Using growth velocity to predict child mortality

Abstract

Background: Growth assessment based on the WHO child growth velocity standards can potentially be used to predict adverse health outcomes. Nevertheless, there are very few studies on growth velocity to predict mortality.

Objectives: We aimed to determine the ability of various growth velocity measures to predict child death within 3 mo and to compare it with those of attained growth measures.

Design: Data from 5657 children <5 y old who were enrolled in a cohort study in the Democratic Republic of Congo were used. Children were measured up to 6 times in 3-mo intervals, and 246 (4.3%) children died during the study period. Generalized estimating equation (GEE) models informed the mortality risk within 3 mo for weight and length velocity z scores and 3-mo changes in midupper arm circumference (MUAC). We used receiver operating characteristic (ROC) curves to present balance in sensitivity and specificity to predict child death.

Results: GEE models showed that children had an exponential increase in the risk of dying with decreasing growth velocity in all 4 indexes (1.2- to 2.4-fold for every unit decrease). A length and weight velocity z score of <-3 was associated with an 11.8- and a 7.9-fold increase, respectively, in the RR of death in the subsequent 3-mo period (95% CIs: 3.9, 35.5, and 3.9, 16.2, respectively). Weight and length velocity z scores had better predictive abilities [area under the ROC curves (AUCs) of 0.67 and 0.69] than did weight-for-age (AUC: 0.57) and length-for-age (AUC: 0.52) z scores. Among wasted children (weight-for-height z score <-2), the AUC of weight velocity z scores was 0.87. Absolute MUAC performed best among the attained indexes (AUC: 0.63), but longitudinal assessment of MUAC-based indexes did not increase the predictive value.

Conclusion: Although repeated growth measures are slightly more complex to implement, their superiority in mortality-predictive abilities suggests that these could be used more for identifying children at increased risk of death.

Keywords: WHO growth velocity standards; anthropometry; longitudinal growth; mortality; prediction.

FIGURE 1

Study profile of the cohort study in Bwamanda, Democratic Republic of Congo: 1989–1991. LFU, lost to follow-up.

FIGURE 2

Observed and predicted (by using a GEE model) probabilities of death within 3 mo for children included in a cohort study in Bwamanda, Democratic Republic of Congo (1989–1991), for weight velocity z score (A; n = 2296), length velocity z score (B; n = 2296), change in MUAC-for-age z score (C; n = 4451), and changes in absolute MUAC (D; n = 4451). GEE, generalized estimating equation; MUAC, midupper arm circumference.

FIGURE 3

Kaplan-Meier survival curve showing survival for all 5657 children under the age of 5 y included in a cohort study in Bwamanda, Democratic Republic of Congo (1989–1991), at different ages (with 95% CIs).

FIGURE 4

Receiver operating characteristic curves for ability to predict death within 3 mo in 2567 children aged 3–24 mo included in a cohort study in Bwamanda, Democratic Republic of Congo (1989–1991), by WVZ and WAZ (A), LVZ and LAZ (B), ΔMUACZ and MUACZ (C), and ΔabsMUAC and absMUAC (D). z Scores are calculated with help of the WHO Child Growth Standard. AUC values of the individual predictors are given in the inserts of each plot. absMUAC, absolute midupper arm circumference; LAZ, length-for-age z score; LVZ, length velocity z score; MUAC, midupper arm circumference; MUACZ, midupper arm circumference z score; WAZ, weight-for-age z score; WVZ, weight velocity z score; ΔabsMUAC, change in absolute values of midupper arm circumference; ΔMUACZ, change in midupper arm circumference z score.