JMIR Publications

ABSTRACT

Background:

Healthcare costs have been continuously increasing in the past few years despite various efforts and policies by the government. The Centers for Medicare and Medicaid Services projects that healthcare costs will continue to grow over the next few years. Rising readmission costs have been a significant contributor to the increasing healthcare costs. Multiple areas of healthcare, including readmissions, have benefited from the application of various machine learning algorithms in several ways.

Objective:

We identify suitable models for predicting readmission charges billed by hospitals. Our literature review revealed that this application of machine learning is still underexplored. We used various predictive methods, ranging from glass-box models (such as regularization techniques) to black-box models (such as deep learning-based models).

Methods:

Readmission with the same major diagnostic category (RSDC) and all-cause readmission category (RADC) are the two ways we defined readmissions. 576,701 and 1,091,580 individuals were identified from the National Readmission Dataset by the Agency for Healthcare Research and Quality for 2013 for each of the identified readmission categories, i.e., RSDC and RADC, respectively. Linear regression, lasso regression, elastic net, ridge regression, XGBoost, and deep neural networks based on multilayer perceptron (MLP) were the six machine learning algorithms we tested for both RSDC and RADC through 10-fold cross-validation.

Results:

Our preliminary analysis using a data-driven approach revealed that 21% of readmitted individuals (regardless of the number of days to readmission) alone contributed to 48% of hospital charges. We found that, within an RADC, the subsequent readmission charge billed per patient was higher than the previous charge in 49% of the cases, and this number is 53% for an RSDC. The top three Major Diagnostic Categories (MDCs) for such instances were the same for both RADC and RSDC. At the hospital level, the subsequent average readmission charge billed was higher than the previous charge for 80% of the MDCs in the case of RSDC, whereas it was only 52% of the MDCs for RADC. Deep neural networks using MLP performed the best for all performance metrics for both categories of readmissions i.e., RADC (MAPE-3.218%; RMSE-0.427; MAE-0.327; RAE-0.412; RRSE-0.423; NRMSE-0.041; MAD-0.032) and RSDC (MAPE-3.298%; RMSE-0.437; MAE-0.335; RAE-0.410; RRSE-0.423; NRMSE-0.043; MAD-0.033).

Conclusions:

Models built using MLP deep neural networks can accurately predict readmission charges billed by hospitals. The MDCs can be used by models to accurately predict hospital readmission charges.