Predictive risk modelling for early hospital readmission of patients with diabetes in India

Abstract

Hospital readmission is an important contributor to total medical expenditure and is an emerging indicator of quality of care. The goal of this study is to analyze key factors using machine learning methods and patients' medical records of a reputed Indian hospital which impact the all-purpose readmission of a patient with diabetes and compare different classification models that predict readmission and evaluate the best model. This study classified the patients into two different risk groups of readmission (Yes or No) within 30 days of discharge based on patients' characteristics using 2-year clinical and administrative data. It proposed an architecture of this prediction model and identified various risk factors using text mining techniques. Also, groups of consistently occurring factors that inference readmission rates were revealed by associative rule mining. It then evaluated the classification accuracy using five different data mining classifiers and conducted cost analysis. Out of total 9381 records, 1211 (12.9 %) encounters were found as readmissions. This study found that risk factors like hospital department where readmission happens, history of recent prior hospitalization and length of stay are strong predictors of readmission. Random forest was found to be the optimal classifier for this task using the evaluation metric area under precision-recall curve (0.296). From the cost analysis, it is observed that a cost of INR 15.92 million can be saved for 9381 instances of diabetic patient encounters. This work, the first such study done from Indian Healthcare perspective, built a model to predict the risk of readmission within 30 days of discharge for diabetes. This study concludes that the model could be incorporated in healthcare institutions to witness its effectiveness. Cost analysis shows huge savings which is significant for any healthcare system especially in developing countries like India.

Predictive Modeling of Hospital Readmission Rates Using Electronic Medical Record-Wide Machine Learning: A Case-Study Using Mount Sinai Heart Failure Cohort

“…One of the first predictive modeling of hospital readmissions using healthcare data from Quebec, Canada by Hosseinzadeh et.al 16 showed that Naïve Bayes models (0.65) performed better than Random Forest models (0.64). Using a diabetes cohort from a hospital in India, Duggal et.al 17 showed that Naïve Bayes (0.67) showed higher readmission associated savings compared to logistic regression (0.67), Random Forests (0.68), Adaboost (0.67) and Neural Networks (0.62). Futoma et.al 30 showed that Random Forests (0.68) and deep learning using neural networks (0.67) have similar accuracy rate with >1 million patients and > 3 million admission.…”

“…There is an immediate need for tools that may be used at the bedside or as part of discharge disposition planning to assess and minimize risk for readmission. Studies led by Hosseinzadeh et.al 16 leverage claims data to predict all-cause readmissions, and Duggal et.al 17 used EMR-derived clinical and administrative data to predict readmission in the setting of a diabetes cohort. To the best of our knowledge, our study is one of the first attempts to use phenome-wide data to identify novel factors driving readmissions related to congestive heart failure and develop EMR-wide prediction models with orthogonal validation to predict the readmission event.…”

Diabetes and data in many forms

“…For instance, it could be subjected to predictive modeling using neural network methods. In the current issue, Duggal et al used machine learning methods to predict readmission of patients who were discharged from a hospital [4]. From 9381 records, random forest was the optimal classifier using area under precision-recall curve to identify risk factors.…”

Predicting Hospital Readmission for Patients With Copd Using Natural Language Processin

“…The earliest applications of NLP is medical domain [10]. Recent researches uses Apache cTAKES to annotate unstructured EHR [11].…”