Predicting risk of treatment non-adherence in patients receiving intense light therapy: development and evaluation of a new predictive nomogram.

BACKGROUND

This study aimed to develop and validate a treatment non-adherence risk nomogram for patients receiving intense pulsed light (IPL) therapy, utilizing internal bootstrap validation and external independent set validation.

PATIENTS AND METHODS

We analyzed data from 316 patients who received IPL therapy between January and December 2021. The dataset was divided into a training set (n = 213, 67.41%) for model development and a validation set (n = 103, 32.59%). Twenty-two variables spanning demographics, clinical consultations/treatments, and personal habits were evaluated. A multivariable logistic regression model was constructed using predictors selected via Least Absolute Shrinkage and Selection Operator (LASSO) regression. Model performance was assessed by discrimination (C-index), calibration (calibration curves), and clinical utility (decision curve analysis, DCA). Internal validation employed 1,000 bootstrap resamples, followed by external validation.

RESULTS

LASSO regression identified seven predictors: number of jewels, local residents, unrealistic demands, undergone plastic surgery, regular physical exercise, awareness of sun protection, and carbonated/sugary drinks. The nomogram demonstrated strong discrimination (training C-index: 0.922, 95% CI: 0.888-0.956; modified C-index: 0.896) and excellent calibration. External validation confirmed robustness (C-index: 0.897, 95% CI: 0.832-0.962). DCA revealed superior net benefit at threshold probabilities of 1-100% (training) and 1-88% (validation), supporting clinical utility.

CONCLUSION

This validated nomogram provides a practical tool for individualized prediction of IPL therapy non-adherence risk, facilitating targeted interventions to improve treatment adherence.