PM10 ISSUES IN THE ANALYSIS OF HEALTH CARE COST DATA

Sunday, October 23, 2016: 2:00 PM - 5:30 PM
Bayshore Ballroom Salon D, Lobby Level (Westin Bayshore Vancouver)
Course Type: Half Day
Course Level: Intermediate

Background: This short course will provide with a critical overview of a number of issues surrounding the analysis of health care cost data, for a number of research applications and type of studies. Typical studies that require this type of analysis are economic evaluations, cost effectiveness analysis and cost of illness studies. Topics will cover properties of cost data distributions, type of estimates of interest, challenges regarding data availability and proposed statistical methods for tackling those issues. Participants will learn how to critically assess the suitability of methods to be applied for specific research problems requiring the analysis of health care cost data, as well as how to appropriately interpret the results from the application of these methods.

Format Requirements: The format of the course will be of an informal interactive type lecture, where the participants will have the opportunity to interact with the presenter in order to ask and answer relevant questions. The discussion will be stimulated by a number of case study examples, where the participants will have the opportunity to comment on the application of the presented methods to real research problems and data. The participants are required to have basic knowledge of statistics and probability, including regression. Previous exposure to economic evaluations and analysis of health care cost data is desired but not required.

Description and Objectives:

Estimating health care costs is required for economic evaluations and cost-effectiveness analyses, being of increasing importance for informing decision and policy making. Abundance of prospective and observational cost data, such as administrative databases, contain very valuable information available to be analyzed by proper statistical methods. These methods though need to comply with distributional characteristics of the data (e.g. non-negative values, skewness, non-constant variance) and to overcome a number of data shortcomings,  including prematurely ended observation of cost accumulation, resulting into censored data. Despite the tremendous work on methods appropriate for the analysis of medical cost data over the last decades, their use into practice remains suboptimal.

The objectives of this course include:

  • To provide with comprehensive overview of issues surrounding the analysis of medical cost data, such as properties of cost data distributions, type of estimates of interest, challenges regarding data availability and proposed statistical methods for tackling those issues.
  • To discuss proper estimation of standard errors and confidence intervals that are used for the quantification of “uncertainty” around the mean cost estimates
  • To present methods such as ordinary least squares regression with transformation, generalized linear models, two-parts models, as well as semi-parametric approaches
  • To discuss considerations regarding the selection and evaluation of the method to be chosen, as well as regarding the validation of the model and the interpretation of the results
  • Emphasis will be given to issues due to “informative censoring” and to proposed methods for the analysis of censored cost data, such as the “direct method”, the “inverse probability of censoring weighting” method, and the phase-based method
  • Illustrative examples and case studies requiring analysis of real health care cost data will be given.
Course Director:

Nicholas Mitsakakis, MSc PhD
Toronto Health Economics and Technology Assessment (THETA) Collaborative