Predicting Intensive Care Unit Readmission with Machine Learning Using Electronic Health Record Data

This article by Rojas and colleagues was published in Annals of the American Thoracic Society during May 2018.
Rationale:  Patients transferred from the intensive care unit (ICU) to the wards who are later readmitted to the ICU have increased length of stay, healthcare expenditure, and mortality compared to those who are never readmitted. Improving risk-stratification for patients transferred to the wards could have important benefits for critically ill hospitalized patients.
Objective:  We aimed to use a machine-learning technique to derive and validate an ICU readmission prediction model with variables available in the electronic health record (EHR) in real-time and compare it to previously published algorithms.
Methods:  This observational cohort study was conducted at an academic hospital in the United States with approximately 600 inpatient beds. A total of 24,885 ICU transfers to the wards were included, with 14,962 transfers (60%) in the training cohort and 9,923 transfers (40%) in the internal validation cohort. Patient characteristics, nursing assessments, ICD-9 codes from prior admissions, medications, ICU interventions, diagnostic tests, vital signs, and laboratory results were extracted from the EHR and used as predictor variables in a gradient boosted machine model. Accuracy for predicting ICU readmission was compared to the Stability and Workload Index for Transfer (SWIFT) score and Modified Early Warning Score (MEWS) in the internal validation cohort and also externally using the Medical Information Mart for Intensive Care (MIMIC-III) database (n=42,303 ICU transfers).
Results:  Eleven percent (2,834) of discharges to the wards were later readmitted to the ICU. The machine-learning derived model had significantly better performance (AUC 0.76) than either the SWIFT score (AUC 0.65), or MEWS (AUC 0.58); p value < 0.0001 for all comparisons. At a specificity of 95%, the derived model had a sensitivity of 28% compared to 15% for SWIFT score and 7% for the MEWS. Accuracy improvements with the derived model over MEWS and SWIFT were similar in the MIMIC III cohort.
Conclusions:  A machine learning approach to predicting ICU readmission was significantly more accurate than previously published algorithms in both our internal validation and the MIMIC-III cohort. Implementation of this approach could target patients who may benefit from additional time in the ICU or more frequent monitoring after transfer to the hospital ward.
To access the full text of these articles via the journal’s homepage you require a personal subscription to the journal.  Some articles may be available freely without a password.  Library members can order individual articles via the Rotherham NHS Foundation Trust Library and Knowledge Service using the article requests online via this link.


Readmission to the Intensive Care Unit: Incidence, Risk Factors, Resource Use, and Outcomes. A Retrospective Cohort Study

This article in the Annals of the American Thoracic Society August 2017 issue was produced by Ponzoni et al.

Rationale:  Readmission to the intensive care unit (ICU) is associated with poor clinical outcomes, increased length of ICU and hospital stay, and higher costs. Nevertheless, knowledge of epidemiology of ICU readmissions, risk factors, and attributable outcomes is restricted to developed countries.

Objectives:  To determine the effect of ICU readmissions on in-hospital mortality, determine incidence of ICU readmissions, identify predictors of ICU readmissions and hospital mortality, and compare resource use and outcomes between readmitted and non-readmitted patients in a developing country.

Methods:  This retrospective single-centre cohort study was conducted in a 40-bed, open medical-surgical ICU of a private, tertiary care hospital in São Paulo, Brazil. The Local Ethics Committee at Hospital Israelita Albert Einstein approved the study protocol, and the need for informed consent was waived. All consecutive adult (≥18 yr) patients admitted to the ICU between June 1, 2013 and July 1, 2015 were enrolled in this study.

Results:  Comparisons were made between patients readmitted and not readmitted to the ICU. Logistic regression analyses were performed to identify predictors of ICU readmissions and hospital mortality. Out of 5,779 patients admitted to the ICU, 576 (10%) were readmitted to the ICU during the same hospitalization. Compared with non-readmitted patients, patients readmitted to the ICU were more often men (349 of 576 patients [60.6%] vs. 2,919 of 5,203 patients [56.1%]; P = 0.042), showed a higher (median [interquartile range]) severity of illness (Simplified Acute Physiology III score) at index ICU admission (50 [41-61] vs. 42 [32-54], respectively, for readmitted and non-readmitted patients; P < 0.001), and were more frequently admitted due to medical reasons (425 of 576 [73.8%] vs. 2,998 of 5,203 [57.6%], respectively, for readmitted and non-readmitted patients; P < 0.001). Simplified Acute Physiology III score (P < 0.001), ICU admission from the ward (odds ratio [OR], 1.907; 95% confidence interval [CI], 1.463-2.487; P < 0.001), vasopressors need during index ICU stay (OR, 1.391; 95% CI, 1.130-1.713; P = 0.002), and length of ICU stay (P = 0.001) were independent predictors of ICU readmission. After adjusting for severity of illness, ICU readmission (OR, 4.103; 95% CI, 3.226-5.518; P < 0.001), admission source, presence of cancer, use of vasopressors, mechanical ventilation or renal replacement therapy, length of ICU stay, and night time ICU discharge were associated with increased risk of in-hospital death.

Conclusions:  Readmissions to the ICU were frequent and strongly related to poor outcomes. The degree to which ICU readmissions are preventable as well as the main causes of preventable ICU readmissions need to be further determined.

The full paper can be accessed by subscribers to “Annals of the American Thoracic Society” via this link.  Some articles may be available freely without a password.  Library members can order individual articles via the Rotherham NHS Foundation Trust Library and Knowledge Service using the article requests online via this link.


Avoiding pediatric readmissions: Quite a challenge!

Journal of Critical Care: Published Online: September 05, 2015

Readmissions in the pediatric intensive care unit rates are an important tool used as a measure of quality. Although many hospitals are benchmarking readmission rates and working to lower them, the use of readmission rates as quality measures has generated controversy, particularly with respect to including readmission rates in public reporting and pay-for-performance programs.

Readmissions impact the hospital budget and might change the patient outcome and length of stay. Besides that, the transference for the pediatric ward exposes the patient to lower levels of care and monitoring, increasing the risks of clinical deterioration and mortality.

via Avoiding pediatric readmissions: Quite a challenge! – Journal of Critical Care.