Ending malnutrition in all its forms requires scaling up proven nutrition interventions and much more: a 129-country analysis

Background Sustainable Development Goal (SDG) 2.2 calls for an end to all forms of malnutrition, with 2025 targets of a 40% reduction in stunting (relative to 2012), for wasting to occur in less than 5% of children, and for a 50% reduction in anaemia in women (15–49 years). We assessed the likelihood of countries reaching these targets by scaling up proven interventions and identified priority interventions, based on cost-effectiveness. Methods For 129 countries, the Optima Nutrition model was used to compare 2019–2030 nutrition outcomes between a status quo (maintained intervention coverage) scenario and a scenario where outcome-specific interventions were scaled up to 95% coverage over 5 years. The average cost-effectiveness of each intervention was calculated as it was added to an expanding package of interventions. Results Of the 129 countries modelled, 46 (36%), 66 (51%) and 0 (0%) were on track to achieve the stunting, wasting and anaemia targets respectively. Scaling up 18 nutrition interventions increased the number of countries reaching the SDG 2.2 targets to 50 (39%), 83 (64%) and 7 (5%) respectively. Intermittent preventative treatment of malaria during pregnancy (IPTp), infant and young child feeding education, vitamin A supplementation and lipid-based nutrition supplements for children produced 88% of the total impact on stunting, with average costs per case averted of US$103, US$267, US$556 and US$1795 when interventions were consecutively scaled up, respectively. Vitamin A supplementation and cash transfers produced 100% of the total global impact on prevention of wasting, with average costs per case averted of US$1989 and US$19,427, respectively. IPTp, iron and folic acid supplementation for non-pregnant women, and multiple micronutrient supplementation for pregnant women produced 85% of the total impact on anaemia prevalence, with average costs per case averted of US$9, US$35 and US$47, respectively. Conclusions Prioritising nutrition investment to the most cost-effective interventions within the country context can maximise the impact of funding. A greater focus on complementing nutrition-specific interventions with nutrition-sensitive ones that address the social determinants of health is critical to reach the SDG targets.

Several risk factors for stunting in children are modelled: birth outcomes (pre-term birth and/or a child being born small for gestational age [SGA]), stunting in a previous age-band, suboptimal feeding practices (age-appropriate breastfeeding and complementary foods), and incidence of diarrhoea (Figure A.2). In addition, anaemia in women of reproductive age is modelled to be a risk factor for suboptimal birth outcomes; birth outcomes and diarrhoea incidence are modelled to be risk factors for wasting; and sub-optimal breastfeeding is modelled to be a risk factor for diarrhoea incidence.
In the model, interventions can improve nutritional outcomes directly or indirectly by reducing risk factors. For example, Figure A.2 shows that changes to breastfeeding practices, perhaps through better education, can directly reduce mortality and diarrhoea incidence. Moreover, in the model this will also lead to an indirect reduction in mortality because a reduction in diarrhoea incidence will lead to a reduction in stunting and wasting, which will subsequently further reduce mortality. Changing the coverage of an intervention among its target population leads to changes in projected outcomes based global estimates of intervention effectiveness.
In previous work [27], the epidemiological component of the Optima Nutrition model has been validated by comparing the effects of scaling up the included interventions to those estimated by the Lives Saved Tool (LiST) [28], and found to be in good agreement. This is not surprising as both models use the same underlying impact pathway [29], based on the latest available evidence. The Optima Nutrition model is newer than the LiST model, however the LiST model has been used and validated against real-world outcomes in sub-Saharan Africa, finding it to be a useful tool despite limitations in data availability [30,31]. Optima Nutrition uses an economic model to translate the amount spent on an intervention to its estimated coverage. For each intervention, this requires a setting-specific input for the unit cost.
A.2 Modelling stunting using Optima Nutrition • The model divides children in each age-band into four height-for-age categories, based on WHO criteria ( Figure A.3), with the two lowest categories (severe and moderate) being considered stunting: • Risk factors for stunting are suboptimal birth outcomes (pre-term birth and/or a child being born SGA), stunting in a previous age-band, suboptimal feeding practices (age-appropriate breastfeeding and complementary foods), and incidence of diarrhoea ( Figure A.2).
• Stunting increases the risk of mortality for children who have diarrhoea, pneumonia, measles and other illnesses.
• Odds ratios and relative risks are model inputs, can be changed, and have defaults based on the literature.

A.3 Modelling wasting using Optima Nutrition
• The weight-for-height distribution is tracked for children in each age band ( Figure A.4, just like for stunting). Children are divided into four categories: o Severe acute malnutrition (SAM): < -3 standard deviations below than the median weightfor-height of the WHO reference population o Moderate acute malnutrition (MAM): < -2 and >= -3 standard deviations below than the median weight-for-height of the WHO reference population o Mild acute malnutrition: < -1 and >= -2 standard deviations below than the median weight-for-height of the WHO reference population o Normal: >= -1 standard deviation below than the median weight-for-height of the WHO reference population • Children are considered to be "wasted" if they are in the SAM or MAM categories.
• Wasting is modelled as an incident (short-duration) condition: o As opposed to stunting, where being stunted in one age band increases the risk of being stunted in the next, wasting distributions are independent in each age band -this means that the distribution (i.e. prevalence) of wasting in a given time period does not affect the distribution of wasting in subsequent periods.
• Wasting increases the risk for mortality for children who have diarrhoea, pneumonia, measles and other illnesses • Diarrhoea incidence and birth outcomes are risk factors for wasting ( Figure   • Children enter the age band (shown from the left), and will be classified as SAM, MAM, mild or normal according to the prevalence of these states from the data • Children in the mild and normal categories can develop MAM (incidence of MAM) • Children with MAM can deteriorate to SAM (incidence of SAM) • When they are in MAM or SAM categories, children have increased risks of death • Children can recover from the SAM and MAM categories due to treatment.
• The incidence rates, probabilities of death and average duration spent in SAM and MAM are calibrated to match country-specific data on prevalence, mortality and treatment numbers.

A.4 Modelling anaemia using Optima Nutrition
Each population in the model is stratified by anaemia status: anaemic (mild, moderate, or severe) or not anaemic ( Figure A.6). The model also includes a setting-specific input for the fraction of anaemia that is severe.
• Anaemia in pregnant women is modelled as a risk factor for maternal mortality (e.g. due to risk of haemorrhage) • Anaemia in pregnant women is also modelled to be a risk factor for suboptimal birth outcomes o This can affect stunting, which in turn can affect mortality in children.   C.3 Impact of diarrhoea on stunting, wasting and anaemia