# Predictors of and hospital-level variation in intensive care unit readmissions in Japan: a nationwide inpatient database study

## TL;DR

This study examines ICU readmissions in Japan, finding that early readmissions are rare and often occur long after discharge, with hospital-level differences suggesting quality-of-care indicators need refinement.

## Contribution

The study provides novel insights into ICU readmission patterns and predictors in Japan, highlighting the limited utility of early ICU readmission as a quality indicator.

## Key findings

- Early ICU readmission occurred in 0.9% of patients, with 32.8% of all readmissions occurring ≥14 days after initial discharge.
- Higher Charlson Comorbidity Index, emergency surgery, and elevated SOFA scores at ICU discharge predicted early readmission.
- Hospital-level variation in ICU readmission rates was substantial (ICC 14–15%) and persisted after adjusting for patient and hospital factors.

## Abstract

Intensive care unit (ICU) readmission is a widely recognized marker of patient outcomes and organizational performance. Early ICU readmission—commonly defined as a return to the ICU within 48 h or 2 full calendar days—has been posited as a quality indicator. However, its appropriateness as a quality indicator remains debatable, and evidence from Japan is limited.

We conducted a retrospective nationwide cohort study in Japan using data from the Diagnosis Procedure Combination database linked with the Hospital Bed Function Report from 2018 through 2023. Adults who were discharged alive after their initial ICU stay, with available Sequential Organ Failure Assessment (SOFA) scores at ICU admission and discharge were included. The primary outcome was early ICU readmission, defined as readmission to the ICU within ≤ 2 full calendar days after the initial ICU discharge. All ICU readmissions (early and late combined) during the same hospitalization were also assessed. Patient- and hospital-level predictors were evaluated using multilevel logistic regression. Between-hospital variability was quantified using intraclass correlation coefficients (ICC) and median odds ratios (MOR).

Of 552,545 eligible patients across 401 hospitals, 22,112 patients (4.0%) underwent ICU readmission; 4728 patients (0.9%) required early readmission. The temporal distribution lacked an early peak, with 32.8% of readmissions occurring ≥ 14 days and 12.2% occurring ≥ 30 days after the initial ICU discharge. Early readmission was associated with higher Charlson Comorbidity Index values, emergency surgery, and elevated SOFA scores at ICU discharge, particularly residual respiratory and circulatory dysfunction. SOFA scores at ICU admission and ICU stay ≥ 7 days were not predictive of early readmission. Intermediate care transfer conferred protection against early but not overall readmission. Hospital-level variation was substantial (ICC, 14–15%; MOR, approximately 2.0) and persisted after adjusting for patient- and hospital-level factors.

In Japan, early ICU readmissions are less frequent than in most Western countries and often occur long after discharge. Discharge severity scores, rather than disease severity at admission, served as the key predictors of readmission. Persistent unexplained hospital-level variations suggest that the ICU readmission rate alone is insufficient as a standalone indicator of the quality of care.

The online version contains supplementary material available at 10.1186/s40560-025-00838-3.