Multivariate regression with sparse data for a few independent variables

In my analysis I have several independent variables. My dataset has 2000 records but for a few variables the data is sparse, i.e., information is available only for about 200 records. When I include these variables the regression runs only on 200 records and I am effectively losing information that is available in remaining 1800 records. How do I overcome this loss of information?

Someone suggested introducing a dummy variable for each of these sparse data variables. Is that a standard practice? Is there any implication that I should know?