Performance Tuning :: How To Verify If A Row Exists (index Range Scan)

In the link below [URL] Thomas kyte has said, use the CBO and select /*+ FIRST_ROWS */ primary_key from table where rownum = 1; it'll read the index and stop at the first row. very fast on a big empty table (as the index is small and empty).

very fast on a big full table as the index is just read to find the first leaf node and then "stop".

It gives faster result if the primary key is used. But what if we have a table with around 1000 million rows and for the predicates there is a index range scan on the table.

What if we have a table say big_table (10000000000 rows) and the sql is something like

select /*+ first_rows */ 1 /* id, attribute_id*/
from big_table
where attribute_name ='Gross Premium'
and value ='10000'
and version_date is null
and rownum=1; --it's taking around 3 min

We observed that in such case there will be a range index scan for the index on the predicates. For a particular id there may be different values for attribute 'Gross Premium' and may have multiple versions.

How I would tune such a query where the purpose is to check if at least 1 records exists in the table for the input?