Assessing Medicare Hospital Scores | DSI | Final Project

Huilan Tang | Candace Haghighi | Nicholas Pantaleo | Carolina Downie

In July 2016, the Centers for Medicare and Medicaid’s “Hospital Compare” program (a consumer-oriented website that gives information about how well hospitals provide care to their patients) rolled out a new Overall Hospital Quality Star Rating, which rated hospitals between 1 and 5 stars based on a variety of measures. This roll out was controversial, with many hospitals criticizing the rankings. One major complaint was that major hospitals that consistently score highly in US News & World Reports rankings and other commercial rankings did not score highly on the Medicare rankings, and instead relatively obscure hospitals received the highest rating (a rating of 5 stars).

For our final Data Science project, we decided to analyze the Medicare data to try to assess some of these claims.

Data Sources

More information about the US News & World Reports methodology for the 2016-2017 rankings can be found here.


Please see our video for an narrated screencast of our analysis.

Questions we were interested in included:

The plots that address these questions can be found here.

The plots that adress these questions can be found here.

The preliminary model we created can be found here.

Conclusions and Future Directions

Our analysis suggests that some of the criticisms of the Medicare Hospital Star rating system may be valid. There was an unexpected geographical distribution of 5-star hospitals (for example, there were more 5 star hospitals in South Dakota than in New York). However, the comparison of the Medicare Hospital rankings and the US News & Reports rankings were more similar than we initially thought they might be, based on the controversy over the Medicare stars.

Future analyses could include: assessing the Medicare hospital ratings in the context of the socioeconomic demographics of the patients they serve, seeing how these ratings change over time, and more rigorous statistical analyses.