Similar to per isakson's suggestion, but with proper indexing on the "Hospital" column:
>> subset = your_ds_array( strcmp( 'B', your_ds_array.Hospital ), : );
Or, if you want to eliminate the strcmp which can be slow if your array is large and you do alot of these queries, convert the hospital column to a nominal array first:
>> your_ds_array.Hospital = nominal(your_ds_array.Hospital);
>> subset = your_ds_array( your_ds_array.Hospital == 'B', : );