In the main study with a sample size of 3  you will be able to say little about hospital differences and what they are related to. If this is important to you  (that is you are interetsed in contexual differences) - you need to increase the number of hospitals- that is you need a clustered design with more hospitals but (due to costs) fewer individuals per hospital.. You would also want to stratify on hospitals.

If you whant to do this clustered design ( an it depends on your research questions) thiis free specialist software

http://sitemaker.umich.edu/group-based/optimal_design_software

I can't give you the citation but I have read about 10% of the sample should be sufficient.

The reference for 10% is:   Hertzog, M.A. (2008). Considerations in determining sample size for pilot studies. Research in Nursing & Health, 31, 180-191.

However, presumably, since you have developed your own questionnaire, the first order of business is to test its reliability and validity, no?  

http://www.youtube.com/watch?v=Nnx7MxQx0Xo  This video walk you through using GPower:  how to install the GPower software (it's free) and then to use it to calculate how many participants you require to achieve significant results, and will tell you the effect size.  This depends entirely on the type of tests you intend to perform on your data.  Presumably you must satisfy an ethics committee for your research and they will want this information.

An excellent reference for both qual and quant research methodologies, including sampling, is Ruth Tappen (2012). Advanced nursing research; From theory to practice.

In the main study with a sample size of 3  you will be able to say little about hospital differences and what they are related to. If this is important to you  (that is you are interetsed in contexual differences) - you need to increase the number of hospitals- that is you need a clustered design with more hospitals but (due to costs) fewer individuals per hospital.. You would also want to stratify on hospitals.

If you whant to do this clustered design ( an it depends on your research questions) thiis free specialist software

http://sitemaker.umich.edu/group-based/optimal_design_software

Rizwana, Charlotte, Kelvyn, Bernard and Gary thank you. I came up with questionnaire alone, since I haven’t found anything existing which would suit our environment. Three colleagues who also work with dysphagia patients have helped me improve the questionnaire. I would like to test it now. Charlotte, the link you have sent me will really help. Afterwards I will conduct a pilot study. In September I will ask ethical committees of hospitals where I wish to conduct study for their approval. Probably there will be more than 3.

After that I have a month of time to finish research, with statistical analysis following.

Thank you again for your input. A have a lot to learn. 

There are quite a few options available to test your measure. You can assess face validity by presenting the measure to patients and ask then to complete it and give you feedback regarding item relevance, wording, complexity etc and then make changes. For this you can use as little as 8 participants because the feedback will be qualitative. After this you can administer the measure to a small sample (10%) and look at reliability and factor structure. 

Thanks Martin,

in some hospitals I need the approval of the ethical committees and in some from group for research in nursing and midwifery care. But the idea for contacting sales representatives is great. I will try to contact Nutricia from Danone group . I will focus on more clinics as Kelvyn suggested and try to involve as many clinics as possible. I hope they will be willing to cooperate.