Meeting the Data Challenge

High-quality observational data are the foundation of value-based health care. Sometimes, putting that data together can be a major challenge. A well-designed data set combines the right variables with an adequate number of observations (both in terms of the number of patients and across time), all with sufficient quality control to ensure reliable clinical reporting and consistency within and across data sets.

The recently established National Cancer Intelligence Network (NCIN) in the U.K. illustrates how clinicians and researchers are meeting the data challenge. In 2007, a consortium of cancer research groups and government officials came together to create a national cancer registry. Although the country had a number of regional cancer registries, they were highly fragmented and tended to focus primarily on patient mortality. The goal of the new effort was to create a truly nationwide registry that would track a broad variety of outcomes and enable clinicians to answer questions ranging from “How can we improve performance after surgery?” to “How do cancer incidence and outcomes vary across regions and clinical practices?”

Cancer is a complex disease with many risk factors, some of which are still poorly understood. There have been rapid advancements in treatments and protocols, but many of these treatments are costly and their long-term impact on patient health is not yet clear, leading to high variation in the types and quality of treatment. Identifying and understanding best practices are important to improving care.

These characteristics of the disease defined the key data requirements for the registry. Unknown future risk factors increased the number of variables that had to be collected. It was also essential to develop longitudinal data—both across time (in order to evaluate outcomes appropriately for individuals and institutions) and across care settings (given the multiple medical specialties and institutions involved). Finally, standards had to be established both for data capture from all providers and for risk adjustment across them.

The existing regional registries captured tumor incidence and mortality data (either directly or by linking to official mortality statistics), but they often lacked detailed information on treatments and risk adjustment factors. And while common standards were available for coding, the regional registries often differed slightly in what they actually captured.

The U.K.’s single-payer system and unique patient identifiers enabled the consortium to link patient and reimbursement information for over 90 percent of all cancer sufferers in the nation. The new system harmonizes and links data across the regional registries to create a patient record with consistent data. And linking to databases outside the regional registries (for example, the cancer data set from the U.K.’s Office for National Statistics, national clinical-audit data, and local-hospital episode statistics) increases the variables available for risk adjustment.

The NCIN is still working to harmonize and improve the data collected by individual registries and will eventually roll out a standard data-collection system to all regions. The new system gives researchers the ability to use aggregated patient information as a basis for secondary analysis—for example, comparing risk-adjusted mortality for individual hospitals to highlight those outside expected limits. NCIN is now able to track the risk-adjusted performance of individual hospitals in terms of 30-day mortality following colorectal cancer surgery, a key indicator of clinical effectiveness. The comprehensive NCIN database allows researchers to adjust for risks based on stage of tumor, age, and comorbidities and then compare performance across a large number of medical institutions to identify the best and worst performers.