By Jennifer J. Salopek
PCCI is a nonprofit research and development organization founded six years ago out of Parkland Hospital in Arlington, Texas. Its signature product is Pieces, a software system that leverages predictive analytics to improve patient care. Vice President of Clinical Services Anand Shah, MD, MS, explained the concept in a session at the AAMC annual meeting in Chicago in November; and provided additional comments in a telephone interview.
“We wondered, Can we take information from the electronic medical record to drive changes in patient care?” he explains. “We want to demystify predictive analytics and how they can be applied to health care, and share what we have learned along the way.”
A great deal of data didn’t exist or wasn’t accessible before the advent and wide adoption of electronic medical records, Shah notes. He and his colleagues at Parkland began their quest by trying to reduce readmission rates for patients with congestive heart failure. Using an approach that became known as e-HPA, Electronic Health Predictive Analysis, electronic data can be used to predict future clinical events for purposes of:
- Stratifying patients according to risk level for a specific outcome
- Allocating scarce clinical resources in a maximally effective manner under real-time demands
- Scheduling follow-up that is tailored to the patient
- Preventing adverse or harmful events by recognizing the highest-risk patients and devoting resources earlier.
“It’s impossible for health care providers to focus on all data points for all patients simultaneously, but the computer has nothing else to do,” Shah notes. “It also can compare my patients to other doctors’ patients to give us the whole picture. e-HPA is intended to augment physician judgment and improve workflow.”
In its initial study, the team at Parkland demonstrated a 26 percent relative reduction in the odds of readmission among CHF patients using e-HPA. Their findings, which were published in the British Medical Journal of Quality and Safety, included concentrated care management efforts on one-quarter of the patients and absolute reduction of five readmissions per 100 index admissions.
The idea grew into a software platform, now known as Pieces, that can be deployed across multiple conditions, all with the goal of reducing readmissions.
“We take the idea that key activities can be plotted along a timeline, but we must identify patients who are at risk,” explains Shah. “If you wait for coding, it’s too late—they have already left the hospital.”
Pieces sits on top of the EMR and uses natural language processing to review notes, identifying particular diagnoses. (“We must identify high-risk patients, but not very high-risk,” says Shah. “We have to target the right subgroup that will be amenable to our interventions or pathways.”) It then creates worklists, using evidence-based parameters, to determine what orders they need. It captures data on follow-up instructions and monitors completions.
There has been little resistance to adoption of e-HPA, and Pieces, says Shah. “It is a hard-wired continuous feedback loop—real-time data capture that generates failure analysis. Although it is a powerful tool that must be used with deep thought and commitment on the part of the hospital, it is not dependent on leadership to work once it is deployed.”
At Parkland, a safety-net hospital, the team has been able to demonstrate a return on investment of $4.59 on every dollar expended; ROI at other hospitals will depend on their payer mix. Pieces is a software-as-a-service platform that can be deployed at adopting institutions within 90 days; the cost varies based on what hospital leadership is seeking to address.
Organizational culture can be key to success, says Shah. “Hospital leadership must demonstrate a clear commitment to improving quality and safety. This approach is not just the IT department doing whiz-bang experimentation. It is truly a tool to drive change.”