How to Harvest Big Data to Reduce Readmissions
By Jeanie Davis
“Big data” is a buzzword in healthcare these days. The term refers to the vast amount of electronic data healthcare providers have accumulated over the years, explains Dina Walker, RN, MSN, national director of case management for Encompass Health.
While the concept can seem pretty abstract, big data is more relevant than ever and potentially at every case manager’s fingertips. You just need the right tools to “harvest” it, Walker says.
Walker has been harvesting Encompass Health’s big data since she joined the company in 2016, and has seen firsthand how the organization reaps results by using data to build tools to help clinicians. In 2017, Encompass Health reduced readmissions to acute care during a rehab stay within the first six months of using their predictive analytic tool, ReACT.
To reach these results, Encompass Health partnered with its electronic medical record (EMR) vendor to create a customized dashboard, data analysis tools, and electronic clinical workflows. The data revealed where changes were needed (in-hospital processes) to reduce readmissions, Walker says.
Her message to case managers: “Look for opportunities to use your big data and you can perform this same type of analysis in your own hospital.”
Phase One: Assessing the Current State
Encompass Health is the largest owner and operator of inpatient rehabilitation hospitals, with 136 facilities in 35 states and Puerto Rico. Last year, they discharged nearly 187,000 patients. They also are the fourth-largest provider of Medicare-certified home health services, with 245 home health and 83 hospice locations.
“That translates into a lot of data that can be analyzed to make performance improvements,” Walker explains. “When you combine internal data from EMRs, for example, with external data like Medicare claims, you can really learn some valuable information about your current state and the possibilities for improvement. It is very exciting.”
Every Encompass Health rehab hospital and home health agency uses the same documentation platform. Combining data from each system afforded the ability to track patient data across the continuum of care through the inpatient rehab stay and the home health encounter. Then, they reviewed claims data from the Centers for Medicare & Medicaid Services (CMS).
“By using all those data, we could slice and dice and analyze the causes of the two types of readmissions that CMS measures us on: those that occur during a rehab stay (we call those ACTs — acute care transfers) and those that occur after discharge,” Walker explains.
“Readmissions that occur during the stay” is a CMS metric unique to rehab hospitals, she adds. For example, a Medicare patient in rehabilitation after a hip replacement might develop a complication like infection. If the infection is severe enough, the patient may have to return to the acute care hospital for treatment. If the surgeon keeps the patient to treat the infection, or performs another surgery, that counts as an “inpatient readmission” and incurs additional healthcare costs that are billed to Medicare.
“In those types of scenarios, inpatient rehabilitation may appear as if it is costly to Medicare, so we have to try to control those costs,” Walker says. “One main way to control costs is to focus on controlling preventable readmissions during an IRF [inpatient rehabilitation facility] stay,” she adds. “In the hip replacement scenario, a predictive tool may have picked up a change in the patient’s condition and potentially avoided that readmission.”
Medicare Spending per Beneficiary (MSPB) refers to the total cost of an episode of care. The episode begins on the day of admission to the IRF and ends 30 days after discharge. It is the cost of the IRF stay, plus all Medicare Part A and Part B healthcare services used within the 30-day post-discharge window.
For example, the average length of stay (ALOS) for an IRF stay is about 13 days. Combining that stay with roughly 17 days of home healthcare equals the 30-day MSPB measure window. Because of this, “we have to effectively manage care delivery during and after the IRF stay,” says Walker. “If we can prevent ACTs and avoid post-discharge readmissions, we will improve the MSPB. We also need to help avoid unnecessary emergency visits that can lead to unnecessary acute care hospitalizations. It’s not just about Medicare spending, though; avoiding these unnecessary types of healthcare transitions is also the right thing to do for the patient.”
The beauty of analytic tools is their ability to learn from historical data, says Walker. That is the function of ReACT.
“The ReACT tool is a model that risk-stratifies patients for risk of an ACT (the readmissions during the IRF stay),” she explains. “We leveraged a robust database of over 90,000 encounters from about 65 sites over a two-year period.”
The model showed statistical significance of about 30 clinical variables associated with higher ACT risk, says Walker. “We tested and validated the predictive algorithm and deployed the ReACT risk assessment and ReACT dashboard across all Encompass Health hospitals at end of 2017.”
In the dashboard view, clinicians can see patient trends over time (past 12 hours, past 24 hours, and past 48 hours) and can clearly see whether the patient’s risk level is improving or becoming more elevated.
The ReACT tool and clinician dashboard indicate when a patient’s risk level is increasing, says Walker. This risk level assessment is based on factors including functional measurements, Braden skin score, patient’s participation in therapy, certain medications, laboratory test results, demographics, pain level, and appetite.
ReACT indicated patients in the Very High Risk level are six times more likely to be transferred back to the acute care hospital than Low-Risk patients, she says.
ReACT was implemented in all Encompass Health hospitals by November 2017. “By March 2018, we began really seeing a nice decrease in the ACT rates,” says Walker.
How ReACT works: The algorithm starts monitoring documented factors upon admission, and as more data are entered in the patient’s EMR throughout the stay. Each patient’s ReACT risk score is displayed on a dashboard in their EMR, along with the list of factors contributing to that risk. Clinicians receive electronic notifications about risk score changes.
As the scores update over time, the trend becomes available. Nursing is electronically tasked to assess the patient if he or she escalates to Very High Risk level. They document the assessment on a custom-built screen in the EMR. Nurses inform the physicians of their assessment findings.
“Historical data can help identify issues in the patient care process,” Walker explains. “When those processes improve, readmissions decrease.”
Phase Two: Clinical Decision Support Tools
When analyzing readmission data, the team discovered opportunities to improve clinical care of chronic conditions like heart failure or chronic obstructive pulmonary disease (COPD).
“We also realized that in order to improve both ACTs and readmissions after discharge, we had to get the management of these conditions during the IRF stay right,” Walker says. “For that reason, we implemented clinical decision support tools for CHF and COPD.”
A business intelligence tool called Beacon Dashboards are the heart of the program. Beacon houses clinical care metrics, program adherence measures, and outcomes that can be monitored on the dashboards. Clinicians use these dashboards to trend results and detect areas of opportunity, as well as success stories.
Some key results:
- The number of opioid prescriptions written at discharge are decreasing.
- Medication reconciliation is performed more timely before discharge.
- The percentage of patients discharged home is increasing.
They also saw “marked improvement” in ACTs by March 2018, within six months of initiating the ReACT tool. “The 2018 average ACT rate across all hospitals was 10.41%. In 2019, it was 10.31%,” says Walker.
“These metrics are critical to Medicare IRF Compare reports, as they may impact consumer/patient perception of the quality of care,” she adds. “Data are transparent now more than ever, so organizations have to be aware of how they appear on websites like IRF Compare.”
Walker advises: “Assess the data, identify the issues, then work to fix them. The true driver for these tools is to improve patient care. Better care results in improved outcomes, and improved outcomes may lead to becoming a trusted, recommended provider.”
Phase Three: Tackling 30-Day Readmissions
Piggybacking on the success with the ReACT program, Encompass Health and its EMR vendor developed a readmission prevention algorithm that monitors both IRF and home health patient data as patients move through the continuum of care.
“The Readmission Prevention predictive model was developed using data collected from over 400,000 patients over a two-year period,” Walker explains.
The model was piloted in eight inpatient rehab hospitals and 10 home health branches. The model continuously monitors more than 40 clinical features or variables. From this, the model calculates a patient’s probability of a post-discharge readmission and reports a “risk score percentage,” she explains.
The Readmission Prevention tool identifies patients with a high probability of readmission to an acute hospital after discharge from the rehab hospital. Although the algorithm monitors clinical factors, patients also may have nonmedical risk factors.
“Assessing for and documenting those factors is critical. They can include social determinants of health, like food insecurity, limited literacy, or lack of caregiver support,” Walker explains. Because of this, a case management risk assessment was built in the EMR.
“Over time, as the case managers become proficient at assessing and documenting these nonclinical factors, that documentation will feed the predictive algorithm,” she says. “The algorithm will then help determine if those nonclinical factors are indeed contributing to readmission risk.”
One issue they discovered during the Readmission Prevention pilot phase was some readmitted patients were missing home health visits. “We discovered those patients were not allowing home health nurses to visit because they ‘didn’t feel like having company’ or they ‘didn’t feel well,’” says Walker. “That is the very time a patient probably needs to see the home health staff.”
They also found patients who did not attend their physician follow-up appointment timely after discharge were more likely to be readmitted.
Changes Based on Data
Now, a case manager explains the significance of the home health visits and reminds the patient to avoid canceling home health visits — even if he or she does not feel well.
“We also have taken a more active approach to scheduling post-discharge follow-up appointments within three to five days of discharge,” says Walker. “We provide information about the types of phone calls a high-risk patient may get after discharge. We ask them to be cognizant of keeping their phone charged, keep it with them, and to answer it — even if the call is from an unknown number.”
The time gap from discharge to the follow-up appointment with the primary provider is a risky period when readmissions are likely to occur, she adds. Encompass Health implemented another unique intervention to tackle that problem: The discharging IRF physician helps bridge the transfer gap from IRF physician to the patient’s own physician.
“The IRF physician will follow the patient with home health and our case manager for a few days post-discharge, until the patient sees their own physician — to bridge that gap,” says Walker.
The readmission predictive algorithm helps determine the level of risk, but staff also were trained on actions and interventions to mitigate that risk. “Those interventions are outlined in a playbook,” says Walker. “Since avoiding post-discharge readmissions is an interdisciplinary process, the playbook includes key actions for all members of the interdisciplinary team, including our home health team.”
She expects to report more information on the Readmission Prevention program and the results they achieve at national case management conferences in 2021.
“After a recent conference presentation, someone asked if they can buy the ReACT or Readmission Risk algorithms. Those algorithms are proprietary and customized for our rehab hospitals,” says Walker. “Hospitals can work with their EMR provider to develop their own tools, assess their current state, and pilot an improvement program. Start with analyzing as much data as you can.”
Bring Clinicians Aboard
“Integrating the ReACT and the Readmission Prevention predictive tools into the clinical workflow required clinician buy-in,” says Walker. “We educated all clinical staff and all physicians on the program and their role.” The team discussed patients’ ReACT and readmission risks in interdisciplinary huddles, focused on the highest-risk patients and those showing worsening trends.
At every step, clinicians were encouraged to embrace notifications and triggered electronic tasks. They could intervene at the earliest point possible when risk is escalating.
Walker advocates using “business intelligence” tools (like their Beacon dashboards) to analyze clinical data to develop targeted clinical decision support tools and interventions. They have been extremely valuable in readmission reviews, retrospective analyses, and process improvement.
“Predictive tools are a form of artificial intelligence and can be a tremendous addition to patient care workflows,” she says. “But they are just that: artificial. Absolutely nothing replaces good clinical knowledge, critical thinking, sound clinical judgment, and the ability to act on what you think is best for the patient.”
Her advice for case managers: “Talk to your IT department leaders and ask what data are currently available for analysis,” says Walker. “Tell them the type of analysis you would like to do and see if they can help pull some reports for you, or ask if there is a plan to work with the EMR vendor or create monitoring dashboards.”
She adds: “It is easy to get overwhelmed with all the predictive tools, the dashboards, and the numbers and feel some sense of ‘analysis paralysis.’ Just remember, behind each number is a patient, and behind each trend is a certain volume of patients who have been impacted by your current processes and the care you provide.”
Be open to what the numbers are telling you about your processes, Walker says. “Use them to justify and support those initiatives or improvements you have long known are needed.”
“Big data” is a buzzword in healthcare these days. The term refers to the vast amount of electronic data healthcare providers have accumulated over the years. While the concept can seem pretty abstract, big data is more relevant than ever and potentially at every case manager’s fingertips if provided with the right tools to harvest it.
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.