If you can assume a parametric distribution, an independent t test with carer's gender as the grouping variable and patient satisfaction as the dependent variable will work. If there are confounders to consider, an ANCOVA (use the GLM option in SPSS) will be better. If the distribution is thought to be non-parametric then a Mann Whitney test will be useful, though you will not be able to take account of confounders using non-parametric approaches unless you stratify your data set.

Dear mark,

Thanks a lot for your response.

Best reagrds,

Maha

Chi-square if patient satisfaction variable has multiple levels.

http://onlinelibrary.wiley.com/doi/10.1046/j.1525-1497.2000.91156.x/full