It is usual to end up a different sample size than was planned. Usually the difference is small. Occasionally the difference is large. The first thing to do if the difference is to work out the reason(s) for the difference. In your case you may have (1) overestimated the prevalence of SAM, (2) overestimated the population in communities, (3) overestimated the proportion of the population aged between 6 months and 5 years, or (4) used a less than exhaustive case-finding procedure. In the case of (1), (2), or (3) then you do not need to worry too much. In the case of (4) you may have a selection bias towards "easy to find" cases. This bias will tend to lead you to overestimate coverage. All surveys have biases. The best surveys identify and report their probable presence and direction with a guess of the magnitude of the bias. What to do? I think you have two options ... (A) Go ahead and and use your n = (18 + 2) sample. You will have a larger effective sample size than n = (18 + 2) because of the information contained / summarised in the prior. This means that your final (posterior) coverage estimate will be a little less precise than expected. (B) Top-up your sample. You could take a sample from the same number of communities with the expectation of doubling your sample size. The new sampling frame would be the list of communities with those already sampled removed. If you suspect issue (4) above than you may want to use a new within-community sampling method to reduce the effect of the bias. I hope this is of some use.