Abstract / Summary
Abstract Background Sarcopenia is closely related to adverse outcomes such as infection, hepatic encephalopathy and death in patients with liver cirrhosis. Although current imaging diagnostic methods like CT/ MRI are accurate, they are costly, complex to operate, and difficult to be promoted in grassroots settings or in scenarios with limited resources. This study sought to develop a simple and low‑cost screening model that can be applied at the bedside for sarcopenia risk in patients with liver cirrhosis. Method This prospective study used the third lumbar skeletal muscle index measured by abdominal CT as the diagnostic reference for sarcopenia. Routinely available clinical variables and anthropometric indicators at admission were collected. An elastic net logistic regression model was used to develop the prediction model. Model performance and clinical utility were evaluated using nested cross-validation, bootstrap correction, and decision curve analysis. Results A total of 201 patients with liver cirrhosis were included, of whom 49 patients had sarcopenia, with a prevalence of 24.4%. The final model included calf circumference-to-height ratio, mid-upper arm circumference-to-height ratio, and sex. After internal validation, the AUROC of the model was 0.773 (95% CI: 0.700–0.844), and the Brier score was 0.157 (95% CI: 0.130–0.185), indicating that the model has moderate discrimination ability and relatively low overall prediction error. The calibration intercept was − 0.017, suggesting that there was no significant systematic overestimation or underestimation overall. Decision curve analysis showed that the model had potential net benefit across the evaluated threshold probability range. Conclusion This study developed a sarcopenia risk screening model based on calf circumference-to-height ratio, mid-upper arm circumference-to-height ratio, and sex. The model uses simple, non-invasive, and easily accessible variables that can be obtained at the bedside. Internal validation showed moderate discrimination, low prediction error, and potential clinical net benefit. This model may serve as an auxiliary tool for early screening and risk stratification of sarcopenia in hospitalized patients with liver cirrhosis.