Lown mortality ranking methods affirmed by new study

Hospital mortality rate (the percentage of patients who die after being admitted to the hospital) is the most widely used and arguably the most important metric for evaluating quality of care. There are many ways to measure hospital mortality rates, so examining the differences between these methods is crucial for creating accurate and fair metrics. 

Lown Institute researchers have published a paper in Health Services Research comparing different mortality ranking methodologies. They find that the methods used in the creation of the Lown Index are just as good, if not better, than the methods used by the Centers for Medicare and Medicaid Services (CMS).

What the study examined

1. The research focused on “hospital-wide” mortality measurement, meaning that all patients are included in a hospital’s result, which is what is reported in the Lown Institute Hospital Index of Social Responsibility. This is also what CMS has started to do recently.
2. Researchers compared two different modeling methods: a single model on all admissions (which the authors call an “all-conditions” approach) and combining results from multiple models, each on admissions for a specific condition (called the “summed-conditions” approach). 
3. Researchers also compared two different risk adjustment tools: The risk stratification index (RSI), which is used in the Lown Index, and Hierarchical Condition Categories (HCC), which is used by government agencies and other payers.

Key findings from the study

The all-condition approach and summed-condition approach provided similar results for hospital-wide mortality. In particular, hospitals with higher volume showed very little differences between the two methods. 

The risk adjustment tool used in the Lown Index (Risk Stratification Index) resulted in better model performance than the tool used by government agencies and other payers (Hierarchical Condition Categories).

Research & policy implications

For researchers and others assessing hospital quality, this study shows that there are more options for measuring hospital-wide mortality. Government agencies and other researchers can consider using an all-condition approach as well as a summed-condition approach, as the 30-day mortality ranking results were similar between these two methods. Deciding on the specific conditions and how to split hospital admissions into multiple cohorts is an investment without too much pay-off, given the similar results achieved using a single model.  

Government agencies and other researchers can consider using the Risk Stratification Index as a risk adjustment tool, as model performance was superior to Hierarchical Condition Categories. Smaller and rural hospitals were more likely to show a difference in ranking between the all-condition and summed-condition models. Government agencies and other researchers should look into novel methods and metrics of quality for these hospitals, as assessing quality for low-volume hospitals remains more challenging. 

Read the full study at Health Services Research!