Evaluating the accuracy and feasibility of gridded population sampling to overcome bias due to missing populations in household surveys