The risk score or risk-adjustment factor for an individual can be calculated by adding up a member’s HCC scores on the Model Output Report, and adding their demographic score.  Then,  divide by the current payment year FFS normalization factor, for 2009, this is 1.03. This should yield the Part C score for the patient as shown on their Member monthly report (MMR).

Here is an example (using Community Model):

Male, 81 years old: demographic score – 0.597
HCC 18- Diabetes with Ophtho – 0.259
HCC 80- Congestive Heart Failure- 0.410
HCC 108 – COPD – 0.399

Total Risk score = 1.665

Divide by normalization factor of 1.03, so risk score = 1.616

Remember, several of the ICD-9s that map to HCCs also automatically map to RxHCCs as well, so the patient’s Part C and Part D risk scores can both go up with the capture of one ICD-9 code.   For example, the ICD-9 for Congestive Heart Failure, unspec. (428.0) maps to both HCC 80 and RxHCC 91, so the patient’s monthly capitated payment for both Part C and Part D will go up.

Both the HCC and RxHCC payment models contain groupings or hiearchical condition categories of related disease states such as diabetes, infection, or cancer.  The model is set up so that only the most severe form of a particular disease is reflected in the patient’s risk score and payment.  The CMS-HCC model contains 70 HCCs in the standard community model (usually covers more than 95% of all patients in a plan), and there are 84 RxHCCs in the Part D model.

The CMS-HCC Risk Adjustment model is actually comprised of four separate models depending on the population: Community, Long Term Institutional, End Stage Renal Disease, and New Enrollee.  The population a patient is in determines the value of their risk adjustment factors or diagnoses.  

New Enrollees are patients with < 12 months of Medicare Part B entitlement.  LTI patients have been in a long term care facility for > 90 days.  ESRD patients are those patients who are receiving dialysis and have been designated as ESRD by physicians using the ESRD Medical Evidence Report (CMS Form 2728).  

The New Enrollees only get credit for demographics/Medicaid status, whereas the other 3 models get credit for diagnosis data as well.

Most beneficiaries fall under the Community model, which along with the LTI model uses the same 70 HCC categories.  The ESRD model excludes the dialysis/renal failure HCCs, and the New Enrollee model has no HCCs.

I finally stumbled upon the most articulate, succinct explanation for what the CMS-HCC risk-adjustment model really is: Comes from the journal Medical Care, Vol 43, Number 1, January 2005, pg. 34.

The “CMS-HCC model uses demographics and a diagnosis-based medical profile captured during all clinician encounters—both inpatient and outpatient—to produce a health-based measure of future medical need.”

The CMS-HCC 70 model is a great, simple, ready-made tool to risk-stratify Medicare patients according to disease burden severity.  The specific codes that CMS chose for inclusion in the model were chosen based on many factors: 1) they represented diseases of sufficient prevalence so that there would be enough cost data to calculate average costs per member, 2) for the most part, they represented diseases that increased costs in an additive way (i.e. not exponential or multiplicative – the exceptions are the 6 diseases interactions that increase costs in a “super” additive way, i.e. by an extra factor beyond just adding them together), 3) In some cases, the ICD-9 codes chosen actually represent an interesting degree of specificity or precision in describing the patient’s condition.  A good example of a relatively precise cluster of related, but differentiated ICD-9 codes are the atherosclerosis of the native arteries codes in the 440.0 to 440.29 range.  Here, the ICD-9 codes differentiate atherosclerosis with symptoms (rest pain, claudication) from atherosclerosis with manifest signs (ulceration, gangrene). The predicted prospective costs of atherosclerosis with signs (and reimbursement to MA plans) is approximately twice that of atherosclerosis with symptoms.

ICD-10, with its increased number of codes and dramatically increased level of detail, will be an enormous boon to informaticists charged with the task of risk-stratifying patients either for capitated reimbursements purposes or for disease management.  Here is a look ahead at the 50+ ICD-10 codes for atherosclerosis. We will be able to further stratify patients based on whether the condition is in the left or right extremity or bilateral. Unfortunately, it may well be 2015 or 2016 before CMS has captured enough data to further risk stratify patients based on more specific ICD-10s.

Make sure your ICD9/HCC recovery efforts include the Part D or RxHCC valued ICD9 codes as well.  Although, individually each RxHCC recovered may be less valuable compared to the value of the HCCs in Part C, in the aggregate their increased prevalence is likely to have an impact.  The Part D payments are a bit more complicated, but they are very much influenced by a patient’s RAF or risk score. The patient’s risk score changes the capitated, direct subsidy portion of CMS’ payment to the plans with Part D benefits. Each patient will have a Part A/B, i.e. “Part C” or “HCC based” risk adjustment factor and a Part D risk adjustment factor – these will be reported separately on the Member Monthly Report.

There are several ICD-9 codes that have both HCC and RxHCC value, however there are over 2,000 ICD-9 codes that have only Part D value.

Here are the fresh MedPac Payment Methodology (Oct 2008) papers for: Med Adv – Part C  Med Adv – Part D
You will need Adobe Reader 8.0 or better, each file is 0.8MB.

The CMS-HCC model remains essentially the same for the 2009 payment model.  It will contain the same 70 hierarchical condition categories (HCCs).  The hierarchy within and between disease groups remains the same.  CMS used a new 2007 denominator of $7,463.14 to adjust the risk factor values somewhat.  So, the model is the same, but the RFVs have been tweaked, i.e. the relative monetary values have been updated.  Keep in mind, the “value” of an HCC or ICD-9 code is different for every plan depending on their plan bids.

The denominator of $7,463.14 is the predicted annual cost of taking care of an average Medicare bene – this is about $622 per month.  The denominator does not change every year because CMS likes to roll in 2 or 3 years worth of data into the denominator before updating it.

A look at the most recently updated list of acceptable ICD-9 codes for each payment year reinforces CMS’ intention to make sure from now on that acceptable ICD-9 codes are also “valid” ICD-9 codes, i.e. not the incomplete or so called “truncated codes” that were accepted  in previous years.  All of the acceptable ICD-9 codes for HCCs in payment year 2009 are not truncated with the exception of a single truncated ICD-9 code, namely 337.0 Idiopathic Peripheral Neuropathy – this may just be an oversight as CMS intends to not use truncated codes going forward, so it would be best to carry the 337.0x to its fifth digit, either 0,1, or 9.

You can always find the current model diagnosis codes at cms.gov, then search for “Risk Adjustment” in the search box on the top right corner.

For everyone who has wondered why they cannot come up a certain patient’s RAF score by just adding the risk factor values of their HCCs and demographic variables, remember that the RAF or risk adjustment score on the Member Monthly Reports (MMRs) has been already divided by the FFS normalization value.  In 2008 this was 1.04.  This downward adjuster accounts for increased coding intensity from year to year and is an effort to keep the average beneficiary’s risk score right at 1.0.  For example, if you calculate a member’s RAF at 1.04, you must then divide by the FFS normalization factor 1.04 to get to the 1.0 that you will see on the MMR.

Helping your providers embrace the challenge of reporting Physician Quality Reporting Initiative (PQRI) measures to Medicare for performance-based bonus payments to the providers may yield dividends for Medicare Advantage plans as well. Many of the ICD-9 codes that serve as population denominators for inclusion into the data set for PQRI reporting also carry RxHCC or HCC value.  Even though the main intent of the CMS-HCC model was to be a prospective predictor of costs, it is by design a reporting system that identifies chronic diseases with considerable prevalence and aggregate cost, i.e. diseases for which there is financial incentive to identify and improve the quality of treatment for.  Here is a list of the 119 PQRI measures for 2008.

Add: 10/29/08 – Of course the “Medicare Patients” that the providers will be reporting on are mostly the Medicare Part B or FFS patients. There is one instance in which providers can use Medicare Advantage patients for their “registry reporting” of consecutive patients – see the CMS PQRI Fact Sheet for further details. My point is, a Medicare Advantage plan should incorporate into their Risk Adjustment education any opportunity to help providers document better and provide better quality of care and better disease management- this will help them both with MAO patients and other patients as well, you may also have more success if you sell them on the idea that better and more accurate coding will help them across the board.