Variation in Identifying Sepsis and Organ Dysfunction Using Administrative Versus Electronic Clinical Data and Impact...

Area of Science:

• Healthcare analytics
• Clinical informatics
• Patient outcomes research

Background:

• Administrative claims data are widely used for sepsis surveillance and quality improvement.
• Coding variations for sepsis and organ dysfunction can distort analyses.
• Accurate sepsis rate and outcome comparisons between hospitals are challenging.

Purpose of the Study:

• To evaluate hospital variation in the sensitivity of claims data versus electronic health record (EHR) clinical data for sepsis and organ dysfunction.
• To assess the impact of data source differences on sepsis outcome comparisons.

Main Methods:

• Retrospective cohort study of 4.3 million adult hospital encounters across 193 U.S. hospitals (2013-2014).
• Sepsis defined using EHR-derived clinical indicators of infection and organ dysfunction.
• Compared sepsis prevalence and mortality rates using claims data versus clinical data.
• Reliability adjustment using hierarchical logistic regression.

Main Results:

• Claims data sensitivity for sepsis was low and variable (median 30%).
• Sensitivity varied for organ dysfunction: acute kidney injury (66%), thrombocytopenia (39%), hepatic injury (36%), shock (66%).
• Moderate correlation between claims and clinical data for sepsis prevalence (0.64) and mortality (0.61).
• 46% of hospitals shifted sepsis mortality quartiles when using clinical data.

Conclusions:

• Inaccurate and variable claims data limit reliable hospital comparisons for sepsis.
• Objective clinical data from EHRs are superior for meaningful hospital performance assessment.
• Improved data accuracy is crucial for sepsis surveillance and quality initiatives.