Importance of postmortem anthropometric evaluation in defining the role of malnutrition as a cause of infant and child deaths in Sub-Saharan Africa and South Asia: a cohort study

Abstract

Objectives: To evaluate how postmortem anthropometric malnutrition (PAM) measures align with expert panel attribution of malnutrition as a causal or significant condition in under-5 mortality (U5M).

Design: Cohort study using data from the Child Health and Mortality Prevention Surveillance network, incorporating clinical records, postmortem anthropometrics, minimally invasive tissue sampling, clinical abstraction and verbal autopsy to determine multiple causes of death.

Setting/participants: 1405 deaths of children aged 1-59 months from six African countries between 2016 and 2023.

Primary and secondary outcome measures: PAM was determined using z-scores from the WHO Child Growth Standards: underweight (weight-for-age<(-2)), wasting (arm circumference-for-age or weight-for-length<(-2)) and stunting (length-for-age <(-2)). Performance metrics (sensitivity (SE), specificity (SP) and positive predictive values (PPV)) were calculated to determine the alignment between PAM and expert panel attribution of malnutrition as a causal or significant condition to death.

Results: Nearly 75% of cases demonstrated moderate-to-severe malnutrition by PAM, while expert panels attributed malnutrition in 41% of cases. Performance metrics varied across anthropometric indices: underweight exhibited the highest SE (89.7%), while wasting based on arm circumference had the highest SP (81.9%) and PPV (76.8%). Discrepancies between PAM classification and expert panel attribution differed significantly by site, age, location of death and preventability of death (p<0.05). Adjusted multivariate regression showed that expert panel attribution was more likely with increasing severity of PAM.

Conclusions: The proportion of U5M attributable to malnutrition ranged between 41% (expert panel attribution) and 74% (PAM). Variability in classification underscores the need for monitoring and quality improvement measures to address discrepancies. Improved alignment between PAM and panel assessments is essential for accurately identifying malnutrition-related deaths and designing effective interventions to reduce U5M.

Keywords: Community child health; Epidemiology; Mortality; Nutrition.

Figure 1. Adjusted odds of malnutrition* recognised by DeCoDe† panel as a causal or other significant condition† by type and severity of postmortem anthropometric malnutrition‡ at death. *The x-axis is shown on the log scale. Associations were examined using ordinal logistic regression models comparing the exposure (severity of various postmortem anthropometric indices) and outcome (expert panel attributing malnutrition as a cause or other significant conditions to death) adjusting site, sex and age. The reference group for the exposure varies by panel. Panel 1 examines the relationship between the severity of underweight based on ZWEI and malnutrition attributed by DeCoDe. ZWEI were classified as at-risk (−2≤ZWEI<−1), moderate (−3≤ZWEI<−2) or severe (−10≤ZWEI<−3). The reference group was not underweight (−2≤ZWEI≤5). Panel 2 examines the relationship between the severity of wasting based on ZAC and malnutrition attributed to DeCoDe. ZAC were classified as at-risk (−2≤ZAC<−1), moderate (−3≤ZAC<−2) or severe (−10≤ZAC<−3). The reference group was not wasted (−2≤ZAC≤5). Panel 3 examines the relationship between the severity of wasting based on ZWFL and malnutrition attributed to DeCoDe. ZWFLs were classified as at-risk (−2≤ZWFL<−1), moderate (−3≤ZWFL<−2) or severe (−10≤ZWFL<−3), with ZWFL to the reference group not wasted (−2≤ZWFL≤5). Panel 4 examines the relationship between the severity of stunting based on ZLEN and malnutrition attributed to DeCoDe. ZLEN were classified as at-risk (−2≤ZLEN<−1), moderate (−3≤ZLEN<−2) or severe (−10≤ZLEN<−3), with ZLEN to the reference group not stunted (−2≤ZLEN≤6). †DeCoDe expert panels analyse all available individual information, including laboratory, histopathology, abstracted clinical records and verbal autopsy findings for each death. Using this information, the site panel ascertains the underlying cause (event that precipitated the fatal sequence of events) and other antecedent, immediate and maternal causes of death in accordance with the International Classification of Diseases, 10th Revision, and the WHO death certificate. ‡Additional details on the type and severity of postmortem anthropometric indices are as follows: ZWEI and severity of underweight missing (n=10), flagged as biologically implausible, those below −10 or above the upper limit of 5 for ZWEI; ZAC and severity of wasting missing (n=246) due to WHO growth standards not able to calculate the z-score for those aged below 3 months or flagged as biologically implausible values, those below −10 or above the upper limit of 5 for ZAC; ZWFL and severity of wasting missing (n=94) due to WHO growth standards not able to calculate the z-score for those with lengths below 45 cm or flagged as biologically implausible values, those below −10 or above the upper limit of 5 for ZWFL; and ZLEN and severity of stunting missing (n=36), flagged as biologically implausible, those below −10 or above the upper limit of 6 for ZLEN. DeCoDe, determination of cause of death; ZAC, mid-upper arm circumference-for-age z-scores; ZLEN, length-for-age z-scores; and ZWEI, weight-for-age z-scores; ZWFL, weight-for-length z-scores.