Predicting and preventing hospital readmissions
The development and implementation of an organization-wide strategy to prevent 30-day hospital readmissions.
Context and Aims
Readmission reduction is a primary focus of health care systems worldwide in efforts to improve quality of care and efficiency across care settings. Inconsistencies in evidence on effectiveness make setting targets for readmission reduction difficult. Nonetheless, readmission reduction is one of the few quality measures that, if implemented properly, can serve as a catalyst for greater care integration between the community and inpatient settings.
As part of these efforts, evidence on the prevention of unplanned hospital readmissions increasingly points to the importance of early in-hospital intervention aimed at addressing patients’ needs as early as possible. This project aims to develop and implement an ongoing, organization-wide strategy to prevent 30-day hospital readmissions.
A three phase organization-wide integrated program has been established:
(I) Predictive modeling
(II) Hospital and primary care intervention
(III) Quality monitoring
(I) To guide the intervention, a prediction algorithm was developed based on pre-admission electronic health record and administrative data (the Preadmission Readmission Detection Model—PREADM). Variables on chronic conditions, prior health services use, body mass index, and geographical location were selected with decision trees and neural network algorithms.
(II) This risk model algorithm was introduced into all hospitals in Israel and to all of Clalit’s primary care clinics’ electronic medical records system to yield a risk score for each patient that is hospitalized and admitted to an internal medicine department.
The PREADM risk score is then used by a Transitional Care Nurse (a role developed by Clalit’s Community Division for this program) to target high-risk patients aged 65+ years in internal medicine departments of all general hospitals. They provide in-hospital coordination, discharge planning, and coordination with primary care clinic nurses for post-discharge follow-up and monitoring (e.g., need for a home visit or drug reconciliation). The PREADM score is used to prioritize outreach efforts to high-risk patients within 72 hours of discharge.
(III) Quality monitoring is performed using both objective and patient reported measures. Readmission rates and post-discharge primary care visits (within 3 days and 7 days) are measured comparatively across all hospitals and regions. Additionally, data from the post discharge nursing assessments is collected to elicit patients’ evaluation of the quality of their post discharge care.
Key Findings and/or Potential Impact
Using data-driven predictive modeling, patients are flagged during a hospital admission as high-risk for readmission. The nursing staff coordinate care prior to and immediately following discharge for high risk patients.
This streamlined prediction tool and intervention is being evaluated for its impact on preventing readmissions. Early results indicate that rates of post-discharge primary care clinic visits have markedly increased.