Drivers of Health in sub-Saharan Africa: A Dynamic Panel Analysis

Abstract

Globally, life expectancy increased while infant mortality reduced substantially between the 19^th and late 20^th century. Although there is relatively mature literature on the drivers behind these gains in life expectancy and reductions in infant mortality, there is a dearth of studies that focus on the drivers of health in sub-Saharan African (SSA) countries. The few studies that do exist do not account for a broader array of determinants such as the quality of access to health services and institutional quality which may have important implications for health policy. We contribute in filling this gap by estimating the effect of a rich set of socio-economic, environmental, health system and lifestyle factors on life expectancy and infant mortality using a panel of 30 sub-Saharan African countries. We employ a dynamic Generalized Method of Moments (GMM) estimator and focus on the period between 1995—2014.

Our findings show that increases in health expenditure, educational attainment, and health care access quality are associated with increases in life expectancy and reductions in infant mortality. Higher HIV prevalence rates are associated with reductions in life expectancy whereas urbanization, per capita income growth and access to clean water are positively associated with life expectancy.

We conclude that increases in life expectancy and reductions in infant mortality can be accelerated by paying particular attention to interventions linked to these drivers, including, health care access quality.

Highlights

• •Drivers of health in Sub-Saharan Africa depend on the health measure used
• •Education has a greater impact on infant mortality compared to life expectancy
• •Better health-care access quality reduces infant mortality rates in the region
• •Urbanization, Per-capita income growth, access to clean water have a positive effect on life expectancy levels.

1. Introduction

1.1. Background

The economic prosperity of nations is strongly linked to their stock of human capital, consisting of education and health [5]. Two of the most commonly used measures of population health are life expectancy and infant mortality [32]. Life expectancy captures the health impact of the disease environment since premature death is the most significant and observable impact of disease [43]. On the other hand, infant mortality is an indicator of the state of the health care system and can signal the need for improved health care services [34].

From the early 19^th century to the late 20th century, the world experienced more significant improvements in life expectancy and reductions in infant mortality than in any other period in global history [39]. However, the distribution of these gains across world regions have been uneven. For instance, while the global average life expectancy more than doubled between 1900 and 2000 [35], a regional breakdown of these gains reveals that there have been substantial inequalities in these population health gains between poor regions and those that are better off [35]. Moreover, the gap between average life expectancy in the more developed North America Region and the least developed Sub-Saharan Africa (SSA), region has fallen by only 28 per cent (from 27 years to 19.1 years) between 1970 and 2015 [46]. To better understand why these improvements are not evenly spread between high income and low- and middle-income countries (LMICs), especially sub-Saharan Africa, we must address two important questions. First, what factors are driving these improvements in population health and second, are the drivers of population health different between high income and low income countries, particularly sub-Saharan Africa?

While the literature on the drivers of life expectancy and infant mortality at the macro level is mature in high-income countries, there is a dearth of studies in sub-Saharan Africa. Macro-level determinants are important and often compliment micro-level determinants, given the importance and pervasiveness of global health interventions. The rise in the global health agenda implies that health interventions at the macro level may not be restricted to national borders as they are of global concern. A more recent example is funding for HIV/AIDS response or health interventions related to climate change. A few exceptions of studies that have looked at Macro level determinants in sub-Saharan Africa are Fayissa [9], Fayissa and Gutema (2005), and Ogunleye [30]. Important factors identified to determine life expectancy and child mortality include income per capita, education, food availability, and alcohol consumption [9,10,30]. However, the biggest challenge with this literature is that it does not carefully account for endogeneity, arising from omitting important variables in the health production model. There are a lot of factors that drive health and it may be impossible to include all these in the models. If these factors are correlated with included variables, then they may bias the results and fundamentally change the conclusions. We contribute to filling this gap by examining the impact of socio-economic, environmental, health system and lifestyle factors on life expectancy and infant mortality for 30 Sub-Saharan African countries for the period between 1995 and 2014. Importantly, we allow for dynamic effects in life expectancy and infant mortality and account for endogeneity by employing an instrumental variable approach.

In high-income countries, there is a rich literature attempting to shed light on the macro determinants of life expectancy and infant mortality. Macro-level factors such as income growth, improved food security and advances in technology and medicine have traditionally been intrinsically linked to improvements in population health outcomes [14,22,38]. The literature shows that increases in income and improvements in individual diets and later, generation and diffusion of specific inexpensive technology and knowledge have been the most important drivers of reductions in mortality and increases in life expectancy in the 19^th and 20^th centuries [22,29]. Further, in the United States of America, 60% of all premature deaths were due to behavioural, social, and environmental circumstances, which rely on public health interventions, while medical care only prevented 10% of premature deaths [28,37].

Our study contributes to the body of literature on the factors influencing population health in sub_Saharan Africa in at least two ways. First, we adopt a dynamic panel data model using a difference Generalized Method of Moments (GMM) estimator, which uses internal instruments and allows us to control for endogeneity in the model without succumbing to the pitfall of selecting weak theoretical external instruments. Second, we include health system variables that influence population health such as measures of the quality of health care and political-institutional quality which have been absent from previous studies on the sub-Saharan Africa region. Including health system indicators provides an additional dimension to the study which could have important health policy implications for the region.

By assessing evidence for sub-Saharan Africa, where institutional factors may differ from other regions and give rise to differences in the relative importance of drivers of population health, we provide evidence for policymakers at the national level and global level.

1.2. Context

After decades of economic stagnation and widespread structural reforms, several countries in sub-Saharan Africa experienced substantial economic growth in the early 2000s with growth rates peaking at 11 per cent in 2004 [21]. Investment in health care also increased considerably with total per capita health expenditure rising from US$41.6 in 1995 to an average of US$ 98.2 by 20141 [46]. There were also remarkable improvements in health indicators, particularly child and infant mortality, which fell by over 50 per cent between 1990 and 2015 [41]. Similarly, life expectancy at birth went from 49.8 years in 1990 to 59.9 years in 2015, a gain of slightly over 10 years in 25 years [46].

While these recent successes in improving health outcomes indicate that the region is on the path towards better health outcomes, the improvements are slower than other regions of the world and considerable inequalities within countries exist. For policymakers and actors in global health to adequately design policies and interventions to accelerate improvements in life expectancy and infant mortality, they must acquire a comprehensive understanding of the influence of various macro-level determinants of health on these health indicators. Understanding the effects of macro-level determinants could provide policymakers with the evidence that they need to design targeted interventions and policies to rapidly improve health outcomes in the population.

2. Methods

2.1. Data and Variables

We use annual data from 302 sub-Saharan countries for the period between 1995 and 2014, which gives us a total of 600 observations. The number of countries and chosen proxies was determined by data availability, specifically, eliminating countries which had extensive missing data. Data was obtained from multiple sources including the World Development Indicators (WDI) of the World Bank [47], Global Health Observatory (GHO) database of the World Health Organization (WHO) [45], Human Development Reports Database of the United Nations Development Programme (UNDP) [40], and Global Burden of Disease (GBD) database of the Institute for Health Metrics (IHME) [20]. The World Development Indicators are compiled by the World Bank using officially recognised international sources and represent current and accurate development data at the national, regional and global level. The World Health Organization compiles its data from multiple sources such as household surveys, routine reporting by health services, censuses and disease surveillance systems and is a reliable source of aggregate data at the national and global level. Appendix A provides a summary of the variables included in the model along with their measurement, mean and standard deviation. Further, a few notes on each variable and its data source is included.

The determinants of health were grouped into socio-economic, environmental, health system and lifestyle variables. For the socio-economic variables, ‘average years of schooling’ was used as a proxy for education attainment whereas ‘GDP per capita’ and the ‘food production index’ measured average real aggregate income and food availability respectively. The environmental indicators included the ‘percentage of the population with access to an improved water source’ which served as a proxy for access to clean water and sanitation and the ‘annual percentage change in the urban population’ which was a proxy for the urbanization rate.

The variable ‘Health expenditure’ which represented investments into health and the ‘Healthcare Access Quality Index’ which is a proxy for health care quality (see [4]) accounted for the health system factors associated with health outcomes. ‘Alcohol consumption’ in litres per capita represented lifestyle factors that influence population health. Given the significance of the HIV/AIDS pandemic on population health in sub-Saharan Africa, the model will control for the influence of HIV during the 20 years using the HIV prevalence rate among individuals aged between 15 and 49 years. We also controlled for the political institution quality using the ‘Political institutional quality index’ (see [23,24]).

2.2. Model

Economic analyses on the determinants of cross country differences in health outcomes usually take the form of either multivariate panel data or cross-sectional regression models. Because we are more interested in the effect of different factors on health outcomes over time that allow past levels of health to determine current levels, we adopt a dynamic panel data regression model.

To show the dynamic panel model that we are interested in, we start with a simple health production function specified as;

$$ln{\mathit{Health}}_{\mathit{it}}={\beta }_0+{\beta }_{1}ln{\mathit{GDP}}_{\mathit{it}}+{\beta }_{2}ln{\mathit{HE}}_{\mathit{it}}+{\beta }_{4}ln{\mathit{FP}}_{\mathit{it}}+{\beta }_5{\mathit{HAQI}}_{\mathit{it}}+{\beta }_6{\mathit{HIV}}_{i,t}+{\beta }_7{S}_{i,t}+{\beta }_8{\mathit{UP}}_{i,t}+{\beta }_9{\mathit{WS}}_{i,t}+{\beta }_{10}{\mathit{AL}}_{i,t}+{\beta }_{11}{\mathit{PIQ}}_{i,t}+{\epsilon }_{\mathit{it}}$$ (1)

Where the subscript ‘i’ represents country and t denotes time measured in years. Thus, lnHealth_it is the natural log of either life expectancy or infant mortality for the i^th country at time t, GDP_it is real GDP per capita, HE_it is total health expenditure, FP_it is the food production index, HAQI_it is the healthcare access quality index, HIV_{i, t} is the HIV prevalence rate, S_{i, t} represents average years of schooling, UP_{i, t} is the percentage change in the urban population, WS_{i, t} is the percentage of the population with access to a clean water source, PIQ_{i, t} is the political-institutional quality index and AL_{i, t} denotes annual per capita alcohol consumption and ε is the error term.

While the simple regression model of the production function in Eq. (1) might provide some insight into the relationship between health and the included variables, it is plagued by endogeneity arising from the correlation between the unobservable characteristics of the population, e.g., cultural factors and attitudes in a particular country, and some of the independent variables such as education and income per capita. Additionally, the health production function might be better represented as a dynamic model since population health in the next period might be dependent on the stock of health in this period [16]. To address these issues, Eq. (1) can be transformed into a dynamic panel regression model by taking first differences and including time dummies to control for trend effects;

$$∆{\mathit{Health}}_{\mathit{it}}=\alpha ∆{\mathit{Health}}_{i,t-1}+\mathit{\beta ∆x}{\prime }_{\mathit{it}}+\mathit{\gamma }{\mathit{D}}_{\mathit{t}}+∆{\epsilon }_{\mathit{it}}$$ (2)

Where; ∆Health_it is the first difference of the log of life expectancy or infant mortality of country i at time t, ∆Health_{i, t−1} is its one-period lag, ∆x′_it is a vector of the first differenced explanatory variables presented in Eq.(1), D_t is the vector of time dummies and ∆ε_it is the first differenced error term for country i at time t. The error term is given by;

$${\epsilon }_{\mathit{it}}={\mu }_i+v_{\mathit{it}}$$ (3)

Where μ_i~iid(0, σ_μ²) and v_it~iid(0, σ_v²) are independent amongst themselves and each other.

The estimation of Eq. (2) poses two main challenges. First, the inclusion of the lagged value of 'Health' means Ordinary Least Squares or Generalized Least Squares (GLS) (such as fixed effects estimation) will not be consistent since by construction, Health_-1 and ε_it are correlated [3]. Secondly, there are several time-varying and fixed un-observable variables that may be correlated with both Health_it and the variables included on the right-hand side. While fixed effects estimation, as many studies have done, is likely to take care of unobserved fixed effects, the problem of endogeneity persists because of the presence of dynamism—the lagged health stock—and unobserved time-varying omitted variables.

To account for these issues, we adopt a difference Dynamic Panel Data (DPD) estimator based on the Generalized Method of Moments (GMM) as expressed by Arellano and Bond (1990). The DPD GMM is particularly suited for our data as it performs well with a relatively large N and small T and thus fits well with our data that has N=30 and T=20. The GMM estimator is an instrumental variable (IV) estimator where endogenous variables are instrumented using their lagged values (internal instruments) and/or other variables (external instruments) that are only correlated with Health_it through the endogenous variables [2].

For the estimation, the study will employ the following specification of the Dynamic Panel Data model;

$$∆ln{\mathit{Health}}_{i,t}={\beta }_1∆ln{\mathit{Health}}_{i.t-1}+{\beta }_2∆ln{\mathit{GDP}}_{i,t}+{\beta }_3∆ln{\mathit{HE}}_{i,t}+{\beta }_4∆ln{\mathit{FP}}_{i,t}+{\beta }_5∆{\mathit{HAQI}}_{i,t}+{\beta }_6∆{\mathit{HIV}}_{i,t}+{\beta }_7{∆S}_{i,t}+{\beta }_8{∆\mathit{UP}}_{i,t}+{\beta }_9{∆\mathit{WS}}_{i,t}+{\beta }_{10}∆{\mathit{AL}}_{i,t}+{\beta }_{11}∆{\mathit{PIQ}}_{i,t}+\mathit{\beta }{\mathit{D}}_{\mathit{t}}+{∆\epsilon }_{\mathit{it}}$$ (4)

Where D_t denotes the vector of time dummies that are included as controls in the model and ε_it is the error term as defined in Eq. (3). For this model, internal instruments will be used for the lagged dependent variable, average years of schooling, per capita GDP and the food production index as they are expected to be endogenous regressors.

We estimate Eq. (4) separately for life expectancy and infant mortality.

3. Results

The results of the model estimation are presented in Table 1. Column 1 of Table 1 shows results for the model using life expectancy whereas column 2 depicts the results for the model using infant mortality.

Table 1.

Estimation results of the Health Production Model with Life expectancy and Infant Mortality

Variables	Coefficient Estimates Dependent Variable- Log Health
	Life Expectancy	Infant Mortality
lnHEALTHt−1	0.854*** (0.0308)	0.850*** (0.0363)
ln GDP per capita	0.045** (0.0202)	0.032 (0.0293)
ln Food Production index	-0.001 (0.0103)	-0.021 (0.0158)
ln Health expenditure	0.013*** (0.0025)	-0.007* (0. 0039)
Political Institutional Quality Index	0.001 (0.0196)	-0.035 (0.0239)
HIV prevalence	-0.004*** (0.0012)	-0.001 (0.0042)
Healthcare Access Quality Index	-0.002* (0.0009)	-0.004** (0.0020)
Average Years of Schooling	0.009* (0.0052)	-0.037*** (0.0117)
Urban Population growth	0.002** (0. 0008)	-0.013 (0.0140)
Access to Clean Water Source	0.0001 (0.0004)	-0.001 (0.0007)
Alcohol Consumption	0.0005* (0.0003)	0.001*** (0.0003)
Time Dummies	Yes	Yes
First Order Serial Correlation	2.08 (0.038)	- 0.71 (0.471)
Second Order Serial Correlation	1.93 (0.053)	-0.88 (0.377)
Sargan Test	12.59 (1.000)	17.15 (1.000)
Wald Test	1.29e+06	296540.35
Number of Obs.	520	491
Number of Instruments	99	87

Note; * p < 0.1, **p < 0.05 and ***p < 0.01, Standard Errors in Parentheses for Coefficients: P-values in parentheses for Sargan test and Serial correlation tests

Life Expectancy – Health stock is measured by Life expectancy

Infant Mortality – Health stock is measured by Infant Mortality

Beginning with the life expectancy, we find that higher per capita GDP, nominal total health expenditure, years of schooling, and urban population growth is associated with increased life expectancy. Specifically, a 10 per cent (or US$225.1 at the sample mean) increase in per capita income is associated with a 0.4 per cent increase in life expectancy or 2.5 additional months added to life expectancy at the sample mean. A 10% rise in the urban population growth rate (0.4 per cent at the sample mean) is associated with a 2 per cent rise in life expectancy or an additional year of life expectancy at the sample mean.

The results further indicated that a one year gain in average years of schooling would result in a 0.9 per cent rise in life expectancy. At the sample mean, this translates into an additional six months of life expectancy. As expected, increases in HIV prevalence rates are associated with reductions in years of life expectancy with a 10 per cent increase in the HIV prevalence rate associated with a 4 per cent decline in life expectancy or a loss of 2 years of life expectancy at the sample mean. This finding was significant at the 1 per cent level of significance. The positive coefficient on the lagged dependent variable indicates that life expectancy depends greatly on the previous period life expectancy. Although the signs for alcohol consumption and health quality were as expected when we ran the model with infant mortality as our population health measure, they are not expected in the model for life expectancy. Nonetheless, the result is only significant at the 10 per cent level.

For the infant mortality model, our results suggest that a unit increase in the Health Care Access Quality Index is associated with a 0.4 per cent decline in the infant mortality rate whereas an extra litre of annual per capita alcohol consumption is associated to a 0.1 per cent increase in the infant mortality rate with both findings significant at the 5 per cent level of significance.

The estimation results further show that an additional year of schooling for individuals aged 25 years and above was associated with a 3.7 per cent ( or 2.7 deaths per 1000 live births at the sample mean) reduction in the infant mortality rate and this was significant at a one per cent level of significance. Also, a percentage rise in the real total health expenditure is associated with a 0.007 per cent decline in the infant mortality rate. Unlike its effect on life expectancy, the effect of health expenditure on infant mortality is quite small and only significant at the 10 per cent level. Similar to life expectancy, the coefficient on the lagged infant mortality indicates that infant mortality is greatly dependent on previous period infant mortality.

Overall, health expenditure, health care access quality, alcohol consumption and the average years of schooling were the main significant determinants of population health for the region in the sample period. However, we note that the magnitude of the effect of health expenditure on life expectancy was greater than that on infant mortality. On the other hand, the association between education and infant mortality was stronger than that life expectancy and infant mortality. The impact of per capita GDP growth on health was dependent on the measure of population health used, while it significantly contributed to the level of life expectancy it was not a significant determinant of the infant mortality rate.

3.1. Robustness of findings

The post-estimations tests of the model are presented in the last six rows of Table 1 above. The consistency of the AB GMM estimator depends on the assumption that there is no second-order serial correlation for the disturbances of the first differenced equation (Baltagi, 2013). As such, the Arellano and Bond Test for serial correlation was used to test for both first and second-order serial correlation. From the results of the tests, we do not reject the null hypothesis that there is no second-order serial correlation at the 5 per cent level of significance for both models. This implies that there was no second-order serial correlation in our model.

Table 1 also reports the results of the Sargan test of over-identifying restrictions. This test verifies the validity of the instruments used in the analysis [36]. The variables that were considered to be endogenous and thus instrumented were average years of schooling, the food production index, GDP per capita and life expectancy (infant mortality for model 2). The instruments used for each variable are presented in Appendix A. The Sargan test does not reject the null hypothesis that the over-identifying restrictions used in both models are valid at the five per cent level of significance. This means that the set of instruments used in both models were valid and there was no over-identification arising from too many instruments.

The analysis also included time dummies to control for year effects which capture the influence of aggregate time-series trends, these are explicitly stated in Appendix A. Further, the Wald test rejects the null hypothesis that the impact of all the variables on life expectancy and infant mortality respectively was equal to zero.

4. Discussion and Conclusion

The central message from the results is that at the macro-level, health system factors and select socio-economic factors play a major role in determining life expectancy and infant mortality in Sub-Saharan Africa in the short term.

The positive impact of education on population health coincides with both the theoretical literature and previous empirical studies [6]. A natural explanation for this effect is that even basic improvements in education and literacy are associated with improvements in health outcomes through improved access to health care and higher earnings which could translate in higher living standards [12,18]. It is noted however that the short-term impact of education is much greater on infant mortality than life expectancy. The reason for this is that the effect of education on infant mortality may be more apparent in the short-term than that on life expectancy. A plethora studies have found direct causal links between education and reductions in infant and child mortality in the short-term whereas the influence of education on life expectancy tends to be long term and varies widely across populations [1,25]. Thus it follows that accelerated reductions in infant mortality in the short term can be achieved by educational policies aimed at increasing average years of schooling and improving literacy rates. Education policies such as the Free Primary Education Policies adopted by the majority of sub-Saharan Africa countries [31] can be extended to encompass secondary education as empirical evidence suggests that health gains increase with levels/years of education [12].

From our findings, health expenditure emerges as one of the main contributors to improvements in population health with its effect being greater for life expectancy than infant mortality. Increased health expenditure over the past few decades has been associated with an increase in the number of health facilities and infrastructures as well as improvements in health care services and delivery [8]. Currently, as the world moves towards universal coverage of health care sub-Saharan African countries are beginning to adopt National Health Insurance schemes to replace previous health financing policies such as the removal of user fees [15]. To derive optimum benefits in terms of improvements in life expectancy and infant mortality from these schemes, there is a need to ensure adequate and consistent funding of the schemes.

An interesting finding from the study was related to the relationship between health care quality, annual per capita consumption of alcohol and health measured by life expectancy. While it is assumed that alcohol consumption would reduce life expectancy and health care quality would have a positive impact on life expectancy, this was not the case for this study. Previous studies done for sub-Saharan Africa countries indicate that alcohol consumption negatively impacts life expectancy [11,30]. The finding could be explained by the intuition that alcohol consumption may have a lagged effect on population health. Since excessive alcohol consumption is associated with non-communicable diseases that usually manifest over longer periods, alcohol consumption may have a delayed impact on life expectancy. For health care access quality, the same is true. Improvements in health care access quality may not have an immediate impact on life expectancy. The negative sign may be due to the very low levels of health care access quality in the region [4].

For our model with health measured by infant mortality, the effect of alcohol consumption on infant mortality may be a reflection of the fact that women and children are more negatively impacted by excessive alcohol consumption [13]. While there are overarching global policies to reduce harmful alcohol consumption led by the World Health Organization3, these policies have not translated into national policies in most sub-Saharan African countries. By 2014, of the 46 African countries that agreed to the global strategy to reduce the harmful use of alcohol in 2010, only 10 countries in the region had implemented alcohol policies [19]. These findings, therefore, make a case for policies aimed at reducing excessive alcohol consumption in the region.

The study further found evidence that HIV prevalence, per capita GDP growth and urbanization were significant determinants of health measured by life expectancy which was consistent with studies in similar contexts [10,26]. The negative impact of the HIV/AIDS prevalence was expected since the disease shortened the life span of the economically active population, who were most affected by the disease [26]. With regards to the influence of urbanization, urban areas often have better infrastructure and thus provide improved access to several social services such as education, health care, sanitation and safer water supply that in turn are associated with improved population health [27]. In the period under consideration, the ‘urban advantage’ exceeded the ‘urban penalty’ that is the health benefits of residing in an urban area outweighed the costs. While there are no specific national health policies in the region targeted specifically to urban populations there is a need to enact policies that anticipate the health challenges urban population growth could pose and plan for them accordingly. Stakeholders must, therefore, acknowledge the importance of urbanization in promoting healthier populations and to plan urban areas to optimize their returns to health outcomes [42].

In line with theoretical literature, increases in per capita GDP were associated with significant improvements in life expectancy [17,33,44]. Intuitively, higher income on a macro level translates into greater investments into the health sector that can, in turn, be used to improve the health status of the population. Higher per capita income is also associated with an overall increase in the standard of living via greater access to adequate sources of nutrition, greater access to health care, improved housing and environmental conditions which increase the likelihood of living longer. An intriguing finding, however, was that per capita income was not a significant determinant of infant mortality. This finding was contrary to similar studies [10]. The finding was however similar to the work by Ensor et al. [7] who found that economic growth did not have a significant impact on infant mortality from 1965 to 2005 for a panel of developed countries. There could be several explanations for this. One important one is that the relationship between economic growth and infant mortality could be unidirectional with infant mortality having a significant impact on economic growth and not the other way around [7].

Overall, the study shows that the socioeconomic, lifestyle, health system and environmental determinants of population health in the region vary with the measure of population health used. This is expected because of the two different measures of health capture different aspects of population health. This presents an area for future study with the analysis extending to additional health measures. Measures such as HIV prevalence and urbanization were sensitive to the measure of population health used. In light of this, the study proposes that efforts to increase the levels of health in the region should focus in increasing the years of schooling, improving health care quality, increasing expenditure on health and reducing the per capita consumption of alcohol. Specific policies targeting these issues can be put into place. For instance, more comprehensive taxation can be used to reduce alcohol consumption, whereas increased budget allocation towards the health sector would increase health expenditure. Similarly, policies aimed at improving the levels of education should be given priority as they tend to have a significant impact on the overall health status of the population

The main limitation of this study is that we are unable to control for differences in technological advancements among countries due to a lack of complete information for the period under consideration. Previous studies have shown that technological progress has a significant impact on improvements in health outcomes and call for policies aimed at facilitating the diffusion of low-cost high-return interventions made possible by improvements in technology [22]. Future research should endeavour to determine the role that technology plays in improving health outcomes in Sub Saharan African countries.

CRediT author statement

Mwimba Chewe: Conceptualization, Methodology, Software, Data curation,

Writing- Original draft preparation, Investigation, Formal analysis. Peter

Hangoma: Supervision, Software, Validation, Writing- Reviewing and Editing,

Appendix A. Supplementary data

Supplementary tables