Table 1_Development of a prediction model for acute kidney injury in critically ill patien

A novel clinical risk prediction model for acute kidney injury (AKI) was developed using white blood cell-related indicators. The model was trained on 981 critically ill patients with advanced colorectal cancer from the eICU database and validated on data from Wuhan Union Hospital ICU. It achieved an area under the curve (AUC) of 0.746 for predicting AKI and 0.788 for predicting in-hospital mortality in the training set.

Use Cases

• Validating a novel AKI prediction model based on white blood cell-related indicators like WNR, WER, MLR, and SII.
• Comparing the performance of this logistic regression-based model against other scoring systems for AKI prediction.
• Assessing clinical net benefit and predictive ability for in-hospital mortality using decision curve analysis (DCA).
• Conducting subgroup analysis to evaluate model consistency across different critically ill cohorts with advanced colorectal cancer.

Strengths

• Model was developed and externally validated on a combined cohort of 981 critically ill patients with advanced colorectal cancer.
• The prediction model achieved an AUC of 0.746 (95% CI, 0.704-0.788) for AKI and 0.788 (95% CI, 0.736-0.840) for in-hospital mortality in the training set.
• The study provides a specific, implementable formula: 1.271×WNR+0.709×Lg(WER)+0.085×MLR+1.225×Lg(SII).

Limitations

• The dataset is very small at 13.0 KB, suggesting limited underlying data or a focus on the model description rather than raw patient records.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale machine learning tasks.

Provenance

Source

Training cohort from the eICU Collaborative Research Database; validation cohort from the ICU of Wuhan Union Hospital.

Collection Method

Retrospective cohort study using multivariable logistic regression analysis.

Freshness

Last updated 2026-04-27 04:14:30; freshness should be verified.

Geography

Data sources suggest coverage includes the United States (eICU) and China (Wuhan).