Malnutrition Among Children Under Five with Neurodevelopmental Disabilities: Identifying and Managing Risk Using a Multimethod Approach

Abstract

Children with neurodevelopmental disabilities (NDD) are at elevated risk of malnutrition compared to children without NDD. Yet they remain notably underrepresented in prevention and intervention research. This dissertation used a multimethod approach to examine malnutrition risk and management among children under five with NDD. The first study used retrospective longitudinal electronic health record (EHR) data from Seattle Children’s Health System to identify baseline factors associated with incident malnutrition outcomes among children under five with NDD. In the primary 12-month wasting outcome cohort, 135 of 2,715 children developed incident wasting following an NDD diagnosis, corresponding to an incidence rate of 8.49 cases per 100 person-years. Children younger than six months had higher hazard of incident wasting compared with children aged 24 to <48 months (aHR, 3.56; 95% CI, 2.33–5.44). Feeding difficulty was also associated with higher wasting hazard (aHR, 2.00; 95% CI, 1.09–3.65), as was having multiple broad NDD domains compared with social and communication NDDs (aHR, 2.42; 95% CI, 1.11–5.26). Secondary analyses showed that underweight and percentage weight loss captured distinct vulnerability patterns: underweight was associated with younger age, area deprivation index (ADI) rank, perinatal factors, respiratory morbidity, hospitalization, and feeding difficulty, whereas hospitalization was most strongly associated with percentage weight loss from baseline. The second study shifted from risk-factor association to prognostic prediction. Using the same EHR source, a prototype point-of-care prediction tool was developed and internally validated to estimate the six-month risk of incident wasting among children with NDD. The prediction cohort included 3,480 children aged 54 months or younger at baseline. During six-month follow-up, 97 children developed incident wasting. A compact Cox model including age, broad NDD type, and respiratory morbidity performed similarly to larger clinical and data-driven models. After bootstrap internal validation, the final model had an optimism-corrected C-index of 0.693, calibration slope of 0.947, IPCW (Inverse Probability of Censoring Weighting) AUC of 0.708 at 6 months, and IPCW Brier score of 0.0288. The model stratified children into groups with distinct observed 6-month cumulative incidence of wasting. The final model was translated into a simple bedside prototype lookup table using age, broad NDD type, and respiratory morbidity. The third study explored health-worker perceptions of the feasibility of implementing a malnutrition risk prediction tool for children with NDD in a Kenyan referral-hospital setting. Sixteen healthcare workers participated in in-depth interviews guided by the Consolidated Framework for Implementation Research (CFIR 2.0). Participants perceived strong potential value in a tool that could identify children with NDD at elevated malnutrition risk, but emphasized that adoption would depend on simplicity, local validation, integration into existing workflows, availability across multiple service points, and clear guidance on actions after risk identification. Health workers also highlighted major barriers, including staffing shortages, gaps in NDD and malnutrition guidelines, weak referral and follow-up systems, poverty, and stigma related to both disability and malnutrition. Together, these studies show that children with NDD have heterogeneous malnutrition risk profiles, that short-term wasting risk can be estimated using a small number of routinely available clinical variables, and that implementation of such a tool in a tertiary LMIC clinical setting will require attention to workflow, resources, local context, and proof of efficacy.