Ever since the ACA authorized Medicare to withhold reimbursement for certain preventable readmissions starting in 2011, hospitals have been looking for ways to identify and manage high-risk patients, lest those providers be stuck with the tab should any patients return within 30 days of initial discharge. For many, addressing this new imperative starts with analytics and a focus on population health.
Carolinas HealthCare System, a 41-hospital network headquartered in Charlotte, NC, had a federal Beacon Communities grant from 2010 to 2013 to address population health. “It allowed us to pilot and test things with a new, fresh eye,” said Chief Innovation Officer Dr. Jean Wright.
By Neil Versel, Contributing Writer
New analytics technologies provide caregivers with the insight necessary to identify patients at risk of readmission prior to discharge, yielding a template for proactive intervention and personalized care.
![]()
Ever since the ACA authorized Medicare to withhold reimbursement for certain preventable readmissions starting in 2011, hospitals have been looking for ways to identify and manage high-risk patients, lest those providers be stuck with the tab should any patients return within 30 days of initial discharge. For many, addressing this new imperative starts with analytics and a focus on population health.
Carolinas HealthCare System, a 41-hospital network headquartered in Charlotte, NC, had a federal Beacon Communities grant from 2010 to 2013 to address population health. “It allowed us to pilot and test things with a new, fresh eye,” said Chief Innovation Officer Dr. Jean Wright.
In 2012, Carolinas HealthCare created a department called Dickson Advanced Analytics. Around the same time, Microsoft introduced Wright to an analytics vendor called Predixion Software, and she subsequently told the analytics department about the company. The leaders of Dickson Advanced Analytics were “intrigued” about the possibility of bringing predictive analytics to the bedside, Wright said. Today, advanced analytics is such an integral part of Carolinas HealthCare that it is featured on the menu of key information about the organization on the “About Us” page of the Carolinas website.
The Predixion software has helped case managers pinpoint the factors that can predict the risk of readmission. “It allowed us to move analytics from the back room to the point of care,” Wright said. “It’s what I call logistic regression for dummies.”
Identifying LOS & Readmission Potential
At the 2014 HIMSS conference in Orlando, FL, Predixion and Carolinas HealthCare announced that they would jointly develop a software tool, called Predixion Length of Stay (LOS) Insight, to assist hospitals in creating individualized care plans for each patient. Wright said that this was the first time Predixion had worked with live, point-of-care data during discharge discussions.
Carolinas HealthCare and Predixion built, ran, and tested their model. “We carry it a step further to the bedside,” Wright explained.
The organization rolled out Predixion LOS to seven hospitals the first year and is in the process of bringing the technology to six more facilities, Wright reported. “It’s almost plug and play,” she said. All along, leadership has tried to simplify the interface so as not to scare off clinicians (Carolinas started with about 200 variables and narrowed it down to about 40).
Designed by a multidisciplinary team, Predixion LOS is meant to identify, in real time, the risk of having a long hospital stay to help organizations mitigate risks and ultimately avoid penalties from exceeding length-of-stay guidelines set by CMS and other payers. Wright called this model the “strongest predictor of readmission” she had ever seen.
The Influence Of Geography On Patient Health
However, length of stay is not the only factor Carolinas considers. Dr. Michael Dulin, chief clinical officer for analytics and outcomes research at the Dickson Advanced Analytics department, is a supporter of “geospotting.” In other words, he believes location (i.e., ZIP code) has a significant influence on the health of a patient population.
“It is always important to marry the analytics with regional data,” Wright said. Different diseases are prevalent in different parts of the large organization’s service area. For example, sickle-cell anemia has long been known to affect African-American communities disproportionately, while chronic obstructive pulmonary disorder (COPD) is common in rural, tobacco-growing parts of North Carolina.
In addition to searching by diagnosis, providers at the more than 900 Carolinas care sites in North and South Carolina can query a database of more than 1.5 million patient records by geographic location and demographics to look for trends and apply predictive analytics. “Being able to drill down into the data helps us reduce variations in care and deliver overall better care, outreach, and coordination,” said Dulin.
Carolinas HealthCare now divides patients into quadrants of risk, identified by color-coded bars on departmental dashboards. The care team can make decisions based on risk as well as each patient’s proximity to discharge. “Through this process, we can clearly see the risk stratification of our patients,” said Wright.
Analytics has helped uncover what Wright calls “sleepers” (i.e., patients for whom risk is not immediately apparent from observation or a quick glance at the medical record). For example, a middle-aged, single man living alone might not always take his medications because he doesn’t have anyone around to remind him.
This type of insight also allows the health system to make best use of its case managers, reserving them for the most complex patients. “Case managers are a very precious resource,” Wright said. “You want them to be working where the payoff is going to occur.”
Predictive Analytics Reduces 30-Day Readmissions For Fragility Fractures
A Carolinas HealthCare System affiliate, the three-hospital Roper St. Francis Healthcare, of Charleston, SC, has been focusing on readmissions for specific patient cohorts — and not just because of the CMS payment policy. “At the start of 2014, we implemented a systemwide fragility fracture program,” said Stacey Seipel, clinical nurse specialist for quality at Roper St. Francis.
The health system turned its attention to 30-day readmissions, postoperative care, and population health management among women 55 and older and men 60 and older with hip fractures that resulted from falls. According to Seipel, clinical improvement outweighed financial considerations. “It was more in terms of improving our population health management for this condition,” she said.
After a year, 30-day readmissions for fragility fractures are down four percentage points when compared to the 2013 baseline. In order to get there, Roper St. Francis had to be able to identify patients in the high-risk category in real time. Roper St. Francis turned to Micromedex 360 Care Insights from Truven Health Analytics.
The technology provides both real-time clinical surveillance for flagging at-risk patients and clinical decision support at the point of care. With 360 Care Insights, Roper St. Francis is capable of conducting surveillance on free text in the EHR.
![]()
“I can identify every patient who has been here in the last 30 days, the minute they hit the door,” said Prudence Mack-Brown, quality data analyst at Roper St. Francis. “This insight has helped us solve the problem of identifying fragility fracture patients the moment they enter the hospital.” A positive identification triggers specific clinical pathways for medical staff to follow.
At the onset of the project, the first thing Mack-Brown did was conduct a one- to two-year retrospective study to identify any commonalities among fragility fracture patients who had to be readmitted. Following the early success of this strategy, the rapid- response team for stroke care asked Mack-Brown to build a similar surveillance tool to alert them the minute the emergency department ordered tissue plasminogen activator (tPA) for acute ischemic stroke. Real-time surveillance also led to the creation of a new medication protocol for COPD in mid-2014.
Prior to 2014, Roper St. Francis had applied analytics to address readmissions for heart failure, sepsis, and pneumonia. In each case, the 360 Care Insights system can send alerts to the smartphone of the appropriate nursing leader. Whenever a patient at risk of fragility fracture returns to a Roper St. Francis facility, Seipel gets an email.
“We’ve got several additional readmissions ideas in place and will potentially roll out many this year,” Mack-Brown said. Many of these initiatives involve care delivered outside clinical environments. “At-home factors, such as medication nonadherence, are contributing to hospital readmissions at an increased rate,” said Mack- Brown. “Health providers need to be able to identify these potential factors and develop an appropriate treatment plan to address these issues.”
“I can identify every patient who has been here in the last 30 days, the minute they hit the door,” said Prudence Mack-Brown, quality data analyst at Roper St. Francis. “This insight has helped us solve the problem of identifying fragility fracture patients the moment they enter the hospital.” A positive identification triggers specific clinical pathways for medical staff to follow.
