Abstract
Objectives
This study aimed to develop and internally validate a nomogram prediction model for differentiating subacute combined degeneration of the spinal cord (SCD) from other neurological disorders, based on established clinical predictors including a symptom scale score, risk factors, and laboratory indicators.
Methods
We retrospectively enrolled 80 patients with SCD (SCD group) and 80 patients with other neurological disorders (non-SCD group) treated at Beijing Yanhua Hospital from March 2015 to March 2025. Demographic characteristics, SCD-related risk factors, SCD symptom scale scores, and laboratory indicators were collected. Eight risk factors for vitamin B12 deficiency were integrated into the Risk Load Score of B12 Deficiency (RLSS). The RLSS, total SCD symptom scale score, and serum homocysteine (Hcy) were included as candidate variables. LASSO regression was used for variable compression and selection. The selected variables were entered into multivariate logistic regression to construct the nomogram prediction model. We evaluated the model’s discrimination, calibration, and clinical utility, and performed internal validation using Bootstrap resampling (1,000 Bootstrap resamples).
Results
LASSO regression identified three non-zero coefficient predictors (total SCD symptom scale score, RLSS, and Hcy) from the seven candidate variables (RLSS, total SCD symptom scale score, Hcy, MCV, age, sex, and disease duration). Multivariate logistic regression showed that RLSS, Hcy, and total SCD symptom scale score were all independent predictors of SCD. The area under the curve (AUC) of the combined model was 0.954. The decrease in the C-index after Bootstrap correction was less than 0.05, indicating good calibration (Hosmer-Lemeshow
χ
2
= 2.326,
p
= 0.969; Brier score = 0.095). Decision curve analysis showed significant clinical net benefit across threshold probabilities of 0.1 to 0.8.
Conclusion
The nomogram model based on total SCD symptom scale score, RLSS, and Hcy demonstrated good discrimination and calibration in differentiating SCD from non-SCD neurological disorders, as confirmed by internal validation. With its simple operation and readily available indicators, this model provides visualized decision support for early clinical identification of SCD. However, external validation is required before it can be recommended for routine use in primary healthcare settings.