VISUALIZING DATA FOR HYPOTHESIS GENERATION USING LARGE-VOLUME HEALTH CARE CLAIMS DATA
Margret Bjarnadottir, PhD, University of Maryland, College Park, College Park, USA; Ebere Onukwugha, PhD, MS, University of Maryland School of Pharmacy, Baltimore, USA; Shujia Zhou, PhD, University of Maryland, Baltimore County, Baltimore, USA
PURPOSE: To identify tools for data exploration and discuss the process of hypothesis generation using published literature and case studies.
DESCRIPTION: Hypothesis-driven analyses of health care utilization data can provide important information regarding health outcomes and associated costs. These studies rely on informed, carefully-developed hypotheses generated from prior data exploration and analysis. Despite the important role that hypothesis generation plays in health services research, there is limited discussion and sharing of practical tools for hypothesis generation and data-driven insight. As growth in health information technology increases the availability of Big Data (i.e., large-volume, high velocity and varied data) for health services research, it will be important to be familiar with tools for data exploration and their role in hypothesis generation. The purpose of this workshop is to discuss such tools based on the published literature and via case studies. We focus on tools that provide intuitive visual representation to explore data, identify systematic patterns (including unexpected patterns), and unlock insights that are not possible through traditional statistical analysis. We describe tools available in the field and present two case studies. The first case study illustrates the use of a grouping algorithm to investigate cost accumulation and identify cost drivers following diagnosis of prostate cancer. The second case study illustrates the use of a visual analytics tool, EventFlow, to explore the impact of decisions regarding gap length and claims overlap on the measure of medication adherence. We will show how to interpret output from the grouping algorithm and the visual analytics tool. We will seek audience participation to draw insight from the output and formulate testable hypotheses. This workshop is appropriate for individuals who work with large-volume observational data (e.g., medical and pharmacy claims data, electronic medical records) and seek to identify practical tools for exploring and learning from their data.
Conference/Value in Health Info
ISPOR 2016, Washington DC, USA, May 2016
Methodological & Statistical Research, Real World Data & Information Systems