Great question Greg!  So the MEI is a perfectly valid ENSO index for certain applications but in operations, we prefer indices that are also provided at monthly mean resolution (the ONI can be split into  monthly Nino3.4 index values which can be used in real time declaration).  Also, as we mention in this older blog post, it can be difficult to dynamically understand what is going on when using combined indices like the MEI.  It is useful to us to see the OLR index by itself, the SOI index by itself, etc, etc.  When the indices are all combined, we can't rule out that one variable is dominating the rest (Example: SSTs are consistent with El Nino, but none of the other atmospheric indices are participating).   Further, the SST datasets have a lot of investment and time put into climate homogenization meaning we can, fairly confidently, compare today's value with one in the past (at least since the 1950s).  I can't say the same about gridded datasets of winds, etc.   And, lastly, SST trends are becoming an issue and we probably need this to be addressed for SST variables in the MEI as well.  Hope that helps.