Unplanned hospital readmissions soon after a person is discharged indicate the poor performance of the healthcare. Previous attempts of readmission prediction pose it as a binary classification problem and largely ignore the previous history. This study proposes a novel neural network architecture called Sequential Readmission Predictor with Multitasking (SRPM), to enhance the existing readmission prediction models. We retain the previous admission history of a patient by learning a latent representation for the patient, which could be used for every new admission by the same patient. Our proposed model uses a multitask neural network model that simultaneously models it as a binary classification problem and as a regression problem that predicts the exact days of readmission. By doing so, the error information from the regression task augments the classification task. The results show a promising improvement of up to 6.59% in AUCROC and 19% in F1 score over four benchmark methods.
Krishnamurthy, Sanjay and Pervin, Nargis, "Generalized Representation of Electronic Health Records for Unplanned Hospital Readmission" (2022). ECIS 2022 Research Papers. 150.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.