Related Experiment Videos
Cardiac surgical mortality: comparison among different additive risk-scoring models in a multicenter sample
J M Pons1, J A Espinas, J M Borras
1Catalan Agency for Health Technology Assessment, Barcelona, Spain. jpons@olimpia.scs.es
Archives of Surgery (Chicago, Ill. : 1960)
|October 28, 1998
Summary
Risk-scoring models can help compare surgical outcomes, but they are not reliable for individual patient predictions. Customizing models to specific healthcare settings is crucial for accurate provider performance evaluation.
Area of Science:
- Cardiovascular Surgery
- Health Services Research
- Medical Informatics
Background:
- Predicting surgical mortality is essential for quality assessment in open heart surgery.
- Existing risk-scoring models may not perform optimally across different healthcare contexts.
Purpose of the Study:
- To evaluate the performance of various risk-scoring models in predicting surgical mortality after open heart surgery.
- To compare the predictive accuracy and calibration of different models across multiple cardiac centers.
Main Methods:
- Prospective observational study involving 1287 patients across seven tertiary cardiac centers in Catalonia, Spain.
- Assessed model discrimination using the c-statistic and calibration via chi-squared tests.
- Evaluated center performance using standardized mortality ratios and logistic regression.
Main Results:
- Models developed externally showed lower predictive accuracy (c-statistics) and poorer calibration.
- No significant differences in hospital performance were found after adjusting for patient risk factors.
- Models showed agreement in ranking centers but poor agreement for individual patient risk predictions.
Conclusions:
- Risk-scoring models can be valuable for comparing provider performance when customized to local contexts.
- Severity-adjusted models can support clinical judgment but are unsuitable for individual patient risk prediction.