Selective Assignments of Disadvantaged Patients to Low Quality Hospitals and Disparities in the Quality of Medical Care

This paper analyzes the mechanisms behind disparities in the quality of medical care among patients in particular demographic groups such as non-white patients and patients under social welfare.  Patients in such demographic groups may be a burden to hospitals for a variety of reasons such as lower rates of reimbursement.  On the other hand, some patients may be systematically assigned to low quality institutions when high quality hospitals are able to use their market power to avoid such patients.

This study attempts to analyze the endogeneity between the quality of medical care and demographic compositions of patients at a hospital in order to understand why some patients receive lower quality care.  Our data set comes from administrative billing data (Uniform Billing), charity care data, and death certificate data between 2008 and 2010 in the State of New Jersey for colon, breast, and prostate cancer patients.  Using the data set, we study disparities between non-white and white patients as well as patients under different insurance types.  We use incidences of death and readmission as dependent variables and formulate a binary outcomes model at the patient-level.

For our estimation, the demographic composition in a hospital’s market serves as an instrumental variable, however, this information is not readily available. It is problematic to exogenously decide market boundaries, therefore, we build a model of patients’ hospital choice and simulate patients’ behavior to estimate the size of the market for each hospital.  This information provides the estimated share of patients for each hospital in our data set, and we use the estimated share as an instrumental variable for the model of binary outcomes. 

We adopt two models to control for the endogeneity. A General Method of Moments model provides an unbiased estimator for the marginal effects of share of patients at a hospital assuming an orthgonality condition between the instrumental variable and error terms whereas the control function approach eliminates the endogeneity by including the difference between the estimated share and observed share of patients at a hospital as an independent variable.  In our analysis, the marginal effects estimated by the two methods agree with each other in almost all cases.

Our results show the importance of controlling for the endogeneity.  When the endogeneity is not controlled for, it is consistently found that a higher share of non-white patients and patients under social welfare at a hospital is associated with more frequent incidences of death and readmission.  However, once the endogeneity is controlled for, it is found that non-white patients are consistently assigned to low quality institutions.  On the other hand, it is found that patients under social welfare could be a burden to hospitals.

These findings carry significant policy implications on two confronting policy tools; disproportionate share hospital payments and Performance-based payments.  Our results indicate that minority-serving hospitals should be provided with incentives to improve their quality, therefore performance-based payments would be more appropriate to improve the quality whereas hospitals serving a higher share of patients under social welfare might require more generous reimburcements.