Reducing diarrhoea deaths in South Africa: Costs and effects of scaling up essential interventions to prevent and treat diarrhoea in under-five children

Background: Diarrhoea is one of the leading causes of morbidity and mortality in South African children, accounting for approximately 20% of under-five deaths. Though progress has been made in scaling up multiple interventions to reduce diarrhoea in the last decade, challenges still remain. In this paper, we model the cost and impact of scaling up 13 interventions to prevent and treat childhood diarrhoea in South Africa. Methods: Modelling was done using the Lives Saved Tool (LiST). Using 2014 as the baseline, intervention coverage was increased from 2015 until 2030. Three scale up scenarios were compared: by 2030, 1) coverage of all interventions increased by ten percentage points; 2) intervention coverage increased by 20 percentage points; 3) and intervention coverage increased to 99%. Results: The model estimates 13 million diarrhoea cases at baseline. Scaling up intervention coverage averted between 3 million and 5.3 million diarrhoea cases. In 2030, diarrhoeal deaths are expected to reduce from an estimated 5,500 in 2014 to 2,800 in scenario one, 1,400 in scenario two and 100 in scenario three. The additional cost of implementing all 13 interventions will range from US$510 million (US$9 per capita) to US$960 million (US$18 per capita), of which the health system costs range between US$40 million (less than US$1 per capita) and US$170 million (US$3 per capita). Conclusion: Scaling up 13 essential interventions could have a substantial impact on reducing diarrhoeal deaths in South African children, which would contribute toward reducing child mortality in the post-MDG era. Preventive measures are key and the government should focus on improving water, sanitation and hygiene. The investments required to achieve these results seem feasible considering current health expenditure.

Modelling was done using the Lives Saved Tool (LiST), a module in the Spectrum software [12]. Version 5.07 was used (downloaded 28/11/2014). LiST is a deterministic mathematical model that compares the effect of various interventions on population level risk factors, as well as stillbirths and maternal, newborn and child deaths [13,14]. Included in the model are more than 60 interventions that have an impact on cause-specific mortality. An intervention can have an impact on single or multiple causes of death and risk factors. The outcome measures (risk factors and cause-specific mortality) change based on the level of coverage of the interventions included in the model. Increasing the level of coverage of one or more interventions can thus lead to a reduction in associated risk factors or cause-specific mortality. LiST therefore enables a user to assess the simultaneous impact of interventions on health outcomes. Intervention impact on mortality can be direct or indirect (through the reduction of risk factors). The direct impact of each of these interventions is modelled by multiplying its effectiveness estimate with the level of coverage, assuming all other interventions are kept constant. For example, an intervention with an effect estimate of 30% can avert 30% of the associated cause-specific deaths if coverage for that intervention is 100%. When LiST analyses multiple interventions, each intervention is applied to the residual deaths from the previous intervention. This prevents double counting the number of lives saved. The model starts by applying the preventive interventions in succession, followed by the treatment interventions on remaining deaths. The total number of deaths prevented is therefore not attributable to specific interventions but rather the full intervention package [15]. LiST includes 14 interventions for the prevention and treatment of diarrhoea. Walker and Walker (2014) describe the interactions between these interventions and the modelling methods used in LiST [15]. There are 12 interventions in LiST that have a direct impact on diarrhoeal mortality. Eight of these are preventive interventions: rotavirus vaccine; vitamin A supplementation; zinc supplementation; and the water, sanitation and hygiene (WASH) programmes that include a water connection in the home or improved water source, improved sanitation hand washing with soap and hygienic disposal of children’s stools. Interventions for breastfeeding promotion, severe wasting and moderate acute malnutrition have an indirect impact on diarrhoeal mortality. The impact of breastfeeding can be modelled either as a risk factor that changes when breastfeeding promotion shifts breastfeeding rates, or as a direct risk factor for death due to the lack of appropriate breastfeeding. (In our analysis, we ramped up breastfeeding according to WHO guidelines, which recommend six months of exclusive breastfeeding and appropriate complementary feeding up to two years). LiST also includes three interventions for treating diarrhoea: zinc treatment, ORS and antibiotics for dysentery. Figure 1 (adapted to reflect the interventions addressed in our analysis) provides an overview of the intervention interactions; Zinc supplementation has been excluded from our analysis because this is not provided in South Africa. LiST interventions that impact diarrhoea mortality. Green shaded boxes = preventive interventions. Blue shaded boxes = treatment interventions. Peach and grey boxes = interventions via a risk factor pathway. WASH = interventions for water, sanitation and hygiene. MAM = moderate acute malnutrition. (Adapted from Walker C and Walker N, 2014). Malnutrition is represented as a risk factor for diarrhoea mortality through the impact of stunting and wasting. Lack of appropriate breastfeeding, vitamin A supplementation and the WASH interventions influence diarrhoea incidence, which in turn affects stunting and subsequent mortality. The effectiveness values of the diarrhoea interventions included in LiST have been reviewed by the Child Health Epidemiology Reference Group (CHERG) [16-20] and are presented in Additional file 1 [15]. The methods used in our analysis are based on similar multi-country assessments in LiST [21,22]. We assessed the impact of increasing the coverage of 13 interventions on diarrhoeal mortality. The baseline (2014) coverage levels of these interventions were reviewed and modified during a one day expert consultation hosted in South Africa. Twenty-three participants were invited from the health sector, including clinicians, academics and others in positions at national and district level. Coverage levels are indicated in Table 1. Breastfeeding prevalence at baseline was input by age group: 8% coverage of exclusive breastfeeding for infants younger than 6 months, 51% coverage of any breastfeeding for infants aged 6 – 11 months and 31% coverage of any breastfeeding for infants aged 12 – 24 months [23]. Coverage for the WASH interventions ranged from 17% for hand washing with soap to 95% for an improved water source [24]. Unchanged default LiST coverage levels have been indicated. Interventions were linearly scaled up from the baseline year 2014 until 2030, with coverage increases starting in 2015.Three scale up scenarios were implemented: in scenario one, we assumed that the coverage of all interventions increased by 10% from their baseline estimate (a fixed 0.7% increase per year); in scenario two, coverage increased by 20% (a fixed annual increase of 1.3%); and in scenario three, coverage for all interventions was increased to 99% (full coverage) in 2030. In the rest of the document, the scenarios are accordingly referred to as scenario one (10% increase), scenario two (20% increase) and scenario three (full coverage). Coverage levels for other maternal and child health interventions were not altered, in order to isolate the impact of the priority interventions for diarrhoea prevention and treatment. Baseline and projected coverage of interventions to prevent and treat diarrhoea *Default coverage level in LiST. The baseline mortality rates used in our analysis were 41 deaths per 1,000 live births for under-five children and 13/1,000 for neonates [3]. The causes of newborn mortality were adapted (South African Medical Research Council: Preliminary estimates for burden of disease in 2010, unpublished) estimates to fit the causal categories in LiST (Figure 2). The categories in LiST differ slightly from those presented by the MRC. For example, neonatal diarrhoea is not reported separately in the MRC BOD, but rather combined with under-five diarrhoeal deaths. Therefore, we separated these using the default proportions in LiST. Causes of death in children under-five years, used in LiST (adapted from MRC, 2010). Modelling of costs was done using the costing module in LiST, with the most recently available data. The module uses an ingredients approach to costing, based on four components: personnel and labour; drugs and supplies; other recurrent costs; and capital costs. Staff remuneration is based on current salary structures of health workers in South Africa. Salary increases were not applied. The unit costs of drugs and supplies are based on international drug prices from UNICEF and the Management Sciences for Health International Drug Price Indicator [25,26]. The unit costs found in LiST were comparable to South African prices of drugs and supplies requested for tender by the Department of Health [27]. The unit costs for WASH programmes are not included in LiST. We estimated these costs using data available from the South African Department of Water and Sanitation [28,29]. A home water connection includes water piped into either the home or yard. This is reflected in the average cost estimate of US$480 per household (adjusted for inflation). The cost for improved sanitation was estimated using the proportion of South Africans with access to dry and wet sanitation, (60% and 40%, respectively) [10]. Wet sanitation includes various types of flush latrines and dry sanitation includes pit latrines (with and without ventilation), chemical toilets and bucket toilets. The average household cost for sanitation was approximately US$900 (adjusted for inflation). Recurrent costs related to hospitalization and outpatient visits were not included. Recurrent costs include personnel training, gasoline, building rent, office supplies and promotional activities [30,31]. These were outside of the scope of the analysis. In addition, costs estimated in LiST exclude infrastructure development, such as building clinics [30]. All costs were adjusted to 2014 US dollars. Per capita costs use the 2014 South African population estimate of 54 million [32]. Intervention impact was measured in terms of diarrhoeal deaths averted. First, we calculated the expected number of deaths (and cases) at the current (baseline) level of intervention coverage. Second, the number of deaths (and cases) was recalculated for the three intervention scale up scenarios. Deaths averted (or additional lives saved) were then estimated by subtracting the numbers of deaths at baseline from the deaths at increased coverage (the same methodology was used to estimate the number of diarrhoea cases averted). Ethical review board approval was not required for this analysis as no human subjects were involved.

• chola, lumbwe
• michalow, julia
• tugendhaft, aviva
• hofman, karen