Table 1 Performance comparison of predictive models by cancer type (C-Index and 95% confidence intervals) on evaluation set.

From: Predicting time-to-first cancer diagnosis across multiple cancer types

Cancer Type

Cox proportional hazards

Survival decision tree

RSF

Lung

0.813 (0.803–0.824)

0.784 (0.773–0.796)

0.802 (0.791–0.813)

Liver

0.759 (0.725–0.789)

0.742 (0.704—0.772)

0.740 (0.701–0.772)

Bladder

0.747 (0.734–0.759)

0.714 (0.701–0.727)

0.724 (0.710–0.737)

Upper GI

0.719 (0.702–0.735)

0.681 (0.666–0.698)

0.694 (0.675–0.709)

Head and Neck

0.696 (0.678–0.716)

0.668 (0.651–0.688)

0.673 (0.653–0.694)

Renal

0.668 (0.645–0.687)

0.628 (0.606–0.649)

0.622 (0.599–0.643)

Pancreatic

0.665 (0.644–0.684)

0.588 (0.562–0.611)

0.616 (0.594–0.639)

Colorectal

0.645 (0.636–0.655)

0.608 (0.599–0.618)

0.632 (0.623–0.642)

Hematological

0.633 (0.624–0.644)

0.610 (0.599–0.621)

0.605 (0.594–0.616)

Biliary

0.618 (0.578–0.660)

0.531 (0.500–0.565)

0.594 (0.562–0.625)

Melanoma

0.593 (0.579–0.606)

0.563 (0.550–0.575)

0.573 (0.558–0.587)

Glioblastoma

0.593 (0.571–0.617)

0.562 (0.538–0.588)

0.549 (0.523–0.569)

Any

0.589 (0.585—0.592)

0.584 (0.580—0.587)

0.573 (0.569—0.576)

Thyroid

0.584 (0.538–0.621)

0.573 (0.532–0.613)

0.544 (0.498–0.580)

  1. The table shows the C-index values and 95% confidence intervals for Cox proportional hazards, Survival Decision Tree, and RSF models across different cancer types on the UKBB dataset. Confidence intervals were computed with 400 bootstrap iterations.
  2. Significant values are in bold.