Northwell Health Lab Team Leverages Data to Add Value

Helps parent organization by finding patients eligible for risk-adjusted payments

Share on facebook
Share on twitter
Share on linkedin
Share on print
Share on email

CEO SUMMARY: Today, insurers get risk-adjusted payments for treating patients who have high-cost health conditions and they make risk-adjusted payments to physicians, hospitals, and other providers. At Northwell Health, the clinical lab saw the opportunity to leverage lab test data with other clinical and demographic data to identify patients who were undiagnosed or whose conditions were undocumented under the hierarchical cost category (HCC) system. Using data in this way allowed the lab’s parent healthcare system to qualify for additional risk-adjusted payments associated with these patients.


WHILE MORE HEALTH INSURERS PAY FOR VALUE-BASED CARE, clinical laboratories are realizing that the insights derived from lab test data may have more value to insurers than the lab-test results themselves.

Some health insurers are recognizing a fact well-known to clinical pathologists and lab scientists for decades: insights derived from patients’ lab-test data have value because a health insurer can use those insights to manage the health of each of its members more effectively and more efficiently than it may have done so previously.

One factor behind health insurers’ interest in value-based care is the increased payment that health plans participating in the Medicare Advantage (MA) program get from the federal Centers for Medicare and Medicaid Services. Those payments are based on the risk scores that CMS assigns to Medicare Advantage plan members. As risk scores rise, CMS adjusts the payments it makes to health insurers.

As a result, increased risk scores translate into increased payments for MA plans and for the physicians and other providers serving those patients under shared-risk arrangements between payer and providers.

“Under shared- and full-risk arrangements, increased payments can flow from CMS to health systems and providers,” explained Liya Lomsadze, the Project Manager in Pathology Informatics for Northwell Health, in New Hyde Park, N.Y.

“This is a positive development for clinical laboratories, because payers, providers, and patients can benefit when risk adjustment works as intended to improve patient outcomes and reduce healthcare costs,” Lomsadze added. “Labs are positioned to benefit because they are often first to know when a patient has a costly disease or health condition.”

In the past, insurers paid clinical labs based on the volume of tests they ran, a practice that sustained clinical laboratories for years. But today, labs seeking to improve care delivery under the clinical lab 2.0 service model understand that health insurers using shared-risk arrangements will pay more to providers who care for patients who have more costly health conditions.

New Lab Revenue Streams

These labs also understand that they can develop new revenue streams when they develop insights from the lab-test data they’ve collected on patients over many years, and then deliver those insights to physicians and other providers who care for high-risk and high-cost patients.

At THE DARK REPORT’s Executive War College in May, Lomsadze gave a presentation on risk scores and risk adjustment, explaining that insurers use risk scores to identify the relative illness burden for each patient or for a group of patients. Her presentation was titled, “Why Risk Adjustment Is Every Lab’s Surefire Way to Add Value: What It Is, How It Works, and Where Savings Are Found.”

She explained the issue in two parts. In the first part, she outlined the basics of risk adjustment and how insurers and providers are paid for managing the financial risk of delivering care to patients.

In this part, she outlined how a clinical laboratory can support health insurers, health networks, physicians, and other providers who are delivering care in value-based payment systems in which all the payers and providers are sharing the financial risk of caring for a panel of patients in a risk-adjusted environment.

In the second part, she provided a case study on how Northwell Health’s clinical laboratory volunteered its capabilities to help the health network’s administration identify patients with increased health risk. Lomsadze explained how the lab used its database of lab-test results to identify patients whose conditions were either undiagnosed or whose conditions were undocumented. Doing so allowed Northwell Health to qualify these patients for increased payments associated with the increased costs.

PART ONE: Understanding Risk Adjustment

“The risk adjustment concept is based on the recognition that patients with more costly and chronic conditions require more intensive care than patients who are relatively healthy,” said Lomsadze. “Thus, in a risk-adjustment arrangement, CMS pays health insurers more for sicker patients than they pay for patients who are healthy. Increased payments are based on the risk scores of the insurers’ members.

“Health insurers assign each patient a risk score based on the patient’s level of illness,” she continued. “Members who have the highest risk scores have the most-costly conditions, and patients with the lowest scores are relatively healthy.

“Risk scores can also be used throughout the health system in value-based contracting so that payers can reward physicians, hospitals, and other providers in an equitable way,” she said. “Thus, if my hospital happens to care for sicker people than your hospital does and we have a higher readmission rate, we can adjust for that so that every hospital gets judged fairly.”

Evaluating Risk and Reward

“Each patient’s risk score is determined based on the predicted cost of delivering care for a year. You can put any data you have on your member population into a regression model to predict the cost of care for each member,” she added. A regression model is used to analyze the relationship among two or more variables and to compare one variable against others.

“In the early days of risk adjustment, CMS had only basic demographic information on patients,” she commented. “But today, more risk adjustment models are incorporating data on medical conditions, the drugs patients take, and what procedures those patients have had.”

When assessing patients, payers use CMS’ hierarchical condition categories (HCCs), a system designed to estimate future costs. “There are two HCC models that apply to two populations,” Lomsadze explained. “One model applies to members of Medicare Advantage plans, and the other applies to members who have individual or small-group coverage under the Affordable Care Act (ACA).”

In her presentation, many of the risk-adjustment examples came from the MA model and others came from Northwell’s CareConnect health insurance program, which provided coverage for members covered under the ACA.

More Costly Conditions

“HCC models are useful for predicting the cost of care,” she noted. “For example, a relatively common condition, such as hypertension, which many individuals have, is, by itself, not a meaningful indicator of higher cost of care. But other conditions, such as diabetes or end-stage renal disease, are much less common and can be more costly.

“Under Medicare, there are 79 different conditions on the HCC list and as much as 60% of the Medicare population will have one or more of these conditions in the United States,” commented Lomsadze. “When people who are under 65 are added, there are about 128 different conditions associated with risk adjustment payments.”

Since 2004, Medicare has assigned risk scores to patients with certain conditions. For diabetes in patients under 65, the HCC score is 1.3, which means it costs about $470 more per month and about $5,645 more per year to care for a patient with diabetes, she explained.

For a patient with sepsis, the HCC score of 10.7 reflects a higher cost of care of $3,739 more per month, and $44,877 more over a year compared with a healthy patient.

“If a health insurance plan thinks it can control the costs of a diabetic population at less than $5,645 per patient per year, it might want to attract diabetes patients specifically and then control their costs down to or under that threshold,” Lomsadze explained. “That happened after the ACA was implemented when Aetna deliberately sought to attract diabetes patients from the health insurance marketplace.

“Labs can assist health insurers in identifying the most-costly patients by delivering data that allow payers to estimate each member’s risk score according to the patient’s level of illness,” she noted.

Northwell Health Lab’s Strategy Designed to Help Its Parent with Risk Adjustment

WHEN DESCRIBING THE STRATEGY taken by the clinical laboratory at Northwell Health in New Hyde Park, N.Y., Liya Lomsadze, the Project Manager in Pathology Informatics, explained how risk adjustment works, using the example of the Medicare Advantage Program. Table A below shows an example of the annual base premium an insurer would be paid, plus the additional premium amounts for the higher risk of selected conditions.

Table A: Risk Adjustments in Medicare Advantage Program

Additional annual payment to insurer for selected conditions

Diabetes without complications $1,058
Breast, prostate; other cancers, tumors $1,490
Diabetes with acute complications $3,251
Drug/alcohol dependence $3,910
Major depressive, bipolar, paranoid disorders $4,039
Lung and other severe cancers $9,904
Metastatic cancer and acute leukemia $26,795

Male, age 70-74
Base annual rate: $3,866
Source: Center for Public Integrity

PART TWO: Northwell Case Study

In the second part of her presentation, Lomsadze explained how the clinical laboratory at Northwell Health uses insights from lab test data to identify patients who have undiagnosed conditions or whose conditions are not documented properly in the electronic health record (EHR) system.

“In 2014, Northwell started a health insurance company called CareConnectto manage patients under the ACA,” she said. “But that plan lost more than $200 million due to risk adjustment, and, in 2017, a decision was made to close the plan.” For Northwell, that result was a significant lesson learned.

“Under the Medicare program, the base rate per member per month is a capitated payment set according to the age and gender of each member, and each additional condition that is documented in the patient’s record represents additional money from Medicare to the insurer,” she said.

“Given all of this, how would a payer prove that its members are as sick as they are?” she asked. “It would need to document all of the different conditions in claims. And there are four requirements that need to be met to document HCCs in claims:

1) The qualified provider needs to have a face-to-face encounter with the patient.

2) The encounter must be during the measurement period.

3) The encounter needs to be documented with an appropriate diagnosis code.

4) The payer needs to pay for that encounter.

“Once a payer follows and documents these four steps, it needs to repeat them annually,” Lomsadze advised.

“Regarding the face-to-face encounter with a provider, let’s say a patient only had a medical laboratory test last year but didn’t see his or her primary care physician,” she said. “That doesn’t count. If the patient got only a drug or a radiology test, that doesn’t count.

Evaluation and Management

“What’s required is an E&M code (for evaluation and management) and it must happen during the measurement period,” she added. “For most plans, that means the calendar year, but the plan year can start at other times as well.

“Also, the provider must actually bill the insurer with the appropriate diagnosis code that maps to the HCC,” she added. “In return, the payer has to pay the provider for that patient’s care.

“Risk adjustment requires the payer to repeat this process every year,” she noted. “But not every patient sees their physician every year.

“Consequently, payers and healthcare systems won’t get that added risk-adjustment payment for patients who have high-cost HCCs and who do not see their physicians every year,” Lomsadze observed. “This problem is just one challenge payers face when seeking to get paid for their patients with high-cost HCCs.

“It’s important for labs to recognize that payers rely on claims data,” she added. “We all know the problems with claims data, including timeliness issues and data gaps. And there is churn among members, meaning they move from one insurer to another.

Approaches to Risk

“Our lab has noticed that payers take one of two ways to approach risk adjustment,” she added. “One is passive, meaning the health system waits for its insured patients to seek the care they need organically. Then, the health system waits for the claims to roll in and maybe at year-end it audits a sample of patient charts to make sure nothing was missed.

“The other way is more active in which the payer seeks out those patients who have not seen a physician during the year. In healthcare, being more proactive in this way is a best practice,” she recommended.

“With this more active approach, a payer will use all the historical data at its disposal to identify HCCs that may have been dropped,” she said. “It may have clinical data to identify its high-risk members, for example. Then, the payer can develop ways to incentivize those patients to get annual health assessments, allowing all of their conditions to be captured every year.

“Once the payer follows these steps, more diligent clinical documentation and greater attention to proper coding can cause the risk scores of a population to go up,” she explained.

Lacking Historical Data

“Lacking the historical data it needed to do risk adjustment properly, Northwell’s CareConnect health plan became a casualty of sizeable unfavorable risk adjustments against it,” she said. “Understanding how to address these challenges became important lessons learned.

“In 2016, Northwell put out a call across the system to improve our ability to do risk adjustment,” she said. “At that time, our clinical laboratory told administration it could help address these challenges.

“In September 2016, our lab team reviewed the list of risk-adjustable conditions in the ACA insurance market,” she commented. “About half of these conditions can be diagnosed or monitored using a lab test, including hemophilia and end-stage renal disease.

“Next, using this list of lab-identifiable conditions, we made up two separate lists,” said Lomsadze. “On one list, we put the health conditions that have the highest risk-adjustment value, and on the other list, we added those conditions that are generally under-documented or under-recognized.

“Using data in our lab database, we used an algorithm to match our patients against the roster of enrolled members from our insurance company,” she said. “This patient-matching method allowed us to identify patients whose lab data indicated the presence of certain high-cost conditions. Then, the lab presented this analysis to our health insurer.

“That’s when the real work began, because our colleagues at CareConnect had to operationalize these clinically-actionable insights,” she added. “For example, they worked to confirm that—for each patient encounter—the diagnosis predicted by lab data was documented on that patient’s chart.

To Identify Patients with HCCs, Northwell Lab Combines Different Data Sets, Uses Algorithm

BELOW IS GRAPHIC B, WHICH LIYA LOMSADZE USED TO SHOW how the Northwell Health lab compares the health insurer’s roster of enrolled members to the patient in the laboratory data warehouse. Algorithms are used to identify patients with conditions included in Medicare Advantage’s hierarchical cost category (HCC). The lab submits these findings to the insurer.

Graphic B: Northwell Health Laboratory’s Steps Used to Identify Patients with Risk

USING THE STEPS ABOVE, TABLE C BELOW SHOWS THE NUMBER OF PATIENTS and dollar value that the Northwell Health Lab was able to identify as having a condition not in the patient’s health record that would increase the risk adjustment payments. Lab’s return on investment was 10 to one.

Table C: Lab Impact on CareConnect in 2016

6% of lab leads operationalized with >10:1 ROI

Manual Chart Review Needed

“If the HCC was not documented on a claim, it could still be supported by clinical documentation and submitted to CMS,” Lomsadze said. “This requires a manual chart review using a retrospective approach.

“Our colleagues compared those patients suspected of having high-cost conditions against the claims filed for those patients,” she explained. “Potential gaps in documentation can be identified in this step.

“This retrospective approach is time-consuming,” she noted. “It requires someone to find the eligible patient encounters, retrieve those patient charts, and then code those charts appropriately using ICD-10 codes. Once CareConnect followed these steps, they could submit these supplemental diagnoses to Medicare.

This Work Is Challenging

“Doing this work is challenging, especially getting the documentation for those patients who received care outside of our health system,” she added. “Anyone who has attempted to get patient charts from an organization that’s not part of the same health system knows how difficult this step can be.

“As you would expect, there was fallout at every step,” commented Lomsadze. “But even with many challenges, we still identified about 600 leads in 2016. From those leads, CareConnect chased down the records to get them coded and submitted properly. We ended up with 41 patients whose diagnoses would not have been found without lab leads.

“While that number was small compared to the plan’s total enrollment, it was both interesting and useful,” she said. “For example, our lab identified a number of diabetic patients who lacked the proper documentation in their claims. Within that same pool were two patients whose transplants had not been documented in claims during the entire year of those patients’ care.”

This data-matching exercise allowed the lab to identify specific patients who had sepsis, or who had cancer, and yet CareConnect might not have been paid the full HCC amount for those patients.

“That was a major wakeup call for us, because the proper coding and documentation for those patients helped the lab generate about $600,000 for the health system,” she recalled.

From the data-matching initiative, Northwell learned that it had documented correctly three patients with a diagnosis of cancer, but failed to capture the diagnosis of 14 other patients who likely had cancer. It had documented 30 patients with diabetes correctly, but missed documenting 219 other diabetic patients.

Retrospective coding for these patients, and those with sepsis, chronic hepatitis, rheumatoid arthritis, seizures, bipolar disorder, and patients who got a transplant, brought in a total of $607,291 in one year to CareConnect.

“This demonstrated the power of lab data to find gaps in documentation, and to find patients whose claims were incomplete for diagnoses that would qualify for higher risk-adjusted payments,” she noted. “Since then, our lab has made regular improvements to our data-matching algorithms.”

As she closed her presentation, Lomsadze emphasized the value of lab data, saying insights derived from lab test results collected over many years have more value than other sources of patient data.

Reflecting Provider Intentions

“From clinical laboratory data, our lab team is the first to know what’s happening with the patient,” commented Lomsadze.

“The laboratory knows where these patients are and how severe their conditions are,” she said. “In addition, our data are discrete, granular, and objective compared with other sources of data, such as claims. Our lab team can use clinical lab test data to improve the accuracy of risk-adjustment efforts, and it can help predict the costs for patients with these conditions.”

Contact Liya Lomsadze at 516-269-1236 or


Leave a Reply


You are reading premium content from The Dark Report, your primary resource for running an efficient and profitable laboratory.

Get Unlimited Access to The Dark Report absolutely FREE!

You have read 0 of 1 of your complimentary articles this month

Privacy Policy: We will never share your personal information.