Neural Networks for Survival Prediction in Medicine Using Prognostic Factors: A Review and Critical Appraisal

Table 4

Summary of the findings from the critical appraisal across the 24 manuscripts.

---

Unclear addressing of missing data (42.9%) or ad-hoc methods (23.8%)

Unclear reporting of hyperparameters (62.5%)

Unclear reporting of the performance criterion for model development (25.0%)

Unclear scaling of prognostic factors (41.7%)

Unclear programming language for SNNs (29.2%)

Large variability and improper performance measures for survival data

External validation for only 4 outcomes (11.8%)

No confidence intervals for the predictive measures (54.2%)

No calibration plots (54.2%)

No interactions in Cox regression or unclear reporting (89.5%)

---