Completion Factors

[oscar peterson]

If you are developing a claims distribution, and do not have complete run-out, one needs to use completion factors to convert paid to fully-incurred claims. This works fairly well in total for some segment or block of business, which is also probably the level at which the factors were developed. Is it appropriate to apply completion factors developed at that high level to annual claims for an individual member?

That is, member X has some claims in May and June; gross up those claim amounts using May and June CFs?

This was suggested to me, but if you have a 1 million dollar claim in the last month of the paid period, without runout that amount could be doubled?

[Zee]

Probably not the answer you want, but is there some reason you can't just use a period that is sufficiently run out? As you've pointed out, there are issues with what you are proposing. The scenario of over-inflating individual (large) claims would certainly occur, assuming you've got any large claims at all in your experience. On the same note, the completion rate of small and large claims is likely different.

[Dr T Non-Fan]

...
1. Is it appropriate to apply completion factors developed at that high level to annual claims for an individual member?

2. That is, member X has some claims in May and June; gross up those claim amounts using May and June CFs?

3. This was suggested to me, but if you have a 1 million dollar claim in the last month of the paid period, without runout that amount could be doubled?

1. No. Not if you plan to make decisions based on that one member. Yes, if you aggregate a number of members' claims together. How many? I don't know off-hand.
2. No. Not if you plan to make decisions based on that one member. That member will have a different and totally unpredictable completion pattern. Besides, completion is more for members whose claims haven't been paid. And, yes, that doesn't make any mathematical sense, either.
3. Sorry, I do not understand this question. Your most recent month's completion factor has such a high variance that it should be unusable (if it's anything like mine). Even an imputed completion factor (paid so far divided by eventual paid, estimated by some other means) isn't very helpful.

Summary:
A. Individual claims paid are usually complete. Not always, but the fact that they were paid would usually mean that the claim is closed.
B. Completion factors are used to estimate the cost of claims that are not yet paid. You see how these two pieces -- claims already paid, and claims yet to be paid -- at an atomic level are independent of each other. I mean, the IBNR generated by a $1 million claim is for other peoples' claims, not for that claimant's.

[oscar peterson]

Thanks -- this is along what I was thinking, but perhaps not as clearly. Someone suggested that it was merely an 'allocation' issue, which is true in some sense. But, as you've said, allocating IBNR to individuals based on their reported claim amounts seems wacko. Txs.

[Dr T Non-Fan]

Glad to help.