Bayesian statistic model for nurse call data considering time-series, individual patient variabilities and massive...

Area of Science:

• Healthcare Analytics
• Statistical Modeling
• Nursing Informatics

Background:

• Nurse call data are crucial for evaluating nursing management and understanding patient needs.
• Traditional statistical methods struggle with the time-series nature, individual patient variability, and excess zeros in nurse call data.

Purpose of the Study:

• To develop and evaluate a novel Bayesian statistical model for analyzing complex nurse call data.
• To address the limitations of traditional frequentist statistics in handling nurse call data characteristics.

Main Methods:

• Proposed a Bayesian model incorporating transition (time-series change), random effects (patient variability), and a zero-inflated Poisson distribution.
• Utilized a dataset of 3324 patients in an orthopedic ward to evaluate the model.
• Compared the proposed model with variations excluding its core elements.

Main Results:

• The model including all three elements (transition, random effect, zero-inflated Poisson) demonstrated the best fit to the nurse call dataset.
• The proposed Bayesian model identified a longer duration of nurse call differences between patient groups compared to other models.
• The model effectively handles the unique properties of nurse call data, including high numbers of zero calls.

Conclusions:

• The proposed Bayesian statistical model offers a robust approach for analyzing complex nurse call data.
• This methodology can enhance the evaluation of nursing management and patient care by providing deeper insights.
• The model's ability to capture temporal dynamics and individual differences makes it a valuable tool in healthcare analytics.