Comparison of usefulness of randomly amplified polymorphic DNA and amplified-fragment length polymorphism techniques in epidemiological studies on nasopharyngeal carriage of non-typable Haemophilus influenzae.