Predicting In-Hospital Mortality in Japan
Author Information
Author(s): Miyata Hiroaki, Hashimoto Hideki, Horiguchi Hiromasa, Matsuda Shinya, Motomura Noboru, Takamoto Shinichi
Primary Institution: University of Tokyo, Japan
Hypothesis
Can we develop accurate in-hospital mortality prediction models for acute hospitalization using administrative data in Japan?
Conclusion
The developed risk models successfully predicted in-hospital mortality with high accuracy using easily accessible administrative data.
Supporting Evidence
- The models achieved c-index values of 0.869 and 0.841, indicating high predictive accuracy.
- In-hospital mortality rates were 2.68% and 2.76% for the preliminary and test datasets, respectively.
- The study included a large sample size of 224,207 patients from 82 hospitals.
Takeaway
This study created models to help hospitals predict how likely patients are to die while in the hospital, using information that is easy to get.
Methodology
The study used administrative records of 224,207 patients, split into development and test groups, and applied multivariate logistic regression analysis to predict in-hospital mortality.
Potential Biases
Potential bias from excluding certain diagnostic categories with low mortality rates.
Limitations
Exclusion of low mortality diagnostic categories may bias model performance.
Participant Demographics
53.2% male, 35.9% emergency admissions, 8.9% used an ambulance, 46.6% under 60 years old.
Statistical Information
P-Value
0.00
Confidence Interval
95% CI: 0.860–0.879
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website