CMS released the 2015 policy year risk adjustment and re-insurance payment notice on Thursday afternoon. Several billion dollars net are changing hands. Roughly 10% of total premiums is changing hands.
Larry Levitt has the smartest take on the cash flows:
The $ changing hands under risk adjustment reflects some insurers getting sicker enrollees, and some being better at operating the system.
— Larry Levitt (@larry_levitt) June 30, 2016
What does Larry mean by “operating the system?”
Experienced insurers know how to optimize their risk scores while inexperienced insurers are still groping forward with limited data. Insurers that have only ever operated in the small group and individual market are at a data disadvantage compared to insurers that operate in individual, small group, CHIP, Medicaid, Medicare, and large employer group markets.
We’ll talk through an couple of examples of how the system is worked.
The first case is pure data mining from only individual market owned QHP claims.
Let’s say hello to Mary. She is 48 years old with Type 2 diabetes that is under control, asthma that is well controlled by drugs, and she had a pacemaker inserted a couple of years ago. Each of those conditions needs to show up on a valid claim every year for her insurer to gain risk adjustment points.
This is the first area of differentiation on operating a system between a good operator and a naive operator. The experienced operator will take a look at a couple of years of claims history and identify that Mary had a Type 2 diabetes diagnosis submitted in 2014, the pacemaker V-code submitted in 2014 and asthma medication scrips get routinely filled on the 24th of the month every other month. This information will run through a scrub program and produce a targeted diagnosis form. Mary’s primary care provider will get the form in her chart. or it will be inserted into the electronic medical record. The next time Dr. Patel sees Mary, he’ll add the relevant for the visit diagnosis code (conjunctivitis/pink eye) as well as the three extra valid and legitimate well documented but not coded diagnoses (Type 2 Diabetes, asthma, pacemaker status). The insurer will then pay him $400 for the coding. The insurer will now receive credit for an extra $10,000 or $20,000 worth of risk adjustment points.
Naive insurers will not have the system set up to aggressively chase from recent claims nor do they have the staff set up to do the chasing. Insurers that have plenty of experience working Medicare Advantage will just add a couple of data geeks to that team and a dozen outreach folks to support Exchange HCC.
Now let’s take Mike. He had his left arm amputated above the elbow in 2007 due to an industrial accident. He was covered by employer sponsored coverage through March of 2015. He then goes on Exchange to get a new policy as he moved across state lines. He does not go to the doctor for the rest of the year. He gets a flu shot in October and visits an urgent care to get stitches in September. For the current year, he is a very low cost member.
If he chooses the local co-op for his Exchange insurance, they see him as a very low cost member and they do not know anything else.
If he chooses the same insurer that covered him in 2007 when he lost his arm, they are able to mine their data and identify that he does not have a left arm. That condition does not go away, it does not change, it does not get better. They flag this and a coding specialist is allowed to go through his chart and submit a retrospective risk coding request supported by the 2007 claim and medical chart data. CMS will accept the change and attribute a $3,000 risk adjustment bump to the incumbent insurer with deep data sets.
The incumbent with the deep data sets and experience at running risk adjustment is not paying any additional money in claims expense for Mike. He is just a bonus revenue center.
That type of data mining is legal, and it is common (it is my 11:00 meeting this morning). However it is a competency that requires a lot of experience, specific and expensive technical knowledge, and deep data sets.
Other factors impact risk adjustment. Narrow networks with restrictive gatekeepers that are priced very low will have large net risk adjustment outflows as they attract the healthy and the young. PPO’s with broad networks are magnets for sick people. Knowing how to maximize risk scores is a key component of the risk adjustment swings but inherent plan design is probably a stronger factor in most cases.
Smedley Darlington Prunebanks (Formerly Mumphrey, et al.)
I’m about to make a doctor’s appointment this morning, now that we have insurance again! O, happy day! Have you settled on where we’re going to meet on Sunday evening?
Richard Mayhew
@Smedley Darlington Prunebanks (Formerly Mumphrey, et al.): next post up :)
Wag
Thanks for the interesting view behind the curtain. As a PCP who directs a large outpatient clinic for a major teaching hospital each summer I begin the process of trading a new batch of IM residents in coding their visits so that they get credit for the complexities of medical decision making with patients like your first example, Mary. If Mary comes in complaining of a cough, her asthma obviously plays into your decision-making. If she really has an asthma exacerbation, you need to take into account her underlying diabetes as well, because the steroids will give her for the asthma may cause her diabetes control to slip. Teaching our residents to recognize these complexities is fairly easy. Getting them to make the next step, and code for each of the diagnoses (and document their thought process) that are influencing their decision-making is more of a challenge.
Richard Mayhew
@Wag: that is actually a conversation or a job shadow I would love to have for a couple of days over the next week or two as the newbies get on the floors for the first time. Good docs that prefer to do minimalist interventions tend to be very good coders and charters as they explain why they are deciding to do or not do something. But the actual act of choosing what factors were involved in evaluation/decision making is something that I know nothing about.
Wag
@Richard Mayhew:
It’s an interesting process. The gut instinct is to focus on just what’s in front of you and to only code the symptoms that comes in front of you. Part of the education process is teaching the residents to recognize the broader context of the patient’s care (including past medical history, as well as social determinants of health, i.e. habits like smoking or drug use, as well as the impact of insurance status and Socio economic issues) and to give themselves credit for taking those things into account.
Mary Jo
I love these “look behind the curtains” posts. This was a good one.
Richard Mayhew
@Wag: That raises a good question — who are our best code closers — does it correlate with geography, specialty, who their educator is and/or experience — something to look into during down time
StringOnAStick
This is why I love Richard’s posts – we learn so much about a system we all interact with but have little understanding of. Thanks!
amygdala
@Wag: This is one of many reasons I wish there had been more practitioner input into EHR design. Good systems can be great teaching tools for students and post-grad trainees. Bad systems can set up sloppy documentation (and worse, reasoning) habits that may well last over an entire career. Distill the best practices of the master clinicians and try to incorporate that into an EHR. Is that too much to ask? Evidently so.
We went through two output EHRs before I threw in the towel and retired. Both times I had to serve as transcriptionist–from paper to pixels the first time and from mostly free-text to ICD codes the second. And each time I was left with notes that made it harder for me to see my patients, much less anyone covering, for patients needing to be seen while I was away.
It’s even worse with EHRs that make it hard for students to use the system. They’re potentially some of the best source for fixes, since they’re technologically more savvy than pretty much everyone above them in the hierarchy. But except for the VA’s CPRS, students are often excluded from the system. It’s so short-sighted.
Prescott Cactus
Say Prescott Cactus is an asthmatic who can control my symptoms with OTC meds, but yeah it’s a lot easier to get by with the good stuff (scripted inhalers). He also has a bum knee which only bothers him after a bit too much weekend softball. I move out of state and get new insurance. If I mention it to my new Doc am I raising my premium next year (or rest of my life).
Like Mike who lost his arm, does his $3,000 risk adjustment bump cost him more for premiums forever ?
Richard Mayhew
@Prescott Cactus: nope, your premiums are the same no matter what your medical history is.
Everything in this post is things you never see as an individual
Bob P
Sorry, but the “deep dataset” theory doesn’t hold up to scrutiny as a core reason for large RA transfers from new or small plans. Yes, the effect exists. No, it is not material. Yes, an effect exists regarding “data naivete” by new or small plans. But any such problem speaks volumes…and not in a good way…about the management team.
New and small plans are GREAT…but it would be lousy public policy to cross-subsidize them in a back-door way, via risk adjustment. My narrow-network plan is our state’s biggest payor this year, and deservedly so.
Richard Mayhew
@Bob P: agreed it would be shitty public policy to back door subsidize new plans via risk adjustment.
Knowing how to run a data intensive risk adjustment revenue maximization program will drive a couple points of revenue over not running that program. A narrow network hmo is guaranteed to run big net RA outflows. My point is aggressive chasing from propriety data can reduce that outflow. The co-ops might have enough data in 2016 to do a half decent chase. My employer had a decent chase system in 2014 and a deep dive for 2016. It helps but won’t dominate plan and network design decisions.