Incidence and predictors of mortality among children with severe acute malnutrition admitted to therapeutic feeding units in Ethiopia: a systematic review and meta-analysis

Abstract

Objective

To estimate the mortality rate and identify predictors of mortality among under-five children with severe acute malnutrition (SAM) admitted to therapeutic feeding units (TFUs) in Ethiopia.

Methods

We searched PubMed, HINARI, Science Direct, Google Scholar and African Journals Online from 1 March to 30 May 2024. The Joanna Briggs Institute checklist was used to appraise the included studies. Heterogeneity was identified using I² statistics. Funnel plots and Egger’s tests were used to determine publication bias.

Results

Out of 1085 studies, 15 were included in this analysis. The pooled mortality rate among under-five children with SAM admitted to TFUs in Ethiopia was 8.32 per 1000 person-days of observation (95% CI: 6.25 to 11.06). The mortality rate has not changed over time. HIV infection (HR: 2.84; 95% CI: 1.25 to 6.42), tuberculosis (HR: 1.86; 95% CI: 1.35 to 2.56), intravenous fluid use (HR: 3.37; 95% CI: 2.39 to 4.75), altered body temperature (HR: 4.47; 95% CI: 1.90 to 10.51), impaired consciousness (HR: 2.91; 95% CI: 1.94 to 4.37), not receiving F-100 supplementation (HR: 4.51; 95% CI: 3.25 to 6.26), shock (HR: 4.20; 95% CI: 2.92 to 6.04), and nasogastric tube feeding (HR: 2.02; 95% CI: 1.67 to 2.44) were predictors of mortality.

Conclusion

The pooled mortality rate in Ethiopia was 8.32 per 1000 person-days, and it has not decreased over time. Most of the identified factors are related to comorbidities and complications of SAM, as well as nutritional therapy. Thus, it is essential to strengthen nutrition policies, programme implementation and healthcare services, which focus on the timely management of SAM complications, integrated care for comorbidities and improved F-100 supplementation.

PROSPERO registration number

CRD42024555014.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Subgroup analysis was done to minimise heterogeneity.

• All the included studies were cohort studies, which better identify cause-and-effect relationships.

• A cumulative meta-analysis was conducted to assess changes in the mortality rate over time.

• The analysis included only seven regions, so it may not adequately represent other regions.

• Due to the lack of published meta-analyses or nationwide studies, it was difficult to compare our findings.

Background

Child malnutrition remains one of the leading causes of morbidity and mortality globally, particularly in developing countries.^{1 2} Globally, in 2022, an estimated 149 million under-five children were affected by stunting, 45 million by wasting and 37 million by underweight.³ Nearly half of all deaths among under-five children are associated with malnutrition; the majority of deaths occurred in low-income and middle-income countries.³ Despite ongoing efforts, progress in tackling malnutrition has been slow, making it difficult to meet the Sustainable Development Goals (SDGs) 2 and 3 (Zero Hunger and Good Health and Well-being).^{4 5} It requires urgent and coordinated action from governments, communities and individuals.⁶

Severe acute malnutrition (SAM) is a severe form of undernutrition characterised by a very low weight-for-height/length (≤−3 SDs) or the presence of nutritional oedema. In children aged 6–59 months, a mid-upper arm circumference of less than 115 mm also indicates SAM.⁷ According to the World Health Organisation and UNICEF, an estimated 19 million under-five children are affected by SAM globally, about 400 000 of them dying each year, and 10%–40% of these deaths occur after hospitalisation.^{7 8} In Ethiopia, 15% of under-five children are affected by SAM.⁹ It accounts for around 20% of paediatric hospital admissions, with mortality rates ranging from 3 to 22 deaths per 1000 person-days in health facilities.^{10 11}

Ethiopia has implemented several initiatives to reduce under-five deaths caused by SAM. These efforts include the adoption of WHO guidelines, setting up therapeutic feeding units (TFUs), training healthcare providers and expanding services for early detection and treatment.^{7 12 13} Nevertheless, SAM remains one of the leading causes of under-five mortality in the country, with over one-fourth of these deaths occurring during hospitalisation.¹⁴ Existing evidence highlights the persistently high mortality rates associated with SAM, particularly during hospitalisation.^{10 15}

The pooled incidence of mortality rate among under-five children admitted with SAM in Ethiopia has not been estimated. In addition, although a systematic review and meta-analysis on the predictors of mortality in children admitted with SAM was conducted before 2019, important factors such as HIV, tuberculosis (TB), shock, level of consciousness, body temperature, the use of F-100 therapeutic milk and nasogastric tube feeding were not investigated.¹⁶ Therefore, this study aimed to estimate the pooled incidence of mortality and identify its associated factors among under-five children with SAM admitted to TFUs in Ethiopia.

Methods

Search strategy

This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA-2020) guidelines (online supplemental additional file 1). To identify relevant articles, we searched PubMed, HINARI, Science Direct, Google Scholar and the African Journals Online database between 1 March 2024, and 30 May 2024. The key search terms included incidence, mortality rate, treatment outcome, time to death, SAM, determinants, predictors, causes, associated factors, risk factors, children, under five, 6–59 months, less than 6 months, paediatrics and Ethiopia. The search terms were formulated based on the Medical Subject Heading Terms and Boolean operators (AND, OR) (online supplemental additional file 2).

Eligibility criteria

The inclusion criteria were as follows: (1) studies conducted in Ethiopia, (2) studies reporting the incidence rate and/or predictors of mortality among under-five children with SAM admitted to TFUs, (3) studies reporting the number of new deaths, (4) studies providing person-time data and (5) studies published in English and available online between 1 March and 30 May 2024. This systematic review and meta-analysis excluded studies reporting predictors other than HRs, conference papers, articles without full-text, anonymous reports, editorial reports and qualitative research.

Data extraction

Data were collected and organised using a customised Microsoft Excel spreadsheet. The extraction process was independently conducted by three authors (AA, GFA and MSA). From each study, the following information was retrieved: author’s name, year of publication, number of child deaths, study region, study design, person-days of observation, follow-up period, mortality rates and predictors of mortality with HRs.

Quality assessment

The Joanna Briggs Institute (JBI) critical appraisal checklist for cohort studies was used to assess the quality of each study. The checklist has Yes, No, Unclear and Not Applicable options: ‘1’ is given for ‘Yes’ and ‘0’ is given for other options. The checklist includes four response options: Yes, No, Unclear and Not Applicable. Two authors (AA and GFA) independently evaluated the studies using the checklist. Any discrepancies between the authors were resolved by averaging their scores. The quality scores were then summed and converted into percentages. Finally, 15 studies with quality scores above 50% were included in this meta-analysis (online supplemental additional file 3).

Patient and public involvement

Patients or the public were not involved in this research’s design, conduct, reporting or dissemination.

Outcome measurement

The first outcome of this systematic review and meta-analysis was the mortality rate among under-five children with SAM admitted to the TFUs. The mortality rate was calculated by dividing the total number of deaths by the total number of follow-up days and multiplying by 1000. Identifying the predictors of mortality among under-five children with SAM admitted to the TFUs was the second outcome of this study.

Statistical analysis

Microsoft Excel was used for data entry, and then the data were imported into R software version 4.3.2 for analysis. Heterogeneity was evaluated using the I² statistic, where values of 25%, 50% and 75% indicated low, moderate and high, respectively.¹⁷ Publication bias was evaluated by a visual inspection of the funnel plot and Egger’s test. A p-value of less than 0.05 in Egger’s test indicated the potential presence of publication bias. A trim-and-fill analysis was also performed to modify the pooled mortality rate and estimate the number of studies that might be missing as a result of publication bias. To explore potential sources of heterogeneity, a univariate meta-regression analysis was performed, taking into account the study region. Additionally, to assess the influence of individual studies on the pooled mortality rate, a leave-one-out sensitivity analysis was carried out by sequentially removing one study at a time.

A cumulative meta-analysis was carried out to track how the mortality rate has changed over time. A random-effects model was used to pool the mortality rate. For the pooled HR, a fixed-effects model was applied to variables such as level of consciousness, F-100 supplementation, IV fluid use, NG tube feeding, shock and TB. However, a random-effects model was used for HIV status and body temperature. The pooled mortality rate was presented using forest plots, and the pooled HRs were summarised in a table with CIs.

Results

Characteristics of included studies

1085 studies were searched from PubMed, HINARI, Science Direct, Google Scholar and the African Journal Online Database. From these studies, 450 articles were excluded due to duplication. An additional 635 articles were removed because their titles and abstracts were not relevant. The remaining 54 studies were browsed for full-text review. Hence, after reviewing the full text, 39 studies were excluded due to non-compliance with the inclusion criteria. Finally, 15 studies met the eligibility criteria and were incorporated into the final study1011 14 18,29 (figure 1). All studies were conducted using a retrospective follow-up study design.1011 14 18,29 Twelve studies reported the mortality rate and its predictors,1011 14 15 18,21 23 while the remaining three studies reported only the predictors of mortality.^{26 28 30} These studies were done in different regions of Ethiopia (Addis Ababa, Dire Dawa, Bnishangul-Gumuz, Amhara, Tigray, South Ethiopia, and Oromia regions) (table 1).

Figure 1. PRISMA 2020 flow diagram showing the screening and selection of studies for systematic review and meta-analysis in Ethiopia, 2024. PRISMA, Preferred Reporting Items for Systematic reviews and Meta-Analyses.

Table 1. Characteristics of studies included in the meta-analysis for the pooled mortality rate and its predictors among children with SAM admitted to TFUs, Ethiopia, 2024.

Author/s, year of publication and reference	Study region	Study design	Age of participants in months	Number of deaths	Follow-up time in days	Total PDO	Mortality rate/1000 PDO
Gebremichael et al, 201423	Tigray	RFS	6–59	60	17	8274	7.25
Jarso et al, 201526	Oromia	RFS	0–59	88	2–66	_	_
Adal et al, 201619	South Ethiopia	RFS	0–59	56	1–54	7389	7.57
Oumer et al, 201625	Dire Dawa	RFS	0–59	47	1–81	6267	7.45
Girum et al, 201728	South Ethiopia	RFS	0–59	51	1–54	_	_
Wagnew et al, 201815	Amhara	RFS	0–59	66	8–17	6333	10.42
Guesh et al, 201810	Tigray	RFS	0–59	21	1–45	6671	3.14
Abiyu et al, 201918	South Ethiopia	RFS	0–59	62	1–39	7920	7.82
Desyibelew et al, 201930	Amhara	RFS	0–59	34	_	_	_
Ashine et al, 202020	Amhara	RFS	6–59	20	_	3688	5.42
Kassaw et al, 202114	Amhara	RFS	0–59	54	46	5936	9.09
Bitew et al, 202121	Addis Ababa	RFS	0–59	61	1–71	10 829	5.63
Oumer et al, 202111	Dire Dawa	RFS	0–59	60	1–47	2727	22.00
Kebede et al, 202224	Benishangul-Gumuz	RFS	6–59	91	5–31	5146	17.68
Alemu et al, 202329	South Ethiopia	RFS	0–59	54	_	6750	8.00

.IDR, incidence density rate; PDO, person day observation; RFS, retrospective follow-up study; SAM, severe acute malnutrition ; TFUs, therapeutic feeding units.

The pooled mortality rate among children admitted for SAM

Twelve studies were included to estimate the pooled mortality rate.1011 14 18,25 29 Accordingly, the mortality rate among under-five children with SAM admitted to the TFUs in Ethiopia was found to be 8.32 (95% CI, 6.25 to 11.06) per 1000 person-days of observation using a random-effect model (showing the screening and selection of studies for systematic review and meta-analysis in Ethiopia, 2024

figure 2). There was high heterogeneity between studies included in the meta-analysis (I²=92%, p<0.001). Therefore, subgroup analysis was done based on the study region. Accordingly, the lowest mortality rate was reported from Tigray (4.87 (95% CI, 2.15 to 11.02) per 1000 person-days of observation), while the highest was reported from Benishangul-Gumuz (17.68 (95% CI, 14.40 to 21.72) per 1000 person-days of observation) (figure 3).

Figure 2. Forest plots of mortality rate among under-five children with SAM admitted to TFUs in Ethiopia, 2024 (n=12). SAM, severe acute malnutrition; TFUs, therapeutic feeding units.

Figure 3. Forest plot of subgroup analysis of mortality rate among under-five children with SAM admitted to TFUs in Ethiopia, 2024 (n=12). SAM, severe acute malnutrition; TFUs, therapeutic feeding units.

Publication bias

Asymmetric distribution was displayed in the funnel plot visual inspection (figure 4). The Egger test also shows a statistically significant publication bias (B₀=−2.53, p-value=0.02). Furthermore, trim and fill analysis was done, and three studies were filled, and the pooled mortality rate became 10.12 (95% CI: 7.37 to 13.89) per 1000 person-days of observation (online supplemental additional file 4).

Figure 4. Funnel plot showing the distribution of studies on mortality rates among under-five children with SAM admitted to TFUs in Ethiopia, 2024 (n=12). SAM, severe acute malnutrition; TFUs, therapeutic feeding units.

Meta-regression and sensitivity analysis

Meta-regression analysis was conducted, considering sample size and study region, to identify potential sources of heterogeneity. Both sample size and study region were identified as possible sources of heterogeneity (online supplemental additional file 5). Sensitivity analysis was performed to assess the impact of each study on the pooled estimates. The analysis revealed that all studies contributed equally to the pooled incidence of mortality among children admitted for SAM (figure 5).

Figure 5. Sensitivity analysis for the incidence of mortality rate among under-five children with SAM admitted to TFUs in Ethiopia, 2024 (n=12). SAM, severe acute malnutrition; TFUs, therapeutic feeding units.

Cumulative meta-analysis

The cumulative meta-analysis indicated that the incidence of mortality among under-five children with SAM admitted to TFUs in Ethiopia has not changed over time with the addition of new studies (figure 6).

Figure 6. Forest plot showing the cumulative meta-analysis of mortality rate among under-five children with SAM admitted to TFUs in Ethiopia, 2024 (n=12). SAM, severe acute malnutrition; TFUs, therapeutic feeding units.

Predictors of mortality among children with SAM admitted to TFUs in Ethiopia

A total of 15 studies1011 14 18,29 were used to identify significant factors associated with mortality among children under 5 years of age with SAM admitted to TFUs in Ethiopia. These factors include HIV, TB, IV fluids, NG tubes, body temperature, level of consciousness, shock and F-100 supplementation. The hazard of death among HIV-infected children was 2.84 times higher as compared with their counterparts (HR: 2.84; 95% CI: 1.25 to 6.42). Children with TB had a nearly twofold higher hazard of death compared with those without TB (HR: 1.86, 95% CI: 1.35 to 2.56). Children treated with IV fluids had a 3.37 times higher hazard of death as compared with their counterparts (HR: 3.37; 95% CI: 2.39 to 4.75). The hazard of death was 4.47 times higher among children who had altered body temperature compared with those with normal body temperature (HR: 4.47; 95% CI: 1.90 to 10.51). Children who had an impaired consciousness level had a 2.9 times higher hazard of death as compared with those who were conscious (HR: 2.91, 95% CI: 1.94 to 4.37). The hazard of death was 4.5 times higher among children who did not feed F-100 than their counterparts (HR: 4.51, 95% CI: 3.25 to 6.26). The hazard of death among children who had shock was 4.2 times higher than those who had not shock (HR: 4.20, 95% CI: 2.92 to 6.04). Children with NG tubes had a twofold higher hazard of death compared with their counterparts (HR: 2.02; 95% CI: 1.67 to 2.44) (table 2).

Table 2. Predictors of mortality among children with SAM admitted to TFUs in Ethiopia, 2024.

Factors	Included studies, year of publication and reference	HR (95% CI)	Pooled HR (95% CI)	I2 With p value
HIV	Kassaw et al, 202114	2.80 (1.24 to 6.36)	2.84 (1.25 to 6.42)	I2=76.8%,p=0.00
	Desyibelew et al, 201930	3.82 (1.3 to 11.15)
	Alemu et al, 202329	1.31 (1.12 to 1.72)
	Oumer et al, 201625	11.57 (2.34 to 57.2)
Body temperature	Girum et al, 201728	7.17 (3.05 to 16.86)	4.47 (1.90 to 10.51)	I2=72.7%,p=0.02
	Adal et al, 201619	6.94 (2.94 to 16.4)
	Abiyu et al, 201918	2.01 (1.01 to 3.91)
Level of consciousness	Kassaw et al, 202114	2.40 (1.08 to 4.67)	2.91 (1.94 to 4.37)	I2=29.4%,p=0.24
	Guesh et al, 201810	6.69 (2.43 to 19.93)
	Jarso et al, 201526	2.60 (1.5 to 4.5)
F-100 therapeutic milk	Bitew et al, 202121	8.33 (4.34 to 16.66)	3.37 (2.39 to 4.75)	I2=0.0%,p=0.55
	Wagnew et al, 201815	3.0 (1.67 to 5.41)
	Abiyu et al, 201918	4.87 (2.75 to 8.63)
	Oumer et al, 201625	3.26 (1.32 to 8.07)
IV fluid	Bitew et al, 202121	5.2 (2.42 to 10.4)	3.37 (2.39 to 4.75)	I2=0.0%,p=0.55
	Wagnew et al, 201815	3.20 (1.75 to 5.88)
	Adal et al, 201619	3.24 (1.54 to 6.8)
	Gebremichael et al, 201423	2.52 (1.25 to 5.07)
NG tube feeding	Oumer et al, 201625	2.33 (1.15 to 4.72)	2.02 (1.67 to 2.44)	I2=10.8%,p=0.33
	Kebede et al, 202224	3.22 (1.65 to 6.26)
	Girum et al, 201728	3.18 (1.18 to 8.57)
	Alemu et al, 202329	1.87 (1.56 to 2.37)
Shock	Bitew et al, 202121	3.20 (1.56 to 6.34)	4.20 (2.92 to 6.04)	I2=22.5%,p=0.27
	Bitew et al, 202121	7.9 (3.7 to 16.7)
	Wagnew et al, 201815	3.15 (1.5 to 6.5)
	Abiyu et al, 201918	4.15 (2.01 to 8.55)
TB	Oumer et al, 202111	2.68 (1.08 to 7.63)	1.86 (1.35 to 2.56)	I2=0.0%,p=0.43
	Ashine et al, 202020	1.62 (1.1 to 2.37)
	Gebremichael et al, 201423	2.45 (1.23 to 4.89)

F-100, formula-100; IV, Intravenous; NG, nasogastric; SAM, severe acute malnutrition; TB, tuberculosis; TFUs, therapeutic feeding units.

Discussion

This systematic review and meta-analysis provided available evidence on the incidence of mortality and its predictors among under-five children with SAM admitted to TFUs in Ethiopia. The meta-analysis revealed that the mortality rate among these children was 8.32 per 1000 person-days of observation. The cumulative meta-analysis also indicated that the mortality rate has remained unchanged over time, despite the addition of new studies. These findings highlight the ongoing challenges in the management of SAM within Ethiopia’s health system.

Regarding the predictors, the hazard of death was higher in HIV-infected children compared with those without HIV. This finding is consistent with previous systematic reviews and meta-analyses in sub-Saharan African countries and Africa.^{31 32} This is due to the fact that children with SAM and HIV often face infections and complications that make treatment harder, slow recovery and increase the risk of mortality.³³ Similarly, children with TB had a higher hazard of death compared with their counterparts. This is consistent with a nationwide study conducted in Zambia.³⁴ This could be due to the effects of TB, which limit dietary intake, weaken the immune system and lead to severe complications and death.³⁵ The association between HIV, TB and child mortality shows the need to improve integrated screening and treat children with SAM in TFUs.

Children treated with IV fluids had a higher hazard of death compared with their counterparts. This could be due to secondary complications of IV fluid, like fluid overload, infection, heart failure and cerebral oedema.³⁶ Similarly, children receiving NG-tube feeding had a higher hazard of death compared with those fed orally, which is consistent with a systematic review and meta-analysis conducted elsewhere.³¹ This may be due to poor management of NG tubes, which increases the risk of complications like aspiration pneumonia and bacterial infections, potentially leading to death. However, it should be noted that the association between IV fluids, NG tube feeding and mortality likely reflects the critical condition of children at admission rather than the direct impact of these interventions.³⁷

This systematic review and meta-analysis also revealed that children with impaired consciousness had a higher hazard of death compared with those who were fully conscious. Impaired consciousness in children with SAM may result from dehydration or infections and can lead to death. This highlights the importance of early identification and appropriate management of unconsciousness. In addition, children experiencing shock had a higher risk of death. This finding is in line with a previous systematic review.³¹ In children with SAM, shock often results from dehydration, infection or sepsis, conditions that are life-threatening and can lead to death.³⁸ This highlights the need for early detection and quick stabilisation of severely ill children with SAM to improve survival. Furthermore, children with abnormal body temperatures (hypothermia or hyperthermia) had a higher hazard of death. In children with SAM, abnormal body temperatures often result from severe metabolic dysfunction or infection, which significantly increases the risk of mortality.³⁹

Finally, the findings emphasise the importance of nutritional management. Children who did not receive F-100 had a higher hazard of death. This might be due to the high calorie and protein content of F-100, which supports weight gain, and tissue repair and improves recovery. This highlights that ensuring the proper provision of F-100 in inpatient care is essential for children’s recovery.

In summary, this systematic review and meta-analysis estimated the pooled mortality rate among under-five children with SAM admitted to TFUs in Ethiopia. The analysis also assessed trends in mortality rates over time, revealing that mortality rates have remained unchanged. This hinders the nation’s progress towards achieving the SDGs and highlights the need to improve nutrition policies, implementation and healthcare service delivery. The meta-analysis also identified key predictors of mortality. Most of the identified factors were complications and comorbidities associated with SAM, as well as nutritional management. These findings indicated the need to strengthen efforts in providing timely and effective management of SAM complications, ensuring integrated care for comorbidities, and enhancing nutritional support services.

Limitations of the study

A key limitation of this study is the high heterogeneity observed across the included studies. Therefore, we encourage readers to interpret the pooled findings with caution and to give particular attention to the subgroup analyses, which provide more specific and contextually relevant insights. Additionally, the analysis included data from only seven regions, which may limit the generalisability of the findings to other areas. Furthermore, comparing our findings with those of other systematic reviews, meta-analyses or nationwide studies is challenging due to the scarcity of published research in this field.

Conclusion

The pooled mortality rate among under-five children with SAM admitted to TFUs in Ethiopia was found to be 8.32 per 1000 person-days of observation. HIV infection, TB, IV fluid use, NG tube use, abnormal body temperature, impaired consciousness and shock were risk factors for mortality. On the other hand, F-100 supplementation was a preventive factor. Policymakers need to re-evaluate existing nutrition policies, as child mortality rates have shown no meaningful improvement over time despite ongoing intervention. In addition, healthcare providers should prioritise targeted interventions that address the identified risk factors.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.