Project Team and Purpose
This project is a collaboration between Duke University, the Health Care Cost Institute (HCCI), and Blue Cross and Blue Shield of North Carolina, generously supported by the Commonwealth Fund and Arnold Ventures, with in-kind contributions from Blue Cross and Blue Shield of North Carolina. HCCI is an independent, non-profit research institute that maintains a comprehensive database of Employer-Sponsored Insurance and Medicare Advantage claims covering 50 million people annually across the U.S., and 100% of Medicare Fee-For-Service claims as a national Medicare Qualified Entity. The project brought together health economists, health services researchers, actuaries, and policy experts to design an approach to help policy leaders and the public better understand some of the basic drivers of and variations in health care costs in North Carolina.
All data used in this project are comprised of “de-identified” summary information, i.e., data that does not contain any information that may lead to researchers to be able to identify any particular person. One of the most challenging technical hurdles this project confronted was designing a feasible and objective way to measure health care costs consistently across very different patient populations with data stored in and governed by different entities. This required making dozens of technical decisions about how to report spending in comparable categories, adjusting for factors expected to influence costs (i.e., age and gender), and addressing other issues to measure costs accurately while also conforming to the data available.
With input from a national advisory board, researchers from multiple institutions collaborated for nearly a year to design, test, and implement a unified “analytic plan” for the project. The plan was implemented on different de-identified datasets. HCCI compiled the resulting summary data and worked with partners to design this set of interactive infographics and data tables as primary outputs.
While the project’s primary aims are to equip policy leaders and the public to better understand some of the basic drivers of healthcare spending, the project also demonstrates the challenge of measuring health care spending using highly fragmented claims data stored in different locations covering different patient populations. We hope this initial project helps inform policy conversations in North Carolina and shed light on ways to make it easier and more efficient to securely monitor health care access, cost, and quality in the future, in North Carolina and around the country.
- Christopher Bush, MPH
- Hilary Campbell, PharmD, JD
- Bradley G. Hammill, DrPH
- Zhen Li, PhD
- Jeannie Fuglesten Biniek, PhD
- William Johnson, PhD
- Daniel Kurowski, MPH
- Julie Reiff
Blue Cross and Blue Shield of North Carolina
- Robert Emerson, PhD
- Patrick Getzen, FSA, MAAA
- Cameron Lucey
- Jay Paulson, MS, MBA
Additionally, we would like to thank The DataFace for creating this website and the interactive data visuals. Finally, we gratefully acknowledge and thank the Commonwealth Fund (especially Lovisa Gustaffson) and Arnold Ventures (especially Hunter Kellett and Alexandra Spratt) for their generous support, collaboration, and vision, as well as Blue Cross and Blue Shield of North Carolina for significant in-kind contributions of time and data collaboration. SM Marks of the Blue Cross and Blue Shield Association, an association of independent Blue Cross and Blue Shield Plans. Blue Cross NC is an independent licensee of the Blue Cross and Blue Shield Association. All other marks are the property of their respective owners.
Below is a high-level overview of the methods used to create the aggregated data explored in this interactive site. Please refer to the Methodology document, published on HCCI's website, for a more in-depth description of the project's methods.
The study sample included enrollment data for individuals who met the following criteria:
- At least one member month of medical coverage in 2016 or 2017 as a resident of North Carolina, defined as membership ZIP code “27XXX” or “28XXX”
- Known and unchanging gender
- Known age
Members were assigned to a county for the duration of the study period based on their county of residence in the first month in which they appear in the data. Members were not required to have prescription drug coverage to be included in the study sample; prescription drug spending was calculated using the subset of the study sample with such coverage.
The variables and codes used to separate claims into inpatient, outpatient, professional, and prescription drug service categories are available in the Methodology document. Note that inpatient claims include those from a variety of non-hospital facilities, such as nursing facilities, skilled nursing facilities (SNFs), and long-term services and supports (LTSS) facilities.
All spending was reported as allowed amounts. Allowed amounts are defined as the amount paid for the service, which is the sum of the insurer payment and the deductible, copayment or cost-sharing amount from the insured. All spending is reported as 2017 USD; all 2016 spending was inflated to 2017 USD using U.S. Bureau of Labor and Statistics inflation data. See the Methodology document for specific information on how Medicare FFS spending was calculated.
Medicaid spending includes only claims payments and does not include supplemental payments to hospitals and other providers (e.g., cost settlements, Disproportionate Share Hospital Payments, Upper Payment Limits). For encounters associated with Medicaid Local Management Entities-Managed Care Organizations (LME-MCOs), allowed amounts are based on Medicaid pricing rules and provide the most appropriate comparison to fee for service claim payments, but the amounts may differ from what the LME-MCOs ultimately paid for those encounters.
Annual per-person spending was calculated by dividing the sum of spending for all claims within each high-level and detailed category by the sum of member months, either for the total population or the prescription drug coverage population and multiplying by 12. To calculate total annual per-person spending, the annual per-person spending for each high-level category (i.e., inpatient, outpatient, professional, prescription drug) was summed.
While this analysis includes a substantial portion of health care spending in North Carolina, it does not include data for the total population of the state. It does not include spending of the uninsured population, those insured through the individual market, those enrolled in Employer-Sponsored Insurance (ESI) with plans administered by issuers other than Blue Cross and Blue Shield of North Carolina or those housed at HCCI, or other types of supplemental insurance, nor does it include those enrolled in Tricare or care delivered through the Veterans Administration. Because these populations may be inherently different than the population included in this analysis, the findings may not be generalizable to the entire state of North Carolina.
The structure of each contributors’ data holdings differs. While close collaboration between the data contributors occurred throughout the analysis to ensure consistency, there are inherent differences in the claims that the contributors could not in all circumstances fully reconcile. For example, Outpatient Prospective Payment System Comprehensive Ambulatory Payment Classifications (OPPS C-APCs) in the Medicare Fee-For-Service claims are unique to Medicare, and other payers may pay for similar services in a different way.
The analysis does not include premium spending or account for manufacturer rebates for prescription drug spending. Additionally, the analysis does not consider benefit design when reporting spending.
Total spending includes medical spending and prescription drug spending. Total spending is a function of utilization and prices. For example, higher total spending in one county or population may be attributed to higher prices, higher utilization, or a combination of both. Thus, it is important to acknowledge that the variation in spending seen throughout this project implicates both of these factors. The specific causes for variation across populations and across the state was not explored in this project.