Data Visualization and Health Econometrics
By Andrew M. Jones, Department of Economics and Related Studies University of York, United Kingdom, andrew.jones@york.ac.uk
Abstract
This article reviews econometric methods for health outcomes and health care costs that are used for prediction and forecasting, risk adjustment, resource allocation, technology assessment, and policy evaluation. It focuses on the principles and practical application of data visualization and statistical graphics and how these can enhance applied econometric analysis. Particular attention is devoted to methods for skewed and heavy-tailed distributions. Practical examples show how these methods can be applied to data on individual healthcare costs and health outcomes. Topics include: an introduction to data visualization; data description and regression; generalized linear models; flexible parametric models; semiparametric models; and an application to biomarkers.
Data Visualization and Health Econometrics
Econometric models for health outcomes and health care costs are used for prediction and forecasting in health care planning, risk adjustment by insurers and public providers of health care, geographic resource allocation, health technology assessment and health policy impact evaluations. Methods for risk adjustment focus on predicting the treatment costs for a particular type of patient, often with very large survey or administrative datasets.
Data Visualization and Health Econometrics focuses on the principles and practice of data visualization and statistical graphics and how these can enhance empirical analysis of health care costs and outcomes, especially for skewed and heavy-tailed distributions. The scope of this review is limited to non-normal but continuous outcomes such as health care costs and biomarkers. The methods and applications used here are limited to cross sectional data.
The review of methods that have been developed for health care cost regressions is complemented by an empirical case study that focuses on objectively measured health outcomes, whose distributions share many of the features of cost data. The case study applies the econometric and graphical methods to blood-based biomarkers as the dependent variables.
