This varies with population distribution per region and the nature of house arrangement in the region of the study. However, doing a multistage sampling method would be more appropriate, where you first randomly sample regions, the clusters in the sampled regions finally sampling the households within the sampled clusters. You must first come with the clear framework for households in order to easily sample them.

thanks somuch Asweto!

multistage stratified sampling technique will be the best method for conducting survey. first regions will be selected randomly then clusters will be selected and finally from the selected clusters the households will be selected with PPS sampling as per requirement of sample size.

Please check the malaria indicator survey (MIS) implemented by the USAID funded MEASURE DHS.

Thanks all for you contributions. its are very helpful!

Consider Lot quality assurance sampling method. It is used to monitor programs.

There's no one-size-fits-approach of sampling methods for all study designs (even surveys), health indicators (like malaria)... This depends on many factors. But since DHS (http://www.dhsprogram.com/) does similar surveys, including malaria surveys (eg.https://dhsprogram.com/pubs/pdf/MIS8/MIS8.pdf), using articulated sampling methods, you can adopt those methods to your study. 

Thank  for this helpful information 

there is no standard formula for the calculation of sample size .but in india NVBDCP Does similar survey You may like to see there docum. ents on house hold surveyand health facility survey : in depth review of NVBDCP

The best approach depends on local or contextual quality indicators  documented such as current malaria incidence, prevalence, mortality, census versus migration population which are valuable in sample size estimation.  particularly, pregnant women population, versus trend in ANC seeking women population, nationwide and locality specific similar should be documented for under 5 years children sleeping or no sleeping under bednet for more accurate sample calculation. The power also should be considered if cross-sectionnal and longitudinal representiveness.