Scholarlink

Purpose <p>As the number of cancer survivors increases, psychological distress has become an important issue. Using nationally representative data, we evaluated mental health outcomes among Korean cancer survivors compared with cancer-free controls and developed models to identify individuals at risk of psychological distress.</p> Methods <p>We analyzed data from the Korea National Health and Nutrition Examination Survey (KNHANES) from 2007 to 2021. Psychological outcomes were assessed using standardized questionnaires, and a composite distress outcome was constructed. Risk stratification models were developed among cancer survivors using logistic regression and machine learning algorithms, including random forest, XGBoost, LightGBM, support vector machines, k-nearest neighbors, and naïve Bayes.</p> Results <p>A total of 88,061 participants were included, comprising 3733 cancer survivors and 84,328 cancer-free controls. Compared with cancer-free controls, cancer survivors had higher odds of depressed mood (OR 1.33; 95% CI 1.18–1.51), suicidal ideation (OR 1.14; 95% CI 1.00–1.31), suicide planning (OR 1.91; 95% CI 1.37–2.65), and mental health counseling (OR 1.36; 95% CI 1.08–1.71). Among cancer survivors, multiple models were evaluated, with logistic regression showing the highest performance (AUROC 0.689), followed by XGBoost (0.686). In logistic regression, longer working hours, depression history, activity limitation, female sex, smoking, employment, low income, and distorted body image were independently associated with distress. SHAP analysis identified activity limitation, sex, and depression history as key factors.</p> Conclusions <p>Cancer survivors experience increased psychological distress across multiple outcomes. Machine learning–based models may help identify individuals at higher risk of psychological distress, supporting risk-based assessment in survivorship care.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Psychological distress in cancer survivors: a population-based analysis and machine learning–based risk stratification

  • 期刊论文

  • 发表日期：

    2026年5月27日

查看全文
  • Jeong Yun Jang,
  • Dawoon Jeong,
  • Hyeon Kang Koh,
  • Kyunghye Bang,
  • Gyurim Kim,
  • Semie Hong

摘要

Purpose

As the number of cancer survivors increases, psychological distress has become an important issue. Using nationally representative data, we evaluated mental health outcomes among Korean cancer survivors compared with cancer-free controls and developed models to identify individuals at risk of psychological distress.

Methods

We analyzed data from the Korea National Health and Nutrition Examination Survey (KNHANES) from 2007 to 2021. Psychological outcomes were assessed using standardized questionnaires, and a composite distress outcome was constructed. Risk stratification models were developed among cancer survivors using logistic regression and machine learning algorithms, including random forest, XGBoost, LightGBM, support vector machines, k-nearest neighbors, and naïve Bayes.

Results

A total of 88,061 participants were included, comprising 3733 cancer survivors and 84,328 cancer-free controls. Compared with cancer-free controls, cancer survivors had higher odds of depressed mood (OR 1.33; 95% CI 1.18–1.51), suicidal ideation (OR 1.14; 95% CI 1.00–1.31), suicide planning (OR 1.91; 95% CI 1.37–2.65), and mental health counseling (OR 1.36; 95% CI 1.08–1.71). Among cancer survivors, multiple models were evaluated, with logistic regression showing the highest performance (AUROC 0.689), followed by XGBoost (0.686). In logistic regression, longer working hours, depression history, activity limitation, female sex, smoking, employment, low income, and distorted body image were independently associated with distress. SHAP analysis identified activity limitation, sex, and depression history as key factors.

Conclusions

Cancer survivors experience increased psychological distress across multiple outcomes. Machine learning–based models may help identify individuals at higher risk of psychological distress, supporting risk-based assessment in survivorship care.