Predict Readmissions

According to CMS, a readmission occurs when a patient is readmitted to the same or another acute care facility within 30 days of an initial hospital stay. Annually, adult patients experience 4.2 million hospital readmissions in the U.S., and among Medicare beneficiaries, one in six are readmitted within 30 days of discharge (1)(2). For older adults with functional impairments, the risk of readmission rises substantially and is 40% higher than the risk for a Medicare patient with no functional impairments (3).

Size of the Problem

• 14,5 k is the average cost of a readmission.
• 83% of general hospitals in the HRRP were penalized by CMS in 2019.
• 40% of patients are discharged with pending test results.
• 4,2 M adult hospital readmissions occur in the U.S. annually.

Readmissions are expensive. Hospital readmissions cost Medicare $26 billion annually with costs for readmissions of commercial payers and Medicaid beneficiaries amounting to $8.1 billion and $7.6 billion, respectively (4).They are also expensive for hospitals. CMS imposes a penalty on hospitals with excessive Medicare readmissions as part of the Hospital Readmissions Reduction Program (HRRP) and in 2019, penalized 2,583 hospitals $564 million for excessive 30-day hospital readmission rates (5).

The conditions that contribute most to readmissions differ for Medicare, commercial payers, and Medicaid, and the first step to managing them is identifying patients with these conditions who are the most likely to be readmitted (6). This also involves pinpointing any other reasons that patients might return to the hospital, which can include inadequate caregiver support, housing instability, food insecurity, or other social determinants of health. Using these insights to proactively work with patients, care teams can better plan transitions from hospital to home. When successful, such programs have been able to reduce readmissions by 34% (7).

1. Integration of Clinical and Social Data with AI: By integrating clinical data with information on social determinants of health, artificial intelligence can identify complex patterns that contribute to readmissions. This integration helps to better understand the full context of the patient, allowing for more holistic interventions that address not only medical aspects but also social factors such as housing insecurity or access to nutritious food.
2. Real-Time Risk Alert Automation: By implementing systems that use AI to continuously monitor patients during their hospital stay, organizations can receive real-time alerts about patients at risk of readmission. These systems can analyze changes in patient status, test results, and other indicators in real time to provide proactive alerts to medical teams.
3. Predictive Model for Hospital Readmissions: We have developed an artificial intelligence model that uses a balanced dataset to predict hospital readmissions. This model considers variables such as patient age, length of hospital stay, number of comorbidities, discharge status, and the presence of pending test results at discharge. The target variable 'Readmission within 30 days' classifies patients as 'Yes' for those who are readmitted and 'No' for those who are not. Using this model allows healthcare organizations to anticipate who may need additional care or personalized follow-up strategies to prevent readmission.

• ADT Records: Data from Admit, Discharge, and Transfer feeds and patient data notification services that synchronize patient demographic, diagnostic, and visit information across healthcare systems.
• Rx Claims: Data extracted from health insurance pharmacy claims with details about each medication and its type, fill dates, days supply, pharmacy location, and prescribing clinician.
• Social Needs Assessments: Self-reported data that identify an individual's specific needs and the acute social and economic challenges they are experiencing.