Causal Hospital reAdmission Risk Prediction (C-HARP)

C-HARP builds on the HARP model by combining estimates of both a patient's risk for readmission and the likelihood that they will benefit from the Transitions of Care (TOC) interventions.

The Lab conducted a Randomization Pilot Study from November 18, 2024, through June 30, 2025. Patient data from individuals who were included in this Pilot Study will be used to develop the C-HARP model.

Using the data collected in the Randomization Pilot Study, the team has submitted a manuscript to JAMA Network Open that evaluates the effectiveness of the TOC interventions on reducing hospital readmissions. The manuscript is currently under review.

Findings

In this study, assignment to the intervention bundle had a low completion rate and a non-significant effect; however, patients who completed it experienced a significant reduction in readmission rates.

A new randomized control trial is being planned for CY2026, which will use causal machine learning to estimate the individualized effects of TCIs on unplanned 30-day readmissions in the Michigan Medicine population.

Principal Investigators

Jenna Wiens, PhD (Computer Science and Engineering)
Vikas Parekh, MD (Hospital Medicine and Office of the CMO
Michael Sjoding, MD (Pulmonary and Critical Care)