Actuarial Outpost Decomposing monthly allowed by Age/Sex factors
 User Name Remember Me? Password
 Register Blogs Wiki FAQ Calendar Search Today's Posts Mark Forums Read
 FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions

 Upload your resume securely at https://www.dwsimpson.com to be contacted when new jobs meet your skills and objectives.

 Thread Tools Search this Thread Display Modes
#1
04-06-2018, 01:27 PM
 wellactuary SOA Join Date: Mar 2018 College: MTSU Posts: 4
Decomposing monthly allowed by Age/Sex factors

I have used claims data to calculate age and gender factors for a given coverage year.

What I am ultimately trying to do is to decompose the trend into some constituent parts. I am only concerned with isolating the age/gender component, however.

My question is this ... How do I apply the factors to the PMPM allowed amount?

For the sake of clarity, I am using an old exercise on the "Sec 30.3" tab to reference the elements I am asking about.

Do I divide/multiply the monthly PMPM allowed amounts (claims experience data) by the sub-totalled annual factor(s) in X13 and X26? Do I combine X13 and X26 into a single factor? Do I need to generate these factors on a monthly, as opposed to an annual basis? Is the trend I am attempting to arrive at simply the trend of the monthly factors themselves?

I just want to know how to properly apply these factors in order to generate a separate monthly demographic component. I lack regional data.

Thanks

Last edited by wellactuary; 04-06-2018 at 01:28 PM.. Reason: typo
#2
04-06-2018, 02:06 PM
 cincinnatikid Member SOA Join Date: Jul 2005 Posts: 2,187

I think it's as simple as calculating a MM weighted factor in both the periods and determining how much your Age/Gender mix has changed between the two years and is driving the change in costs.

Ex: Base year weighted A/G factor = 1.00
Current year weighted A/G factor = 1.02

If your year-over-year allowed PMPM trend is 5%, 2% of this is driven by change in change in demographics; the other 3% is driven by other factors (utilization, unit cost, mix of services, geography...).
#3
04-09-2018, 04:32 PM
 wellactuary SOA Join Date: Mar 2018 College: MTSU Posts: 4

Hi, I really appreciate it. I'm sorry my initial post is such a jumble. Your answer provides a lot of guidance.

If you do not mind, I would like to ask a few other questions ... I apologize for the awkward way I am asking them, as I am trying to make clear any assumptions I have regarding what I'm asking.

1) After calculating the two final weighted factors (for both male and female), these two are simply averaged together (as they have already been weighted by MM), in order to arrive at the single factor to be used?

2) If I want to represent the age-gender mix on a monthly basis (annualized/YoY, of course, when it comes to trend), my assumption is I would calculate age-gender factors on a monthly basis?

3) This is all experience-based. If I were attempting to forecast this out, I would probably rely upon any given demographic assumptions for out-of-sample factors, correct?

4) I also want to use morbidity factors. My understanding is that demographic factors should be taken out first. Can I create morbidity factors using something along the lines of a comorbidity index, and apply them in a similar way?
#4
04-11-2018, 01:41 PM
 cincinnatikid Member SOA Join Date: Jul 2005 Posts: 2,187

Quote:
 Originally Posted by wellactuary Hi, I really appreciate it. I'm sorry my initial post is such a jumble. Your answer provides a lot of guidance. If you do not mind, I would like to ask a few other questions ... I apologize for the awkward way I am asking them, as I am trying to make clear any assumptions I have regarding what I'm asking. 1) After calculating the two final weighted factors (for both male and female), these two are simply averaged together (as they have already been weighted by MM), in order to arrive at the single factor to be used? 2) If I want to represent the age-gender mix on a monthly basis (annualized/YoY, of course, when it comes to trend), my assumption is I would calculate age-gender factors on a monthly basis? 3) This is all experience-based. If I were attempting to forecast this out, I would probably rely upon any given demographic assumptions for out-of-sample factors, correct? 4) I also want to use morbidity factors. My understanding is that demographic factors should be taken out first. Can I create morbidity factors using something along the lines of a comorbidity index, and apply them in a similar way?
1) You would weigh them together based on total M and total F MMs.

2) I think technically yes, but at that point you are probably using math to imply precision that isn't there. Ex: a female who is 44 yrs 11 months old doesn't suddenly get 50% more expensive when they turn 45.

3) Yes, projected costs would be relative to forecast period demographics.

4) I would tend to view morbidity as more of a scalar for a particular sample relative to the population, but be very careful in how you develop and apply these factors as morbidity and demographics are heavily co-mingled.
#5
04-11-2018, 02:15 PM
 Dr T Non-Fan Member SOA AAA Join Date: Sep 2001 Location: Just outside of Nowhere Posts: 93,889

Aren't all these morbidity factors of some sort? Perhaps you can tell me what you think morbidity factors are, so I can try to help you?
__________________
"Facebook is a toilet." -- LWTwJO
#6
04-12-2018, 03:28 PM
 wellactuary SOA Join Date: Mar 2018 College: MTSU Posts: 4

Thank you guys for the responses. I really appreciate it. Just to reiterate my intentions, I am attempting to generate a trend out of the PMPM allowed which accounts for the effects of a few basic components (at the bare minimum, age-sex factors and morbidity). I unfortunately lack regional data at this time.

Quote:
 Aren't all these morbidity factors of some sort? Perhaps you can tell me what you think morbidity factors are, so I can try to help you?
My understanding of morbidity is that it is simply a measure of disease prevalence/severity. So I would like to create a type of morbidity index in order to be able to isolate it's effect from the allowed PMPM trend. I've extracted a per-visit diagnostic history of my population, and I'm using R to generate DX-based comorbidity indices. I'm also messing around with CCS groupings and DRGs and attempting to create factors. Something tells me there are probably superior ways of building a morbidity factor, though.

Quote:
 4) I would tend to view morbidity as more of a scalar for a particular sample relative to the population, but be very careful in how you develop and apply these factors as morbidity and demographics are heavily co-mingled.
I'm taking care to not double count the effect of the [work-in-progress] age-sex factors, by first removing applying the age-sex factor to the PMPM allowed amount, then sequentially applying any other [working] factors.

As an example, for a given month:

PMPM Allowed: \$255.36
Demographic (age/sex) factor: .98
Morbidity factor: 1.19

\$255.36 / .98 = \$260.57
\$260.57 / 1.19 = \$218.97

Of course, when converting to an annual trend, I do so on a month-over-month basis.

Thoughts?
#7
04-12-2018, 03:44 PM
 Dr T Non-Fan Member SOA AAA Join Date: Sep 2001 Location: Just outside of Nowhere Posts: 93,889

Ah. You are "normalizing" the PMPM Allowed to get to \$218.97. Then, after the identifiable noises are filtered out, voila the trend.

I'd say that is one way to do it. Not the only way to do it. Big issue will be the unidentifiable noises. And no, not all of that is trend. Identifying is the researcher's job.
That's you.

Seems to be an interesting concept, though. I'd like you to change the name "morbidity factor" to something else, but I can't think of anything. The "Wellactuary Factor"?

Is your morbidity factor based on CMS Risk Adjusters?
__________________
"Facebook is a toilet." -- LWTwJO
#8
04-12-2018, 04:32 PM
 wellactuary SOA Join Date: Mar 2018 College: MTSU Posts: 4

Quote:
 Ah. You are "normalizing" the PMPM Allowed to get to \$218.97. Then, after the identifiable noises are filtered out, voila the trend. I'd say that is one way to do it. Not the only way to do it. Big issue will be the unidentifiable noises. And no, not all of that is trend. Identifying is the researcher's job. That's you. Seems to be an interesting concept, though. I'd like you to change the name "morbidity factor" to something else, but I can't think of anything. The "Wellactuary Factor"? Is your morbidity factor based on CMS Risk Adjusters?
Well, It's good to know that I'm not completely in my own ballpark with regard to the overall methodology of removing the effects.

I'm messing around with everything from the Charlson comorbidity index (via my experience data and the 'icd' package in R) and simply trying to calculate factors over major diagnostic groupings using the PMPM allowed for each DX group.

While I need the output to be in a monthly time series format (factors and all, for secondary purposes), should I be generating annual factors and somehow converting them into monthly factors? I have read a lot of stuff about this (obviously I still have gaps) and I came across the notion of trending forward. It was a post related to trend-normalization and the use of factors.

 Tags age, claims, health, trend, trend factor

 Thread Tools Search this Thread Search this Thread: Advanced Search Display Modes Linear Mode

 Posting Rules You may not post new threads You may not post replies You may not post attachments You may not edit your posts BB code is On Smilies are On [IMG] code is On HTML code is Off

All times are GMT -4. The time now is 01:44 AM.

 -- Default Style - Fluid Width ---- Default Style - Fixed Width ---- Old Default Style ---- Easy on the eyes ---- Smooth Darkness ---- Chestnut ---- Apple-ish Style ---- If Apples were blue ---- If Apples were green ---- If Apples were purple ---- Halloween 2007 ---- B&W ---- Halloween ---- AO Christmas Theme ---- Turkey Day Theme ---- AO 2007 beta ---- 4th Of July Contact Us - Actuarial Outpost - Archive - Privacy Statement - Top