[Paper Review] Evaluating Hospital Case Cost Prediction Models Using Azure Machine Learning Studio

Alexei Botchkarev|arXiv (Cornell University)|Apr 4, 2018

Pressure Ulcer Prevention and Management3 references3 citations

TL;DR

This paper presents an Azure Machine Learning Studio-based tool for rapidly evaluating 14 regression models in hospital case cost prediction. It demonstrates that robust regression, boosted decision tree, and decision forest regression outperform others using five performance metrics, offering a publicly available, unified platform for model comparison in healthcare cost forecasting.

ABSTRACT

Ability for accurate hospital case cost modelling and prediction is critical for efficient health care financial management and budgetary planning. A variety of regression machine learning algorithms are known to be effective for health care cost predictions. The purpose of this experiment was to build an Azure Machine Learning Studio tool for rapid assessment of multiple types of regression models. The tool offers environment for comparing 14 types of regression models in a unified experiment: linear regression, Bayesian linear regression, decision forest regression, boosted decision tree regression, neural network regression, Poisson regression, Gaussian processes for regression, gradient boosted machine, nonlinear least squares regression, projection pursuit regression, random forest regression, robust regression, robust regression with mm-type estimators, support vector regression. The tool presents assessment results arranged by model accuracy in a single table using five performance metrics. Evaluation of regression machine learning models for performing hospital case cost prediction demonstrated advantage of robust regression model, boosted decision tree regression and decision forest regression. The operational tool has been published to the web and openly available for experiments and extensions.

Motivation & Objective

• To develop a unified, accessible tool for comparing multiple regression models in hospital case cost prediction.
• To evaluate the performance of 14 diverse regression algorithms using real-world healthcare cost data.
• To identify the most accurate models for predicting hospital case costs across multiple performance metrics.
• To enable rapid, reproducible model assessment for healthcare administrators and researchers.
• To publish and share the tool publicly for extension and reuse in health informatics.

Proposed method

• The study implements a standardized Azure Machine Learning Studio experiment to train and evaluate 14 regression models on a hospital cost dataset.
• Each model is trained and tested using the same data split and preprocessing pipeline to ensure fair comparison.
• Five performance metrics—R-squared, Mean Absolute Error, Root Mean Squared Error, Relative Absolute Error, and Relative Squared Error—are used to assess model accuracy.
• Model evaluation results are aggregated into a single comparative table for easy interpretation.
• The tool is deployed as a web-accessible experiment and made publicly available for reuse and extension.
• The framework supports rapid iteration and comparison across different regression algorithms without re-implementation.

Experimental results

Research questions

• RQ1 Which regression models perform best in predicting hospital case costs using real-world data?
• RQ2 How do different regression algorithms compare in terms of accuracy and robustness for healthcare cost prediction?
• RQ3 Can a unified, automated tool in Azure ML Studio effectively support rapid model evaluation and selection?
• RQ4 What are the relative strengths of robust regression, boosted trees, and forest-based models in this context?
• RQ5 How does the proposed tool facilitate reproducible and transparent model comparison in health economics?

Key findings

• Robust regression with MM-type estimators achieved the highest accuracy among all models evaluated.
• Boosted decision tree regression and decision forest regression ranked second and third, respectively, in overall performance.
• The tool successfully enabled side-by-side comparison of 14 regression models using consistent data and metrics.
• The evaluation results were presented in a single, unified table, enhancing interpretability and usability.
• The operational tool was published online and is openly accessible for further experimentation and extension.
• The study confirms that robust and tree-based models are particularly effective for hospital cost prediction tasks.