Purpose: The early detection of major post-stroke depression (PSD) is essential to optimize patient care. A major PSD prediction tool needs to be developed and validated for early screening of major PSD patients.
Patients and Methods: A total of 639 acute ischemic stroke (AIS) patients from three hospitals were consecutively recruited and completed a 3-month follow-up. Sociodemographic, clinical and laboratory test data were collected on admission. With major depression criteria being met in the DSM-V, 17-item Hamilton Rating Scale For Depression (HRSD) score ≥ 17 at 3 months after stroke onset was regarded as the primary endpoint. Multiple imputation was used to substitute the missing values and multivariable logistic regression model was fitted to determine associated factors with a bootstrap backward selection process. The nomogram was constructed based on the regression coefficients of the associated factors. Performance of the nomogram was assessed by discrimination (C-statistics) and calibration curve.
Results: A total of 7.04% (45/639) of patients were diagnosed with major PSD at 3 months. The final logistic regression model included age, baseline NIHSS and mRS scores, educational level, calcium–phosphorus product, history of hypertension and atrial fibrillation. The model had acceptable discrimination, based on a C-statistic of 0.81 (95% CI, 0.791– 0.829), with 71.1% sensitivity and 78.6% specificity. We also transformed the model to a nomogram, an easy-to-use clinical tool which could be used to facilitate the early screening of major PSD patients at 3 months.
Conclusion: We identified several associated factors of major PSD at 3 months and constructed a convenient nomogram to guide follow-up and aid accurate prognostic assessment.
Keywords: major post-stroke depression, nomogram, calcium–phosphorus product, C-statistic