Individual stroke risk

26 preoperative inputs. Multicentre retrospective model.

Calculated in your browser
No input values are uploaded

Research prototype. Use the original dataset's coding and units. ASA I–IV correspond to source codes A–D, as confirmed by the data provider. FAR scale requires confirmation. No prospective clinical utility evaluation.

Preoperative profile

MODEL ESTIMATENo profile submitted

—%

Recorded perioperative stroke outcome

Complete the profile to calculate a model estimate.

Research risk group / 研究风险分组

Not calculated / 尚未计算

Fixed development reference cutoff: 0.677816%. Labels use the unrounded probability.

A research grouping, not a validated treatment threshold.

Individual explanation

SHAP approximation · log-odds

Increases model estimateDecreases model estimate

Feature contributions will appear here.

SHAP explains the model, not causation. The estimate is not a treatment recommendation.

Model scope, verification and limitations

Source population

Development: First Medical Center, Chinese PLA General Hospital (223,415 records). Evaluation: Henan Provincial People's Hospital (48,719) and Nanfang Hospital, Southern Medical University (37,568).

Technical verification

Browser implementation compared with the locked native model using source-record inputs and split-boundary cases. Full results accompany the figure package.

Interpretation

Multicentre retrospective prediction model study. Original missing-data handling and selected source codes require verification. Cross-hospital calibration and the Henan CAD coding anomaly limit clinical interpretation. Automatic high/low research groups use the development out-of-fold 95th-percentile cutoff (probability 0.006778159165645454). The cutoff is not a validated treatment threshold and does not select exactly 5% at other hospitals. General assessment prompts are provided to both groups; no personalised drug, dose or treatment recommendation is generated.