Malnutrition related child morbidity and mortality: a space-time based analysis using Kilifi County Hospital Data 2002 to 2015
Background: Globally malnutrition is an underlying cause of death and accounts for over 45% of under-5 mortality mainly resulting from diarrhoea and pneumonia. The post-2015 era has seen, more than 25% of Kenya’s population being food insecure, with considerable geographicandtemporaldisparities. Ourprimaryaimwastounderstandthedeterminants ofmalnutritionrelatedmorbidityandmortalityintheruralKilifiHDSS,withaspecialfocus on children admitted in Kilifi County Hospital (KCH) during 2002-2015. Methodology: Our study participants were all the children between the ages of 6 months to 15 years who were admitted two times or more at the KCH. The outcomes were derived from malnutrition-related admissions based on wasting (WHZ<-2) and oedema and the discharge outcome whether alive or died. There were 3114 children with a total of 7620 admissions for children with more than one admission. In the exploratory data analysis, temporality and seasonality were determined using SARIMA time series models. Morans I index was used to investigate for the presence of spatial autocorrelation. SatScan was used to identify the spatial clusters of malnutrition related admissions and mortality. To understand mortality patterns, geo-additive logistic models were fitted to the KCH data. Mixed effects negative binomial models with separate space and temporal random effects were fit using the Maximum Likelihood and Bayesian Estimation procedures. The Bayesian methods were used to estimate the spatial parameters using Markov Chain Monte Carlo (MCMC) assisted with either Metropolis Hastings or Integrated Nested Laplace Approximations (INLA). iii ABSTRACT Results: Therewere17,740childrenobservedovertheperiodofstudyand4.01%ofthosedied. A total of 23,347 admission events were observed of which 7,128 were malnutrition related. Outofthe17,740childrenadmitted, 3,114hadoneormoreadmissionevent. A seasonal hike in the May to July month was identified for malnutrition admission. Children with morethanoneadmission,(7620admissions)~24%(n=1858)hadamalnutritioneventand 6.24%ofthemdied. SpatialhotspotsclusterswereidentifiedintheNorthandSouthofthe creek and areas near Kilifi Town was identified as cold spots. Children with two or more severe diseases are more likely to have a malnutrition admission event and females are less likely to be admitted with malnutrition. There was a protective effect as the children grewolderandalsoastheirbodyweightsincreased.Themaleshadahigherriskofdeath compared to the females and a year increase in age reduced the risk of death by 15%. Conclusion: Abetterunderstandingofthefactorsthatcontributetomalnutritionattributableadmission and mortality can be used to advocate for and develop earlier and more appropriate responses. Additionally, this can provide an indication of future trends and the potential impact of interventions.Importantly, including spatial and temporal random effects biostatistical modelling can help reduce bias reporting and help understand better the patterns of morbidity and mortality. Campaigns providing food and/or vitamin or other supplements can contribute to reducing morbidity and ultimately deaths in Kenyan childrenandbuildingmorehealthfacilitiestoreducethedistanceoftraveltocareishighly recommendable.
A research report submitted to the school of public health, University of the Witwatersrand, Johannesburg, partial fulfillment of the requirements for the degree of Masters of science in biostatistics.
Wambui, Kennedy Mwai, (2017) Malnutrition related child morbidity and mortality : a space-time based analysis using Kilifi County Hospital Data 2002 to 2015, University of the Witwatersrand, Johannesburg, https://hdl.handle.net/10539/24663