorithm for decision-making on management /treatment of illnesses as evidenced by another Kenyan study (32).
There was overall improvement in the proportion of pregnant women who attended at least one ANC visit during pregnancy and those that completed the four focused ANC visits in the intervention sites. This is recommended towards identi cation of danger signs in pregnancy and planning for individual birth plan that contributes to maternal health monitoring, early identi cation maternal and fetal complication, and eliminating preventable maternal and perinatal morbidities and mortalities (50). World Health Organization recommends community-based health systems interventions with speci c focus on packages of interventions that include household and community mobilization and antenatal home visits (51). These recommendations aim to improve antenatal care utilization, women's experience of care and perinatal health outcomes, particularly in rural settings with low access to health services. This study demonstrated that referral of pregnant women for antenatal care alone is not enough. Follow up on defaulters referred for ANC helps to ensure that the pregnant women access skilled antenatal care attendance, a feature well illustrated in the intervention CUs and health facilities with eventual improvement on antenatal care coverage. Health messages on the need for ANC attendance by the trusted CHVs in the community plays an integral role in improving antenatal care coverage in these hard-to-reach communities.
The proportion of deliveries conducted by skilled birth attendants, now known as skilled health personnel, in the intervention sites doubled. This can be attributed to the focused demand creation activities to improve access to quality lifesaving interventions from frontline community providers. Through community mobilization and advocacy for skilled birth attendance by the CHVs during their household visits and sensitizations, it can be acknowledged that the community becomes responsive to the bene ts associated with skilled birth attendance as highlighted by a multi-country study on newborn care practices in Malawi, Nepal, Bangladesh and Uganda. In that study, women from Malawi, Nepal and Bangladesh who were visited by a CHV three or more times during pregnancy were more likely to report use of selected life-saving newborn care practice (24). Implementation of iCCM in addition to the MNH package with focused demand creation activities can improve access to quality lifesaving interventions anchored on the community health strategy (24,41).
Newborn visitation in the community in the early days of life is crucial in screening for life-threatening neonatal danger signs. Despite improvements in newborns visited at home within 48 hours of birth in the control community units, there was no change in the numbers with danger signs referred to health facilities for appropriate care. However, in the intervention sites, there was a signi cant increase in the numbers of newborns identi ed with danger signs and referred to health facilities for appropriate care although no change in the newborns visited within the same period. Community health volunteers trained in iCCM are equipped with knowledge and skills to identify danger signs that could be lifethreatening in U5s including the neonatal period and referred them to health facilities for appropriate care. It must be acknowledged that children face the highest risk of dying in their rst month of life, at a global rate of 18 deaths per 1,000 live births. A third of all neonatal deaths tend to occur on the day of birth and close to three quarters die in the rst week of life (52,53). Therefore, this evidence suggests that focusing on the critical periods immediately following birth is essential to saving more newborn lives for realization of SDG 3 target by 2030.
Immunization coverage increased signi cantly in all antigens at the intervention health facilities with no change at the control health facilities. This study demonstrated that despite a signi cant improvement in the number of immunization referrals made by the CHVs, it is the overall improvement in defaulter tracking to identify immunization defaulters and referral to the health facilities that is key in these hard-to-reach communities. This nding is similar to the study ndings evaluating the effect of enhanced defaulter tracing in the hard-to-reach communities in Migori county (54). This therefore means that efforts should be made to strengthen defaulter tracking mechanisms in the communities to improve immunization coverage and utilization in rural hard-to-reach populations. This is in line with the WHO's reaching every child/district (RED/REC) strategy that advocates for partnering with communities to promote and deliver immunization services which best t local needs and reach all eligible populations (11). Through integration of community health services, there is an opportunity to reduce inequities in immunization coverage, key areas for immunization programs in Africa underscored by the Global Vaccines Action Plan 2011-2020 (55).
Service utilization at both intervention and control sites is optimal based on the Pentavalent 3 and Measles/Rubella 1 drop-out rates of less than 10%. However, access to the services is suboptimal at the control sites (52.8% coverage for Pentavalent 1 vaccine), lower than the expected minimum of at least 80% as recommended by WHO (11). This implies that demand creation strategies must be intensi ed through the community health platforms and iCCM provides the earliest opportunity to achieve the expected coverages. However, it should be noted that there were over 100% immunization coverages in the post-intervention phase in the intervention health facilities for most antigens. It should be acknowledged that the facility expanded program on immunization (EPI) targets assigned by the MOH departments are usually mere estimates based on available demographic data that may not be very accurate hence performance as per the results may be over 100% for some of the antigens.
This study has strengths and weaknesses. To the best of our knowledge, this is the rst published study in Kenya on the effect of iCCM on maternal and newborn health. The choice to have a comparison group with similar characteristics as the intervention sites makes the ndings credible and change observed in the intervention sites is not just part of a secular trend -similar changes occurring in those kinds of facilities because of other contextual factors. This is also supported by the fact that all the study CUs had less than 40 percent coverage for 4th ANC and skilled birth attendance before the intervention. There were no changes in the proportion of CHVs reporting before and after the intervention in both sites. This therefore means that the selected intervention and control CUs had equal chances of over 80% of the CHVs reporting monthly. Besides, the fact that no stipends were provided by the project for the CHVs to carry out their community health responsibilities strengthens the sustainability of the model in hard-to-reach communities. Data quality challenges (timeliness, accuracy and completeness) are common with the use of DHIS2 data. The project team supported the subcounty teams conduct data veri cations with individual health facilities and CUs before reporting in the health information system. The small samples of intervention mean that these ndings should be interpreted within the local context.

Conclusion
Training and implementation of iCCM to build the knowledge and skills of CHVs in community case management using simple algorithms can lead to improved identi cation and treatment of malaria, diarrhoea, suspected pneumonia and malnutrition in the hard-to-reach communities. Beyond strengthening knowledge and skills in common childhood illnesses identi cation and management, iCCM broadly improved competencies and motivation of the CHVs which essentially contributed to improved access and utilization of skilled birth attendance (antenatal care, skilled births in health facilities, early neonatal care danger signs identi cation and referral for treatment) and childhood immunization. The CHVs are further involved in demand creation for maternal child health services, though delivery of targeted health messages at household level for improved health seeking behavior. Routine mentorship, coaching and support supervision of CHVs reinforced their knowledge and skills in identi cation of antenatal and immunization defaulters and referral to health facilities in hard-to-reach communities. Enhanced defaulter tracking by CHVs for antenatal and immunization care provides the platform under which facility utilization of health services can be strengthened.

Recommendations
Key recommendations from the ndings of this quasi-experimental study are: governments (national and county) should invest more resources in strengthening the community health systems so that CHVs are motivated and retained to carry out demand creation for maternal, newborn and child health and immunization services in hard-to-reach communities. This can be achieved through formulation of community health sensitive policies that will ensure funding, CHV motivation and provision of essential supplies. Regular mentorship and coaching should be integrated in community health interventions to reinforce the knowledge and skills acquired for realization of improved performance in community maternal, newborn and child case identi cation and management. In addition, community-facility linkages should be strengthened through data demand and information use to identify service gaps, coverage inequities and prioritization of strategies to improve community maternal, newborn and child health and immunization services utilization. No institutional review board determination was sought for the study because the Kenya DHIS2 data are publicly available (37), and the use of program reports in aggregate form was not human subjects' research (56).

Consent for publication
Not applicable Map showing the iCCM intervention and control health facilities in Migori County. Map is author's own generated using the QGIS software. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.