RESEARCH

The questions we're working on — and the ones we genuinely don't have answers to yet.

Longitudinal data doesn't just improve monitoring. It opens up research questions that cross-sectional surveys structurally can't answer. Here's what we're investigating, and where the evidence currently stands.

OPEN QUESTIONS

Four research questions longitudinal data can actually answer

Despite decades of investment in treatment, recurrent acute malnutrition remains one of the least understood problems in child nutrition — largely because most systems aren't built to see it. A child who recovers and stays healthy, and a child who recovers and relapses six months later, are typically recorded as two separate, unrelated episodes.

What's documented

Relapse after acute malnutrition treatment is common enough to matter at a programme level, and recovery durability varies widely between children in similar circumstances.

What's still open

Why some children relapse and others don't; how much household conditions, illness, or seasonality drive recurrence; how early relapse risk could realistically be flagged.

Children sharing the same caregiver, food supply, and household conditions frequently experience markedly different nutritional outcomes. Conventional, population-level surveys are poorly suited to even surface this pattern, let alone explain it — they're built to compare across households, not within them.

What's documented

Within-household variation in nutritional outcomes is a consistent, observable pattern across many different contexts and conditions.

What's still open

Whether the divergence is driven by individual child factors, caregiving allocation within the household, illness exposure, or something not yet identified — and whether it's predictable in advance.

Food security systems and nutrition systems both generate valuable data, largely in parallel. Almost no system currently links a specific household shock — a livestock loss, a displacement, a cut in food assistance — to what happens next in a specific child's growth trajectory.

What's documented

Household-level shocks are broadly associated with worse child nutritional outcomes at a population level, across many studies and contexts.

What's still open

The timing and size of the effect at an individual child level — how soon a shock shows up in growth data, and which households absorb similar shocks without measurable impact on their children.

This is our long-term research direction: using growth velocity, illness history, seasonality, and household conditions together to flag rising risk before a child crosses a clinical threshold. The aim isn't to replace clinical judgement or existing screening — it's to test whether routinely collectable data can identify which children would benefit from earlier attention.

WHERE THE EVIDENCE IS HONEST ABOUT ITS LIMITS

Prediction is more tractable for slow-building decline than for sudden, shock-driven malnutrition. Predictive models have, to date, performed unevenly on acute, fast-onset cases. We see this as the genuine frontier of the work, not a solved problem — and we'd rather be upfront about that than oversell where the field currently stands.

THE FULLER AGENDA

Additional research directions

Maternal & Child Nutrition Across the First 1,000 Days

How maternal nutrition, early growth, and infant feeding practices shape the trajectories that follow.

Standardising Relapse Measurement

Working toward a consistent definition of "relapse" across contexts, so findings from different settings can actually be compared.

Trend-Based vs. Threshold-Based Risk Indicators

Testing whether indicators built from trajectories outperform single-point screening criteria at identifying real risk.

Community-Based Longitudinal Monitoring Systems

What it actually takes, operationally, to sustain repeated household-level monitoring over months and years.

SYNTHESIS

Most of these questions can't be answered with a survey taken once.

They require the same children, watched over time. That's why longitudinal data isn't just a better monitoring method — it's a different kind of evidence, and one the field doesn't yet have enough of.