Wrapping Things Up

My seven week fellowship concluded last Friday. As I have discussed previously, I spent the last week examining  racial differences in cardiac procedure utilization. My initial analysis involved running t-tests on black and white utilization rates for the years 1997, 2000, 2003, 2006, and 2008. I found that there were significant differences in the rates, although the the magnitude of the difference was relatively small and has been closing in recent years.

My next step was to  run a regression modeling utilization of a given cardiac procedure. Essentially, my goal was to see what factors influence  how likely a heart attack patient is to receive one of the three cardiac procedures of interest (cardiac catheterization, CABG, PTCA). I was most interested in how the race of the patient affects utilization, but I also included several other independent variables in order to control for various factors.  My race variables were dummy variables set up to identify whether a patient was black, Hispanic, white, or none of the above.

Additionally, I set up dummy  variables for the type of insurance  a patient has. I used insurance as a proxy for an individual’s income, figuring that those with more money will be more likely to be able to afford and undergo a cardiac procedure.  I also included the patient’s age in my model; elderly may be deemed physically unfit for a major heart procedure, and thus less likely to undergo the cardiac procedures I am analyzing. Finally, I included a variable for the sex of a patient,  in order to control for differences in utilization by gender.

Even after controlling for these other factors, the race of a patient played a significant role in determining the utilization of cardiac procedures. Both African Americans and Hispanics were less likely than whites to undergo cardiac procedures.  After seeing this result, I ran a model that included the facility  a patient was treated at, in order to look for differences within the same hospital. The race variables still remained quite significant , indicating one of two things. It is possible that race, for an unidentified reason, truly does have an influence on whether a patient receives a certain procedure. It is also possible that there is an omitted variable in my model which would explain the link between race and utilization. For instance, maybe if I had a better way to account for income differences amongst patients, the significance of the race variables would be diminished. Further research with more precise control variables is necessary to make any concrete conclusions.