Relapse of severe acute malnutrition post-discharge from outpatient programs among children recovered from complicated SAM in a conflict-affected setting: a matched cohort study, Sana’a, Yemen

Abstract

Background

Little evidence exists on post-discharge outcomes and the sustainability of recovery among children recovering from complicated severe acute malnutrition (SAM). This study assessed the incidence and predictors of SAM relapse and patterns of nutritional status over time among children discharged from outpatient therapeutic programs (OTPs) in Sana’a, Yemen.

Methods

A prospective cohort study followed 148 children (74 post-SAM and 74 community controls) at 3, 6, and 9 months post-discharge from OTP. Incidence rates of SAM and moderate acute malnutrition (MAM) were calculated. Poisson regression with generalized estimating equations estimated incidence rate ratios (IRRs) for SAM and MAM. Among post-SAM children, competing-risk regression was performed, treating first MAM as a competing event for SAM relapse, to estimate sub-hazard ratios (SHRs).

Results

At 6 months, 36% of post-SAM children experienced SAM compared with 9% of controls, rising to 45% versus 14% at 9 months. MAM occurred in 58% of post-SAM children versus 20% of controls. Adjusted Poisson regression showed higher SAM incidence among post-SAM children (IRR = 2.57; 95% CI: 1.11–5.93), with incidence peaking at six months (IRR = 2.20; 95% CI: 1.12–4.33). Lower baseline weight-for-height z-score (WHZ) was protective (IRR per 1-unit increase = 0.70; 95% CI: 0.51–0.96), while fever increased SAM incidence (IRR = 2.60; 95% CI: 1.59–4.20). MAM incidence did not differ significantly between groups (IRR = 1.60; 95% CI: 0.86–2.97) but was predicted by fever (IRR = 2.20; 95% CI: 1.43–3.37) and diarrhea (IRR = 1.70; 95% CI: 1.20–2.48). Competing-risk regression confirmed lower baseline WHZ (SHR = 0.67) and fever (SHR = 2.41) as predictors of SAM relapse. The hazard of relapse to SAM was significantly lower at nine months compared with three months (SHR = 0.27).

Conclusion

Children recovering from complicated SAM remain vulnerable to recurrent acute malnutrition after discharge. These findings support extended post-discharge monitoring, early management of acute illnesses, and better integration of MAM services within routine care, particularly in fragile settings. Further research is needed to guide optimal discharge criteria and post-discharge support strategies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40795-026-01299-z.

Introduction

Severe acute malnutrition (SAM) remains a critical global public health concern, affecting an estimated 12–14 million children under five, according to the latest Joint Malnutrition Estimates [1]. Management of SAM follows differentiated care pathways based on clinical presentation. Children with complicated SAM, those presenting with medical complications, severe edema, anorexia, or failing an appetite test, require inpatient stabilization at therapeutic feeding centers (TFCs). Once stabilized, they are referred to outpatient therapeutic programs (OTPs) for continued treatment of uncomplicated SAM, defined as the absence of medical complications and the ability to pass the appetite test. After discharge from OTPs, children who do not meet full anthropometric recovery criteria, or who remain at risk of relapse, may be referred to Targeted Supplementary Feeding Programs (TSFPs), where supplementary rations are provided to manage or prevent moderate acute malnutrition (MAM) [2]. While programmatic advances have improved short-term survival and recovery, the post-discharge period remains marked by high nutritional vulnerability.

Evidence from diverse settings, including Pakistan, Nigeria, Ethiopia, Malawi, Zimbabwe, and Nepal indicates relapse after recovery from uncomplicated SAM, with rates ranging from 7.4% to 37%. Definitions of relapse vary across studies, and reported predictors include suboptimal anthropometric status at discharge, rural residence, inadequate infant feeding practices, and socioeconomic constraints [3–7]. A recent systematic review and meta-analysis reported substantial variability in post-treatment SAM relapse, with cumulative incidence across 26 studies ranging from 0.3% to 75.8%, highlighting wide heterogeneity in risk across settings and programs [8]. However, evidence on relapse following treatment for complicated SAM, particularly in fragile and conflict-affected contexts such as Yemen, remains limited.

Previous work among children admitted to TFCs in Sana’a and later managed in OTPs revealed gaps in recovery measurement. Using WHO discharge criteria, which define recovery as either weight-for-height z-score (WHZ) ≥ − 2 or and mid-upper arm circumference (MUAC) ≥ 12.5 cm, 96% of children were classified as recovered, yet only 38% achieved full recovery on both indicators without edema [9]. A longitudinal follow-up showed a median time of 14 weeks (range 12–18) to full recovery, with longer durations among children with concurrent WHZ < − 3 and MUAC < 11.5 cm at admission, those not receiving antibiotics, or those experiencing diarrhea during OTP care [10]. Building on this cohort, the present study assessed relapse to severe and moderate acute malnutrition at 3, 6, and 9 months post-discharge, identified key predictors, and aimed to provide evidence to guide post-discharge care among children initially admitted to TFCs for complicated SAM and subsequently discharged from OTPs.

Method

Study design and setting

This prospective matched cohort study was conducted in Sana’a City, Yemen. It followed two cohorts of children over a 9-month period. The post-SAM group included children previously treated for complicated SAM according to WHO guidelines, who received inpatient stabilization at TFC followed by OTP care and achieved full anthropometric recovery. The control group consisted of age- and sex-matched children from the same communities with no prior history of acute malnutrition in the previous year. Both groups were followed for 9 months, with assessments at 3, 6, and 9 months post-index date to evaluate nutritional status, relapse to moderate or severe acute malnutrition, and sustained recovery. The index date for children in the post-SAM group was defined as the date of OTP discharge, while for control children it was the date of enrollment, which occurred within one week of the corresponding post-SAM child’s OTP discharge.

Study population

Post-SAM group

The post-SAM group comprised children previously enrolled in a cohort study assessing time to full anthropometric recovery among those admitted with complicated SAM to TFCs and subsequently managed in OTPs in Sana’a City. Children with edema at TFC admission were excluded to avoid potential bias. For each child, the same anthropometric indicator used at TFC admission (WHZ or MUAC) was maintained throughout TFC treatment and at OTP admission, in line with WHO guidelines, which do not specify a preferred indicator for children transferred from TFCs. Children continued care at OTPs until discharge per program criteria, and Targeted Supplementary Feeding Programs (TSFPs) services for MAM were not operational/available during the study period [10].

For this study, children were included in the post-SAM group if they had completed both TFC and OTP care, were discharged as cured, were free of edema, achieved full recovery of both anthropometric indicators at OTP discharge (WHZ ≥ − 2 and MUAC ≥ 12.5 cm), resided in the catchment areas of the study hospitals, and had caregivers who consented to follow-up.

Control group

The control group consisted of well-nourished children from the same communities with no prior history of acute malnutrition in the previous year. Well-nourished status was defined as WHZ ≥ − 2, MUAC ≥ 12.5 cm, and absence of edema at baseline. The absence of prior acute malnutrition was confirmed primarily through caregiver recall. Each control child was individually matched to a post-SAM participant by sex, age (within two months), and neighborhood (defined as the catchment area of the study hospitals) to control for key demographic and environmental factors influencing malnutrition risk. Controls were recruited with the assistance of case caregivers and community outreach. Baseline anthropometric measures for controls were obtained at enrollment, which occurred within one week of the matched post-SAM child’s OTP discharge, ensuring comparability of the index date between groups. Children in either group were excluded if they had chronic illnesses (e.g., congenital heart disease, metabolic disorders) or disabilities affecting growth or relapse risk.

Sample size calculation

Sample size was calculated to detect a difference in relapse incidence between post-SAM and control groups, assuming a relapse rate of 27% in the post-SAM group and 8% in controls, based on findings from Somalia, the geographically closest context [11]. Using a two-sided alpha of 0.05, 80% power, and 1:1 matching, a minimum of 62 children per group was required. To account for an estimated 20% non-respondents or loss to follow-up, the sample size was increased to 74 per group (148 total participants).

Participant recruitment and follow-up

Children in the post-SAM group were recruited at the time of OTP discharge as cured, after confirming full anthropometric recovery and absence of edema. Trained research staff conducted in-person interviews with caregivers to confirm eligibility, residence within the study catchment area, and obtain informed consent. Matched controls were recruited within one week of the corresponding post-SAM child’s OTP discharge. Eligible children were identified through community outreach and review of health records, and enrollment occurred only if baseline anthropometric and health criteria were met.

Follow-up assessments were conducted at 3, 6, and 9 months post-index date. Caregivers were asked to bring enrolled children to the OTP facility or study site, with reminders issued via scheduled phone calls. For families unable to attend, assessments were conducted at the nearest OTP facility or through home visits by trained field staff, typically within two weeks of the scheduled follow-up date. At each visit, anthropometric measurements were obtained using standardized procedures, and caregivers were interviewed about recent morbidity, including fever, diarrhea, and cough, using a 2-week recall period. Children who could not be traced after two additional contact attempts or refused participation were considered lost to follow-up.

Study outcomes

The primary outcome of this study was the incidence of severe acute malnutrition (SAM) during follow-up among children who had been stabilized for complicated SAM in therapeutic feeding centers (TFCs) and subsequently discharged as cured from outpatient therapeutic programs (OTPs). Nutritional status was assessed at 3, 6, and 9 months post-discharge from OTPs. SAM was defined as a weight-for-height z-score (WHZ) < − 3 and/or mid-upper arm circumference (MUAC) < 11.5 cm, and/or the presence of bilateral pitting edema. Among children in the post-SAM group, incident SAM represented relapse, whereas among community controls it represented new-onset SAM.

Secondary outcomes included the incidence of moderate acute malnutrition (MAM) during follow-up at 3, 6, and 9 months post-discharge or post-enrollment. MAM was defined as WHZ ≥ − 3 and < − 2, or MUAC ≥ 11.5 cm and < 12.5 cm, in the absence of nutritional edema. In addition, the study assessed predictors of incident SAM, including baseline anthropometric measures (WHZ and MUAC), morbidity indicators (fever, diarrhea, and cough), demographic characteristics (age and sex), and household-level factors such as socioeconomic status, household food security, breastfeeding status, and vaccination status.

Baseline variables

For post-SAM children, baseline WHZ and MUAC were measured at OTP discharge (index date), while for controls, baseline values corresponded to enrollment (within one week of the matched post-SAM child’s OTP discharge).

Baseline child and household characteristics considered as potential predictors of SAM relapse included anthropometric measures, morbidity, feeding, vaccination, and socioeconomic status. Anthropometric measures at discharge included weight-for-height z-score (WHZ) and mid-upper arm circumference (MUAC). Morbidity during follow-up encompassed fever, diarrhea, and cough.

• Breastfeeding status: Children currently receiving breastfeeding who had been breastfed for at least the first six months were classified as breastfed; others were considered not breastfed.

• Vaccination status: A child was considered fully vaccinated if all doses per the national immunization program were received; children who missed any dose or received no vaccines were considered not fully vaccinated.

• Household Socio-Economic Status (SES): Determined using the modified WAMI index (Water/Sanitation, Assets, Maternal Education, and Income) adapted by FANTA [12].

• Household Hunger: Assessed using the Household Hunger Scale (HHS), which includes three questions on food shortage, skipping meals, and going to bed hungry [13].

Statistical analysis

Data were analyzed using Stata version 14.0. Continuous variables were summarized as means (SD) and categorical variables as frequencies and percentages. Baseline characteristics were compared between post-SAM and control groups using t-tests or Wilcoxon rank-sum tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables, as appropriate.

Incidence rates (IRs) of SAM and MAM at 3, 6, and 9 months were calculated per 100 person-months of observation, with person-time defined as the cumulative months contributed by children at risk during each interval. Proportions of children with SAM or MAM at each follow-up visit represent point prevalence. IRs reflect new episodes occurring during each interval; children could contribute multiple episodes if they recovered and subsequently relapsed, but only the first episode within each interval was counted for IR calculation.

Poisson regression models with generalized estimating equations (GEE) and robust standard errors were applied to the full matched cohort (post-SAM and control children) to estimate incidence rate ratios (IRRs) for SAM and MAM, accounting for repeated measurements within individuals. Models included group (post-SAM vs. control), baseline WHZ and MUAC, morbidity (fever, diarrhea, cough), breastfeeding, vaccination status, household food security, and socioeconomic status.

To evaluate predictors of SAM relapse, Fine and Gray competing-risk regression was conducted among post-SAM children only, treating incident MAM as a competing event. Baseline WHZ and MUAC were defined as OTP discharge values for post-SAM children and enrollment values for controls and included as predictors in both Poisson GEE and Fine–Gray models.

Two-sided p-values < 0.05 were considered statistically significant.

Ethics approval and consent to participate

The prospective cohort study was approved by the Research Ethical Committee at the Faculty of Medicine and Health Sciences, Sana’a University, Yemen, (Letter no.389, date 31/7/2023). Official approval from health authorities and hospital administration was obtained. Written informed consent was obtained from all subjects and/or their legal guardian(s). all data were anonymized to ensure patient confidentiality. Children identified with SAM or MAM relapse during follow-up were referred to the nearest outpatient therapeutic program or appropriate health facility. The authors confirm that all methods performed following relevant guidelines and county regulations.

Results

Participants characteristics (baseline)

A total of 148 children were enrolled, including 74 in the post-SAM group and 74 in the control group. Overall, 68 children (46%) were males, with equal proportions in both groups. The mean (± SD) age was similar between the post-SAM and control groups (15.9 ± 8.3 months vs. 16.6 ± 7.9 months; p = 0.593). At baseline, children in the post-SAM group had significantly lower mean height and weight compared with controls (73.2 ± 6.7 cm and 8.7 ± 1.4 kg vs. 76.5 ± 6.5 cm and 9.7 ± 1.4 kg; p < 0.05 for both). MUAC was also significantly lower in the post-SAM group (12.9 ± 0.5 cm vs. 13.9 ± 0.8 cm; p < 0.001). No significant differences were observed between groups for the remaining baseline variables (Table 1).

Table 1.

Baseline characteristics of participants in post-SAM and community control groups, Sana’a City

Characteristics	Post SAM n = 74	Control n = 74	P Value
Gender (Male) n(%)	34(46%)	34(46%)	0.100
Age months (Mean SD)	16.6 ± 7.9	15.9 ± 8.3	0.594
Height in cm (Mean SD)	73.2 ± 6.7	76.5 ± 5.9	< 0.001*
Weight Kg (Mean SD)	8.7 ± 1.4	9.7 ± 1.4	< 0.001*
WHZ (Mean SD)	-0.83 ± 1.5	-0.13 ± 1.4	0.004*
MUAC (Mean SD)	12.9 ± 0.5	13.9 ± 0.8	< 0.001*
Breastfeeding (Yes)n(%)	40(54%)	33(45%)	0.250
Ful Vaccination (Yes) n(%)	46(62%)	45(61%)	0.866
Fever (Yes) n(%)	18(24%)	16(22%)	0.696
Diarrhea (Yes) n(%)	18(24%)	17(23%)	0.847
Cough (Yes) n(%)	20(27%)	21(28%)	0.854
SES index score (Mean SD)	15.7 ± 5.0	14.7 ± 5.3	0.224
Hunger index score (Mean SD)	1.2 ± 1.6	1.2 ± 1.5	0.9157

MUAC Mid-upper arm circumference, WHZ Weight-for-height z-score, SAM Severe acute malnutrition, SD Standard deviation, CI Confidence interval. Fever, diarrhea, cough assessed using a 2-week caregiver recall period

*Significant at p < 0.05. Height presented in centimeters (cm); weight in kilograms (kg); MUAC in centimeters (cm). Recent morbidity (fever, diarrhea, cough) was assessed using a 2-week caregiver recall period

Malnutrition occurrence over 9 months

During the 9-month follow-up, 72 children experienced at least one episode of acute malnutrition (MAM or SAM). Among post-SAM children, 51 (69%) experienced at least one episode, compared with 21 (28%) in the control group. MAM occurred in 58 children overall, including 43 post-SAM (58%) and 15 controls (20%), while SAM occurred in 43 children overall, with 33 post-SAM (45%) and 10 controls (14%). Of these, 29 children (25 post-SAM, 4 controls) experienced both MAM and SAM, reflecting overlap between moderate and severe forms; this overlap explains why the total number of affected children (72) is lower than the sum of MAM and SAM episodes (58 + 43 = 101).

At the end of the 6-month interval, 36% of post-SAM children met criteria for SAM relapse compared with 9% of controls, increasing to 45% versus 14% at 9 months. Similarly, MAM prevalence among post-SAM children increased from 43% at 6 months to 58% at 9 months, compared with 15% and 20% among controls. These figures represent the point prevalence of acute malnutrition at 6 and 9 months, illustrating the dynamic transitions between MAM and SAM among children during follow-up (Supplementary Table 1).

Incidence rates of acute malnutrition

During the 9-month follow-up, incidence rates (IRs) of acute malnutrition were calculated per 100 person-months observation. For post-SAM children, the IRs of MAM) were 6.8, 9.6, and 8.7 per 100 person-months at 3, 6, and 9 months, respectively. The IRs of SAM were 3.6, 9.6, and 4.3 per 100 person-months at the same time points. For controls, the IRs of MAM were 2.7, 2.5, and 2.1 per 100 person-months, while IRs of SAM were 1.4, 1.9, and 1.5 per 100 person-months at 3, 6, and 9 months, respectively. Over the entire 9-month period, the total IR for post-SAM children was 8.2 (95% CI: 6.1–11.0) for SAM and 5.9 (95% CI: 1.9–8.0) for MAM, compared with 2.4 (95% CI: 1.5–4.0) for SAM and 1.6 (95% CI: 0.9–2.0) for MAM in controls. Differences between groups were Statistically significant (Table 2).

Table 2.

Severe and moderate acute malnutrition incidence rates among post-SAM children and control groups, Sana’a City

Groups	Time (months)	Severe Acute Malnutrition			Moderate Acute Malnutrition
		Person Time	No.	IR (95%CI)	Person Time	No.	IR (95%CI)
Post SAM	3	222	15	6.8 (4.1–11.2)*	222	8	3.6 (1.9–7.2)*
	6	177	17	9.6 (6.0–15.4)*	198	19	9.6 (5.9–15.0)*
	9	126	11	8.7 (4.8–15.8)*	141	6	4.3 (1.9–9.0)*
	Total	525	43	8.2 (6.1–11.0)*	561	33	5.9 (3.9–8.0)*
Control	3	222	6	2.7 (1.2–6.0)	222	3	1.4 (0.9–4.0)
	6	204	5	2.5 (1.0–5.9)	213	4	1.9 (0.9–5.0)
	9	189	4	2.1 (0.8–5.6)	201	3	1.5 (0.9–4.0)
	Total	615	15	2.4 (1.5–4.0)	636	10	1.6 (0.9–2.0)

IR calculated as the number of new episodes per 100 person-months of observation. Post-SAM and control children were followed for 3, 6, and 9 months post-index date. Person-time represents the cumulative months contributed by children at risk at the beginning of each follow-up interval. Children who recovered and subsequently relapsed could contribute multiple episodes, but only the first episode within each interval was counted for IR calculation

IR Incidence rate per 100 person-months, CI Confidence interval, SAM Severe acute malnutrition, MAM Moderate acute malnutrition

* Statistically significant difference between Post-SAM and Control groups (p < 0.05)

Unadjusted and adjusted incidence of acute malnutrition

During the 9-month follow-up, post-SAM children experienced a higher incidence of both SAM and MAM compared with controls. Crude incidence rate ratios for SAM ranged from 2.5 (6.8/2.7 at 3 months) to 4.1 (8.7/2.1 at 9 months), while for MAM they ranged from 2.6 (3.6/1.4 at 3 months) to 5.1 (9.6/1.9 at 6 months) among post-SAM children (Table 2).

In adjusted Poisson regression models accounting for repeated measures, post-SAM children had a significantly higher incidence of SAM compared with controls (IRR = 2.57; 95% CI: 1.11–5.93; p = 0.027), whereas MAM incidence did not differ significantly between groups (IRR = 1.60; 95% CI: 0.86–2.97; p = 0.138). Lower baseline WHZ was associated with an increased risk of subsequent SAM, with each unit increase in WHZ reducing the incidence of SAM (IRR per 1-unit increase = 0.70; 95% CI: 0.51–0.96; p = 0.029). Other variables, including baseline MUAC, cough, household SES, food security, breastfeeding, and vaccination status were not significantly associated with SAM or MAM incidence.

Time-dependent effects showed that SAM incidence was significantly higher at six months compared with three months post-discharge (IRR = 2.20; 95% CI: 1.12–4.33; p = 0.023) and remained elevated at nine months, although this difference did not reach statistical significance (IRR = 1.90; 95% CI: 0.96–3.81; p = 0.065). In contrast, MAM incidence increased significantly at nine months (IRR = 1.90; 95% CI: 1.23–3.06; p = 0.005), predicted by fever (IRR = 2.2; 95% CI: 1.43–3.37; p < 0.001) and diarrhea (IRR = 1.7; 95% CI: 1.20–2.48; p = 0.003). These results are detailed in Supplementary Table 2.

Predictors of SAM relapse (competing-risk regression accounting for MAM)

Competing-risk regression using the Fine and Gray sub distribution hazard model was performed among post-SAM children only, treating the first occurrence of MAM as a competing event for SAM relapse. Children who progressed directly from normal to SAM without a preceding MAM episode were considered to have experienced the primary event. Controls were not included in the relapse analysis, as any SAM occurring in this group represents incident SAM rather than relapse.

In these models, lower baseline WHZ at discharge was significantly associated with increased risk of SAM relapse (SHR = 0.67 per 1-unit increase; 95% CI: 0.46–0.99; p = 0.045), indicating that children with lower WHZ at recovery were more vulnerable to subsequent deterioration. Fever during follow-up was a strong predictor of relapse (SHR = 2.41; 95% CI: 1.12–5.16; p = 0.024). The risk of SAM relapse was significantly lower at 9 months compared with 3 months (SHR = 0.27; 95% CI: 0.10–0.74; p = 0.010), whereas no significant difference was observed at 6 months. Baseline MUAC, cough, diarrhea, household food security, SES, breastfeeding status, and vaccination status were not statistically significant predictors (Table 3).

Table 3.

Predictors of SAM relapse among post-SAM children accounting for MAM as a competing event, Sana’a City, Yemen

Variable	SHR	95% CI	p-value
Baseline WHZ (per 1-unit increase)	0.67	0.46–0.99	0.045*
Baseline MUAC (per 1 cm increase)	1.57	0.84–2.94	0.161
Fever (Yes vs. No)	2.41	1.12–5.16	0.024*
Cough (Yes vs. No)	1.79	0.82–3.92	0.143
Diarrhea (Yes vs. No)	1.22	0.56–2.65	0.615
Period – 6 months vs. 3 months	1.07	0.45–2.55	0.87
Period – 9 months vs. 3 months	0.27	0.10–0.74	0.010*
Food security score (per 1-point increase)	1.04	0.86–1.26	0.708
SES total score (per 1-point increase)	0.93	0.87–1.00	0.063
Breastfeeding (No vs. Yes)	0.91	0.43–1.89	0.794
Vaccination Partial vs. Full	1.69	0.81–3.50	0.161
Vaccination Unvaccinated vs. Full	0.89	0.18–4.34	0.89

SHR Sub-distribution hazard ratio, MUAC Mid-upper arm circumference, SES Socioeconomic status, WHZ Weight-for-height z-score, SAM Severe acute malnutrition, CI Confidence interval

*Statistically significant (p < 0.05)

Discussion

In this study, we found that children who had initially presented with complicated SAM and subsequently transitioned from TFCs to OTPs remained at substantially higher risk of SAM compared with community controls. This highlights that post-SAM children constitute a particularly vulnerable group, even after being classified as recovered at discharge. Although both post-SAM and control groups were classified as non-acutely malnourished at enrollment, post-SAM children had consistently lower anthropometric indicators than controls, reflecting residual deficits from prior complicated SAM. Similar associations between lower anthropometric status at discharge and relapse have been reported in other settings [11, 14–16].

Our findings show that at 6 months, SAM occurred among 36% of post-SAM children compared with 9% of controls, increasing to 45% versus 14% at 9 months. Although relapse was substantially higher among post-SAM children, the occurrence of SAM in controls, despite no documented prior episode, highlights the intensity of underlying risk in this setting. Children in both groups were exposed to the same fragile food, health, and sanitation environment, including limited functional MAM programming and constrained access to preventive services, which may partly explain the non-negligible burden observed among controls. Controls were matched by age, sex, and neighborhood, but not by household SES or other risk factors. Additionally, anthropometric measurements were collected using standard procedures; children classified as SAM met WHZ < -3 and/or MUAC < 11.5 cm criteria. These findings highlight the high vulnerability of children in this setting to acute malnutrition even among those without prior SAM exposure. These children had initially presented with complicated SAM and were discharged from OTPs, which may partly explain the higher incidence compared with children recovering from uncomplicated SAM in other settings. Reported post-discharge SAM relapse or incidence rates among children treated for uncomplicated SAM are generally lower, including 22% in Somalia, 30% in Mali [17], 24% in Nigeria [18], and 7–8% in Pakistan [3].

Recent evidence from Mali further demonstrated substantial post-discharge vulnerability, with a high burden of recurrent acute malnutrition despite programmatic recovery, emphasizing that relapse risk persists well beyond discharge even in settings with active nutrition services [15]. Nonetheless, the rates observed in our study were lower than the 63% reported from South Sudan, a context characterized by extreme food insecurity and limited post-discharge support.

Considering that our study population consisted exclusively of children recovering from complicated SAM underscores their heightened vulnerability and the need for targeted post-discharge interventions. The wide variability in reported relapse or incidence across studies likely reflects important differences in case severity (complicated vs. uncomplicated SAM), discharge criteria, follow-up duration, and contextual factors such as food security, disease burden, and health system capacity. This variability is consistent with findings from the recent meta-analysis which documented substantial heterogeneity in post-discharge outcomes across settings and highlighted the absence of standardized post-SAM care as a critical gap in current nutrition programming [8, 19].

In this study, episodes of MAM were more frequently observed than SAM in both groups at the descriptive level, with higher proportions among post-SAM children compared with controls. Notably, there was substantial overlap between MAM and SAM in the post-SAM group, reflecting recurrent transitions between nutritional states. In the absence of targeted supplementary feeding programs (TSFPs) for MAM children in the community, these transitions likely represent the natural course of post-SAM nutritional vulnerability rather than programmatic response [5].

Temporally, the relapse to SAM among post-SAM children increased from three months and peaked at six months post-discharge, while MAM incidence remained persistently elevated over the follow-up period. Although SAM incidence showed a tendency to decline by nine months, this reduction was not statistically significant. The apparent stability of MAM over time likely reflects dynamic bidirectional transitions, whereby some children with MAM deteriorated to SAM while others recovered to normal anthropometry, resulting in a relatively stable net MAM burden. This pattern underscores ongoing nutritional fragility among post-SAM children in the absence of structured post-discharge monitoring and supplementary interventions and contrasts with the consistently lower and more stable rates observed among controls.

Adjusted Poisson regression models accounting for repeated measures indicated that children in the post-SAM group had a significantly higher incidence of SAM compared with controls (IRR = 2.57), whereas the incidence of MAM did not differ significantly between groups after adjustment. SAM incidence was significantly higher at six months compared with three months (IRR = 2.20), while MAM incidence increased significantly at nine months (IRR = 1.90). Together, these findings reinforce evidence from Ethiopia and other contexts that children recovering from SAM remain nutritionally fragile for prolonged periods following discharge [11, 14].

Fever emerged as a significant predictor of both SAM and MAM, while diarrhea was associated primarily with MAM. These findings reflect the complex interplay between infection and nutritional status. Acute illnesses can impair nutrient intake, increase metabolic demands, and induce systemic inflammation, oxidative stress, and gut barrier dysfunction, thereby compromising nutritional recovery and facilitating regression to malnutrition [20]. Persistent gut inflammation and epithelial damage following SAM may further exacerbate vulnerability, particularly among children who transition through MAM during follow-up [21, 22].

Our competing-risk analysis revealed that lower WHZ at discharge significantly increased the risk of SAM relapse, highlighting that children who recover with lower anthropometric reserves remain vulnerable to deterioration. This highlighting the role of residual nutritional deficits at discharge in shaping post-recovery vulnerability. This finding aligns with evidence from a multi-country prospective cohort study showing that poor growth status is associated with repeated relapse to acute malnutrition [11], and underscores the importance of critically revisiting recovery criteria and strengthening post-discharge strategies to mitigate long-term risk [15]. Recent fever within the two weeks preceding follow-up visits was independently associated with an increased risk of SAM relapse, reinforcing the role of acute illness as a proximal trigger for nutritional deterioration among post-SAM children. This finding is consistent with studies from Sudan and other contexts, where febrile illness has been identified as a predictor of relapse following recovery from SAM [15, 17]. Although the etiology of fever was not systematically classified in this study, likely causes in Sana’a include respiratory infections, gastroenteritis, or sepsis rather than malaria, given the local ecological context. These findings highlight that acute infections of various origins may exacerbate nutritional vulnerability in children recovering from SAM, emphasizing the importance of preventive and responsive strategies such as high vaccination coverage, prompt treatment of infections, and early care-seeking behavior to reduce relapse risk.

The result indicated that the hazard of SAM relapse decreased markedly by 9 months post-discharge compared with 3 months, indicating that the early post-discharge period represents the highest-risk window for relapse. This temporal pattern underscores the importance of close monitoring in the first few months after OTP discharge.

In contrast, other factors, including baseline MUAC, cough, diarrhea, breastfeeding, vaccination status, household food security, and SES, were not statistically significant, though SES showed a borderline protective effect. These findings suggest that while structural and socio-environmental factors remain important, individual nutritional status and acute morbidity may be the strongest determinants of relapse in this conflict-affected setting. This is consistent with findings from a systematic review indicating that while numerous factors have been proposed as predictors of SAM relapse, the strength and consistency of associations remain highly context-specific and inconclusive [17]. It is also important to note that both fever and diarrhea were assessed using caregiver recall over a two-week period, which may be subject to misclassification and recall bias, particularly for diarrhea, where symptom recognition and reporting can vary. These measurement limitations may partly explain the weaker or null associations observed for some morbidity indicators.

Strengths and limitations

The prospective cohort design allowed us to track both SAM and MAM episodes over nine months, capturing dynamic nutritional transitions and relapse patterns. The use of multiple anthropometric indicators, including WHZ, MUAC, and the presence of edema, enabled a more comprehensive assessment of recovery and relapse than reliance on a single measure. This approach reflects the multidimensional nature of SAM relapse, which is not adequately captured by a single indicator. Conducting the study in a conflict-affected, resource-limited setting adds further value, as the findings are directly relevant to humanitarian programming and post-discharge care strategies.

However, the study has several limitations. Children with edema at TFC admission were excluded, which may bias the sample toward less severe initial disease. Only children who achieved full recovery on both WHZ ( ≥ − 2) and MUAC (≥ 12.5 cm) at OTP discharge were included, which may underestimate relapse risk compared with routine programs that discharge based on a single indicator. Follow-up was restricted to 3, 6, and 9 months post-discharge, so relapses occurring outside these intervals were not captured, potentially affecting comparisons with other studies. Certain potential predictors, such as dietary intake, micronutrient status, and caregiver practices, were not assessed and may have influenced relapse risk. Reliance on caregiver recall for fever, diarrhea, and prior acute malnutrition in control children could introduce reporting bias. Mortality was not included as an outcome, as no deaths occurred among post-SAM children with full anthropometric recovery, though deaths were observed among children discharged prematurely from TFCs/ OTP. Classification of controls as having “no prior acute malnutrition” was based on caregiver recall, which may be imperfect. Despite strong neighborhood matching, residual confounding from unmeasured household-level factors such as variability in food access, water, sanitation, hygiene (WASH), and caregiving practices, may persist. These factors could slightly inflate or underestimate the observed contrasts in SAM and MAM incidence between post-SAM and control children, and should be considered when interpreting the findings.

Conclusion

Children recovering from complicated SAM and discharged from OTPs remain at high risk of relapse, particularly to SAM, during the first six months post-discharge. Post-SAM children exhibit persistent anthropometric deficits and transitions between MAM and SAM, reflecting ongoing nutritional fragility. Key predictors of relapse include lower baseline WHZ and the presence of acute illnesses such as fever, while household and sociodemographic factors appeared less influential.

These findings provide supporting evidence for extended post-discharge monitoring and integrated illness management, particularly in the first six months. Although children with lower WHZ at discharge appear more vulnerable, further research is needed before any specific changes to discharge criteria or anthropometric targets can be recommended. This underscores opportunities for future studies to optimize post-discharge care strategies in fragile and conflict-affected settings.

Abbreviations

CI

Confidence Interval

Cm

Centimeter

GEE

Generalized Estimating Equations

IRR

Incidence rate ratios

Kg

Kilogram

MUAC

Mid-upper arm circumference

NAM

Non-Acute malnutrition

OTP

Out-Patient Therapeutic Program

SAM

Severe Acute Malnutrition

SD

Standard deviations

SES

Socio-Economic Status

SHR

Sub-hazard ratios

TFC

Therapeutic Feeding Center

WHO

World Health Organization

WHZ

Weight for height Z score

Authors’ contributions

MAA contributed to the conception, design study, methodology, and data analysis, and wrote the draft manuscript. KA contributed to the results interpretation and writing the draft manuscript. YAR reviewed the overall study, interpreted the results, and revised the final manuscript. All authors read and approved the final manuscript.

Funding

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Data availability

All relevant data are presented in this paper, and more information can be provided upon reasonable request from the corresponding author.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the principles of the Declaration of Helsinki and relevant national regulations. Ethical approval was obtained from the Research Ethics Committee of the Faculty of Medicine and Health Sciences, Sana’a University, Yemen (Letter No. 389, dated 31 July 2023). Written informed consent was obtained from all participants and/or their legal guardians. All data were anonymized to ensure confidentiality.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.