Enhancing the performance of predictive models for Hospital mortality by adding nursing data

Area of Science:

• Healthcare Quality Assessment
• Clinical Informatics
• Patient Prognosis

Background:

• Hospital mortality is a key quality indicator, but accurate risk stratification is challenging.
• Standard electronic health records often lack comprehensive nursing data, such as nursing diagnoses.
• Nursing diagnoses are crucial for a complete understanding of patient condition.

Purpose of the Study:

• To determine the independent predictive value of nursing diagnoses for hospital mortality.
• To assess if incorporating nursing diagnoses enhances existing medical risk adjustment models.

Main Methods:

• Prospective observational study at an Italian university hospital.
• Data collected from nursing information systems and hospital discharge registers over six months.
• Developed logistic regression models incorporating age, sex, admission type, comorbidity (CCI), and nursing dependency index (NDI) alongside medical data (APR-DRGw).

Main Results:

• The inclusion of NDI alone increased explained variance by 20%.
• A comprehensive model with APR-DRGw, CCI, and NDI achieved high accuracy (c-statistic = 0.89).
• Nursing diagnoses demonstrated independent predictive power for hospital mortality.

Conclusions:

• Nursing diagnoses are significant independent predictors of hospital mortality.
• Integrating nursing data into predictive models improves risk adjustment accuracy.
• Standardized nursing data should be incorporated into electronic health records to enhance patient care assessment.