Supplements

Prediction models for psychological distress in patients with malignant tumors show high reported performance but lack external validation and have high risk of bias

This scoping review of 13 studies and 26 prediction models found that models using XGBoost, Random Forests, and Artificial Neural Networks reported high performance (AUCs 0.673-1.000, sensitivities 0.518-0.968, specificities 0.651-1.000). However, all studies had high risk of bias, none underwent external validation, and performance is likely inflated due to small sample sizes and low events per variable.

Last updated: Jul 11, 2026•0 RCTs•📖 Read as article →

Evidence Score

Evidence Score32/100
Human RCT☆☆☆☆☆
Meta-analysis☆☆☆☆☆
Mechanism★★★★★
Safety★★★★☆
Confidencelow

Study Evidence

Study 1. Prediction Models for Psychological Distress in Patients With Malignant Tumors: A Scoping Review.

observational

Chen H, Zhang W, Li X, Li J, Ning X, Yang X · Psycho-oncology (2026)

Participants: N/A
Duration: N/A
Intervention: Prediction models (Logistic Regression, Random Forests, XGBoost, Artificial Neural Networks) for psychological distress
Outcome: Model performance metrics: sensitivity (0.518-0.968), specificity (0.651-1.000), AUC (0.673-1.000); predictors included tumor stage, sleep quality, pain degree, age, financial problems, coping style
Effect Size: N/A
Population: Patients with malignant tumors (cancer patients)

Result:

PubMed: 42417204DOI: 10.1002/pon.70540

Mechanism Graph

Models developed using machine learning algorithms (XGBoost, RF, ANN)
↓
Predictors such as tumor stage, sleep quality, pain, age, financial problems, coping style included
↓
Models presented as nomograms or web-based calculators
↓
Result: High reported performance but likely inflated due to methodological flaws

Limitations

  • ⚠All 13 studies had high risk of bias
  • ⚠No external validation performed on any model
  • ⚠Small sample sizes and low events per variable (EPV) likely inflated performance

Frequently Asked Questions

What prediction models were most commonly used for psychological distress in cancer patients?▼

Logistic Regression, Random Forests, eXtreme Gradient Boosting (XGBoost), and Artificial Neural Networks were the most common development methods.

How well did these prediction models perform?▼

Reported sensitivities ranged from 0.518 to 0.968, specificities from 0.651 to 1.000, and AUCs from 0.673 to 1.000, but these are likely inflated due to methodological issues.

What are the main limitations of current prediction models for psychological distress in cancer patients?▼

All studies had high risk of bias, none underwent external validation, and many had small sample sizes and low events per variable, making performance estimates unreliable.

What predictors were frequently included in these models?▼

Frequently included predictors were tumor stage, sleep quality, pain degree, age, financial problems, and coping style.

Products

Affiliate links coming soon. We only recommend products that match the doses and forms used in the cited research.

References

  1. 1.Chen H, Zhang W, Li X, Li J, Ning X, Yang X. "Prediction Models for Psychological Distress in Patients With Malignant Tumors: A Scoping Review.." Psycho-oncology, 2026. PMID: 42417204 DOI: 10.1002/pon.70540
Disclaimer: This content is for educational purposes only and is not medical advice. Evidence scores reflect the quality and quantity of available research, not clinical recommendations. Always consult a healthcare professional before starting any supplement or intervention.