DxCG Methodology
Transforming Data into KnowledgeGiven the different constituencies within a health care organization (actuarial, underwriting, clinical management) and across the various players in the health care market (government agencies, employers, payers and providers), it is important they share a common language. With this in mind, DxCG predictive models are built from two basic classification systems - Diagnostic Cost Groups (DCG) for diagnoses and RxGroups® for pharmacy data.
DxCG models work by classifying administrative data into coherent clinical groupings and applying hierarchies and interactions (as described in peer-reviewed literature) to create an aggregated, empirically valid measure of expected resource use. The measure, called a "relative risk score," is calculated at the individual patient level and quantifies the financial implications of the patient's "illness burden" or morbidity.
Aggregating individual scores by groups of interest (disease cohorts, employer groups, benefit levels, regions, physician practice panels, etc.) creates predictive model results specific to many financial and medical management applications.
DxCG researchers have devoted more than 20 years to DCG model research. DxCG develops models using panels of physician experts, as well as empirical analyses using nationally representative databases for Medicare, Medicaid and privately insured populations of several million covered lives. Unlike models offered by other organizations, DxCG models explicitly consider turnover in the population, which contributes greatly to illness burden.
DxCG offers both concurrent (historical) models, whereby utilization and expense data can be case-mix adjusted, as well as prospective models, for high-cost case identification, disease management reconciliation, budgeting and pay-for-performance. DxCG models are offered at a variety of reinsurance thresholds to accommodate "stop loss" as well as other models that do not trim outliers from the analyses. As a result, DxCG models are able to identify the top ½% cases that contribute substantially to total costs. In contrast, other model developers only offer models with stop loss, thus missing high-cost cases for case identification.
One of the hallmarks of the "DxCG Difference" is that our algorithms are open and not a "black box." The classification systems are freely available, evidence of our commitment to broadly distribute information about the models and support the educational efforts of user communities. Because the models are open, employers, government agencies, health plans and providers accept them, and clinicians, financial managers, actuaries and underwriters all use them.