Evaluating Intervention Allocation Policies for Reducing Hospital Readmissions at Michigan Medicine

NCT07690657 | Not Yet Recruiting FDA Device

• 1 other identifier: HUM00282075
• Study Type: interventional
• Target: 5,000
• Locations: 1 country
• Sites: 1

Brief Summary

The researchers are investigating if using a risk-based prediction score or benefit-based prediction score to allocate transition of care (TOC) interventions is more effective in reducing the rate of unplanned hospital readmissions or death within 30 days of hospital discharge.

Conditions

Transition of Care

Keywords

Hospital readmission; death

Outcome Measures

Primary Outcomes (1)

• The number of 30-day all-cause unplanned readmission or death

  An unplanned readmission is defined as any not canceled inpatient, outpatient, or observation hospital admission with an admission type not recorded as "scheduled." Deaths are determined using the Michigan Death Index and matched to patient records using Social Security numbers.

  30 calendar days following the index discharge

Secondary Outcomes (5)

• The number of 90-day all-cause unplanned readmission or death

  90 calendar days following the index discharge

• Time to first 90-day all-cause unplanned readmission or death

  90 calendar days following the index discharge

• TOC telephone call bundle completion (care navigator call)

  5 calendar days after discharge.

• TOC telephone call bundle completion (clinical pharmacist call)

  14 calendar days after discharge

• TOC telephone call bundle completion (both calls)

  14 calendar days after discharge

Study Arms (2)

Control Arm/Standard of care

ACTIVE COMPARATOR

The Control Arm, in which patients with a LACE Index score of 9 or greater on the day of hospital discharge will receive TOC interventions. This is the currently implemented allocation policy at Michigan Medicine. In this setting, approximately 60% of patients receive the TOC intervention. Thus, we will use a threshold of the 40th percentile for assigning the TOC telephone call bundle. Note that to maintain the 60% intervention rate throughout the study, the score threshold (LACE 9 or greater) will be monitored and dynamically adjusted. The LACE Index has four components: Length of Stay (L), Acuity of the Admission (A), Comorbidities (C), and Emergency Department Visits (E), and estimates the risk of a patient having an unplanned hospital readmission after being discharged from their current encounter.

Other: The LACE Index TOC assignment

Experimental Arm

EXPERIMENTAL

The Experimental Arm, in which Causal-Hospital reAdmission Risk Prediction Model (C-HARP) scores will be used to allocate TOC interventions. C-HARP is a machine learning model that leverages routinely collected and stored patient data in the electronic health record (EHR) to estimate how much a patient will benefit from receiving MM's TOC telephone call bundle. The assignment threshold will be defined as the 40th percentile of C-HARP scores within the most recent 100 days of the target cohort, so that approximately 60% of patients are assigned the telephone call bundle. Based on current retrospective data, this threshold corresponds to a C-HARP score of 18.

Device: C-HARP TOC assignment

Interventions

C-HARP TOC assignment DEVICE

The Experimental Arm will allocate TOC interventions based on scores generated by the Causal-Hospital reAdmission Risk Prediction Model (C-HARP), C-HARP is a linear model that leverages routinely collected and stored patient data in the electronic health record (EHR) to estimate how much a patient will benefit from receiving Michigan Medicine's (MM's) TOC telephone call bundle.

Experimental Arm

The LACE Index TOC assignment OTHER

The LACE Index has four components: Length of Stay (L), Acuity of the Admission (A), Comorbidities (C), and Emergency Department Visits (E), and estimates the risk of a patient having an unplanned hospital readmission after being discharged from their current encounter.

Control Arm/Standard of care

Eligibility Criteria

• Age: 18 Years+
• Sex: all
• Healthy Volunteers: No
• Age Groups: Adult (18-64), Older Adult (65+)

You may qualify if:

• Patient meets the primary care established rule at Michigan Medicine (MM), AND
• Patient's primary care physician (PCP) Department is one of our General Medicine or Family Medicine or Medical-Pediatrics departments, AND
• Inpatient class is Inpatient or Observation or Obs Greater than 48 hours or Outpatient in a Bed or Extended Recovery or Hospital Care at Home Inpatient or Hospital Care at Home Observation, AND
• Discharged from General Medicine or Family Medicine or Medicine Observation services, AND
• Disposition of home or home with home care

You may not qualify if:

• Physician Organization of Michigan Accountable Care Organization (POM ACO)
• Sepsis patients (had a diagnosis of sepsis during admission, active or resolved on the hospital problem list)

Sponsors & Collaborators

• University of Michigan: lead

Study Sites (1)

The University of Michigan

Ann Arbor, Michigan, 48109, United States

Location

MeSH Terms

Conditions

Death

Condition Hierarchy (Ancestors)

Pathologic Processes; Pathological Conditions, Signs and Symptoms

Study Officials

• Jenna Wiens, PhD

  University of Michigan

  STUDY CHAIR

Central Study Contacts

Stephanie Shepard, PhD

CONTACT

734-647-1098; sdokeefe@umich.edu

Study Design

• Study Type: interventional
• Phase: not applicable
• Allocation: RANDOMIZED
• Masking: TRIPLE
• Who Masked: PARTICIPANT, CARE PROVIDER, OUTCOMES ASSESSOR
• Masking Details: Population Health Coordinators, Care Navigators, and Clinical Pharmacists will see a flag indicating whether or not a patient is assigned to receive MM's TOC interventions. To maintain appropriate masking, they will not have knowledge of whether or not this flag is determined by the Control Arm policy or Experimental Arm policy, in some cases they may access the LACE score but will not have access to the C-HARP scores. Patients will only be aware of TOC interventions if they are deemed to be at high risk for an unplanned hospital readmission and are assigned to receive them, which is the current standard of care and operational process at MM. Given the setting of this trial, there are no circumstances in which unblinding will occur.
• Purpose: PREVENTION
• Intervention Model: PARALLEL
• Sponsor Type: OTHER
• Responsible Party: PRINCIPAL INVESTIGATOR
• PI Title: Associate Professor of Electrical Engineering and Computer Science

Study Record Dates

• First Submitted: July 1, 2026
• First Posted: July 8, 2026
• Study Start: July 1, 2026
• Primary Completion (Estimated): August 1, 2027
• Study Completion (Estimated): September 1, 2027
• Last Updated: July 8, 2026

Record last verified: 2026-07

Data Sharing

• IPD Sharing: Will not share

Locations

US (1)