Six Reasons Your Risk Scores are Inaccurate

By Kameron Gifford, CPC

As we enter the third quarter of 2020, healthcare organizations have been struggling for months to balance priorities and resources while navigating new technologies and processes to keep employees and patients safe during the coronavirus.

The AAFP and MGMA both recently reported, “Medical group practices of all sizes and specialties have felt the direct and indirect financial impact... On average, patient volumes have dropped 60% nationally since the start of the pandemic attributing to a 55% decrease in fee-for-service revenues.  

Medical practices are not alone, hospital revenue is dropping by an average of $1.4 billion per day as COVID-19 continues to impact patient volumes, according to Crowe RCA Benchmarking analysis.

Risk Scores and Value-Based Payments

As more and more healthcare organizations are moving away from traditional fee-for-service payment models, how will this decrease in utilization impact risk scores and value-based payments in the future?

According to Avalere, the deferral of care has resulted in fewer claims and diagnoses among Medicare Advantage (MA) enrollees, which will likely lead to a 3%–7% reduction in 2021 risk scores and lower plan payments.

Mitigating the Impact to Risk and Quality

Inaccurate risk scores not only impact payments to Medicare Advantage plans, but also skew the costs in ACOs and hinder performance in value-based contracts. This further underscores the need to capture an accurate health status on every patient.

What steps can organizations take today to achieve accurate risk scores and mitigate future losses?

Common Errors Leading to Inaccurate Risk Scores

The 2020 ICD-10-CM code set includes 72,184 diagnoses and the 2021 ICD-10-CM code set includes 72,616 diagnoses. With less than 14% of ICD-10-CM codes mapping to an HCC, lack of specificity is the most common cause of inaccurate risk scores.  

Review the six most common coding errors below that lead to inaccurate risk scores and payments.

E11.9 – Type 2 Diabetes without Complications

According to the most recent data released by MedPAC on July 17, 2020, 28.2% of Medicare beneficiaries had a diagnosis of diabetes on a claim in 2019. Roughly 70% of these mapped to HCC 18, Diabetes with chronic complications, while 30% of these mapped to HCC 19, Diabetes without complications.

How does your coding for diabetes compare to MedPAC data? What percentage of patients are coded as E11.9 by primary care providers? What percentage are coded as E11.9 by specialists such as hospitalists and endocrinologists?

Per ICD-10 Guidelines, approximately 30 conditions have an assumed relationship with type 2 diabetes. Meaning, they are always coded as a complication unless the medical record explicitly states otherwise.

Examples include:

• Type 2 Diabetes with CKD, E11.22
• Type 2 Diabetes with dermatitis, E11.620
• Type 2 Diabetes with foot ulcer, E11.621
• Type 2 Diabetes with gastroparesis, E11.43
• Type 2 Diabetes with hyperglycemia, E11.65
• Type 2 Diabetes with hypoglycemia, E11.649
• Type 2 Diabetes with mononeuropathy, E11.41
• Type 2 Diabetes with myasthenia, E11.44
• Type 2 Diabetes with nephropathy, E11.21
• Type 2 Diabetes with neuralgia, E11.42
• Type 2 Diabetes with neuropathy, E11.40
• Type 2 Diabetes with PAD, E11.51
• Type 2 Diabetes with periodontal disease, E11.630
• Type 2 Diabetes with polyneuropathy, E11.42
• Type 2 Diabetes with retinopathy, E11.319

Review 5 – 10 encounters per provider. What percentage of encounters coded as E11.9, had a complication documented in the medical record? Target education, prospective chart checks and pre-billing review per the results.

How it Happens

Several factors contribute to the high error rate related to coding for E11.9. The two most common reasons are failure to update the diagnosis as the disease progresses and failure to follow the ICD-10 Guidelines for “with”.

Why it Matters

The 2020 CMS-HCC RAF for HCC 19 is 0.105 and the 2020 CMS-HCC RAF for HCC 18 is 0.302. That is a loss of 0.197 per error. This adds up quickly across populations, accounting for annual average losses of $60,000 - $130,000 per 1000 MA beneficiaries.

F32.9, Major Depression, Single Episode, Unspecified

According to the most recent data released by MedPAC on July 17, 2020, 11.3% of MA beneficiaries were diagnosed with a condition mapping to HCC 59, Major Depressive, Bipolar or Paranoid Disorders.  

According to CMS, mood disorders (mainly MDD and bipolar disorder) are the second leading cause of disability in Medicare patients under the age of 65. Depression is a major predictor of the onset of stroke, diabetes, and heart disease; it raises patients’ risk of developing coronary heart disease and the risk of dying from a heart attack nearly threefold.

Overall, the economic burden of the disease is significant to managed care organizations, with direct medical costs estimated at $3.5 million per 1000 plan members with depression.

Identify Errors and Opportunities

From a coding perspective, MDD is classified by episode, severity, and remission.

According to AMJMED, 75% to 90% of patients experience >1 episode of depression. This suggests that only 10 - 25% of MDD diagnoses would be assigned to F32 with the remaining 75 – 90% being classified as F33.

Analyze your coding for MDD. What percentage of MDD diagnoses are single episodes (F32.x) vs. recurrent episodes (F33.x) What percentage of single episodes are classified as “unspecified” (F32.9) when a PHQ-9 was completed and/or the documentation supported a more specific diagnosis? What percentage of patients taking an SSRI or other antidepressant have a current diagnosis to support medical necessity?

Review 5-10 encounters per provider with a diagnosis of F32.9. Was the correct diagnosis assigned? Target education, prospective chart checks and pre-billing review per the results.

Why it Matters

From a risk adjustment perspective, F32.9, is the only MDD diagnosis that does not map to an HCC. The 2020 CMS-HCC RAF for HCC 59, Major Depressive, Bipolar and Paranoid Disorders is 0.309. Missed opportunities relating to the use of F32.9 average 20% across populations accounting for an average annual loss revenue of $77,500 per 1000 members.

How it Happens

Two factors contribute to the high use of this code. First, the GEM files mapped the ICD-9 code 311 to the ICD-10 code F32.9 and these files were widely used by EHR vendors. The second factor involves the number of boxes providers must check in their EHR to get to the more specific MDD diagnosis.

One way to avoid these extra clicks is by typing the diagnosis code directly into the search box of your EHR. For example, typing F32.0 (MDD, single episode, mild) vs depression will reduce clicks from 13 to 4 and reduce search results from 800+ to 1. Searching for F33.0 (MDD, recurrent, mild) vs recurrent depression will save even more clicks with the same results.

I25.9, Chronic Ischemic Heart Disease and I25.10, CAD without Angina

According to the NIH, an estimated 10 million adults in the United States carry the diagnosis and ischemic heart disease remains the number one cause of death for male as well as female patients. Furthermore, the increasing survival with the use of modern therapies has produced an aging population where more than 20% of women and 35% of men above the age of 80 have coronary artery disease.

Identify Errors and Opportunities

From a coding perspective, chronic ischemic heart disease is classified to category I25 and CAD is further classified as with or without angina.

Analyze your coding of chronic ischemic heart disease (I25.9) and CAD without angina (I25.10). Depending on your results you may also want to include old MI (I25.2) and chest pain (R07.9) in your search.

Review 5-10 encounters per provider. How many of these patients had evidence of angina documented, history of CABG and/or a current prescription for nitroglycerin? Target education, prospective chart checks and pre-billing reviews per the results.

How it Happens

The term stable ischemic heart disease (SIHD) is often used synonymously with chronic coronary artery disease (CAD) and encompasses a variety of conditions. Many EHR’s include an IMO to assist providers in searching for codes. This “tool” adds multiple code descriptions for each ICD-10 code and can increase search results by 70%. Many providers do not have the time to search dozens of code descriptions for multiple diagnoses prior to closing their note. This often results in the selection of the first or second result, even when a more specific diagnosis is supported by the documentation.

Why it Matters

From a risk adjustment perspective, I25.9 and I25.10 are included in the Rx-HCC Model V05, but not in the CMS HCC Model V24. However, CAD with Angina (I25.110 – I25.119) and Angina (I20.0 – 120.9, I23.7) are all included in the CMS HCC Model V24. The 2020 RAF for HCC 87 is 0.195 and HCC 88 is 0.135.

According to the CMS Chronic Disease Warehouse, 10,238,321 (or 17.1%) Medicare beneficiaries had a diagnosis of ischemic heart disease. While the most recent MedPAC data published on July 17, 2020 reveals only a 4% prevalence rate among MA beneficiaries in the same year.

Missed opportunities relating to the use of I25.9, I25.10, I25.2 and/or R07.9 average 25% across populations accounting for an average annual loss of $64,638 - $93,366 per 1000 MA members. 

I49.9, Cardiac arrythmia, unspecified

According to the most recent data released by MedPAC on July 17, 2020, 11.4% of Medicare Advantage members had a diagnosis that mapped into HCC 96, Specified Heart Arrythmias.

In the CMS-HCC Model V24, 18 ICD-10 codes are mapped into HCC 96.

Examples include:

• AV Block, Complete, I44.2
• SVT, I47.1
• Paroxysmal A. Fib, I48.0
• A. Flutter, I49.92
• Sick Sinus, I49.5

Identify Errors and Opportunities

Analyze your coding for cardiac arrythmias. What percentage of encounters/claims are coded with I49.9, Cardiac arrhythmia, unspecified when the medical record supported a more specific diagnosis? 

You may also want to include the use the ICD-10-CM code Z95.810, Presence of automatic (implantable) cardiac defibrillator, in your analysis. Target education, prospective chart checks and pre-billing reviews per the results.

How it Happens

AHA Coding Clinic recently updated their guidance on coding for sick sinus syndrome treated with a pacemaker. This change in guidance has led to an increased number of opportunities identified in HCC 96. Additional opportunities are identified from diagnostic test results and specialists’ reports.

Why it Matters

I49.9, Cardiac arrhythmia, unspecified is not included in the 2020 CMS-HCC Model V24.

Missed opportunities relating to HCC 96 average 20% across populations accounting for average annual lost revenue of $67,214 per 1,000 MA beneficiaries.

N18.9, CKD, unspecified

According to the CMS Chronic Condition Warehouse, there were 9,360,944 Medicare beneficiaries (15.6%) with a diagnosis of CKD on a claim in 2018. However, a review of the most recent data released by MedPAC on July 17, 2020, does not include CKD, meaning the prevalence for MA members in the same year was less than 1.5%.

Why Is Chronic Kidney Disease Important?

CKD and ESRD are very costly to treat. Nearly 25% of the Medicare budget is used to treat people with CKD and ESRD.

The total Medicare spending on both CKD and ESRD patients was in excess of $120 billion in 2017. For identified CKD (not ESRD) the total Medicare expenditure was $84 billion.

Identify Errors and Opportunities

Analyze your coding for CKD. What percentage of encounters/claims are coded with N18.9, CKD, unspecified, vs. a more specific code such as N18.3 and/or N18.4?

You may also want to include the ICD-10 code N28.9, disorder of kidney and ureter, unspecified, in your analysis.

Review 5-10 encounters per provider. What percentage of encounters/claims are coded with an unspecified diagnosis such as N18.9 and/or N28.9, when a more specific diagnosis is supported by the medical record? Target education, prospective chart checks and pre-billing reviews per the results.

How it Happens

There are several factors that contribute to this large opportunity. Lack of documentation is the most common reason. CKD must be staged by the provider. Pasting a copy of the patient’s most recent labs into the current encounter supports the provider’s medical decision making but does not replace the need for the stage to be documented.

The fluctuating nature of the disease also contributes to the lack of specificity in coding, as providers are less likely to update.  

Historically, multiple terms have been applied to chronic kidney disease (CKD), eg, chronic renal insufficiency, chronic renal disease, and chronic renal failure, the National Kidney Foundation Kidney Disease Outcomes Quality Initiative™ (NKF KDOQI™) has defined the all-encompassing term, CKD. This recent change in terminology also contributes to the size of the opportunity. The IMO search tool in EHR’s will lead providers using older terminology such as, renal insufficiency, to select a diagnosis of N28.9

CKD stage 3 was removed from the HCC model in 2014 and this likely contributed to the decrease in coding by MA plans as well. CMS reversed course in PY 2019, and added HCC 138, CKD stage 3, back into the model.

Why it Matters

Missed opportunities relating to HCC 138, CKD stage 3, average 60% across MA populations accounting for annual average lost revenue of $73,625 per 1000 members.

Want to learn more? Visit www.erm365.org and www.ermconsultinginc.com

ERM Consulting Inc. works with healthcare organizations across the country to optimize their risk adjustment operations.

·