Though consequence severity can (and is) used to decide when a LOPA is needed, the Best trigger for doing a LOPA is when the HAZOP team is confused. More than likely the HAZOP team is Not confused even for high severity. For instance, we were on a high severity ($100,000,000 loss potential, multiple death) consequence today in our HAZOP (Crude Unit) of a refinery and it was very clear they needed one more layer of protection; it was clear a SIL 1 was the right choice. The PSM department rep suggested it should go to LOPA, and the rest of the team looked at him and said “Why?”. I agreed; Why? We will not suggest a LOPA in this case.
Most of the time, the scenario is not confusing and so why send to another analysis? There are times where a LOPA will be more defensible to some managers than a HAZOP conclusion might be, so that is one reason (though not necessary a great reason, longterm) to use LOPA. LOPA is just another method and not necessarily a better method… I like LOPA, of course… but it is not superior.
So, the best rule we have found is to do LOPA whenever the PHA/HAZOP team is confused, and not do LOPA just because the severity is at or beyond a given threshold. BUT, if the HAZOP teams or HAZOP leaders are weak, and if you cannot fix that issue right away, then LOPA can be a partial bandaid and in those instances using LOPA for high consequence severity makes sense (and has help to correct weak HAZOPs/PHAs).
LOPAs should be done after the HAZOP is complete, but we make exceptions when the scenario is limited in scope (when we are sure the scenario will not change by the end of the HAZOP). Again, only about 1-3% of our HAZOP scenarios go to LOPA (if the client follows our suggested rules for when to require LOPA).