MAKIPS

MAKIPS predicts the risk of hospital-acquired acute kidney injury (AKI) in general inpatients using variables extractable from electronic clinical records to enable early risk stratification.

Key Features:

• Comprehensive Data Utilization: Model development used electronic clinical records from 47,466 patients collected over a two-year period in a tertiary care general hospital.
• Automated Risk Estimation: The score is based on 23 variables automatically extractable at admission, including recent abdominal, cardiovascular, or urological surgery and pre-existing congestive heart failure.
• Advanced Statistical Methods: Predictor selection and modeling used step-wise regression and Bayesian model averaging, with penalized logistic regression via the least absolute shrinkage and selection operator (lasso) for calibration.
• Internal Validation: Bootstrap resampling techniques were used for internal validation of the prediction score.
• Predictive Performance: The score demonstrated an area under the receiver operating characteristic curve (AUC) of 0.811 for prediction at admission.

Scientific Applications:

• Clinical risk stratification: Identifying inpatients at elevated risk of developing hospital-acquired AKI to inform targeted monitoring and preventive interventions.

Methodology:

Data were collected from electronic clinical records with applied inclusion criteria; model development employed step-wise regression and Bayesian model averaging, calibration used lasso penalized logistic regression, and internal validation was performed with bootstrap resampling.