Self Funded Account IBNR Estimate

[Elston Gunn]

I am being asked to develop an IBNR estimate for a fairly large (1800 mbrs, 300K in monthly paids) Self Funded HMO account. I have looked at claim lags (IP,OP,Prof) for the past 36 months and, given the significant monthly variance, I am hesitant to develop completion factors using the groups data. In addition, the IBNR estimate will need to be for claims incurred through 6/2003 even though June data is not yet available to me. My questions are as follows:

1. Is it appropriate to simply use the CF's which were developed for our entire block of Commercial HMO business (perhaps with some margin built in for conservatism)? I have no reason to think this groups claims are being adjudicated any differently than those of the commercial groups. I would be using these factors the IBNR for all but the most recent 3 or 4 months.

2. As for the most recent three months (Mar-May), I would develop an Incurred PMPM estimate by looking at the groups (or maybe the entire commercial block again?) average 3 month completion rate over the past 12-18 months. Assuming this approach makes sense, I would assume June 2003's IBNR would simply be the PMPM estimate (loaded for a month of trend) times June 2003 enrollment? OR...

Perhaps I'm spending too much time thinking about this. It seems to me that whatever approach I take, I won't have a whole lot of confidence in the end result. Any suggestions would be appreciated.

[Dr T Non-Fan]

Your conclusion says it all. And, it does depend on what the results will be used for.

Oh, and could you explain to me how you will have an liability estimate for the month of June without knowing what's been paid so far?

Also, although I know you don't control this but I have to shout, it's the fifth working day of the month, plus a long weekend for the Operations area to compile claims! Your company needs to get into the 90's, dude!

Enough beeyotching from me.

1. Check some past IBNR runs of this group vs the commercial block. See how the CFs differ after a few months' runoff. If they don't differ, feel free to use the commercial block's CF with some confidence.
2. Estimate using as many ways as possible: group's own experience, commercial block, PMPM estimate of most recent months, etc. Then find someone who will take responsibility for deciding what to report.
3. The group is likely to have different PMPMs than the commercial block, for a number of reasons: age/sex, benefits, area, industry, etc. So using commercial block's PMPM as an estimate would require adjustment. The group's experience would probably be better. Perhaps a blend.
4. Your answer should be a range that you are comfortable with, such that if the group decides on a number within the range (probably either the high end or the low end, depending on their current financial position), you are confident that the eventual IBNR at 6/30, after about three additional months of knowledge will be close enough (within, say, 20%, for a group of this size), barring any unusual event.
5. If this is being used as some kind of terminal liability with downside risk to your company (the group pays the amount is is not liable for paid claims exceeding that amount), then by all means go conservative.

[Elston Gunn]

Oh, and could you explain to me how you will have an liability estimate for the month of June without knowing what's been paid so far?

That was a big part of my dilemma. Looks as though I'll have to proceed as though I was doing an estimate through May and then use my May PMPM as a proxy for June.

Also, although I know you don't control this but I have to shout, it's the fifth working day of the month, plus a long weekend for the Operations area to compile claims! Your company needs to get into the 90's, dude!

In fairness, claims are out on our mainframe systems. Being a SAS/APL/Etc neophyte, I have to wait until the June data is loaded to our Corporate DataMart. This doesn't happen until next week (around the 15th). I could attempt to get the June data but by the time I do it will be next week.

1. Check some past IBNR runs of this group vs the commercial block. See how the CFs differ after a few months' runoff. If they don't differ, feel free to use the commercial block's CF with some confidence.

Good suggestion.


3. The group is likely to have different PMPMs than the commercial block, for a number of reasons: age/sex, benefits, area, industry, etc. So using commercial block's PMPM as an estimate would require adjustment. The group's experience would probably be better. Perhaps a blend.

Perhaps I didn't do a good job explaining. What I think I can do is look at the historical completion of a 3 month block of claims and use that to back into PMPM's for the most recent 3 months. On a rolling basis going back 12-18 months I would add paids for Month 1 (w/3 months of runout), Month 2 (w/ 2 months R/O) and Month 3 (w/ 1 month R/O) and divide this by the sum of Ultimate Incurreds (based on whatever completions I decide on) for the same months. I would then be able to get to an 'average' historical completion for for the more volatile recent months. Does this seem reasonable?

5. If this is being used as some kind of terminal liability with downside risk to your company (the group pays the amount is is not liable for paid claims exceeding that amount), then by all means go conservative.

The purpose is for the group to be able book a liability to their year end financials (I suppose their year end is June 03).

Thanks a lot for the input DTNF. If you have additional (based on my responses) I'd be interested to hear.

[Dr T Non-Fan]

You're welcome.

Perhaps I didn't do a good job explaining. What I think I can do is look at the historical completion of a 3 month block of claims and use that to back into PMPM's for the most recent 3 months. On a rolling basis going back 12-18 months I would add paids for Month 1 (w/3 months of runout), Month 2 (w/ 2 months R/O) and Month 3 (w/ 1 month R/O) and divide this by the sum of Ultimate Incurreds (based on whatever completions I decide on) for the same months. I would then be able to get to an 'average' historical completion for for the more volatile recent months. Does this seem reasonable?

Yes, it's reasonable. That's what I'd do, and I'm a fairly reasonable person.

You'd have to use your whole month estimate of June as part of your IBNR. The fact that some has been paid, even though you don't know it, would mean that your total IBNR estimate will be high. Not that much, though. Maybe 1%, and we're dealing with a possible +/- 15% range. (Don't know for sure. Just guessing, conservatively and unconfidently.)
IBNR = estimate incurred - paid. You have two unknowns, one equation, right?

You should try to get a different system of obtaining data. I suggest daily loads instead of monthly. The claims are paid every day, and there is probably a database with that information. I suggest getting all the data from one system (or ten), and then YOU figure out what to include and exclude. You know, actuaries are a very important department in an insurance company. And data is your main, if not sole, need. Having data readily available to you is an asset to the company, not an expense (as the budgeteers might insist). A corporate datamart is adequate for corporate work, but this is more of a business unit issue.
Perhaps the group can wait?

Good luck!

[Patience]

We can discuss DTNF's as a reasonable person later.

I work for a reinsurer, so we are almost aways working without knowing paids for the current month.

We usually set up the paids looking at historical trends, including seasonal trends and company payment paterns (some may make more payments if an extra Thursday in the month). I would complete the triangle using the prior months payment incurral pattern and Junes estimated paid claims.

Then for IBNR for the last three month (actually the last five months) we look at a combination of CF, loss ratio and the pattern of Incurred PMPM trend over the prior year. It should be a pattern which you could continue adding conservatism

After actual numbers are received we rerun the current month to see how estimates compare with the actual numbers. We also check each month against the most current run every time it is done.

[Dr T Non-Fan]

Where's my dueling glove....

[other]

Your most recent reserve estimate plus trend will always be your best estimate for next month's reserve.

Your block of business has nothing to do with this large group. The group should be large enough to be credible, but you may want to look at large claims separately.