Machine learning enables construction of a nomogram based on risk factors for adverse emotions in patients with diabetic foot infection.

OBJECTIVE

To identify risk factors and construct a nomogram model using logistic regression to predict mood disturbance in patients with diabetic foot infection.

METHODS

We retrospectively analyzed 313 patients with diabetic foot infection who received treatment at our hospital between October 2020 and January 2023. Patients were grouped based on their post-treatment Self-Rating Anxiety Scale (SAS ≥50) and Self-Rating Depression Scale (SDS ≥53) scores into two groups: 134 patients with adverse mood and 179 with stable mood. The patients were divided into a test group (n=220) and a validation group (n=93) at a 7:3 ratio. Clinical data and laboratory indicators were collected to screen characteristic factors using four machine learning models. Common risk factors were screened using logistic regression, visualized, and incorporated into a nomogram. The clinical value, accuracy, and predictive value of the model were evaluated using receiver operating characteristic curves (ROCs), calibration curves, and decision curve analyses (DCAs).

RESULTS

Analysis identified Wagner classification, comorbidities, glycated hemoglobin (HbA1c), gender, and history of diabetes as common features across four machine learning models. Multifactorial logistic regression confirmed that Wagner classification, comorbidities, HbA1c, gender, and history of diabetes were independent risk factors for adverse mood in patients with diabetic foot infection. We constructed a nomogram based on the five characteristic factors. ROC curve analysis yielded an area under the curve (AUC) of 0.829, indicating high predictive accuracy for mood disturbances in the test group. Calibration curve and DCA analysis demonstrated the model's stability and clinical relevance, further supported by external validation.

CONCLUSION

This study enhanced the predictive accuracy for mood disorders in patients with diabetic foot infections by leveraging machine learning to identify and visualize significant risk factors through a nomogram. This may be a valuable tool for clinical assessments and intervention.