American Association for Clinical Chemistry
Better health through laboratory medicine
The Race to Reduce Readmissions

CLN Banner Logo

April 2013 Clinical Laboratory News: Volume 39, Number 4


The Race to Reduce Readmissions
Can Lab Tests Help Predict Who Will Return to the Hospital?

By Bill Malone

Medicare is dealing painful cuts to more than 2,000 hospitals across the country as punishment for too many readmissions within a 30-day window, flexing new authority under the Affordable Care Act. With millions of dollars in reimbursement at risk, many are now scrambling to find answers to the problem. For often perplexing reasons, too many patients who seem well enough to go home are discharged, only to experience crises days or weeks later that bring them back to the hospital. But with budgets already tightened, most hospitals can't afford to give a lot of extra help to every patient: they need to focus their efforts on those at risk.

This is where the laboratory is critical. Biomarkers like natriuretic peptides, galectin-3, and ST2 are showing unique promise in predicting readmissions and mortality for heart failure (HF) and acute myocardial infarction (MI)—two conditions related to large numbers of readmissions. At the same time, familiar workhorse assays such as complete blood counts and metabolic panels are powering a new breed of risk algorithms designed specifically to predict readmission risk.

Early success stories suggest that predicting risk with laboratory data can be powerful. At Parkland Health and Hospital System in Dallas, an early innovator, heart failure readmissions have dropped by more than 30% in 3 years. Parkland's software, which hinges on laboratory and other data already present in the electronic medical record (EMR), helps case managers identify which patients need extra attention, explained George Oliver, MD, PhD, vice president of clinical informatics at the Parkland Center for Clinical Innovation (PCCI), a non-profit affiliated with the health system. "Our work supports a public safety net hospital, where there always seems to be unlimited needs and limited resources to meet them," Oliver said. "Our predictive models enable us to better identify high-risk patients at the time of admission and to initiate more efficient workflows."

Putting Lab Results to Work

For hospitals hit by the CMS readmission penalties or simply trying to avoid them, finding resources to get back on course has become progressively more difficult as their reimbursement becomes less and less secure. Hospitals have had to swallow repeated cuts from healthcare reform and last-minute budget deals brokered between President Obama and House Republicans that show few signs of improving. Over the next 10 years, hospitals are projected to lose at least $155 billion in reimbursement to the Affordable Care Act and $15 billion for the so-called fiscal cliff deal signed into law in January. They also are projected to lose about $10 billion a year, every year, through 2021 unless Congress acts to reverse the broad federal sequestration that the fiscal cliff deal delayed only 2 months.

In addition, Congress had yet to deal with the next federal debt ceiling scheduled to hit May 18. This could represent an opportunity for even more healthcare cuts. Other reform efforts, such as Accountable Care Organizations, don't approach readmissions directly, but reward hospitals that can save money while improving quality—and readmissions are very expensive.

Frustrated hospital administrators are looking for ways to use existing measures to hold the line on readmissions without compromising other aspects of care. Parkland's model of harnessing EMR data—similar to that of other pioneers in the field—seeks to better allocate resources to tackle the readmissions problem. An example is special teams of nurses focused solely on helping patients transition from the hospital to their home or another care setting.

Healthcare systems for a long time have employed laboratory data in software and algorithms to predict risk. What's new is the special focus on readmissions and the quest to find the right mix of tests and other data. For example, researchers from Intermountain Healthcare in Salt Lake City published data in 2009 that showed risk scores combining complete blood count and basic metabolic profile components were highly predictive of death (Am J Med 2009;122:550–8). Familiar tests included hematocrit, red cell distribution width, mean corpuscular volume, platelet count, mean platelet volume, mean corpuscular hemoglobin, and total white blood cell count, as well as sodium, potassium, bicarbonate, creatinine, glucose, and calcium.

Percent of Patients Readmitted Within 30 Days Following Medical Discharge Among Hospital Referral Regions 2010

Click for map figure

From The Revolving Door: A Report on U.S. Hospital Readmissions
Copyright: Robert Wood Johnson Foundation
Source: Dartmouth Atlas Project

Led by Benjamin Horne, PhD, director of cardiovascular and genetic epidemiology at the Intermountain Medical Center Heart Institute, the researchers found that the c-statistics were strong, between 0.82 and 0.90 for 30-day, 1-year, and 5-year mortality. The unexpected power of an algorithm based on a handful of tests many clinicians take for granted drew a lot of attention, and Horne's team offers their free calculator, called Intermountain Risk Score, on their website.

To tackle the readmissions problem, Horne's group has developed a derivation of Intermountain Risk Score that focuses on predicting readmissions for MI, using many of the same laboratory, treatment, and demographic variables. At the American Heart Association meeting last November, the research team presented an abstract that showed this new algorithm strongly stratified 30-day readmissions in acute MI patients. The risk score showed differences of 3–5 fold relative risk and greater than 25% absolute risk (Circulation 2012;126:A16794).

The next step is figuring out how clinicians can best use the algorithm to manage care, Horne said. "The risk score is right there in the EMR when clinicians begin planning discharge, so they can take extra care for high-risk patients, such as reviewing medications," he explained. "The hospital also has a program for electronic discharge instructions for patients, but they're not sure how they want to integrate this information yet."

Horne noted that the readmission risk score did not prove to be as powerful as the score for mortality, but that this was likely due to the unique factors that play into readmission. "Clearly there are socio-economic and other issues going on with readmissions that make this quite different from predicting mortality," he said. "There are patients who are not readmitted because they die, so in some cases, readmission actually means a better outcome for the patient. And many patients return to the hospital for reasons not related to their hospitalization. Unfortunately, the penalties under the Affordable Care Act do not distinguish among readmissions due to a car accident versus something actually related to a patient's initial diagnosis of heart failure, for example."

In fact, the mortality-readmission connection is complicated and controversial. Some researchers and advocacy groups like the American Hospital Association have raised concerns that these rates might have an inverse relationship, such that hospitals with lower mortality rates are more likely to have higher readmission rates. However, a study using vast amounts of Medicare data, led by Harlan Krumholtz, MD, director of the Yale School of Medicine's Center for Outcomes Research and Evaluation (CORE), found that mortality and readmission rates in hospitals have a very weak association (JAMA 2013;309:587–93). The authors concluded the data should not deter efforts to avoid readmission, but allow them to be refined. "Our findings indicate that many institutions do well on mortality and readmission and that performance on one does not dictate performance on the other," the authors wrote.

CORE researchers published concurrent research that showed patients are often readmitted for problems not directly related to their initial diagnosis. For example, they found that the proportion of patients readmitted for the same condition was only 35.2% after an initial hospitalization for HF. The authors suggested that these often older, sicker patients return to the hospital due to frailty that actually gets worse from a hospital stay. "The broad range of acute conditions responsible for readmission may reflect a post-hospitalization syndrome—a generalized vulnerability to illness among recently discharged patients, many of whom have developed new impairments both during and after hospitalization," the authors wrote. "Inpatients frequently experience loss of strength and mobility and develop new disabilities and difficulties in performing activities of daily living."

CORE has also developed its own readmissions calculators that combine lab results with patient age, sex, and related diagnoses, available online and as a free smartphone app (See Box, below). They offer three calculators—for pneumonia, MI, and HF, and each uses a handful of lab tests like glucose and creatinine, as well as physiologic and demographic data.

Lab-Based Readmissions Calculators Online and on Your Smartphone

Yale Medical School's Center for Outcomes Research and Evaluation has developed three readmissions risk score calculators that are available online and on the Apple App Store.

Click for readmissions risk calculators

For your iPhone: search the iTunes App Store for "(CORE) Readmission Risk Calculators."

At Parkland, which has shown one of the largest drops in readmissions after using a risk prediction model, success seems to hinge on how well the risk model becomes integrated into nurses' and physicians' workflows, according to Oliver. PCCI's software doesn't only examine the EMR for predicting readmissions. Using natural language processing it helps identify patients with specific illnesses, predicts a variety of adverse clinical events, and offers tools to help coordinate care at the individual and population levels.

"Our goal has been to build real-time risk models that incorporate as many risk elements as possible that are available from hospital EMRs, as close to real-time as possible. The risk models incorporate real-time lab data as one of those elements to improve both the timeliness of the predictions and the predictive power of our e-Models," Oliver said. "We've tried several different means of delivery, but in its present state, a list of patients is sent to a team of case managers tasked with reduced readmissions at Parkland. They review that list and select the high-risk patients for enrollment in disease pathway-specific projects. This has proven to be an efficient system for the case managers."

At Geisinger Health System in Danville, Pa.—long known for being a leader in both information technology and care coordination—reducing readmissions is part of an overall plan that rewards physicians based on outcomes and best practices. To begin with, Geisinger prompts clinicians to order certain bundles of tests via best practice alerts in the EMR. As data on patient outcomes feeds back into the brain of the system, reports connect the dots between tests, treatments, and outcomes, explained Jay Jones, PhD, director of chemistry and regional laboratories.

"All of the best practice alerts on a given patient visit are ordered with a single click on the EMR encounter screen. This way we get consistent, standardized information within these clinical classes," Jones said. "Outcome data, including readmission and complication rates, are extracted as quality parameters for several disease classes, and these are reported down to the provider level."

While Geisinger doesn't have a readmissions risk score per se, Jones emphasized the tight connection with lab tests. "Lab results are certainly used to make clinical decisions for avoiding readmission. For example, if a septic patient's lactate or procalcitonin are elevated, the patient remains on antibiotic and is not discharged, as we know a readmission would be more likely," he said. "Our information system, called the Clinical Decision Intelligence System, helps create the evidence which gets translated into clinical decision support. Use of lab data will be increasingly relied upon for real time clinical decision support, including avoiding readmissions."

Targeting Related Sites of Care

The pressure to reduce readmissions is also being felt outside acute care settings as hospitals lean hard on nursing homes and assisted living facilities that care for large numbers of aging, vulnerable patients already at high risk.

Hospitals and Medicare often have competing perspectives about the factors behind 30-day readmissions after hospitalization, according to Gary Milburn, PhD, chief technical officer of Medlab, an independent lab that specializes in serving long-term care organizations.

"When patients from a long-term care facility are readmitted within 30 days, the conclusion, according to Medicare, is that the hospital released the patient too soon and the patient was not stable," Milburn said. "On the other hand, the hospital's take is that, if they're referring patients out to a long-term care facility that can't keep patients healthy, the hospital is the one taking the direct financial hit, so it better find another long-term care facility to refer its patients to."

To help long-term care facilities improve care and maintain good relationships with hospitals, Medlab is now developing lab-based algorithms to identify high risk, recently hospitalized patients. That way the long-term care facility can stabilize these patients before they reach the point of needing to be rehospitalized.

"The long-term care facilities are enthusiastically interested in this," Milburn said. "When I talk to our long-term care customers, we talk about how currently we flag critical results for individual lab tests. But what we're proposing is that instead of flagging individual results, we flag the patients who are at high risk, based on their frailty and a handful of important laboratory results. Right now there are so many patients in these long-term care facilities, and so many pieces of paper flying around, we want to offer an automated solution that can give a heads-up to the medical director and nursing staff so they can intervene and prevent these patients from being readmitted."

As Medlab works to validate its algorithm, Milburn emphasized that it's in the best interests of all providers—not just hospitals—to find ways of keeping patients healthy and avoiding readmissions. "Just the transport from a long-term care facility back to the hospital is about $800, and everything related to readmissions costs the whole healthcare system a lot," he said. "If we can work together to keep patients out of the hospital, patient care improves, and everyone saves money."

For Hospitals, Time to Pay Up

When a hospital admits a patient, generally the goal is to resolve an acute problem so the patient can safely return home and get back to normal life. However, for many who are older and sicker, things quickly go awry after discharge. They end up back in the hospital within a short period of time, and a vicious cycle can begin. According to the Centers for Medicare and Medicaid Services (CMS), nearly one in five Medicare patients returns to the hospital within 30 days, costing the program $26 billion annually.

Laying blame for the problem squarely on hospitals, the penalties levied by CMS aim to curb excess readmissions for common, expensive conditions. Since October 2012, when the government's 2013 fiscal year began, CMS has been cutting up to 1% of a hospital's total reimbursements if its readmissions for heart failure, myocardial infarction, and pneumonia rise above a target based on national averages for Medicare patients. Despite protests from groups such as the American Hospital Association, which objects to how CMS compares readmission rates, CMS plans to expand the program with deeper cuts and a longer list of diseases and conditions. According to an analysis of CMS records by the Henry J. Kaiser Family Foundation, CMS will penalize 71% of the hospitals whose readmission rates it evaluated.

So far, CMS data seem to show that the penalties are working as the agency intended. CMS Director Jonathan Blum told the Senate Finance Committee at a February 28 hearing that he already sees progress. "Though the payment adjustments took effect only recently, hospitals have been preparing for this program for some time and results suggest it is already having a positive impact," Blum told the committee. "After fluctuating between 18.5 percent and 19.5 percent for the past five years, the 30-day all-cause readmission rate dropped to 17.8 percent in the final quarter of 2012. This decrease is an early sign that our payment and delivery reforms are having an impact."

The program will slash reimbursement further as time goes on, taking up to 2% of a hospital's total reimbursement from Medicare starting October 2013 and then 3% the next year. In 2015, the list of diseases and conditions considered in the excess readmissions formula will expand as well. The Affordable Care Act gives the Secretary of Health and Human Services final say over which new conditions should be added, but suggests guidance should come from a 2007 Medicare Payment Advisory Commission report. This report identified seven diseases and conditions that accounted for nearly 30% of what CMS considers preventable readmissions: heart failure; chronic obstructive pulmonary disease; pneumonia; acute myocardial infarction; coronary artery bypass graft surgery; percutaneous transluminal coronary angioplasty; and other vascular procedures.

New Value for Emerging Biomarkers

Laboratorians have been nodding their heads as researchers discover that "simple" blood tests in complete blood counts and metabolic panels can have far more value than previously imagined. However, researchers are also forging ahead and experimenting with ways to make use of the predictive value in powerful biomarkers like natriuretic peptides, and even cutting-edge markers that have been cleared by the U.S. Food and Drug Administration, but are not yet available in many labs, such as galactin-3 and ST2.

Researchers have known for some time now that natriuretic peptides can predict outcomes, according to Christopher deFilippi, MD, a cardiologist and associate professor at the University of Maryland Medical Center in Baltimore and leading cardiac biomarker investigator. deFilippi noted that as early as 2004, researchers found that variations in N-terminal pro-brain natriuretic peptide (NT-proBNP) during hospitalization and pre-discharge predicted readmission and death within 6 months (Circulation 2004;110:2168–74). "Compelling evidence has accumulated for heart failure that shows that the discharge level of natriuretic peptides—or the change between the initial and the discharge level —can be very predictive of ultimately who will be readmitted over the next 30 to 180 days," deFilippi said. "I think that, in part, this has to do with which patients have truly been compensated, despite an improvement in symptoms. Most of the hospital's treatment for heart failure will be diuresis of some sort. However, there may be individuals who diurese, but whose cardiac condition is still poor, and natriuretic peptides could be a harbinger of other comorbidities that drive readmissions in these patients."

Using natriuretic peptides to predict readmission is part of a larger effort to understand the many potential uses of these markers. In fact, deFilippi is taking part in a large, randomized, multi-center study that will examine how NT-proBNP can be useful in managing medication for HF patients. Funded by NIH, the Guiding Evidence Based Therapy Using Biomarker Intensified Treatment (GUIDE-IT) trial will investigate whether NT-proBNP can help physicians adjust HF medications better than clinical judgment alone. The trial is still recruiting participants (ClinicalTrials.gov Identifier: NCT01685840).

Meanwhile, deFilippi and other researchers see promise in other biomarkers for readmission, such as galectin-3. Researchers have found galectin-3 to be involved in inflammation and fibrosis, both known to be intimately related to the progression of HF. At the same American Heart Association meeting in November where Horne presented data, deFilippi and his colleagues presented research showing the utility of a simple risk score for readmissions using galectin-3 (Circulation 2012;126:A17458). The score combined galectin-3 level with ejection fraction, glomerular filtration rate, diabetes diagnosis, and New York Heart Association classification. Using data from three prior studies of hospitalized HF patients, the researchers found that galectin-3 was independently associated with HF hospitalization or death. The risk score identified patients at approximately three times the risk for 60-day readmission or death, a significant improvement compared to the traditional clinical variables.

Galectin-3 makes sense as a marker for readmission, in part because it appears to be sensitive to the increase in overall frailty associated with hospitalization for HF, according to deFilippi. "Galectin-3 offers cardiac information, as would a natriuretic peptide, but it appears to also capture something about overall health status," he said. "We know that fibrosis and inflammation would translate into a lack of resiliency and increased frailty for a patient in the hospital."

Another new marker, ST2, is also showing promise. Since 2007, research has focused on how this marker can be used for prognosis in HF patients, since researchers demonstrated its strong association with HF-related 1-year mortality (J Am Coll Cardiol 2007;50:607–13). A soluble protein expressed by the heart in response to disease or injury, ST2 levels also change relatively quickly in response to a patient's condition, potentially helping physicians to be more nimble in adjusting treatment. Critical Diagnostics, which has an FDA-cleared ST2 assay, published a white paper in which the authors employed computer models that predicted ST2 could reduce HF-related readmissions by 17%.

For both galectin-3 and ST2, however, prospective studies are still needed, deFilippi said. "I'm a big biomarker enthusiast, but we are still dealing with mostly post-hoc analyses for these markers, and we need to evaluate them further in prospective studies," he explained. "I think they may ultimately provide an added role, perhaps targeting patients for specific therapies, but at this point the natriuretic peptides are the most mature as a tool for preventing readmissions."

Interactive Digital Edition