Tenet hospitals use predictive analytic tools to support patient transitions from hospital to community care services. This article will provide an overview of the tools used, how the need for the tools was identified, how they were implemented, and lessons learned.
Long gone are the days when hospital care coordinators assessed patient discharge needs using a combination of chart review, knowledge of payer benefits and community resources, plus their own years of care coordination experience, to magically come up with a post-hospital plan. Foundational tools today include structured assessments with core content conducted via patient/caregiver interviews and dialogue to address patient goals and preferences. But are those enough? Some level of mystery remains in the black box when moving from assessment and evaluation to plan for transition from the hospital to community setting.
"The probabilities are not meant to be prescriptive but rather provide a reference point for how often similar patients were discharged with a given disposition"
Tenet Health is a community built on care, which means that we want to do everything we can to partner with patients, caregivers and providers to identify and support successful transition from the hospital to the next care setting. We sought a more reliable way to guide care coordinators and patients as they planned for the most appropriate level of care when transitioning from the hospital. Our goal was to augment and guide the care coordinator assessment process with predictive analytics that utilize multiple data sources and encounters to inform the decision-making process with objective data.
Tenet data scientists created a Readmission Risk Report (RaR) which aggregates data from the electronic medical record as well as administrative and financial systems. The objective of the RaR Report is to take a broad set of patient information and distill it down to a score depicting potential for an extended length of stay and readmission risk. The report also provides important contextual information related to a patient’s Emergency Department and Hospital utilization within the prior 90 days.
The RaR Report includes a Discharge Planning Probability indicator for each patient that gives an estimate of how often patients with characteristics similar to the subject patient are discharged to a Skilled Nursing Facility (SNF) or Rehabilitation (RHB) facility (Rehab/SNF Probability), discharged with home health (Home Health Probability), or discharged home without additional services (Home Probability). These estimates are derived from a machine learning model that emerged after considering over 100 patient variables and developed using more than 300,000 previous inpatient discharges. The scores are then presented along with contextual information that helps clinicians provide targeted interventions for the patient. Broadly these variables can be classified into nine categories including: Diagnoses, Procedures, Labs, Vitals, Medications, Social Determinants, Nursing Assessment, Current Admission and Past Encounters.