Identifying Increased Risk of Readmission and In-hospital Mortality Using Hospital Administrative Data: The AHRQ...

Area of Science:

• Health Services Research
• Medical Informatics
• Biostatistics

Background:

• Comorbidity measurement is crucial for adjusting healthcare outcome models.
• Existing Elixhauser Comorbidity Measures are widely used but can be complex to implement.
• Predicting in-hospital mortality and 30-day readmissions is vital for quality assessment.

Purpose of the Study:

• To develop and validate two novel indices for predicting in-hospital mortality and 30-day readmissions.
• Indices are based on the established Elixhauser Comorbidity Measures.
• To simplify the incorporation of comorbidity data in administrative datasets.

Main Methods:

• Utilized large-scale, all-payer hospital administrative data from the Healthcare Cost and Utilization Project.
• Employed backward stepwise logistic regression with bootstrapped replications to derive index weights.
• Validated model performance using c-statistics and assessed stability across subgroups.

Main Results:

• The developed index scores demonstrated comparable performance to using all 29 individual Elixhauser variables.
• Achieved c-statistics of 0.777 for mortality and 0.634 for readmissions.
• Indices showed stability across various demographic and clinical subgroups; adding covariates improved discrimination modestly.

Conclusions:

• The new indices provide effective, simplified methods for assessing mortality and readmission risk.
• These tools are particularly valuable for administrative data with limited clinical details.
• The indices are beneficial for studies with small sample sizes where detailed comorbidity adjustment is challenging.