SNF-at-Home: Expanded Penn Medicine At Home Services Allowing Patients to CHOOSE HOME After Hospitalization

Susan Keim, PhD, MSN

Principal Investigator: Susan Keim, PhD, MSN | School of Nursing

Matching patient preferences for hospital discharge to home, this novel Penn Medicine at Home (PMAH) “Skilled Nursing Facility (SNF)-At-Home” program provides eligible patients who qualify for a post-acute care facility stay with an option to recover at home with expanded services instead. Based on pilot findings, this bundled program includes early and intensive home health care clinician visits coupled with virtual case management, and supplemental services (e.g., meals, transportation) tailored to meet patients’ needs. Targeted populations for this 1-year study include approximately 200 patients with mild-to-moderate functional deficits post-stroke, or those with post-elective or emergent lower major joint surgery, discharged from downtown Penn Medicine (PM) hospitals to PMAH. Patients in the program will be matched to similar patients receiving usual care after discharge from other PM sites or from the same sites but the year prior to implementation of the intervention. The overarching objective of this innovative SNF-at-Home model is to optimize efficiency and value in health care delivery by demonstrating that eligible patients with diverse social needs can safely and effectively recover at home, experience greater satisfaction and less adverse outcomes—significant cost savings to payers compared to facility costs. The team will study the impact of the SNF-at-Home program—compared to usual care—on clinical outcomes including functional status and on healthcare utilization, e.g., hospital length of stay, readmission, return emergency department visit data, including financial analyses.

Evaluating the Impact of Audit and Feedback on Reduction in Excess Hospital Days Using a Novel Text and Web-based Platform with Unblinded Peer Comparison

Ryan Greysen, MD, MHS

Principal Investigator: Ryan Greysen, MD, MHS | Perelman School of Medicine

Penn Medicine hospitals and hospitals around the US are under unprecedented strain from length of stay (LOS) that is longer than expected, expressed operationally as “excess days.” Hospitalists are in a unique position to reduce excess days through workflow improvements, however feedback at the individual provider level that is accurate and actionable is often lacking and difficult to scale across large practices. The team will deploy an automated system (Agathos) for audit and feedback of hospitalist practice patterns at the individual level that leverages behavioral change insights, including peer comparison and provides individuals with the ability to explore their own data on their smartphone. Data on LOS will be regularly delivered to all providers via Agathos regularly along with process measures that can influence LOS including laboratory utilization, physical therapy (PT)/occupational therapy (OT) consultation, and timing of discharge orders. Division leadership will support faculty at two hospitals (the Hospital of the University of Pennsylania (HUP) and Penn Presbyterian Medical Center (PPMC)) to engage weekly with the platform during a 6-month intervention period. Experience with this peer comparison approach will inform future work to include scaling of the platform to other inpatient provider groups if the pilot is successful. In addition, feedback from faculty on the delivery of individualized data will be useful even if desired outcomes are not achieved with this specific platform.

Human-AI Collaboration to Improve Efficiency and Equity of Prescreening for Cancer Clinical Trials at Penn Medicine

Ravi Parikh, MD, MPP

Principal Investigator: Ravi Parikh, MD, MPP | Perelman School of Medicine

Clinical trials offer patients with cancer access to novel diagnostics and therapeutics and are key to the research mission of Penn Medicine. The identification of eligible patients for clinical trials–“prescreening”–is a key component of the clinical research enterprise, but it is inefficient. Currently, identifying eligible patients relies on manual chart review by clinical research coordinators (CRCs), which is time-consuming, prone to human error, and reinforces racial, ethnic, and geographic disparities in trial enrollment. As a result of suboptimal prescreening, under 10% of patients with cancer nationally and 20% of patients at Penn Medicine are offered participation in a diagnostic or therapeutic clinical trial. The integration of artificial intelligence (AI) into trial prescreening may streamline and engage greater numbers of patients in clinical trials. The team has designed and retrospectively validated a natural language processing (NLP)-based platform to extract common trial eligibility criteria from unstructured text, including radiology and progress notes. The team has also developed a human-AI collaborative workflow to maximize efficiency of CRC prescreening. They are proposing a prospective validation of this human-AI workflow in the context of a large phase III clinical trial that is enrolling at Penn Medicine over the next two years. This work is also supported by the Dean’s Innovation Fund, established by members of the Council for Discovery Science to support emerging discoveries with promise to improve human health.