Seasonality and risk factors of relapse to acute malnutrition: a multi-country prospective cohort study in Mali, South Sudan, and Somalia, 2021–2022

Abstract

Despite the effectiveness of community-based management of acute malnutrition (CMAM) in treating severe acute malnutrition (SAM), recent data show that up to 60% of treated children relapse into acute malnutrition (AM) within six months. A key gap in understanding relapse is the limited evidence on seasonal timing and drivers. We analyzed rolling study enrollment data from April 2021 to November 2022, tracking 1,689 recently recovered SAM children for six months, compared to 1,060 same-sex, age-matched counterparts without AM in the past year. The results show that AM and relapse are highly seasonal, with the highest prevalence during transitions between dry and rainy seasons, and rainy and harvest periods. The seasonal AM peaks occurred earlier, at greater amplitude, and for longer durations for recently recovered SAM children than their matched counterparts, highlighting their increased vulnerability and need for targeted follow-up prior to known seasonal peaks to maximize cost-efficiency amid limited funding.

Supplementary Information

The online version contains supplementary material available at 10.1057/s41271-026-00648-3.

Key messages

• Acute malnutrition (AM) and relapse to AM are highly seasonal with peaks occurring during the transition periods between the dry and rainy seasons and the rainy and harvest seasons allowing for more time-targeted programs.

• Previously SAM children tended to become AM about one and a half months earlier than children who have not been previously AM, indicating their greater vulnerability to shock.

• Seasonal peaks of AM are more intense and last longer for children who have previously experienced SAM, suggesting vulnerability remains elevated even when the general population benefits from improved nutritional conditions.

Supplementary Information

The online version contains supplementary material available at 10.1057/s41271-026-00648-3.

Introduction

The community-based management of acute malnutrition (CMAM) model integrates inpatient care, home-based treatment, and ready-to-use therapeutic foods to address acute malnutrition (AM) in children. Due to CMAM’s success, the World Health Organization (WHO) endorses CMAM as the standard of care for treating AM [1]. However, despite its effectiveness, children recovering from severe acute malnutrition (SAM) under CMAM remain vulnerable to relapse to AM, often requiring additional treatment. Studies report relapse rates from 0 to 63%, depending on the context and program [2–5]. Recent multi-country research found that children who have experienced SAM are up to six times more likely to develop AM than those who have not had AM in the past year [2].

While it is well established that AM is seasonal, there is limited evidence on the seasonality of relapse to AM. Studies from India and Malawi show that relapse rates increase during the rainy seasons [6, 7]. However, time within relapse studies is mainly analyzed as months since discharge or as a categorical season (dry vs rainy) rather than as a continuous function of the calendar year, which can obscure finer seasonal trends [8]. Recent studies on AM have uncovered unexpected seasonal peaks, challenging traditional views of AM seasonality. A study in Chad as well as secondary data analysis across the Sahel identified two annual peaks of AM—one at the end of the dry season and another at the beginning of the harvest period [9–11]. Additionally, research on the drivers of relapse-related AM compared to AM cases without a previous episode of AM is scarce. The most recent study identified severe household food insecurity as measured by the Household Hunger Scale (HHS) and having diarrhea in the past 7 days at time of follow-up as significant risk factors for relapse [2]. Few other individual or household-level risk factors at the time of recovery or discharge were associated with subsequent relapse [2].

This study aims to identify the seasonal peaks of AM relapse in children discharged from CMAM and compare them to the timing of AM in children who have not experienced AM in the past year. Additionally, the study explores how these patterns align with food insecurity and diarrhea seasonal trends, examining their relationship to AM.

Data and methods

Study setting and AM services

The study was conducted in three protracted humanitarian contexts: Mali, Somalia, and South Sudan.

Data collection in Mali occurred in Kayes which is generally secure, but experiences frequent flooding, droughts, as well as price inflation. A 2020 survey in Kayes showed the global acute malnutrition (GAM) prevalence at 5.6, with 0.3% consisting of SAM [12]. The food insecurity situation was assessed as Phase 1 (Minimal) by the Integrated Food Security Phase Classification (IPC) Analysis in June 2022 [13].

The rural South Sudan site was in Aweil East County in the state of Northern Bahr el Ghazal. Aweil East had GAM and SAM prevalences of 13.1 and 2.6%, respectively [14] in 2021. The county is routinely classified as IPC Phase 3 (Crisis) or IPC Phase 4 (Emergency), signifying chronic widespread food insecurity and high rates of acute malnutrition. The region was classified as IPC Phase 3 in June 2022 [13].

The Somalia data come from an urban internally displaced camp in the capital Mogadishu. Somalia experienced near-famine conditions that led to wide-scale displacement with Mogadishu housing approximately 500,000 internally displaced persons (IDPs). In 2020, GAM prevalence was 9%, and SAM prevalence was 1.3% [15]. The IPC analysis in June 2022 classified the Banadir Region, where Mogadishu is located, as Phase 4 [16].

Acute malnutrition services in Mali were provided by the Ministry of Health; in South Sudan they were provided by Action Against Hunger; and in Somalia, services were provided by both. Clinics in Mali provided general primary care such as vaccinations, growth monitoring, treatment of illnesses, and so on. The South Sudan and Somalia clinics were traditional nutrition facilities that provided care within the CMAM package, including care for malaria, pneumonia, and diarrhea, but provided no additional support. In all three contexts, other assistance programs were available around water and sanitation, food security, and livelihood support.

SAM treatment services for outpatient treatment (OTP) provided ready-to-use therapeutic food (RUTF). Management of moderate acute malnutrition (MAM) services were also available in every country with some differences in product. In Mali, RUTF was used for MAM as well, while in South Sudan, management of MAM included a combination of ready-to-use supplementary food (RUSF) or super cereal (CSB); and in Somalia, it included only RUSF. For SAM, follow-up visits occurred weekly, while for MAM, they occurred fortnightly. The criteria for admissions for SAM treatment and MAM management also slightly varied by country. Somalia was a mid-upper arm circumference (MUAC)-only country with a child admitted for SAM treatment if they had MUAC < 115 mm and for MAM management if they had a MUAC between 115 and 124 mm, inclusive. In Mali and South Sudan, the child could either be admitted on MUAC (same criteria as Somalia) or on weight-for-height Z-scores (WHZ), with children who have WHZ < − 3 admitted to SAM treatment and children with WHZ between − 3 and − 2 for MAM management.

Study design and data

Sixteen health facilities (9 in Mali, 6 in South Sudan, and 1 in Somalia) were purposefully selected based on high SAM caseloads and accessible CMAM programs. In each country, uncomplicated SAM children who were never admitted to inpatient treatment and discharged as recovered from CMAM between the ages of 6–47 months with no chronic or congenital disease/disability were recruited and enrolled in the study. We refer to these children as the ‘previous SAM recovery’ cohort. This cohort was matched based on age, sex, and location to children who had not had an episode of AM in the prior year and were referred to in this paper as the ‘no AM in the year prior’ cohort. Recruitment of the control children occurred through referrals from trained community health workers (CHWs) performing regular nutrition screening in the communities.

Children were assessed at enrollment and in six monthly follow-up visits. At each visit, enumerators collected information on the household and child, including anthropometry (MUAC, WHZ, and edema), morbidity, and food insecurity. The primary outcome of the study was whether a child became AM, defined as having a WHZ less than or equal to − 2, MUAC below 125 mm, or the presence of nutritional edema. Thus, if a child met this criterion in the ‘previous SAM recovery’ cohort, they were defined as ‘relapsed’ and a child who met this criterion in the ‘no AM in the year prior’ cohort was defined as having an initial AM episode. Food insecurity was assessed via the HHS which measures experiential food insecurity over the past month [17]. For the seasonality models, we reclassified it as a binary variable indicating food insecurity if the household was moderately or severely hungry vs little to no hunger over the past month. For morbidity, we used whether the child had diarrhea in the past seven days. During follow-up, children who developed AM were referred for treatment but remained in the study.

The primary aim of the original study was to calculate the relative risk of relapse [2]. The final sample size included 1,689 SAM children discharged from CMAM and 1,060 children who had not had AM in the past year. A detailed description of the study protocol is available [18]. Because of rolling enrollment, data collection spanned almost two years from April 2021 through November 2022. The start and end of data collection and sample size varied slightly by country and for the two cohorts (Table S1). Other than small differences in timing, data collection was identical across cohorts.

We separately incorporated area-averaged time-series data for the study period on daily accumulated precipitation [19], daily temperature [20], and a monthly greenness fraction measure as a proxy for vegetation from the study period [21].

Nutritional status was classified at clinics for z-score-dependent indicators and confirmed using the WHO’s 2006 Child Growth Standards via the zscore06 Stata package [22, 23]. For any children who missed a follow-up visit during months 2–5, we used proximal linear interpolation between non-missing data points for height, weight, and MUAC measurements [24].

Ethical approval was provided by Solutions Institutional Review Board (#20200310), London School of Hygiene and Tropical Medicine’s (LSHTM) Research Ethics Committee (#18059), Ministry of Health Somalia (MOH&HS/DGO/0429/03/202), the Université Des Sciences, Des Techniques Et Des Technologies De Bamako (2020/202/CE/FMOS/FAPH) in Mali, and the Ministry of Health of South Sudan (MOH/ERB6/2020). Written informed consent was obtained from all participants’ caregivers. All caregivers were provided compensation at each completed visit.

Statistical analyses

As a sensitivity analysis, we looked at WHZ and MUAC as continuous variables, as well as a sub-analysis including only children who experienced at least one episode of AM during the study period. To identify seasonal peaks of AM in these outcomes, we included four harmonic terms with 2π and 4π sine and cosine terms (Model 1) [26–28]. The use of the sine and cosine terms allows us to capture the presence of repeated patterns in our outcome measure, while the use of both the 2π and 4π terms allows us to test the presence of multiple and non-symmetrical patterns within the same year. To confirm the seasonal pattern, we ran the same model, including only the climatic variables (independently given the possibility of multicollinearity)—daily precipitation (mm), daily temperature (Celsius), and monthly vegetation—without the sine and cosine terms (Model 2).

To identify whether food insecurity and diarrhea were associated with AM, first, we ran Model 1, but with binary food insecurity and diarrhea as the outcome to identify seasonal patterns for these variables. Second, we ran a third model, including diarrhea and food insecurity as independent variables with respect to AM, controlling for child age (in months) and child sex (Model 3). Third, we ran Model 3, but with one-round lagged terms for food insecurity and diarrhea to test the hypothesis that there is a month lag between these variables and child AM (Model 4). We controlled for Round (1–6) and child- and clinic-level fixed effects in every model (labeled location). All analyses were stratified by the ‘previous SAM recovery’ versus the ‘no AM in the year prior’ cohort. We did not test any other drivers. While all models were run for the two years of data, for simplicity, visualizations were made aggregating across one year using categorical descriptions of seasons. Relationships are described as significant if the p-value is less than 0.05. For details on the model formula and components, see Supplemental Methods.

Results

Climate seasonality

Mali and South Sudan have one precipitation peak, followed by a one-month lag in the vegetation peak (Fig. 1). Two temperature peaks surround the rainfall peaks. South Sudan’s rainy season is longer than Mali’s, corresponding to greater vegetation and an earlier and longer harvest. In Somalia, there are two smaller rainfall peaks but limited seasonal variability with two short harvests.

Fig. 1.

Twenty-year (2002–2022) normalized average monthly temperature, vegetation, and precipitation for the study area in Mali, South Sudan, and Somalia with respect to the FEWSNET livelihood calendar and categorical seasonal calendar

Seasonality of relapse and acute malnutrition

While the distribution of child-level observations across months and seasons was not uniform, over the two years we had a more even distribution by month (Table S1 and Fig. 2). The distribution was best for Somalia, followed by South Sudan. In Mali, we had the biggest discrepancy in observations, with slightly fewer July–October.

Fig. 2.

Heatmap of the distribution of individual-level child observations by month and country

For the ‘previous SAM recovery’ cohort, different seasonality emerged across the three countries (Figs. 3 and S1, Tables 1 and S2). In Mali, there was one significant peak (highest prevalence) in AM (41%) in the middle of the rainy season with one significant nadir (lowest prevalence) at the end of the harvest and the start of the dry season (5%). A higher risk of AM corresponded with rainfall and vegetation, while hotter (within-year) temperature was associated with a significantly lower risk. In South Sudan, there were two significant peaks, with the bigger peak (41%) occurring during the transition between the dry and rainy seasons and a smaller peak (37%) occurring at the end of the rainy season. The nadir (27%) was at the end of the harvest. South Sudan also had a nadir between the two peaks corresponding to the middle of the rainy season (28%). In Somalia, there were two significant equal-sized peaks right at the start (12%) and the end (11%) of the Gu rains.

Fig. 3.

Seasonality of acute malnutrition for the ‘no AM in the year before study enrollment’ and ‘previous SAM recovery’ cohorts with respect to rainfall (aggregated across the two years) and categorical seasons

Table 1.

Mixed effects logit model on AM for the ‘no AM in the year before study enrollment’ and ‘previous SAM recovery’ cohorts by country to identify seasonality

		No AM in the year before study enrollment				Previous SAM recovery
		Coefficient	p-value	95% Confidence Interval		Coefficient	p-value	95% Confidence Interval
Mali		n = 2385				n = 2342
	Days since start of study	0.01	0.25	0.00	0.02	0.00	0.06	− 0.01	0.00
	sin2π	− 0.61	0.09	− 1.30	0.08	− 0.39	0.04	− 0.76	− 0.03
	cos2π	0.15	0.72	− 0.68	0.99	− 0.83	0.00	− 1.12	− 0.54
	sin4π	− 0.73	0.00	− 1.22	− 0.23	0.10	0.41	− 0.14	0.34
	cos4π	−0.53	0.07	− 1.10	0.03	− 0.24	0.05	− 0.49	0.00
	Round (1–6)	0.02	0.89	− 0.28	0.32	0.21	0.01	0.06	0.37
	Constant	− 8.89	0.00	− 13.24	− 4.55	− 2.40	0.00	− 4.05	− 0.75
South Sudan		n = 1649				n = 3029
	Days since start of study	0.00	0.76	− 0.01	0.01	0.00	0.04	0.00	0.00
	sin2π	− 0.88	0.02	− 1.62	− 0.15	0.07	0.42	− 0.11	0.26
	cos2π	0.87	0.07	− 0.07	1.80	− 0.17	0.07	− 0.36	0.02
	sin4π	− 0.62	0.05	− 1.23	− 0.01	− 0.03	0.67	− 0.17	0.11
	cos4π	− 0.16	0.61	− 0.77	0.45	− 0.20	0.01	− 0.34	− 0.05
	Round (1–6)	0.53	0.02	0.10	0.95	0.15	0.00	0.06	0.24
	Constant	− 9.98	0.00	− 14.61	− 5.35	− 0.87	0.04	− 1.70	− 0.04
Somalia		n = 2212				n = 4562
	Days since start of study	0.00	0.92	− 0.01	0.01	0.00	0.13	− 0.01	0.00
	sin2π	0.56	0.01	0.13	0.98	0.24	0.10	− 0.04	0.52
	cos2π	0.02	0.92	− 0.42	0.46	− 0.41	0.00	− 0.68	− 0.13
	sin4π	0.20	0.23	− 0.13	0.53	0.10	0.33	− 0.10	0.31
	cos4π	0.33	0.05	0.00	0.67	0.14	0.21	− 0.08	0.35
	Round (1–6)	0.29	0.01	0.08	0.50	0.24	0.00	0.11	0.38
	Constant	− 7.16	0.00	− 9.54	− 4.77	− 5.43	0.00	− 6.79	− 4.08

Country-specific seasonal trends were observed for the ‘no AM in the year prior’ cohort. In Mali, there were two peaks: the mid-rainy season (4%) and the transition between the dry and rainy seasons (4%). Climate analysis partially confirms this, with temperature approaching significance (p = 0.098) and positively associated with increased AM odds.

In South Sudan, the ‘no AM in the year prior’ cohort showed a trend like the ‘previous SAM recovery’ cohort, with two important differences. In the ‘no AM in the year prior’ cohort, the larger peak happened at the end of the rains (8%), while the smaller peak occurred during the dry to rainy season transition (3.2%), the opposite order from the relapse cohort. Both cohorts shared the time of year with the lowest AM prevalence—mid-rainy season (2.3%)—but the ‘no AM in the year prior’ cohort’s lowest AM prevalence happened at the end of the dry season (1.4%). In the ‘no AM in the year prior’ cohort, higher rainfall was associated with lower odds of AM. In Somalia, the seasonal pattern of both cohorts was almost identical, peaking at the start of the rains (8 and 9%, respectively). However, only in the ‘no AM in the year prior’ cohort higher vegetation was associated with significantly lower odds of AM.

We carried out sensitivity analysis (Fig. 4) using WHZ and MUAC for children who had at least one episode of AM during the follow-up. In South Sudan and Mali (Table 2), there was consistent seasonality across the three outcomes. However, in both Mali and South Sudan, the ‘previous SAM recovery’ cohort became malnourished one–two months before the’no AM in the year prior’ cohort. Additionally, the ‘previous SAM recovery’ cohort did not experience the same size of improvement in the amplitude of AM between peaks as was observed in the’no AM in the year prior’ cohort. There was approximately a one-month lag between when children experienced the highest prevalence of AM and the shift of the entire distribution of WHZ and MUAC to the left. In Somalia, there was little consistency across the outcomes.

Fig. 4.

Sensitivity analysis: seasonality of AM for only children who experienced at least one episode of AM (panel 1), the mean WHZ (panel 2), and the mean MUAC (panel 3) for all children with respect to rainfall and categorical seasonal calendar for the ‘no AM in the year before study enrollment’ and ‘previous SAM recovery’ cohorts (aggregated across two years)

Table 2.

Summary of seasonal timing (months/day*) of predicted AM, mean predicted WHZ, and mean predicted MUAC by country and for the ‘no AM in the year before study enrollment’ and ‘previous SAM recovery’ cohorts

		No AM in the year before study enrollment			Previous SAM recovery
		AM	WHZ	MUAC	AM	WHZ	MUAC
Mali	Highest AM/lowest WHZ & MUAC	10/26	10/27	11/8	8/23	10/4	10/30
	Lowest AM/highest WHZ & MUAC	2/6	6/7	7/3	1/6	3/2	6/9
	Second highest AM/lowest WHZ & MUAC	4/30
	Second lowest AM/highest WHZ & MUAC	7/18
South Sudan	Highest AM/lowest WHZ & MUAC	11/10	11/16	11/12	4/17	5/2	5/16
	Lowest AM/highest WHZ & MUAC	7/12	7/28	3/2	12/30	1/15	1/25
	Second highest AM/lowest WHZ & MUAC	5/4	4/14	6/15	9/21	7/23	10/11
	Second lowest AM/highest WHZ & MUAC	3/11	2/12	8/26	7/13	10/4	8/11
Somalia	Highest AM/lowest WHZ & MUAC	2/3	3/5	11/11	6/29	5/19	9/3
	Lowest AM/highest WHZ & MUAC	10/13	10/9	7/8	11/30	10/14	1/27
	Second highest AM/lowest WHZ & MUAC	6/23				3/18
	Second lowest AM/highest WHZ & MUAC	4/27				3/29

*While the model predicts specific days, we would expect that to vary across years

AM acute malnutrition, WHZ weight-for-height z-score, MUAC mid-upper arm circumference

Seasonal drivers of relapse and acute malnutrition: diarrhea and food insecurity

In Mali, diarrhea was significantly associated with greater odds of AM in both cohorts (9 × the odds versus 3 × the odds in the AM vs no previous AM cohort) (Table S3). Diarrhea peaks corresponded to the post-rain peaks in AM in both cohorts (Fig. 5). No significant relationship with food insecurity was found; however, 99.7% of respondents in Mali reported no food insecurity throughout the study. Boys and older children (in the relapse cohort only) were also significantly more likely to become AM.

Fig. 5.

Normalized seasonality of AM, diarrhea, and food insecurity across the ‘no AM in the year before study enrollment’ and ‘previous SAM recovery’ cohorts with respect to rainfall (aggregated across the two years)

In South Sudan, diarrhea and food insecurity were significantly associated with AM in the ‘previous SAM recovery’ cohort only. In this cohort, food insecurity and diarrhea from the previous round were correlated with AM in the next. Every additional month post-discharge was associated with a 17% increase in the odds a child becomes AM.

In Somalia, AM was significantly associated with severe food insecurity for both cohorts, with no relationship to diarrhea. In the ‘previous SAM’ cohort, every additional month post-discharge was associated with an 18% increase in the odds of a child becoming AM, each additional month in age was associated with a 10% increase in the odds, and if the child is a girl, they have 25% lower odds of being AM.

Discussion

Relapse to AM and the first episode of AM after not having AM in the past year were highly seasonal in the unimodal rural Mali and South Sudan context, like recent research on AM [6–11]. However, while we observed significant seasonality in the Somalia cohorts, the differences in timing of high and low AM prevalence were minimal (one percentage point). The muted seasonality in Somalia likely stemmed from the context. The Somalia sample came from an urban IDP camp in Mogadishu with a bimodal rainy season. Previous research has shown that seasonal fluctuations in weight for age and WHZ tend to be smaller in urban settings compared to rural areas [29]. Additional findings also indicate that seasonal variability was smaller in regions with bimodal rains [30, 31] or multiple harvests per year [32]. The drivers of acute malnutrition, like food insecurity, might also be less seasonal in urban areas [33]. The IDP setting could further dampen differences by season as assistance depends on funding availability more than climate.

The patterns of relapse and AM across Mali and South Sudan were fairly similar, with an end of the rainy season/start of the harvest season peak. In South Sudan and in the Mali ‘no AM in the year prior’ cohort, an additional peak occurred at the end of the dry season/start of the rainy season. The twin peaks phenomenon has been observed in the region across several studies [9–11]. In the Sahel, the transition between the dry and rainy seasons is a very short but difficult season, when access to water is limited and animals are at their leanest, affecting milk consumption [11]. This research highlights how AM is most sensitive during transitional periods, suggesting that using categorical seasons may obscure seasonality.

For both Mali and South Sudan, 41% of children in the ‘previous SAM recovery’ cohort were AM during the main peak in each country: post- and pre-rainy season, respectively. However, in Mali, there was a 36-percentage point difference between the peak and nadir, while in South Sudan, it was only a 14-percentage point difference. This smaller variation in South Sudan was likely associated with a greater vulnerability related to protracted conflict. A previous AM study observed the same pattern and variation in South Sudan [11]. These differences have implications for seasonal targeting. In Mali, intensifying screening during peak times would efficiently identify the most affected children, improving the use of limited humanitarian resources. In Somalia and South Sudan, more regular screening would be required. Prevention programs should be targeted immediately prior to AM peaks.

Besides the country level differences, we also observed differences between the ‘previous SAM recovery’ and’no AM in the year prior’ cohorts in both Mali and South Sudan. There was a consistent pattern where the peak timing of AM for the ‘previous SAM recovery’ cohort occurred one to two months earlier as compared to the’no AM in the year prior’ cohort (August vs October in Mali; April vs May and September vs November in South Sudan). Thus, children with a previous episode of AM are much more vulnerable and deteriorate faster in response to shocks than children who did not have AM in the year prior. We also see a lag between the different indicators. The lowest points for WHZ and MUAC occurred about a month after the peak timing of AM, particularly in the relapse cohort in Mali and South Sudan. This suggests that while seasonal factors affect the entire population, children who fall below the −2WHZ threshold tend to lose weight earlier than the general population. The ‘previous SAM recovery’ population also did not experience the same decline in AM prevalence between peaks as the’no AM in the year prior’ cohort. Thus, children who had previously relapsed should be targeted for earlier support. Rising relapse rates may also reflect broader community stress and these increases could serve as an early warning metric for monitoring nutritional stress, guiding necessary responses.

Food insecurity was significantly correlated with higher odds that a child would experience AM in the ‘previous SAM recovery’ cohort in South Sudan and in both cohorts in Somalia. The lack of a relationship in Mali was likely due to the extremely low food insecurity. While visually there does appear to be a seasonal lag between food insecurity and acute malnutrition, when tested in the model, lagged food insecurity was only correlated to acute malnutrition in the ‘previous SAM recovery’ cohort in South Sudan. Thus, while food insecurity was extremely important, it followed a slightly different seasonal pattern, sometimes peaking when acute malnutrition was at its lowest.

Diarrhea was associated with higher odds that a child would experience AM in Mali and in the ‘previous SAM recovery’ cohort in South Sudan. In South Sudan, similar to food insecurity, experiencing diarrhea in the month prior was associated with a significant increase in the odds of AM in the ‘previous SAM recovery’ cohort. However, diarrhea behaved almost identically across the cohorts thus likely not helpful in explaining the difference between the two. From a seasonal perspective diarrhea—unlike food insecurity—did consistently mimic the overall trend of AM, with two peaks in South Sudan and one peak in Mali.

In line with previous literature [34], girls had significantly lower odds of AM, but only in Mali. However, unlike the previous literature [35], older kids in the ‘previous SAM recovery’ cohort in Mali and Somalia had a greater risk of relapsing. This age-related trend might be attributed to the accumulation of exposures to shocks with age, similar to findings that show that stunting is a result of the accumulation of multiple previous wasting episodes [36]. In addition, the age-related difference could be a product of using WHZ in the definition, which tends to identify older kids.

The study has several limitations. Given that the study had rolling admission, we do not have an equal number of observations by month, with a particular bias in Mali. Furthermore, the’no AM in the year prior’ cohort does not remain a true control, as some children may experience multiple episodes of AM during the follow-up period, making them technically part of the ‘previous SAM recovery’ cohort. There are many measures of food insecurity, which are frequently correlated, but do not always identify the same households. We specifically used HHS in this study, but other food insecurity variables might have a different relationship with AM and seasonal patterns. We also note that, given the high level of between-year climatic variability, the particular times of year identified in this study are likely to move. However, this can be adjusted for by looking at the climatic variables directly as opposed to predetermined categorical seasons. To establish seasonal patterns, more years of monthly data would be preferable.

Conclusions

This study shows that despite recovery from SAM, seasonal peaks of AM relapse occur earlier, are higher, and last longer than seasonal AM patterns for children without previous experience of AM. Thus, children recovered from SAM require additional support and targeting post-discharge to keep them from relapsing in the first place. In addition, given limited humanitarian funding, intensifying screening for AM immediately prior to seasonal AM peaks, with a particular focus on the transition periods, would more cost-effectively identify a large burden of children who are AM. Identifying relapse patterns may also help flag emerging community-level nutrition challenges and inform broader response strategies and early warning.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

AM led the statistical analysis. HS, AM, FL, and SK contributed to analysis and interpretation of results. AM wrote the first manuscript draft. SC, FM, and NL contributed to data collection and quality. All authors had full access to all data in the study, participated in critical revision of the manuscript and had final responsibility for the decision to submit for publication.

Funding

The funding was provided by USAID.

Data availability

The study’s deidentified dataset is available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

The authors declare no competing interests.