The healthcare system extends across patients and providers, over the nation’s physical space, and through time. Analyzing the whole system means going beyond observing one person, one location, or one event to understanding how the social and physical context of healthcare is unfolding. That’s a tall order, but as national healthcare priorities shift to managing cost and quality through system structures, we must find ways to provide the evidence policy-makers need to make the right choices.
The healthcare claims data we already collect are a rich source of the information we need. New analytic approaches and tools for dealing with big data efficiently can give us insights into the provider system, the geographic system, and the time system to shape private and public healthcare decisions.
In this talk, we describe ways to think about healthcare systems measurement and provide three specific examples of analyses we have done to inform the design of healthcare delivery systems using social network analysis to examine care delivery structures, geographic disparity analysis to examine the impacts of physical provider distribution on healthcare fairness, and a “burstiness” analysis to identify anomalies over time in the payment system.
Using claims data to optimize provider network structure
Changing the healthcare delivery model will be a major part of healthcare payment reform. New proposed models limit panels of providers and focus on the primary care workforce and on close coordination between providers. In theory, this approach could improve patient relationships, reduce redundancy, and increase consistency and transparency. On the other hand, emphasis on closed networks could lead to consolidation-related price increases, and tight coordination of care could lead to cronyism or groupthink.
It is possible to use healthcare claims to identify and describe the best naturally occurring provider networks using formal social network analysis. To demonstrate the technique, we describe an analysis of 2009 Medicare claims from which we derived the naturally occurring provider networks. For each patient, we measured that patient’s network “connectedness” (the degree to which the patient’s providers work together) and the total size of the network patient base. The results show that provider connectedness, or coordination of care, was more strongly and consistently related to how many different claims patients had rather than how much their care cost once they were there.
We discuss strategies for implementing analyses like these and the implications of these findings and others for ideal provider network structure and size.
Using claims data to support healthcare systems deployment choices
We know that healthcare takes different forms in different places, but historically we have had little objective data about how the distributions of healthcare providers affected real healthcare outcomes. For instances, disparities in healthcare and in health are a major public policy priority. Despite increasing focus over the past two decades, however, we have made little progress toward reducing disparities in part because we do not have a very good idea about what is causing them. They appear to vary by location, and those geographic differences could be associated with lack of physical access to physicians and facilities. Unfortunately, it is quite difficult to tell whether access to local providers is driving disparities because the prevailing measurement approach creates an aggregate measure or disparity and lacks the variability we need to relate healthcare disadvantage to other variables.
It is possible to use claims data to measure and understand healthcare disadvantages at the individual, local level by leveraging the sheer volume of these data to create individual comparison samples for every record. To demonstrate the value of the approach, we describe an analysis of 2009 Medicare claims from which we derived a personal estimate of relative healthcare disadvantage for each African-American beneficiary in the sample compared to the 100 nearest White American beneficiaries. We also derived each person’s functional access to facilities and office-based providers by counting the local providers with claims in the area.
The results show that while access to a variety of healthcare facilities is important, access to office-based doctors is not a major driver of preventive healthcare disparities among older African-Americans. Instead, how African-Americans choose to utilize the healthcare that is already available affects their healthcare disadvantage.
We discuss strategies for implementing this kind of granular geographic estimation and its applications in estimating the real potential value of interventions designed to make healthcare more available to everyone.
Using claims data to drive change over time in healthcare payment structures
Changes in health and healthcare over time are already the focus of much attention. Researchers are used to measuring incidences of procedures and conditions over time. But claims data can also reveal hidden changes in healthcare and the health market. For instance, we know that innovation decreases the costs of individual procedures. For instance, a handheld electrocardiogram reduces the provider’s cost for an electrocardiogram substantially. Presently, however, payers have no way to detect these changes and modify their reimbursement practices.
It is possible to use claims data to flag procedures that may have become less costly or easier to perform using a modeling approach that searches for “bursts” of procedure billing activity. To demonstrate the approach, we used 2009 Medicare claims data to examine the billed procedures for evidence of sustained increases in the frequencies of specific procedures. The result is a list of procedures to be considered for payment modification.
We discuss strategies for implementing this kind of risk scoring and the uses for the scores in detecting natural market changes as well as fraud.
Evidence-based care goes beyond individual interventions and patients. The major changes we are making to healthcare delivery and payment systems should also be evidence based. With new and improved analysis, we can use data we already have—claims data—to measure the impacts of provider, geographic, and payment systems on overall healthcare cost and quality.
Frederica Conrey specializes in statistics and big data analytics for policymakers. Her background includes a PhD in social psychology and more than ten years of experience in measurement and advanced statistics.
For information on exhibition and sponsorship opportunities at the conference, contact Sharon Pierce at (203) 304-9476 or email@example.com
For information on trade opportunities with O'Reilly conferences contact mediapartners
For media-related inquiries, contact Maureen Jennings at firstname.lastname@example.org
View a complete list of Strata Rx contacts