An Analysis of Variation in Hospital Billing Using Medicare Data

Ankit Agrawal and Alok Choudhary
Northwestern University

New: We now have results for Chicago Hospitals!

Health care costs in the United States are known to be the highest per person as compared to any other country in the world. According to the World Health Organization (WHO), the total health care spending in the US was 17.9% of its GDP in 2011, the highest in the world. Several reasons are given for such extravagant costs, like the unit costs for medical goods, unnecessary use of high-cost services (sometimes as defensive medicine), administrative waste, inefficiently delivered services, excessive variation in service prices, fraud, and so on.

Hospitals determine what they will charge for items and services provided to patients and these charges are the amount the hospital bills for an item or service. We analyze the recently released data from the Centers of Medicare and Medicaid Services (CMS), which includes hospital-specific charges for over 3,000 hospitals across the US for 100 most common inpatient discharges, during Fiscal Year 2011. The goal of this work is to study the variation in the amount that hospitals charge for a given diagnosis-related group (DRG), which is a group of similar diagnoses and clinical procedures, hereafter also referred to as condition.

We also introduce the concept of normalized hospital billing that attempts to eliminate the component of cost variance that could possibly have arisen due to legitimate factors, such as geographical location, whether the hospital is a teaching hospital, whether it serves an older and/or sicker patient base. Of course, what normalization factor to use is an important question to consider. Here, we use the average Medicare payments to a hospital as a fraction of average Medicare payments across the entire US, to be the normalization factor. The premise for this choice is that the variation in Medicare payments across hospitals and geographical locations can be thought of as arising due to Medicare's taking into account the above mentioned legitimate reasons for cost variation.

The inter-state variation in hospital billing, medicare payments, and normalized hospital billing (taking into account the variation in medicare payments) for a given condition are presented visually as heatmaps (depicted at the bottom of this page), for the ease of identifying states where these values are significantly higher or lower than the average values over the entire US. The above analysis was done for the following 5 classes of DRGs: most common, most expensive, highest variation in average hospital billing, highest variation in average medicare payments, highest variation in normalized average hospital billing. The results reveal interesting insights, raising important questions as to how hospitals determine the value of their services.

The following insights may be deduced from our analysis:

  1. Hospital billing variations are much higher than Medicare payment variations.
  2. Normalization helps reduce the variation in hospital billing.
  3. Although normalized hostpital billing does not vary as much as hospital billing, the variations are still substantial.
  4. Normalization can change the relative position of a state in hospital billing with respect to the US average.
  5. Hospital billing variations are not correlated with medicare payment variation.

A detailed analysis behind the above insights may be found in our paper presented at KDD DMH 2013.

U.S. State Heat Maps Depicting Inter-State Variation in Health Care Costs For a Given Diagnosis-Related Group (DRG)

Hospital Billing

Medicare Payments

Normalized Hospital Billing