• Balancing Population Health with Individual Patient Needs

    Karen Wagner Jul 15, 2015

    Northwestern Memorial Hospital is testing a new model for preventing readmissions that balances treatment based on patient populations with treatment based on individual needs.

    Northwestern Memorial Hospital staff discuss patient readmissions

    Lisa Lui-Popelka, RN; Jessica Genz, LCSW; Katherine Allen, pharmacist, George Chiampas, MD; and Jude Kieltyka, MD, Northwestern Memorial Hospital, gather during a daily meeting to assess patients at risk for readmissions. (Photo: Courtesy of Northwestern Medicine, Chicago)

    Healthcare organizations are increasingly turning to data to improve clinical and financial outcomes. By analyzing large amounts of data, physicians and analysts can identify patient population trends and develop more cost-effective ways to treat these populations.

    However, focusing on the “big picture” comes with a risk: missing the small picture.

    For healthcare providers, the challenge is determining how to better understand the needs of patient populations while also being responsive to individual patients.

    “What you [the patient] want from us is that when you show up, you get what you need, not what a national predictive model might think you need,” says Cynthia Barnard, vice president of quality at Northwestern Memorial HealthCare (NMH), a four-hospital system based in Chicago. “However, we’re also obligated to keep our eye on those bigger-picture issues, both financially and as a responsible steward of resources.” Barnard shared these thoughts during a presentation at The Executives’ Club of Chicago, a business forum representing leaders across industries.

    Analyzing the Risk

    Although Barnard noted that NMH has been a long-time user of internal data to improve care for patient populations, she emphasized the importance of clinician input in addressing the needs of individual patients. Barnard referred to NMH’s performance improvement initiative that uses predictive analytics and clinician assessments to determine patient risk for avoidable 30-day readmissions.

    The predictive readmissions strategy is one of several strategies, such as improving the standard discharge process, NMH is employing to reduce its all-payer, all-cause readmission rate by 1.5 percent for its anchor hospital, Northwestern Memorial Hospital, an 894-bed academic medical center.

    For its fiscal year ending Aug. 31, 2014, the hospital had an all-payer, all-cause 30-day readmissions rate of 18 percent—or about 3,200 readmissions out of 18,000 inpatient stays, according to Pradeep Sama, director of analytics. (Northwestern Memorial Hospital’s self-reported readmission rate does not include readmissions to other healthcare facilities that are not part of the health system. Readmission rates reported by the Centers for Medicare & Medicaid Services track readmissions to external health systems.)

    Sama and his team of analysts designed a predictive model that generates a daily risk assessment score for each patient. The model includes about two dozen risk variables, which were chosen through a review of existing medical literature, discussions with clinicians, and exploratory data mining of the clinical data from the electronic health record, Sama says. Examples of variables that are part of the predictive model include:

    • Past healthcare utilization (e.g., the number of admissions, emergency department visits, and clinic visits in the preceding six to 12 months)
    • Whether patients have a primary care physician on record
    • Patients’ zip codes to specify their proximity to the hospital as well as income levels—another indicator of health risk

    The predictive model also incorporates concurrent data, which are tracked as care is administered during patients’ hospital stays—a sizable amount of data. In one day, a patient can have more than 10,000 different clinical documentations, which are gathered in the electronic health record and stored in a data warehouse, Sama says.

    Patients with a daily risk assessment score of 30 percent or higher receive a bedside medication reconciliation with a pharmacist and meet with a social worker to discuss socioeconomic factors, such as inadequate transportation for traveling to follow-up care appointments, that may also affect the risk for readmission, says Luke Hansen, MD, MHS, staff hospitalist and unit medical director in hospital medicine, who is a clinician-leader for the predictive readmissions model initiative.

    Providing Individual Assessments

    More importantly, these patients also are given individual multidisciplinary assessments during morning rounds and daily patient care meetings attended by a physician, pharmacist, unit nurse manager, and social worker, says Hansen.

    By focusing on patients’ specific risk areas, this team helps troubleshoot potential barriers to maintaining wellness after discharge. “A good example would be nutrition,” says Hansen. Patients who are vulnerable to a readmission because of poor nutrition may be given intensive one-on-one time with a nutritionist. One of the program’s goals is to determine how to fit additional risk-targeted nutrition counseling into that patient visit, says Hansen.

    Based on the team’s assessment, additional care may include patient psychiatric consults, coaching on medication adherence, or addition of in-home caregivers, Hansen says.

    “The ideal is to prevent readmissions by achieving optimal discharges for all patients, but no one-size-fits-all discharge plan is appropriate for all patients,” says Hansen. Providing these services to all patients would be a misallocation of resources, he says. The goal of the predictive model initiative is to understand how to best use limited resources in a fashion tailored to individual patient need. The analytics tool helps focus clinical resources more effectively, says Hansen.

    Incorporating More Data

    The long-term goal is to develop a more comprehensive predictive analytics model that would include such information as geospatial data, comprised of characteristics or demographics of neighborhoods that may affect the risk of readmission, says Hansen. The challenge is figuring out a way to incorporate such “non-hospital” data into an analytics solution. One promising approach is coordinating with agencies outside conventional medical environments to gather such data and develop more powerful analytics tools, says Hansen.

    Still, while large amounts of medical data can shine a light on the big picture of the patient population, the focus must remain on the individual patient. “The analytics can take you only so far. Readmission is a hard problem,” Hansen says. “Unfortunately, there is no silver bullet.”

    “I think by pairing a [predictive] model with a clinician’s assessment—maybe a medical clinician and a social work clinician—you would get many domains of interest and expertise. That’s how you optimize your ability to identify patients who are most likely to readmit,” says Hansen.

    Karen Wagner is a freelance healthcare writer and member of HFMA’s First Illinois chapter.

    Quoted in this article:
    Cynthia Barnard, MBA, MSJS, CPHQ, is vice president, quality, Northwestern Memorial HealthCare, Chicago.

    Luke Hansen, MD, MHS, FACP, is staff hospitalist and unit medical director, hospital medicine, Northwestern Memorial HealthCare, Chicago.

    Pradeep Sama is director, analytics, Northwestern Memorial HealthCare, Chicago.

    CommentsPlease login to add your comments

    Add Comment

    Text Only 2000 character limit