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.
This article is automatically generated from the structured evidence profile behind the claim above. Scores reflect the quality and quantity of available research, not clinical advice.
Prediction models for psychological distress in patients with malignant tumors show high reported performance but lack external validation and have high risk of bias The current body of evidence comprises 1 study. EvidenceHub rates the overall confidence at 32/100 (low).
The Claim
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 conclusion is most relevant to: Patients with malignant tumors (cancer patients).
What the Research Shows
The conclusion draws on 1 linked study. Highlights from the cited literature:
- ▸Prediction Models for Psychological Distress in Patients With Malignant Tumors: A Scoping Review. (Psycho-oncology, 2026) —
How It Works
The proposed biological pathway:
- ▸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
Who Might Benefit
Evidence fit by population:
- ▸Patients with malignant tumors (cancer patients)
Recommended Dose
N/A
Limitations & Caveats
Important context when interpreting this evidence:
- ▸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.
References
- 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