Predictive Modeling for Comfortable Death Outcome Using Electronic Health Records

Area of Science:

• Health Informatics
• Nursing Informatics
• Data Science in Healthcare

Background:

• Electronic Health Record (EHR) systems generate vast amounts of complex data, posing significant big data challenges for analysis.
• Mining sparse and multi-dimensional EHR data is difficult, necessitating advanced modeling techniques.
• Death anxiety is a critical issue for dying patients, impacting their quality of life and end-of-life care.

Purpose of the Study:

• To develop predictive models using nursing EHR data to identify factors influencing death anxiety in patients.
• To create both coarse-grained (end-of-hospitalization) and fine-grained (end-of-shift) models for predicting patient outcomes.
• To assess the accuracy and utility of these models in improving end-of-life care.

Main Methods:

• Utilized a nursing EHR system to collect and analyze patient data.
• Applied various existing modeling techniques to construct predictive models.
• Developed coarse-grained models for predicting hospitalization outcomes and fine-grained models for shift-level outcome trajectories.

Main Results:

• Achieved significantly accurate predictions using the developed models.
• Demonstrated the effectiveness of models on relatively noise-free nursing EHR data.
• Identified key factors impacting death anxiety.

Conclusions:

• The predictive models offer valuable insights into patient outcomes and death anxiety.
• These models can guide the development of effective treatments and personalized end-of-life care strategies.
• Implementation of these models has the potential to lower healthcare costs and enhance the quality of end-of-life care.