No, it is the number of change vectors applied to current mode blocks. Because each logical change to a block typically involves more than one physical change to the block, the value of this statistic will always be considerably greater than the number of blocks dirtied in the cache.

Please help me understand this statistic. The way I understand it is that it is the total number of blocks dirtied in the buffer cache by insert, update, or delete statements, committed or uncommitted. Am I right in my assumption?