A practical nomogram for the early prediction of hemiplegic shoulder pain in stroke patients: a development and validation study

Abstract

Abstract Objective To facilitate early clinical risk stratification for hemiplegic shoulder pain (HSP) after stroke, this study aimed to identify potential risk factors and develop a predictive nomogram. Methods This retrospective study included 450 inpatients with stroke admitted to Dongguan Hospital of Guangzhou University of Chinese Medicine between May 2022 and October 2025, among whom 118 developed HSP. The patients were randomly divided into a training cohort ( n = 315) and a testing cohort ( n = 135) at a ratio of 7:3. To ensure predictive validity, all baseline predictors were collected after admission and prior to the onset of HSP. The primary outcome was defined as the new onset of HSP during hospitalization. The least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation was used to screen candidate variables, followed by a multivariable logistic regression to establish a prediction model, based on which a nomogram was constructed. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curve, decision curve analysis (DCA), and clinical impact curve (CIC). Results The LASSO-logistic regression analysis identified abnormal sensation, osteoporosis, sleep disorder, and CRP as independent predictors of HSP. The nomogram incorporating these four predictors demonstrated good discriminative performance in both the training cohort 0.930 (95%CI: 0.895–0.966) and the testing cohort 0.883 (95%CI:0.818–0.948). Calibration curves demonstrated favorable concordance between predicted and actual results, whereas DCA and CIC suggested potential clinical utility. Conclusion The nomogram constructed in this study can serve as a preliminary risk stratification tool for HSP. Although it exhibits favorable discriminative performance in this single-center cohort and holds promise for guiding early intervention in high-risk populations, prospective multicenter studies are warranted to validate its generalizability and clinical utility.

The authors' abstract, as published at the source. BMC Medical Informatics and Decision Making, 2026 · DOI ↗