Seasonal food insecurity in Haydom, Tanzania, Is associated with low birthweight and acute malnutrition: Results from the MAL-ED study

In rural agricultural communities in Africa, particularly those with a single annual harvest, the preharvest period has been associated with increased food insecurity. We estimated the association between seasonal food insecurity and childhood malnutrition in Haydom, Tanzania. Children enrolled in a birth cohort study were followed twice weekly to document food intake and monthly for anthropometry until the age of 2 years. Household food insecurity was reported by caregivers every 6 months. We modeled the seasonality of food insecurity and food consumption, and estimated the impact of birth season on enrollment weight and subsequent malnutrition. Finally, we described the seasonality of admissions for acute malnutrition at a local referral hospital (Haydom Lutheran Hospital) from 2010 to 2015. Food insecurity was highly seasonal, with a peak from December to February. Children born during these 3 months had an average 0.35 z-score (95% CI: 0.12, 0.58) lower enrollment weight than children born in other months. In addition, weight-for-length z-scores measured in these months were on average 0.15 z-scores lower (95% CI: 0.10, 0.20) than that in other months, adjusting for enrollment weight and seasonal infectious diseases, and this disparity was sustained up to the age of 2 years. Correspondingly, the number of admissions with acute malnutrition at the local hospital was highest at this time, with twice as many cases in December–February compared with June–August. We identified acute and chronic malnutrition associated with seasonal food insecurity and intake. Targeting of prenatal care and child-feeding interventions during high food insecurity months may help reduce child malnutrition. The Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) study design and methods,15 and the site in Haydom, Tanzania,10 have been previously described. Ethical approval was obtained from the National Institute for Medical Research in Tanzania and the Institutional Review Board of the University of Virginia. Written informed consent was obtained from the parent or guardian of every child. Briefly, children were enrolled within 17 days of birth and followed until the age of 2 years. Child food intake was recorded by a 24-hour food recall and illnesses were recorded by a maternal report at twice-weekly home visits. Diarrhea was defined as maternal report of three or more loose stools in 24 hours or one stool with visible blood. Acute lower respiratory infection (ALRI) was defined as cough or shortness of breath with a rapid respiratory rate determined by fieldworkers (defined by the average of two measurements per day that were > 60 breaths per minute when the child was < 2 months old; > 50 breaths per minute at age 2 months to 1 year; and > 40 breaths per minute at age ≥ 1 year).16 Anthropometry was measured monthly and converted into weight-for-age z-scores (WAZ), length-for-age z-scores (LAZ), and weight-for-length z-scores (WLZ) using the 2006 World Health Organization (WHO) child growth standards.17 Household food insecurity was assessed every 6 months with the question, “In the past 4 weeks, did you worry that your household would not have enough food?” We considered any frequency of worry (rarely, sometimes, or often) in response to this question as a report of food insecurity because responses of “sometimes” or “often” were uncommon (7.5% and 1.1%, respectively). Despite the subjectivity of this measure, the question was asked in the same way and in the same population over time such that relative differences by season are meaningful. Socioeconomic status (SES) was summarized as a score based on water access, assets, maternal education, and income and was averaged over four biannual surveys.18 Haydom Lutheran Hospital (HLH) is a rural 450-bed referral hospital situated in the town closest to the MAL-ED study area. The hospital has a catchment area of 74 villages and towns and serves approximately two million people,19,20 including all children in the Haydom MAL-ED cohort. We reviewed all hospital discharges from January 2010 to December 2015 among children under the age of five years for diagnoses of malnutrition (defined as malnutrition, acute malnutrition, severe acute malnutrition, kwashiorkor, or marasmus), diarrhea (defined as gastroenteritis, diarrhea, dysentery, acute watery diarrhea, giardiasis, or amebiasis), ALRIs (pneumonia), and all other diagnoses. Age, gender, and mortality associated with these admissions were also collected. The seasonality of the prevalence of food insecurity was modeled using Poisson regression for the number of reports per month. Highly variable crude monthly rates across the years of the study period were smoothed with linear and quadratic terms for the month of the year (m), and the terms sin (2πm/12), cos (2πm/12), sin (4πm/12), and cos (4πm/12) based on optimal fit by the Akaike information criterion. We modeled child food intake patterns using log binomial regression for the intake (yes/no) of certain foods by month. We modeled diarrhea and ALRI incidence by calendar month using pooled logistic regression for incident episodes from birth to the age of 2 years. We similarly assessed the seasonality of anthropometric outcomes by using linear regression to model average WLZ, WAZ, and LAZ by month. To estimate differences in food insecurity, food intake, and anthropometry across seasons, months were split into quarters that capture variation in food insecurity: December–February, March–May, June–August, and September–November. We used general estimating equations and robust variance to account for correlation between measurements within children, and adjusted for the incidence of seasonal infectious diseases: diarrhea and ALRI. Heterogeneity by gender, SES score, and number of siblings was assessed by the likelihood ratio test. We assessed long-term disparities in child health based on seasonal birth cohorts. We used linear regression to estimate the associations between season of birth and WAZ at enrollment and WAZ, LAZ, and WLZ at age 2 years. The seasonality of malnutrition-related and other admissions to HLH and mortality among children aged less than 5 years were modeled using Poisson regression for the total number of cases per month and the number of cases stratified by gender to assess the relative rate of admissions and mortality by season. The seasonality of diagnosis-specific case fatality rates was modeled using Poisson regression for the number of deaths per month with an offset for the number of diagnosis-specific admissions in that month. Analyses with the subset of admissions among children less than the age of two years were consistent with the analysis of all children less than the age of five years (not shown).