Introduction

Under-5 mortality (U5M) has been reduced globally from 93 deaths per 1,000 live births in 1990 to 38 per 1,000 live births in 2019. Nevertheless, 1 in 13 children continue to die before their fifth birthday in sub-Saharan Africa. According to the World Health Organization, most of the deaths are due to preventable and treatable diseases [1].

Liberia ranks among the worst nations globally in child health outcomes, with U5M rate estimated at 93 per 1000 live births [2]. Remote areas are facing the greatest burden of mortality because of poor access and utilization of health care services [3]. The primary causes of U5M in the region are malaria, diarrhea and acute respiratory infections (ARI) [4, 5], which are all preventable and treatable with health interventions that are feasible for implementation in low-income countries [6].

Many countries have adopted integrated Community Case Management (iCCM) strategies to address the causes and reduce U5M. Community health workers (CHWs) are at the core of many iCCM strategies. They act within their local community and provide services such as treatment of uncomplicated childhood diseases and referring complicated cases to health facilities [7]. iCCM interventions have been shown to reduce U5M due to an increased treatment of malaria, ARI and diarrhea [8–14].

In June, 2016, the Liberian Ministry of Health (MOH) initiated a national professionalized, incentivized, and clinically supervised community health worker cadre known as Community Health Assistants (CHA). These CHAs are embedded within the public-sector health workforce, with the goal of extending the reach of the country’s primary health care system to an estimated 1.2 million people through a standardized national community health model. The CHAs provide a package of health care services and epidemic surveillance within communities and households on an equitable basis.

Although numerous studies have demonstrated improvements in child outcomes with iCCM interventions, evidence around the effects of government-run, CHW-delivered healthcare on child mortality in rural areas of sub-Saharan Africa is limited. Our central hypothesis is that the National Community Health Assistant (NCHA) program in Liberia reduced U5M in rural Liberia. To test this hypothesis, we will conduct a large, population-representative survey across a major county in Liberia, and use a mixed effects Cox proportional hazards model to retrospectively compare mortality in the pre- and post-implementation periods in order to estimate the effect of the CHA program on U5M.

Method

Community health worker intervention

The program.

As part of the post-Ebola investment plan for building a resilient health system in Liberia 2015–2021, the MOH prioritized the establishment of a robust cadre of community health workers in remote communities that were situated more than 5 km from the nearest health facility. Through the NCHA program, members of remote communities select their CHA. Those selected by their communities have to fulfil eligibility requirements, including being permanent residents in the community, aged 18 to 50 years, possess basic writing and math skills, be physically and mentally fit and must be fluent in the dialect spoken in their community. Once selected, CHAs complete a standardized training curriculum in community health and surveillance, child health and maternal and neonatal health. Services are covering treatment of diarrhea, ARI, and malaria following iCCM protocols, as well as family planning and birth preparedness. Since 2020 COVID-education and -referral has also been included in the CHAs work. Everyone in the communities more than 5 km from the nearest health facility is eligible to use the program directly or indirectly. CHAs are compensated with a monthly cash incentive of US $70 for about 20 hours of work per week. Program implementation is preceded by community engagement activities, which involves obtaining the support and advice of community leaders.

Implementation.

The Liberian NCHA program was initiated in a stepwise manner across the eight districts of Grand Bassa County by the government of Liberia beginning in September 2018. It is planned to conclude in June 2022 (Fig 1). We will take advantage of this staggered implementation to form a study design similar in structure to a stepped wedge cluster randomized trial. However, the order of implementation was not randomized; judged by the Liberian MOH districts with more need for health interventions were prioritized in the intervention roll out. The staggering of implementation was conducted for operational and logistical reasons, rather than to facilitate an evaluation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. National Community Health Assistant (NCHA) program implementation schedule in Grand Bassa, Liberia.

Source: Dominik Jockers based on data from Humanitarian Data Exchange [16] under a CC BY license.

https://doi.org/10.1371/journal.pone.0272172.g001

Discussion

This study will estimate the impact of Liberian NCHA program including iCCM and perinatal care support in remote Liberia on U5M. The staggered implementation of scale up within Grand Bassa offers a unique opportunity to evaluate the effectiveness of this program on a population scale. The analysis will allow us to evaluate the implementation of a critical healthcare delivery program by the Liberian Ministry of Health on U5M within the target communities.

There are multiple unique aspects of this study which will enhance both its scope and generalizability. The total population census with a program coverage of 1,753 communities and 23,702 households in the Grand Bassa county of Liberia, we will take advantage of a notably large sample size. The fact that the NCHA program is being implemented staggered across eight study health districts in four phases also gives us the opportunity to use a mixed effects Cox proportional hazards model with a random effect by community and district fixed effects to compare the intervention effect both within and between geographic regions. Finally, the use of U5M as the primary outcome, as opposed to intermediate health care delivery indicators and process measures, will allow us to measure the population level impacts of such a CHW-based child health care program at scale.

There are limitations to the design of this study. First, the order of implementation by district was non-randomized. Because the districts with the worst health outcomes were targeted by the intervention first, our estimate of the intervention effect estimate may be biased. Nonetheless, the stepped approach to implementation, availability of pre- and post-implementation data in the districts, and our use of district fixed effects and community covariates all enhance the causal inference of this study design. Second, the random effects at the community level rely on an assumption of minimal correlation of unobserved confounders with NCHA program exposure. Third, the data for our outcome relies on endline survey data, thus selective migration by intervention implementation date potentially causes bias in the main estimates. To address this, we will include a sensitivity analysis that limits the analysis only to those mothers who report living in the implementation area for the full study period. Third, U5M is assessed in the survey design by recall of such events in the last four years. There may be bias associated with mothers not remembering the correct birth/death dates of children. We will assume that this recall bias (and related biases, e.g. mothers’ reluctance to discuss child deaths) is minimal and does not differ significantly by geography or intervention status date. To mitigate to this source of bias and increase the validity of death reporting, we will employ a landmark event calendar, which has been previously shown to improve precision of event reporting [17]. Finally, the fidelity of implementation is not a component of this impact study, but is the focus of ongoing program evaluation to measure dose, coverage and effect of contextual factors including health facility readiness and the COVID-19 pandemic.

Prior studies have demonstrated improvement in process measures from implementation of small scale governmental CHW programs in similar settings including Liberia [18, 19]. Whereas Nyqvist, Guariso [11] were able to randomize program exposure across communities, the program was different from our evaluation in that it is was operated by a non-governmental organization. Our study will add to the literature by providing evidence from a large scale, governmental-led CHW program.

In summary, this study will provide invaluable insight on the impact of large-scale, publicly operated CHW programs on child health. Results will be directly applicable both to the Liberian Ministry of Health, as well as to other governmental entities supporting similar CHW-based rural child health programs in the region. Because such rural areas are known to face the largest U5M burden in many sub-Saharan countries, our study results will directly respond to the need to understand the optimal methods of improving health care among many of the most vulnerable populations in the region.

Supporting information