The Health Care Research Unit (HCRU)
Section of General Internal Medicine
Boston University School of Medicine
Personnel:
Jennifer Fonda, MS
Amresh Hanchate, PhD
Madhuri Palnati, MS
Michael A. Posner, MS (PhD Biostatistics candidate)
Jeanne Speckman-Wishengrad, MSc
Projects:
Overview: Our major projects are large database studies in the area of health services research.
Special areas of expertise include: familiarity with Medicare, Veteran’s Administration
and Military (TRICARE) databases; use of sophisticated “risk-adjustment” tools
for understanding differences in costs, mortality and other outcomes of healthcare in populations;
exploring racial/ethnic differences in receipt of preventive services, aggressive surgeries and end-of-life care; and developing health-based payment models (to distinguish high costs due to inefficiency from those due to sicker populations). We also collaborate on projects that require primary data collection and address a range of clinical and health policy issues, including projects that support the research interests of medical faculty and fellows in the Division of General Internal Medicine and elsewhere at Boston University’s Schools of Medicine and Public Health.
Brief Descriptions of Selected Current Projects
1. Risk Assessment of Military Populations to Predict Health Care Cost and Utilization
When assessing health insurance plans for cost or quality effectiveness it is important to account for the initial health status profile of the enrolled population (risk adjustment). If the patients seen within one health care system cost more, or fare less well, than average, it is not clear whether this is due to system inefficiencies or simply to its patients being sicker than average. Several risk adjustment models are marketed. In this project, funded by the Department of Defense (DoD), we evaluate four such models as to how well they predict costs incurred by enrollees in the DoD’s TRICARE Prime health plan. One additional study will measure and seek to understand the causes of small area variations in health care utilization for selected diagnoses. Another will examine issues relevant to implementing prospective payment in TRICARE.
2. Racial Disparities in Healthcare Services for Medicare Beneficiaries
Many studies show that African-Americans and Hispanics receive fewer health care services and have lower health care costs. However in previous studies in Massachusetts and California, we found African-American and Hispanic Medicare beneficiaries to have significantly higher costs in the last year of life. Funded by the NIH, this project aims to quantify and explain these racial/ethnic disparities in a stratified random sample of nearly 1,000,000 Medicare beneficiaries, over-sampled for decedents and non-whites, and followed for their utilization in 2001 and 2002.
3. Using Claims Data to Examine Mortality Trends Following Hospitalization for Heart Attack in Medicare
This research was conducted for the Center for Medicare and Medicaid Services (CMS) to help them understand an increase in one-year mortality following acute myocardial infarction (AMI) that occurred between 1995 and 1999 among Medicare beneficiaries enrolled in the “traditional” (fee-for-service) program. We hypothesized that increased burden of illness during the later years might explain the increase in mortality. We examined Medicare utilization data (including MedPAR, Outpatient, and Carrier files) for over 1.5 million 1995-1999 AMI discharges. We used diagnostic data for the year preceding and the year following the AMI in each time period. We fit logistic regression models to predict 1995 mortality and applied them to the 1996-99 data, thus producing, for each case in each year, the predicted probability of death for a person with that risk profile and the same relationship between risk and mortality that obtained in 1995. Each model’s predictions are averaged in each year to produce an expected death rate based on that year’s comorbidity burden and the 1995 risk/mortality relationship. We used both a “core model” (age & sex) as well as three other models using different risk classification methods (Charlson, DCG and CCS). We compare each model’s ability to predict mortality and use each to calculate risk-adjusted mortality in 1996-99.
The DCG and CCS models led to more accurate predictions than the Charlson, which in turn dominated the core model (validated C-statistics: 0.81, 0.82, 0.74 and 0.66, respectively). Using the core model for risk adjustment reduced, but did not eliminate, the mortality increase. In contrast, adjustment using any of the morbidity models produced essentially flat graphs.
