Thursday, March 29, 2018

Risk Adjustment and Administrative Waste


BLUF:

Risk Adjustment (RA) is a complex system that adds unnecessary administrative expense to the US healthcare system in the form of many millions of dollars spent for medically useless home health visits to beneficiaries each year. A simple change in RA methodology would obviate the need for some of this wasteful spend.


About once a year we get a reminder of the absurd and growing cost of healthcare in the United States. 2018's annual memento mori was in JAMA last week, where researchers from LSE report that our per capita spend was $9,403 in 2016 [1]. For reference, it was closer to $7,500 in 2010 [2].

Trying to dissect the reasons for this trend can be academically thrilling but practically useless. It is hard to agree on causality: is it labor? Is it administrative spend? Drug pricing? As alluring as the forensic investigation is, the far more useful question to try and attack is what we can do to reduce unnecessary spending.

One of the most complex and flawed systems in American healthcare is our program of Risk Adjustment (RA) for Medicare Advantage (MA) and Marketplace plans. I don't know how much of your money it evolves every year through administrative kinetic friction but the answer is probably “quite a bit". And I think there is possibly a fairly straightforward way to recover many millions in unnecessary cost each year.

For concision, I'll explain the concept through Medicare Advantage, but the same idea basically applies to Marketplace as well. I'll begin with some background on RA and the problem and then build to the solution.

What is Medicare Advantage?

When you turn 65 and qualify for Medicare, you can say:
  • Option A (Normal Medicare): "Hey, federal government, please act as my insurance company for me."
  • Option B: (Medicare Advantage / Medicare Part C): "Hey, federal government - give the money you would have spent on me to Humana, or BCBS, or some other private insurer so that they can provide insurance for me instead."

Option A: Fee For Service (FFS) Medicare reimbursement, Option B: Medicare Advantage

In this case, Humana would get a lump sum payment from the government for each month you're using them to administer your Medicare benefit. 

For more on why a beneficiary might elect Medicare Advantage coverage, see Appendix at bottom.

OK, what is Risk Adjustment?

People vary in age, illness, etc -- it wouldn't be fair to give Humana the same amount of money for a 75 year old marathon runner vs. someone with late stage cancer.

RA solves this problem: quantifying individual risk based upon member age and medical diagnoses and adjusting the monthly payment from Center for Medicare and Medicaid Services (CMS) to a private insurer accordingly. In 2017 a diabetic earned a factor of 0.104 - about an extra $1,000 per year; someone with metastatic cancer is worth a factor of 2.625 or an additional $32,000 annually. Payment grows as risk increases.

Obviously, this matters a great deal to you as an insurer. Risk Adjustment can be the difference between massive losses or a healthy profit on any given member.

And how does Risk Adjustment work, mechanistically?

Over the course of a year, every time you receive a service your provider submits a claim with a diagnosis (ICD10) and service codes (CPT, HCPCS, etc) to your insurer to get paid. For example - if you are a diabetic who goes to see your primary care doctor for a well visit, your claim might contain the following information:
    • Diagnosis: E11.9 - Type 2 diabetes mellitus without complications
    • Procedure: 99211 - Level 1 Established Office Visit
If the current year is 2017 then we look back at all of your claims from 2016 and we pull out the diagnosis codes from those claims. These diagnosis codes are fairly specific - there are probably 30+ codes describing diabetes for example. These individual codes are then rolled up to Hierarchical Clustering Codes (HCCs) - so all 30 diabetes diagnoses for instance become a single HCC 19, "Diabetes without Complication". It is these HCCs that map to distinct payment factors.


Example HCCs and corresponding factors from 2017. Note that in different populations (e.g. Medicare / Medicaid dual eligibles; qualification for Medicare on age vs. disabled status) the factors have some variation.

At the end of the day, these various HCC factors are combined, along with a factor related to age and gender, to get your overall payment multiplier. (NB: there is a bit of complexity here that I am glossing over but this is effectively how it works).

An example: a member has claims for diabetes and COPD in 2016. These claims confer HCC 19 and HCC 111 respectively. These factors are added with a age/gender factor (0.70 here for a male age 85) and then multiplied by a baseline monthly payment ($1,000 here) to the your risk-adjusted monthly payment for this member.
Thus, your payment stems from your members’ diagnoses in the previous year that were recorded in medical claims.

This system sounds logical. What's wrong with Risk Adjustment?

What if you have a member with a specific diagnosis that for one reason or another doesn’t get recorded in a medical claim? You saw that Mr. Smith had type 1 diabetes in 2015, but when 2016 rolls around there is no type 1 diabetes claim. If you are an insurer in that situation, you end up leaving a lot of dollars on the table. And for a number of reasons, this happens all the time. A couple of examples:
    • (A) Patient seeks care for complications of disease or associated pain, but never the underlying diagnosis (e.g. the amputee never actually came in complaining about the amputation itself and it wasn’t recorded).
    • (B) Diagnoses reflect symptoms but don’t name specific disease (difficulty swallowing or dysphagia vs. esophageal cancer).
    • (C) Patient receives medications for a disease, but never actually sees a doctor for that diagnosis.
    • (D) The patient doesn’t see a doctor at all in a given year.

The waste comes when insurance companies hire home health agencies to visit members' homes, conduct exams, and capture diagnoses. For our aforementioned Mr. Smith, you do the ROI calculation in January 2016: if I get a claim with a diabetes diagnosis, that claim is worth $1,000 to me - so I send someone to his home to make sure we get the diagnosis written down. I would imagine that sometimes there is a veil of patient interest, but I am doubtful that many high-value medical interventions get delivered solely on the basis of these visits.

When you consider that perhaps a third of beneficiaries in a given calendar year won’t use their insurance at all, there end up being a lot of diagnoses that don’t get captured in claims. Diagnoses that mean thousands to tens of thousands of dollars to insurance companies.

This assessments may cost the insurer a few hundred dollars, but even for the HCCs with the smallest factors, the payoff exists. Sometimes the insurer will even throw in a $25 gift card for the member as well.

In this case, an insurance company is generating several hundred dollars in pure administrative waste per member to receive higher reimbursement from CMS. This cost probably gets passed along to all of us in the form of higher premiums and tax payments [4]. Meanwhile, home health agencies like Censeo Health (which by their own admission holds “HCC-yield” as a goal) have become businesses with hundreds of millions in annual revenue [5]. 


From the Censeo Health website.

I do not think that insurance companies are bad actors. Neither are the home health agencies. They are responding to a highly complex system that has resulted in some unintended consequences. I think there may be a relatively simple solution however.

How can we reduce the administrative waste associated with home health visits?

What if we could give credit for diagnoses we knew that members had to have regardless of whether or not that service got recorded in a claim? You wouldn’t need to pay for home visits. Insurance companies would be guaranteed to get credit for the diagnosis.

A partial solution is to assign HCCs to members based on not just the previous year's diagnoses, but a member's entire medical history, for those chronic diagnoses that do not change year to year.

Many of the diagnoses that confer HCC status and add multipliers are chronic diseases that generally don’t go away — genetic diseases, T1D, COPD, CHF, amputation, multiple sclerosis, etc — and we could automatically give insurance companies credit for these diagnoses regardless whether the claim occurred in 2013, 2014, 2015, or 2016. Due to increased diagnosis capture, CMS would probably need to dial back per-HCC reimbursement multipliers to keep reimbursement rates steady.

A proposal: for chronic diagnoses that generally do not go away, maintain HCCs from multiple years to avoid wasteful home visits. In this case: if a member had a recorded COPD diagnosis in 2015, do not require another claim in 2016 to get credit for the diagnosis in the 2017 payment year.

This would not completely do away with administrative home visits, but it would obviate the need for a good number of them.

If this works, the net effect would be less administrative spend by insurance companies which could mean higher profitability or lower premiums or ideally both. Insurance companies that are not as effective at HCC capture will have the playing field leveled with the bigger players. Everyone can win. The biggest losers in this scenario would be the home health agencies - though hypothetically can still be hired for home visits so long as it bears some medical utility.

Any concluding thoughts?

I concede that there is a fair amount of inference here and some of this assessment could be wrong. I would welcome criticism or alternative ideas.

I do think there is a real problem with the status quo. No neutral observer would look back and call this a highly rational way of doing things.

Complex systems can bring about unintended consequences. You can’t really recognize those consequences until you actually try to build the system. This is why we should constantly be changing variables, seeing what changes, and learning from and adapting to those results.



[1] Papanicolas et al. (2018) “Health Care Spending in the United States and Other High-Income Countries” JAMA March 13, 2018 Volume 319, Number 10
[4] Money being fungible, I suppose this could also be “lower executive bonuses” or “lower company profitability”. Either way it is an economic waste. Humanity is spending money on activities that net us very few John Stuart Mill points.
[5] https://www.forbes.com/companies/censeohealth/


Appendix: Why would anyone choose coverage through a Medicare Advantage plan over FFS Medicare in the first place?


Simplest answer: you have to pay some premium for normal Fee for Service (FFS) Medicare coverage, and for many MA plans you don't. By offering fewer benefits, a narrower network, etc, it's sometimes possible to cut some cost out of the equation. There may be more to it than this but my understanding is that for many folks this is the key driver.

No comments:

Post a Comment