Secondary data analysis of the distribution and determinants of maternal and child health outcomes across Kenya’s 47 counties

Introduction

The end of the Millennium Development Goals (MDGs) era (2000–2015) witnessed significant improvements in maternal and child health (MCH) in sub-Saharan Africa (SSA). For example, the average maternal mortality ratio in the region declined by 45% since 1990 and under-five (U-5) mortality 2000–2015 reduced at a rate of 4.1% per annum, more than twice faster compared with the decade before.¹ The commencement of the Sustainable Development Goals (SDGs) is to build on the gains made in improving maternal and child health outcomes. Despite these efforts, there are still major inequalities in the outcomes within countries that are driven by specific determinants of health including education, income, age, housing conditions, food and nutrition, physical environment, ethnicity/race, and place of residence.^2–5 However, there are other macro- and micro-level factors that influence maternal and child health outcomes and perpetuate inequity between and within nations. These include health insurance coverage, population per health facility, public nurses/100000, doctors/100000 people, poverty headcount rate,⁶ and gender inequality.⁷ These factors play important roles in influencing outcomes such as maternal mortality, under-five stunting, under-five mortality, vaccine coverage (DPT3), and incidence of child diarrhea.⁶

Health inequities present an unjustifiable challenge to mankind because they are the result of economic and social conditions that ultimately determine human health and wellbeing.⁸ Hence, there is a greater justification for progressive and inclusive policies designed and implemented by governments to respond to a cross section of the economic and social determinants of health.⁹ However, such policies need to be formulated in ways that they best target areas with the greatest inequities while addressing the most impactful root causes for the benefit of the whole society.^10,11 This is in line with UN SDGs that pledged to “leave no one behind”.^12,13

This review set out to answer two important questions- the one is to demonstrate the how inequities around maternal and child health outcomes compare with each other (using the Lorenz curve), and the other to assess the extent to which economic and social determinants drive MCH outcome inequities in the Kenyan counties. The analysis is meant to guide better targeting of resources to achieve greater impact in MCH outcomes.

Kenya devolved the public health sector after the 2010 Constitution was inaugurated. The new constitution recognizes 47 semi-autonomous counties that are directly in-charge of financing and delivery of healthcare. Although there are reports^14,15 demonstrating how the counties differ in socio-economic status, disease burden and health outcomes, such reports often fail to show where the greatest equity and inequity lie, and what actually drives them. Friedman, Gostin et al¹⁶ state that aggregated data mask deep unfairness in the distribution of good health, much as a growing gross domestic product can mask highly unequal distribution of wealth.

Kenya has made significant strides in improving MCH outcomes^17,18 which include reductions incidence of diarrhea, increasing vaccine coverage, lower respiratory infections, increased skilled birth attendant, and increasing insurance coverage through the free maternal care program- Linda Mama. However, achieving equitable distribution of outcomes across the counties remains a challenge.¹⁹ This may explain the fact that MCH outcomes such as mortality remains unacceptably high^17,18,20,21 because resources may not be properly targeted. Knowledge about the drivers of inequity is of particular interest to policy makers for deciding where to intervene to improve, especially MCH, and move closer to the goals of universal health coverage (UHC).^22,23 Consequently, studies on the major variations in maternal and child health outcomes are necessary to inform policy and target resources in the most impactful areas. The choice of measurement variables in this review is based on key social and economic conditions considered to have high impact on maternal and child health outcomes.^6,7

Methods

This was a rapid review of national level data on maternal and child health. Secondary data sources were purposively selected for purposes of this study; i.e., only reports with quantitative information were targeted. The Institute of Health Metrics and Evaluation’s (IHME) local burden of disease²⁴ provided data on key health indicators including U-5 stunting, vaccine coverage, U-5 mortality, diarrhea and years of education for women 15 – 49 years of age for the year 2015. The second data source included summaries of various national survey reports in the County Health Fact Sheets by the Health Policy Project,⁶ 2014 and 2015 data. Data collected from this summary included the following: county government financing per capita (USD); insurance coverage by the National Hospital Insurance Fund (NHIF), number of people per facility by county (calculated), public doctor to population ratio and public nurses to population ratio. Data on maternal mortality rates were captured from a 2015 report by UNICEF.²⁵ Data on poverty headcount rates were sourced from the Open Institute (data for 2015).²⁶ Data for human development index (HDI) and gender inequality index (GII) were mined from a UNDP/Government of Kenya (GOK) report (2017).⁷ The data represented all 47 counties in Kenya (Figure 1).

Figure 1. Map showing the 47 Kenyan counties.

(Source: Wikipedia Copyrighted free use)²⁷

A Lorenz curve and Gini Index were constructed to assess the inequalities within counties based on the following indicators recommended by the Health Policy Project as the key measures of inequity between counties⁶: immunization coverage (DPT3), U-5 mortality, child diarrhea, maternal mortality rates (MMR), county government financing per capita (converted to 2015 USD), insurance coverage by the National Hospital Insurance Fund (NHIF), poverty head-count (food), public nurses to population ratio and number of people per facility.

In addition, the distribution of selected variables was demonstrated using the principal component analysis (PCA). Finally, a multivariate whole model stepwise regression analysis was conducted to determine the role of the following key independent variables: years of education for women 15 – 49 years, county health financing per capita, NHIF insurance coverage, population per facility, public nurses/100000, doctors/100000 people, poverty headcount rate, and gender inequality index (GII) in determining a number of health outcomes. The outcomes are: stunting (U-5), U-5 mortality, MMR, vaccine coverage (DPT3) and incidence of diarrhea.

Results

The Lorenz Curve (Figure 2) shows levels of inequality by various indicators.

Figure 2. The Lorenz curve.

Vaccine coverage with a Gini Index (GI) of 0.063 (Table 1) is the most equitably distributed outcome in the country followed by U-5 mortality rate (GI=0.124). Both vaccine coverage and U-5 mortality rate are closely related; i.e. low vaccine coverage potentially leads to high U-5 mortality rates. Maternal health has the highest inequity (GI=0.381) followed by of the distribution of nurses in the public health sector in each county (GI=0.317). Both vaccine coverage and maternal mortality ratios are often dependent on level of education of women of reproductive age (15 – 49 years). County government funding of health per capita also shows wide variations between counties (GI= 0.230) suggesting different levels of expenditure and prioritization of the health sector by each of the 47 county governments.

Figures 3a to 3i confirm how some of the outcomes and the determining factors are distributed across the 47 counties.

Figure 3.  

a: Vaccine coverage (DPT3). b: Under-5 mortality rate (U5MR). c: Maternal mortality rate (MMR). d: Population per health facility. e: Under-5 stunting. f: Incidence of diarrhoea. g: Gender inequality index (GII). h: Educational achievement. i: Coverage by the NHIF.

The figures show that vaccine coverage (Figure 3a) and U-5 mortality (Figure 3b) are the most evenly distributed across the counties. On the other hand, MMR (Figure 3c) and population per public sector health facility (Figure 3d), are the most unevenly distributed. The distribution of mean U-5 stunting (Figure 3e), incidence of diarrhea (Figure 3f), and gender inequality index (GII) (Figure 3g) tend to have normal distributions across the 47 counties, suggesting that all counties require the same level of attention when it comes to these variables. The educational attainment of women of reproductive age (Figure 3h), and coverage by the public insurance agency, the NHIF (Figure 3i) tend to have similar distribution suggesting that a correlation between insurance coverage educational attainment.

Table 2 presents the results of a multivariate whole model stepwise analysis.

Discussion

Maternal mortality rate (MMR) is globally the most inequitably distributed health outcome indicator.¹ For instance, in SSA about 1000 women die per 100,000 live births, compared to 24 deaths per 100,000 live births in European countries. In Kenya, this type of inequity is worse at the sub-national level where more than 2000 deaths per 100,000 live births are reported in some counties against less than 200 deaths per 100,000 live births in others.²⁵ The Kenya reproductive, maternal, neonatal, child and adolescent health (RMNCAH) investment framework aims to ensure “there are no preventable deaths of women, new-borns or children and; no preventable still-births, where every pregnancy is wanted, every birth celebrated and accounted for …”.²⁸ This investment framework is strengthened by the assurance that as the country makes progress towards UHC, health inequities including maternal mortality, U-5 mortality, immunization coverage, child stunting, and incidence of diarrhea across the country, would diminish.

The results from this study reflect the global trends in MMR as the most inequitable outcome where, across the counties, the distribution matrix showed that most of the burden rests on a few counties, particularly those in the remote northern region, and the lake region which has a high HIV/AIDS burden. It needs emphasis that populations experiencing vulnerabilities and poor access to healthcare including those in urban slums, marginalized populations, refugees, and displaced people, need to be brought into the health system,²⁹ to address health inequities.

Whilst a positive outcome in MMR is strongly and expectedly linked with the level of education of women of reproductive age, perhaps one of the most interesting findings is the significant roles that gender inequality (GII) and congestion at public sector health facilities play in increasing MMR. Investments targeting both areas would be recommended, based on the results, to help address MMR outcomes particularly in northern Kenya, and the lake region. The GII is particularly of interest as it has significant influence on all the outcomes. It is linked with women’s empowerment both culturally and economically to be able to make decisions on when and how to access care, and whether or not to get pregnant. Recent reports indicate that women in many parts of Kenya are significantly disempowered and lack decision making power.³⁰

As the findings suggest, outcomes such as full immunization coverage with DPT3, is already trending toward the line of equity, and as the results show, is driven largely by improved levels of education of women of reproductive age, reduced congestions at health facilities, and public health insurance coverage, which together tend to improve access to health services. Aalemi et al.,³¹ had similar findings in their study of factors driving immunization coverage in Afghanistan. However, despite the progress made in DPT3 coverage, a few counties including Mandera (36%), West Pokot (54%), Wajir (60%), and Migori (65%) record the lowest coverage, and therefore, require research and more focused intervention than currently practiced to improve their immunization coverage.

The U-5MR in Kenya is also relatively equitable even though some counties such as Homa Bay (75 deaths per 1000 live births), Migori (73/1000), Busia (63/1000), Kisumu (58/1000), Vihiga (58/1000) and Kakamega (56/1000) are below the national average (44/1000). It is noteworthy that all these counties registering unacceptably high U-5MR are based in the western/lake region of the country where the prevalence of HIV/AIDS is quite high. There is a direct correlation between prevalence of HIV/AIDS and MCH outcomes.^32–34 All things constant, it could be fair to conclude that child survival in these counties and a few others, would be greatly improved with more interventions focused on reducing the incidence of HIV infections. In addition, authors such Pillai and Maleku³⁵ suggest that other factors including availability of diverse sources of water, age at first birth, wealth status and urban residence are critical to improvements in child health outcomes. All these factors need to be considered in designing interventions to improve under-5 mortality in the affected counties.

There are counter-intuitive results; e.g., increasing poverty head count rate relate positively with reducing MMR. This can be attributed to national interventions such as the Linda Mama health insurance initiative that provide free maternal and child health services and particularly targets vulnerable women. Such outcomes can also be partially attributed to efforts by organizations such as UNICEF that target improvements in MCH in poor, hard-to-reach counties. These interventions have tended to improve equity and it is evident from the results that the NHIF coverage for example, tends to be one of the closest to the line of equity because all MCH services are eligible for coverage. In addition, the results show that counties with less educated women record lower stunting rates, which indicates the role of external interventions in improving child nutrition in marginalized counties. Programs such as Nutrition and Health Program Plus, and those under Nutrition International, tend to target highly vulnerable communities to eliminate nutrition-related stunting, and promote child survival.

Counter-intuitive results are not a rarity. In Ghana, research shows that achieving UHC was associated with increased unmet need for family planning.³⁶ Moreover, fertility levels associated remained high, with limited improvements among young women only.³⁶ As such, UHC is insufficient for achieving socially accessible family planning care unless the social constraints to the reproductive autonomy of women are addressed.

Beyond the macro-level data, it is important, as recommended by Achoki et al.,¹⁵ for decision makers to use evidence on burden of disease at sub-national level to target health interventions and address inequities in MCH outcomes. In addition, the preponderance of inequities in the distribution of MMR across the 47 counties of Kenya calls for a re-think of investment decisions in MCH services in Kenya. One recommendation is the use of technologies such as Lumify because it is known to improve MCH outcomes, particularly in resource-poor, hard-to-reach settings. In Rwanda for example, Royal Philips and PURE (Point-of-care Ultrasound in Resource-limited Environments) rolled out a unique tele-ultrasound mentorship program to provide much needed diagnostic ultrasound training to health workers in Rwanda.³⁷

Thus far, the analysis reveals some of the most inequitably distributed socioeconomic conditions, and outcomes among counties, which are often not apparent to decision makers. In this context, health interventions need to have an equity focus that supports the delivery of MCH services through the lens of intra-county variations in outcomes such as MMR, and socio-economic conditions including GII, congestion at health facilities, availability of nurses in public sector health facilities, burden of infectious disease, and poverty headcount rate. Promising approaches and interventions to reduce inequity do exist. These approaches include UHC whose focus should be on reducing geographical, economic, sociocultural, and gender barriers to healthcare.