Trend

[T-royBoy]

I am also doing a study on trend. I started out with just your basic Per Member Per Month (PMPM) for 2003 and divided it by 2002. I was asked then what was the "NET" trend - Meaning after benefit buydowns and change in mix - did you ever hear of such a thing?

What would I apply to 2002 PMPM's to get them to the similar 2003 PMPM's.

By the way, I think I should change my handle - It is my real name - do you guys have any ideas?

[Dr T Non-Fan]

Somebody seems to have applied a very efficient, though physically impossible, sunblock to his/her head.

What you have IS the net trend. It doesn't tell you much about the future, since all those factors -- benefit mix change, demograhic mix change, etc. -- are embedded in the data, and can't be assumed to be the same in the future. You can calculate a negative net trend, if there is a massive mix change from expensive plans to cheap plans. I've seen it myself.

What they really want is called the "true" or "underlying" trend.

Did you just come from the future? (2003 over 2002.) What's it like?

Don't change your name. It gives the impression of insecurity.

[zzzman29]

Has anyone done any looking at Billed PEPM trends (prior to network discounts)? We have been trying to do this for years, but are getting wide variance depending on what area of the company we are looking at. We have seen about 17% over the past year, and 21% in years prior. This seems astounding (although isn't everything about health care costs pretty astounding?) We've been trying to find others to reality check this against.....

[Dr T Non-Fan]

Billed is worthless. Try going for Covered Charges, then Discounted Covered Charges. First see if there's a change in the average discount. That's somewhat useful. Then go from there.

[T-royBoy]

I did mean 2002 over 2001 - sorry - it is the price we pay for expediency -

I am afraid to ask now that i am insulted, but maybe somebody that enjoys enlightening others with their professionalism will help me out. What is "Underlying" trend? How would I calc it beyond the data that I have. If I expect claim PM to cost $200 next year (2003 DTNF) and 2002's is $185 but we expect to have a totally differnt mix, then how would I go about calculating an "Underlying" trend.

There has got to ba a way to explain it even though we have a different mix of business, demographics, etc...

[C8 Guy]

I don't know of any official defintion of underlying trend. If you just divide 2002 clms by 2001 clms you will not learn anything. Here's a quick brainstorm list of components that will affect the diffecence between the 2 numbers:

1. technology trend
2. utilization
3. deductible leveraging
4. cost shifting
5. benefit buydown
6. rating characteristic shifts
7. pure cost inflation
8. aging of the block (underwriting wear-off)

Take your pick of things to account for to come up with "underlying trend"

[Dr T Non-Fan]

If I can enlighten and insult in the same post, then I've done my job. I was trying to insult the person who asked you the question, BTW.

Actually it's very difficult to determine underlying trend. Difficult, but not impo$$ible. $ee what I'm trying to $say?

It's a Holy Grail of sorts. The underlying trend allows one to forecast all the little blocks of business, then create a summarized forecast from them. If you happen to forecast premium as well, then any change that affects both similarly will not affect the difference as much (which is most important).

I will not enlighten you with my professionalism (if it exists). You'll have to go to a dry place for that. I've got knowledge that any actuary should have.

Unfortunately, much of this knowledge might be somewhat proprietary in the eyes of a company attorney, so I can only give clues.

Try to isolate all the variables known to affect claims. Make sure not to make the resulting data too not-credible. Normalize what you can, based on best estimates of factors. You should have estimates of these factors at your company someplace.

As to your example: how do you get the $200? Is it just a straight-line from 2001-2002 trend? Is that with or without demographic (including benefit mix) changes? (Hint: it includes them if the trend used includes them.)
If you have the underlying trend, then the next nearly impo$$ible ta$k is to guess the changes in demographic and benefit mixes. But I wouldn't go any further. You'll start getting into an anti-productive, perfectionist mode.

[Polly Nomial]

Here are a few suggestions that you might try. None of them will give you the "right" answer but should shed light of the issue.

1) Before crunching the numbers adjust the claims based on the cost of the different plans. For example if plan B is 10% richer than plan A, add 10% to all of the claims from plan A. This reduces the effect of plan shift.

2)After adjusting the premium to a common point in time, examine the trend of the loss ratios. Assuming that the premium is adjusted for age, (and/or other variables) this should reduce any shifts in those variables.

3)Unless you are dealing with a very large block of business, one year divided by the previous year is risky in terms of normal variance in the data. Look at month to month data. Use 12 month moving totals to smooth it out. Fit lines and expontial curves to the data.

[zzzman29]

Ah, Dr. T Fan, but you are naive. :duh: Billed trend would seem worthless, however, as providers continue to move their contracts from fee schedules and per diems to straight discounts this becomes the discounted charges trend. (I do believe we have excluded non-covered services). Therefore, if trying to project where things are going, we need to see what the billed trend is, and adjust that by the percentage of contracts based on straight discounts vs. fee schedules.

[Dr T Non-Fan]

Excluded charges are often included in the Billed Charges. I usually build up from paid, discounts, and member paid to get covered.

[zzzman29]

From our claims system we can usually throw out the excluded stuff up front...usually.

[Soccerboy]

I don't know of any official defintion of underlying trend. If you just divide 2002 clms by 2001 clms you will not learn anything. Here's a quick brainstorm list of components that will affect the diffecence between the 2 numbers:

1. technology trend
2. utilization
3. deductible leveraging
4. cost shifting
5. benefit buydown
6. rating characteristic shifts
7. pure cost inflation
8. aging of the block (underwriting wear-off)

Take your pick of things to account for to come up with "underlying trend"

Can someone please explain to me what Benefit Buydowns are?

[Dr T Non-Fan]

I don't know of any official defintion of underlying trend. If you just divide 2002 clms by 2001 clms you will not learn anything. Here's a quick brainstorm list of components that will affect the diffecence between the 2 numbers:

1. technology trend
2. utilization
3. deductible leveraging
4. cost shifting
5. benefit buydown
6. rating characteristic shifts
7. pure cost inflation
8. aging of the block (underwriting wear-off)

Take your pick of things to account for to come up with "underlying trend"

Can someone please explain to me what Benefit Buydowns are?
This is when Individuals or Groups decide to lower their benefits by increasing deductibles or copayments.
So, when looking at a general 12MMA PMPM claim change, if enough policyholders do this, it lowers the trend determined by the formula.
The way to account for this is to perform separate analyses on each benefit design, and then draw a inference from the sets of analyses.

As an example, if a group has a $250 deductible policy until 12/31/2004, changed their benefits to a $500 deductible policy starting 1/1/2005, then the claims trend year-over-year will appear to be lower than it actually is, and using it for pricing of, the group effective 1/1/2007 would be a bit off.
In this example, the benefit differential needs to be estimated and added (multiplied?) to the formula trend.

[Soccerboy]

Thanks, still on trends, (this actually came from a course 8 exam but coincidentally it is something Im trying to wrestle at work as well)

There is a situation where you have hospital trends of say 16% (PMPM-TY = 116; PMPM-LY = 100)
We are told 50% of hospitals will adopt 10% discount, the rest not
We are also told that these 50% will be responsible for 1/3 of the facility care
What is the revised trend

The solution in the exam is
Discounted = 100*.9*1.16 * 1/3 = 34.8
Non Discounted = 100*1*1.16*2/3 = 77.3
Total = 112.13 for a trend of 12.1%

My question is, why is the 50% of hospitals not factored into this equation..or is it embeded somewhere that I can't spot it?

[Polly Nomial]

The number of hospitals is irrelavant since it is the volume of claims that you want. So the claims divide up 1/3 and 2/3. If all the hospital were the same size then it might be 1/3 of the hospitals