Panel Paper: Identifying Fraud in Home Health Agencies: Using HHA Compare Data to Predict Fraudulent Medicare Payments

Thursday, November 3, 2016 : 1:55 PM
Holmead West (Washington Hilton)

*Names in bold indicate Presenter

Justin Bullock, Texas A&M University and W. David Bradford, University of Georgia


Throughout the 2000’s the number of for-profit home health agencies (HHAs) has increased dramatically while the number of nonprofit HHAs has declined. This growth has been most notable in the states of California, Florida, Mississippi and Texas - states which also do not have state certificate-of-need laws. In the 2013 Report to the Congress: Medicare Payment Policy, MEDPAC reports that more than 400 agencies had unusually high rates of home health episodes per beneficiary. Similarly, the Office of Inspector General cited 257 agencies for providing an unusual number of therapy visits, which increase payments to the HHA. Eighty percent of the cited agencies are from California, Florida, Michigan, and Texas.  Concerns about fraud in the home health industry have only grown in recent years, with a particular focus paid to agencies in states with relatively lax regulatory environments.

As one example, there is evidence that for-profit HHAs are growing most rapidly, and incurring unusually high rates of expensive care in that states that do not have certificate-of-need laws. In this paper, we utilize the Center for Medicare and Medicaid Services’ Home Health Agencies Compare data from 2003 to 2015 to identify systematic signals for fraudulent behavior and assess how those signals vary with key state home health regulatory policies. We do this in two ways.  First, we identify the relationship among several health care outcomes and home health utilizations variables in states that do not have a high rate of suspected fraud (according to OIG reports) and then test whether those relationships hold for HHAs in states and cities that have been a focus of fraud investigations. This aspect of the study uses a combination of factor analysis, outlier analysis, and stochastic frontier regression analysis to identify characteristics of potentially fraudulent behavior.  Second, we repeat the analysis using a set of HHA that have been found guilty of, or entered into plea arrangements to resolve, fraud charges brought by the Department of Justice.  Characteristics of those agencies will be compared to characteristics of randomly selected HHAs from states with low detected fraud rates.  The goal of both analyses will be to assess the degree to which state policies are effective at decreasing fraudulent behaviors.  The results from this work could aid both analysts and policy-makers in using the HHA compare data to guide the search for fraudulent behavior, and ultimately lower the amount of Medicare improper payments.