A complete sampling frame (CSF) is needed for the development of probability sampling structures; utilisation of a spatial sampling frame (SSF) was the objective of the present study. We used two sampling methods, simple random sampling (SRS) and stratified random sampling (STRS), to compare the prevalence estimates delivered by a CSF to that by a SSF when applied to self-reported hypertension and diabetes mellitus in a semi-urban setting and in a rural one. A CSF based on Geodatabase of all households and all individuals was available for our study that focused on adults aged 18-69 years in the two settings. A single digitized shapefile of solely household regions/structures as SSF was developed using Google Earth and employed for the study. The results from the two sampling frames were similar and not significantly different. All 95%CI calculations contained the prevalence rates of the two medical conditions except for one occasion based on STRS and CSF. The SRS based on CSF showed a minimum 95% CI width for diabetes mellitus, whereas SSF showed a minimum 95% CI width for hypertension. The coefficient of variation exceeded 10.0% on six occasions for CSF but only once for SSF, which was found to be as efficient as CSF.