Evidence designed around the decision.

Severe acute malnutrition remains a high-consequence clinical and policy problem, while its evidence is split across treatment trials, household surveys and inpatient trajectories.

Systematic review, causal analysis of population data and planned longitudinal modelling are being developed as one triangulated programme.

The review and population-analysis workstreams have defined protocols, reproducible analytical pipelines and explicit causal assumptions.

This is active doctoral research. Planned inpatient modelling and final conclusions are not yet complete.

Biomarker patterns may help describe biological heterogeneity, but retrospective datasets can easily support claims beyond their design.

Reproducible statistical reanalysis with explicit separation between association, discrimination and clinical prediction.

The analytical pipeline has passed internal scientific checks conditional on the stated design limitations.

Retrospective case-control data cannot establish absolute-risk calibration, prospective validity or clinical utility.

Evidence syntheses often flatten causal assumptions, study design and uncertainty into prose that is difficult to inspect.

Causal structures, evidence provenance and uncertainty are represented visually without replacing expert adjudication.

Reusable mapping and figure-production workflows have been developed for research outputs.

A causal map organises claims and assumptions; it does not by itself establish causality.

Nutrition and social-protection programmes are often simplified to blanket rules. That is operationally attractive, but it can conceal distributional and equity effects — who is missed, and when, by season, severity and setting.

Three linked pieces: a policy viewpoint making the conceptual case; a life-course child-nutrition microsimulation with cost-effectiveness analysis across country archetypes built from open data; and a systematic review with meta-analysis and limited individual-participant analysis.

A completed viewpoint draft, a working microsimulation engine with calibration against published anchors, and a built review protocol. The model carries an internal parameter audit that separates sourced values from remaining scenario assumptions.

Nothing in this programme is submitted, peer reviewed or published, and no cost-effectiveness number should be read as a finding. Several parameters remain scenario assumptions rather than fitted estimates, and the audit records them as such.

National commitments on gender equity and nutrition frequently fail to reach the public markets where food is actually traded, because those markets are governed by council bylaws written largely for revenue and order.

A curated corpus of national and municipal policy instruments across Malawi and Kenya, analysed through a structured coding framework alongside transformer-based text representation, with independent second-coder reliability checks and explicit retrieval-status tracking for every source.

A curated policy-document manifest with per-source retrieval status, an operational codebook, a second-coder reliability report, and two drafted comparative manuscripts.

Both manuscripts are recorded internally as revise-before-submission; neither is submitted, peer reviewed or published. This is policy-architecture analysis, not measured effects on diets, prices or markets. Collaborator and institutional attribution is not published until each co-author and institution has confirmed permitted wording.