Analytics Used to Predict Patients Likely to Be Readmitted Unplanned hospital readmissions are a serious matter for patients and a quality and cost issue for the healthcare system of every country. For example, in the United States, during 2011, nearly 19 percent of Medicare patients were readmitted to the hospital within 30 days of their initial discharge, running up an additional $26 billion in healthcare costs. Hospitals are seeking more effective ways to identify patients at high risk of readmission—especially now that Medicare has begun reducing payments to hospitals with high readmission rates. Identifying patients at high risk for readmission is important so that hospitals can take a range of preventative measures, including heightened patient education along with medication reconciliation on the day of discharge, increased home services to ensure patient effective at home convalescence, follow-up appointments scheduled for soon after discharge, and follow-up phone calls to ensure an additional level of protection. Several studies have attempted to identify the key factors that indicate a high risk for unplanned hospital readmission. One study was based on the analysis of the Belgian Hospital Discharge Dataset. This data set contains patient demographics, data about the hospital stay (date and type of admission and discharge, referral data, admitting department, and destination after discharge), and clinical data (primary and secondary diagnoses). Since 1990, Belgium has required the collection of this data for all inpatients in all acute hospitals. The data is managed by a commission that controls the content and format of patient registration, the data collecting procedures, and the completeness, validity, and reliability of the collected data. In addition, the quality of the data is audited by the Belgium Ministry of Public Health in two ways. First, a software program checks the data for missing, illogical, and outlier values. Second, by regular hospital visits, a random selection of patient records is reviewed to ensure that data were recoded correctly. Key factors for hospital readmission based on analysis of the Belgian Hospital Discharge Dataset included: (1) chronic cardiovascular disease, (2) patients with chronic pulmonary disease, (3) patients who experienced multiple emergency room visits over the past six months, (4) patients discharged on a Friday, and (5) patients who had a prolonged length of hospital stay. The study also found that patients with short hospital stays were not at high risk for readmission. The research highlighted the need for healthcare providers to work with caregivers and primary care physicians to coordinate a smoother transition from hospital to home, especially for patients discharged on Friday, to reduce unplanned readmissions. The mission of Penn Medicine Center for Evidence-based Practice (CEP) is to support healthcare quality and safety at the University of Pennsylvania Health System (UPHS) through the practice of evidence-based medicine. Established in 2006, Penn Medicine’s CEP is staffed by a hospital director, three research analysts, six physician and nurse liaisons, a health economist, a biostatistician, administrator, and librarians. A study conducted by a team at the CEP examined two years of UPHS discharge data and found that a single variable—prior admission to the hospital two or more times within a span of one year—was the best predictor of being readmitted in the future. This marker was added to UPHS’s EHR, and patient results were tracked for the next year. During that time, patients who triggered the readmission alert were subsequently readmitted 31 percent of the time. When an alert was not triggered, patients were readmitted only 11 percent of the time. A group of physicians conducted yet a third study using data from a 966-bed, teaching hospital during a five-month period in 2011. Their objective was to determine the association between a composite measure of patient condition at discharge, the Rothman Index (RI), and unplanned readmission within 30 days of discharge. Software employing the Rothman Index tracks the overall state of health of patients by continuously gathering 26 key pieces of data, including vital signs (temperature, blood pressure, heart, blood oxygen saturation, and respiratory rate), nursing assessments, cardiac rhythms, and lab test results from a patient’s EHR to calculate the Rothman Index, a number between 1 and 100. A patient’s Rothman Index is updated continuously throughout the day. A high score indicates a relatively healthy patient, whereas a low score indicates the patient warrants close monitoring or immediate assistance. A physician or nurse can quickly grasp the condition of the patient based on both current score and the trend in the score. The software also draws a graph of the patient’s Rothman Index over time that can be displayed in the patient’s room, on a central nursing station screen, or on a care provider’s mobile device. The software can even send mobile phone alerts to doctors and nurses when a patient’s condition warrants attention. The Rothman Index study included clinical data from the hospital’s EHR system as well as from a patient activity database for all adult discharges. There was a total of 12,844 such cases. The researchers excluded encounters that were readmissions within 30 days of a previous discharge (2,574), patients who were admitted for observation only (501), patients with length of stay less than 48 hours (3,243), and patients who died during the hospital stay (189)—yielding a sample of 6,337 eligible inpatient discharges. From this sample, 535 additional patients were eliminated due to missing clinical data, for a sample of 5,802 patients, or 92 percent of all eligible inpatient discharges. Sixteen percent of the sample patients had an unplanned readmission within 30 days of discharge. The risk of readmission for a patient in the highest risk category (Rothman Index lower than 70) was more than 1 in 5 while the risk of readmission for patients in the lowest risk category was about 1 in 10
Critical Thinking Questions
1. Three different analytic studies by three experienced and highly respected groups of researchers yielded three similar but somewhat different results. Do you believe that the results of these studies are consistent? Why or why not?
2. Do you think the findings of these studies can be applied broadly to all hospitals and medical centers across the United States and around the world? Why or why not?
3. A hospital specializing in the care of patients with various forms of heart disease is attempting to determine the cause of readmission of its patients. Should it rely of the results of general studies such as those described here or should it gather its own data, perform an analysis and draw its own conclusions? Support your recommendation.