I'm doing KAP based study of parents knowledge, attitude and practice towards diabetes vulnerability in their genetically predisposed offspring in rural province of my country. The province is well known for married couples having diabetes (father or a mother). Now the issue I'm having is there has been no study done before on this topic. The prevalance of diabetes in a country is 8.6 % and Prediabetes 9.6%. Initially I though of involving both the parents in data collection, but then I realised prediabetic parents can change the outcome of the assessment, as they are more on a denial phase. It's obvious they should be having more mean KAP than diabetic parents themselves. Well that makes it for another research.

My main question is what can be done if I want to save the resources and time by not doing a pilot study? Should I go with 50% of diabetic prevalance. It seems fair because most of the parents are from 21-60 age group. Age 21 because that’s the legal to marry and age 60 because that’s the age where parental KAP will have minimum influence their children daily lives- at that age mothers, or parents in general, are not involved in their children decision making and even cooking (main factor) in many culture.

Please share any formulas or answers in your knowledge that could be of help. I'm even confusion in methodology part as there's not must research based on what I've proposed.

Dear Loki Dawnwright:

I'm not sure I understand all the intracacies of your question. I do not know exactly what indicator you want to compare between what groups, so I will have to provide a more general answer. Assuming a prevalence of 50% for a dichotomous variable results in the largest sample size. Using the assumption of 50% prevalence may result in a study sample size much bigger than you actually need, and collecting data from the unnecessarily large number of subjects may cost more than the cost of a pilot study. For example, in a survey to compare the prevalence of a condition between two groups selected with simple random sampling, if you assume a prevalence of 10% and want to see statistical significance in a difference between groups of 10 percentage points, the sample size is 199. If you assume a prevalence of 50%, the sample size is now 387. If collection of data on the extra 188 subjects is expensive, it may be worth doing a pilot study to help formulate your assumptions for calculating sample size. I hope this helps.

Answered:

1 year ago