Benchmarking Inpatient Mortality Using Electronic Medical Record Data: A Retrospective, Multicenter Analytical...

Area of Science:

• Medical Informatics
• Health Services Research
• Clinical Epidemiology

Background:

• Benchmarking hospital mortality is crucial for quality assessment.
• Traditional prognostic models often require manual data collection, limiting their widespread use.
• Electronic medical record (EMR) data offers a rich, accessible resource for developing predictive models.

Purpose of the Study:

• To develop and validate a novel model for benchmarking mortality in hospitalized patients.
• To utilize readily available EMR data for risk stratification and outcome prediction.
• To create a model that is applicable beyond intensive care units (ICUs).

Main Methods:

• Developed a multivariable logistic regression model using EMR data from adult inpatients.
• Incorporated three indices: Physiology, Comorbidity, and Support, alongside primary diagnosis.
• Validated the model on independent cohorts, including a subset of ICU patients.

Main Results:

• The model accurately predicted hospital mortality in validation and revalidation cohorts.
• Achieved high discrimination (Area Under the Curve ~0.88-0.90) and good calibration.
• Demonstrated effectiveness in both general ward and ICU settings.

Conclusions:

• The developed EMR-based model effectively benchmarks patient mortality across diverse hospital locations.
• This model eliminates the need for manual data collection, unlike traditional methods like the Acute Physiology and Chronic Health Evaluation (APACHE).
• Further prospective testing in a larger, representative sample of hospitals is recommended to assess its utility for performance benchmarking.