Subnational estimates of under-five mortality rates (U5MRs) are a vital statistic for the United Nations to reduce mortality inequalities between high-income and Low-and-Middle Income Countries (LMICs). Current methods of modelling U5MR in LMICs smooth across trends in age and year of death, but not birth-cohort, to reduce uncertainty in estimates caused by data-sparsity. Using survey data from Kenya, we innovatively apply an Age-Period-Cohort model which accounts for spatial trends and the complex survey design of the data to estimate subnational U5MRs in Kenya. After validating our results against current methods, the inclusion of cohort can provide new insights into U5MRs. We ensure our method is flexible and can be applied to other LMICs.