Menu
Menu
Menu
Menu
Background: KwaZulu-Natal province in South Africa has the largest population of children under the age of five and experiences the highest number of child births per annum in the country. Its population has also been ravaged by the dual epidemics of HIV and TB and it has struggled to meet targets for maternal and child mortality. In South Africa’s federal system, provinces have decision-making power on the prioritization and allocation of resources within their jurisdiction. As part of strategic planning for 2015-2019, KwaZulu-Natal provincial authorities requested an assessment of current mortality levels in the province and identification and costing of priority interventions for saving additional maternal, newborn and child lives, as well as preventing stillbirths in the province. Methods: The Lives Saved Tool (LiST) was used to determine the set of interventions, which could save the most additional maternal and child lives and prevent stillbirths from 2015-2019, and the costs of these. The impact of family planning was assessed using two scenarios by increasing baseline coverage of modern contraception by 0.5 percentage points or 1 percentage point per annum. Results: A total of 7,043 additional child and 297 additional maternal lives could be saved, and 2,000 stillbirths could be prevented over five years. Seventeen interventions account for 75 % of additional lives saved. Increasing family planning contributes to a further reduction of up to 137 maternal and 3,168 child deaths. The set of priority interventions scaled up to achievable levels, with no increase in contraception would require an additional US$91 million over five years or US$1.72 per capita population per year. By increasing contraceptive prevalence by one percentage point per year, overall costs to scale up to achievable coverage package, decrease by US$24 million over five years. Conclusion: Focused attention on a set of key interventions could have a significant impact on averting stillbirths and maternal and neonatal mortality in KwaZulu-Natal. Concerted effort to prioritize family planning will save more lives overall and has the potential to decrease costs in other areas of maternal and child care.
KwaZulu-Natal (KZN) province is the second most populous province in South Africa, and has the largest population of children under five years old with 1.3 million in 2014 [7] or 22 % of the children in the country. The province also has the largest population dependent on the public health sector for services [8]. KZN has been the most severely impacted by the country’s HIV/AIDS and tuberculosis epidemics. It is the province with the highest HIV prevalence in antenatal clients, at 37.4 % [9] and the highest incidence of tuberculosis in the general population, at 971 per 100,000 population in 2012 [10]. Since 2010, however, maternal and child health indicators have been improving. Maternal mortality in healthcare facilities in the province has decreased from a peak of 209 per 100,000 live births in 2010, down to 160 per 100,000 live births in 2012 [11, 12]. The figures are likely an underestimate, however, as recording and reporting on maternal deaths outside of healthcare facilities remains a challenge in the country [13]. Under five deaths in the province have decreased by about 46 % from 13,000 in 2007 to 7,000 in 2011 [14]. Having achieved considerable success in coverage of prevention of mother to child transmission of HIV (PMTCT) as well as improvement in access to antiretroviral treatment, the province is in a position to consider the potential impacts of other key interventions for saving the lives of mothers and children. The analysis was undertaken using the Lives Saved Tool (LiST) within the Spectrum software (Futures Institute, now Avenir Health, United States of America), which was developed to support decision making in the health sector [15, 16]. The Spectrum program consists of several modules which interact with one another to address a variety of issues in demography and population health. The demographic projection model (DemProj) forms the basis for any Spectrum projection and requires inputs on various determinants of population dynamics, including fertility rates, mortality rates and population age distribution. The modelling for maternal and child health impacts is done in the LiST, while the Family Planning (FamPlan) Module examines the impacts of changing contraceptive use prevalence and methods mix. FamPlan requires inputs on total fertility, and contraceptive prevalence, as well as the proximate determinants of fertility including the proportion of women of reproductive age married or in a sexual union, duration of post-partum insusceptibility, abortion rates and degree of pathological sterility in the population [16]. The analysis was conducted using Spectrum version 5.06. The functionality of LiST has been described in detail in previous publications [17–19]. LiST is a linear mathematical model which describes fixed relationships between inputs, which are primarily levels of coverage for a set of interventions, and outputs in terms of changes in population level risk factors and causes of mortality in children, neonates, pregnant women and stillbirths [18]. LiST is capable of analyzing the impact of the scale up of multiple interventions, which may be effective in reducing risk and mortality from several causes. For example, one particular scenario modelled can calculate the impact of a package of interventions as they are scaled up from a baseline coverage level to forecast the impact of various projected levels of coverage over time. LiST preloads national-level data for health status, mortality rates, and coverage of approximately 70 interventions, and their effectiveness in relation to specific causes of death in mothers, neonates and children under 5 years of age for a national level analysis [17, 20]. The Spectrum software allows for sub-national analyses to be constructed using a country-based model. In order to complete the province-specific analysis, a subnational population scenario was developed within a national level baseline model [3]. This involved collection of province-specific information on population numbers and growth rate, total fertility rates, contraceptive prevalence rate and HIV prevalence information for KZN province [5, 7, 9]. The LiST default settings do not carry all possible interventions, thus several additional interventions were included in the modelling, in response to existing policies and burden of disease. Given the significant burden of HIV and TB in the province, treatment of childhood tuberculosis [21], treatment of tuberculosis in pregnant women [22], as well as early detection and treatment of HIV in pregnant women were included. In response to the Free State province’s success in reducing maternal mortality with improved interfacility transport for obstetric emergencies, this was also included as a potentially impactful intervention [23]. In children between 1 and 4 years old, non-natural causes are responsible for 17.4 % of total deaths, with external sources of accidental injury as the source of 70 % of these non-natural deaths [24]. Thus treatment of childhood injuries was also included as an intervention to save child lives between the ages of 1 and 59 months [25]. These interventions were assigned to the causes of death which they could potentially affect. Intervention effectiveness was drawn from the literature and expert input from a national Millennium Development Goal Countdown analysis in LiST which was undertaken for South Africa in 2013 [3, 21, 23, 26]. Details of the effect sizes and affected fractions which were applied to the model are given in Additional file 1: Sheet 1. Adequately modelling the provincial-level impact of different interventions required the following: Data to populate the Spectrum Suite were drawn from a number of sources available on HIV, maternal and child health indicators, under-5 and maternal deaths and stillbirths in the province, as well as estimates of the coverage levels of interventions in LiST but not routinely monitored or reported in South Africa. Burden of disease estimates for 2011 for maternal, neonatal and child deaths were derived from the National Committee on Confidential Enquiries into Maternal Deaths (NCCEMD) data, and vital registration data [14, 27]. The causes of maternal, neonatal and under-5 deaths were adapted in consultation with the provincial clinical team to fit the causal categories provided by LiST. Mortality rates were determined for 2011 or the closest year. Many of the interventions in LiST are not tracked on a consistent basis within the health monitoring systems. The baseline coverage for LiST indicators was determined using national data from the District Health Information System [28] where possible. Where estimates were not available from standard database sets, the provincial clinical and management team was asked to estimate the coverage for particular interventions, based on their experience and knowledge of the health services in the province. This team consisted of the head of the maternal and child health programme in the province as well as two clinical specialists, one in obstetrics and gynaecology and the other in paediatrics. Consultation with the team was a deliberative process during which baseline coverage levels (Table 1) as well as the interventions, their meaning and application in the provincial context was discussed and explored. Baseline intervention coverage based and achievable levels for LiST interventions as adapted and modified by provincial experts (July 2014) Using 2011 as baseline, we used a provincial maternal mortality rate of 187 deaths per 100,000 live births [27], a neonatal mortality rate of 13.6/1,000 [29] and an under-5 mortality rate of 37.08/1,000 (estimated from vital statistics data, adjusted for underreporting using national level methods) (Additional file 1: Worksheet 2) [14]. Baseline stillbirth rate was 23/1 000 live births [28]. The causes of maternal mortality [27] and neonatal and child mortality [14] were adapted from their source estimates to fit the causal categories in LiST, and adjustments were validated by the clinical provincial team by comparison to nationally reported estimates [3] and provincial estimates where these were available [12, 14] (Fig. 1). Proximate Causes of a newborn deaths, b deaths in children 1–59 months of age and c maternal deaths in KwaZulu-Natal in 2011. Source: Adapted from Stats SA and Vital Registration details of newborn deaths in 2010 and 2011 and the National Confidential Enquiries on Maternal Deaths Interim Report 2011 and 2012 The provincial team was also asked to project the level of coverage for each intervention that they estimated KZN would reach by 2019, if concerted effort were realistically applied and considering existing policy changes, resource inputs and observed local impacts of existing services (Table 1), i.e., the level of coverage attained by realistically scaling up essential interventions. In this paper, this level of coverage is referred to as ‘achievable coverage’. The LiST was used to analyze the number of deaths of mothers, newborns and children and stillbirths that could be averted by scaling up all interventions in the model from their baseline coverage estimates to: The “full coverage” scenario aimed to show what would be possible in terms of reducing the numbers of deaths, if a universal coverage was achieved for all interventions, whereas the “achievable scenario” was considered to be what could be realistically achieved with the prevailing conditions in the province. The achievable scenario was also used for the subsequent costing exercise to inform provincial planning. Family planning also contributes significantly in reducing the numbers of overall deaths of mothers, infants and children, and reducing the number of stillbirths. Averting unwanted pregnancies through improved family planning and birth spacing can potentially have a substantial impact on reductions in maternal and perinatal deaths [30]. Thus, in addition to the two base scenarios, two analyses were completed for contraceptive use prevalence (using the FamPlan Module of Spectrum): The FamPlan method mix was determined at baseline for 2014 using information from the provincial DHIS, and main changes to the mix were an increase in the uptake of the newly introduced etonorgestrel implant, with ongoing decrease of the percentage use of intrauterine devices and injectable options – a trend which was already evident from the contraceptive utilization data (Additional file 1: Worksheet 3). The effect of scaling up the maternal and child intervention coverage was modelled to estimate the deaths averted, overall and by each intervention. Intervention costs were calculated from a provider perspective, using the costing module within the LiST. The module uses an ingredients approach to costing, based on four main components: personnel and labour; drugs and supplies; other recurrent costs and capital costs. The first year of costing was 2014. Staff remuneration was based on current salary structures for healthcare workers in South Africa, as published, and adjusted for annual increases, at 5.6 % per annum [31]. For unit costs of medicines and supplies, the default values in LiST were used. These unit costs are based on international prices from UNICEF and the Management Sciences for Health International Drug Price Indicator Guide 2011 [32]. Unit costs in Spectrum for medicines and supplies were generally comparable to the South African prices of drugs and supplies on various national tenders in most instances, although changes were made specifically to the costs for family planning commodities, since the costs of the sub-dermal implant had decreased substantially [33]. Recurrent costs related to hospitalization and outpatient visits were not included. Recurrent costs include personnel training, gasoline, building rent, office supplies and promotional activities. These were outside of the scope of the analysis. Capital costs such as infrastructure expenditure were also excluded from the analysis. Thus the costing reflects purely the scaling up of the delivery of each intervention in terms of costs of staff time and costs of medicines and supplies. To judge the relative additional number of lives saved against the relative additional costs incurred to scale up interventions, a ratio cost per life year saved was determined. This was done by multiplying the lives saved by the expected life expectancy for each life saved. The average provincial life expectancy in 2014 of 54.4 years [7] was used for child and neonatal deaths and stillbirths prevented. For maternal lives saved, we used the Reproductive-Aged Life Expectancy – RALE [34], estimated to be 27 years, or half of the average life expectancy at birth.
