Analysing repeated hospital readmissions using data mining techniques

Abstract

Few studies have examined how to identify future readmission of patients with a large number of repeat emergency department (ED) visits. We explore 30-day readmission risk prediction using Microsoft's AZURE machine learning software and compare five classification methods: Logistic Regression, Boosted Decision Trees (BDTs), Support Vector Machine (SVM), Bayes Point Machine (BPM), and Two-Class Neural Network (TCNN). We predict the last readmission visit of frequent ED patients extracted from the electronic health records of their 8455 penultimate visits. The methods show differential improvement, with the BDT indicating marginally better AUC (area under the ROC curve) than logistic regression and BPM, followed by the TCNN and SVM. A comparison of BDT and Logistic Regression results for correct and incorrect classification highlights the similarities and differences in the significant predictors identified by each method. Future research may incorporate time-varying covariates to identify other longitudinal factors that can lead to readmission risk reduction.

Comparison of Back-Propagation Neural Network, LACE Index and HOSPITAL Score in Predicting All-Cause Risk of 30-Day Readmission

“…[15][16][17][18] Nonetheless, existing dedicated efforts employing ML for predicting 30-day risk of readmission are mostly tailored to particular cohort or to a specific disease, such as congestive heart failure, 15,[19][20][21] chronic obstructive pulmonary disease (COPD), 22 patients discharged from intensive care unit, 23,24 emergency readmissions. 25 As has been noted in previous studies, predicting risk of readmissions for a general cohort is a completely different medical and data mining problem involving large, heterogeneous patient population sizes compared to disease-specific cohorts. 16,26 It has been pointed out that there is a lot of value in having readmission models that are not tied to a specific disease for patients who do not belong to any of the well-studied cohorts, or for incoming patients for which we do not know which cohort they belong to.…”

A Guided Neural Network Approach to Predict Early Readmission of Diabetic Patients

“…These approaches allow researchers to analyze large datasets and identify complex patterns and relationships between variables. By applying machine learning algorithms to electronic health records, researchers can identify risk factors for readmission or simply predict readmission likelihood for individual patients, and develop personalized intervention strategies to reduce readmission risk [13], [29].…”

Building Prediction Models for 30-Day Readmissions Among ICU Patients Using Both Structured and Unstructured Data in Electronic Health Records

“…Data selection was based on features found in similar studies that predict ICU readmission and mortality in the ICU (Ben-Assuli & Padman, 2017;Futoma et al, 2015). Patient discharge notes were also included in this study.…”

“…EHRs contain fine-grained information about patient care including demographics, laboratory test results, medications, procedures, etc. The potential of using EHR for predicting 30 days readmission to ICU has been shown previously (Ben-Assuli & Padman, 2017;Futoma et al, 2015;Lin et al, 2019). These studies used demographic information, lab results, and chart events as predictors for machine learning models to predict if an individual will be readmitted to the ICU within 30-days.…”

“…Most current studies only used demographic information, results from lab tests, chart events such as heart rate measurements and diagnoses on a patient's chart. They did not consider how discharge notes, which contain rich information about patients and their care, could correlate to a patient's likelihood of being readmitted to the ICU (Ben-Assuli & Padman, 2017;Futoma et al, 2015). For example, both Ben-Assuli and Padman (Ben-Assuli & Padman, 2017) and Futoma et al (Futoma et al, 2015) limited the scope of their dataset to only the structured EHR.…”