So... I've been thinking about how to use our existing data repositories for measuring cancer outcomes and for comparative effectiveness. QOPI is a great start (here)—it is oncologist-driven and adaptable; the measurement process is scalable and can be accomplished in private practices and in large cancer centers. Because data extraction is labor intensive—taking about an hour per chart (here)—we and others have been thinking about how to automate not only the QOPI process, but also other quality measurement and improvement efforts.

Two articles in the recent Annals of Internal Medicine describe the DARTNet—a federated network linking health data from 8 organizations representing more than 500 clinicians and more than 400,000 patients—and the key attributes of a National Distributed Health Data Network (here)—which supports both observational and interventional studies, and should enable local data holder control over access and uses of data—address approaches to aggregating clinical data for outcomes and comparative effectiveness research.

In an accompanying editorial, hopefully subtitled "The end of the beginning," the editorialists point out: "The muliple goals of population, health, pharmaceutical surveillance, comparative effectiveness research, and other major initiatives being advanced to address the needs of our health care and research enterprises have created a growing need for access to high-quality, patient-level health information. One approach to enabling such access is the creation of centralized repositories to which data can be transferred and then readily accessed to answer questions."

In oncology, we may be ahead of the game, as we have a well developed hospital-based, tumor registry system for identification of patients whose outcome of interest—survival—takes years or even decades to measure. Using even this well developed system poses some challenges.

My colleagues at the University of Michigan, Phil Hampton, Pat Shalis, and Jeff Cowall, have devoted some time to cataloging the tumor registry data sources, methods and collections used in the state of Michigan. Their impetus was BlueCrossBlueShield's desire to understand the various registries, and to hopefully realize economies of data collection, storarag and reporting, for at least some of BCBSM initiatives, including MiBOQI and MOQC.

They found that there are multiple users of the data, for educational, public health, and quality improvement purposes. However, the path to these users resembles the Tokyo subway map or the multiply redundant intracellular signalling pathways—it is very complex to generate data with the purpose of showing that the right treatment was administered to the right patient at the right time.

Their summation of the current registry environment is complex enough to make your head spin, and is shown below. It's obvious that the multiplicity of data sources, duplicative data abstractions (represented in the beige data collection box—how many times must the patient's gender, date of birth or diagnosis date be found in the medical record and be entered into a database?), and the multiple storage sites (represented in the purple data storage box) are redundant, introduce multiple opportunities for error, require updates and maintenance, and are clearly wasteful.

A simpler and more elegant solution is necessary.

If you're interested in reading further, contact Phil Hampton (phampton@med.umich.edu). I've encouraged Phil and his team to make their white paper more widely available.