Hospital mortality risk adjustment for heart failure patients using present on admission diagnoses: improved classification and calibration

Abstract

Background and objective: Heart failure patient mortality rates are a focus of hospital quality assessment. This study examines whether comprehensive use of diagnoses identified as present on admission improves methods for comparing hospital mortality rates.

Research design: California hospital mortality rates are assessed using the Agency for Healthcare Research and Quality Inpatient Quality Indicator for heart failure, which uses selected diagnoses identified as present on admission. These rates are compared with rates calculated using comprehensive adjustments for diagnoses identified as present on admission. Models are compared to assess the accuracy of classification and to compare differences in hospitals identified with lower or higher than expected mortality.

Subjects: The study included 91,511 discharge records for patients with heart failure from 365 California hospitals for patients discharged in 2007.

Results: Every aspect of statistical model performance (discrimination, classification, calibration, and explanatory power) was improved by using more comprehensive adjustments for diagnoses identified as present on admission. The number of hospitals originally identified with higher than expected mortality was reduced by 50%.

Conclusions: More comprehensive use of diagnoses identified as present on admission improves the performance of mortality risk adjustment methods, and these improvements meaningfully change the results of hospital mortality rate comparisons.