Machine-learning prediction of cancer survival: a retrospective study using electronic administrative records and a cancer registry
Problem
Predicting cancer patient survival accurately using routinely collected digital records, as an enhancement to conventional prognostic methods.
Why it matters
Accurate survival prognosis guides oncology treatment decisions and care planning.
What the AI did
Machine-learning models trained on a purpose-built cancer registry (ECO, 869 patients) and electronic administrative records predicted survival at 6, 12, and 24 months, validated on 94 patients and compared to five oncologists.
Results
ECO model AUC 0.87 (95% CI 0.848 to ...) for survival prediction; abstract truncated
Impact
ML on registry data predicted cancer survival with AUC 0.87 in a regional Australian cohort.
Record
| Venue | BMJ Open |
|---|---|
| Published | 2014 |
| Work type | article |
| Authors | Sunil Gupta, Truyen Tran, Wei Luo, Dinh Phung, Richard Lee Kennedy, A. Broad, David Colin Campbell, David Kipp et al. |
| Citations | 118 (OpenAlex, at import time) |
| References | 34 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1136/bmjopen-2013-004007 |
| Open access | gold |
| Keywords | machine learning, cancer survival, AUC, cancer registry, electronic records |
Importance rating
significant — thousands to millions affected
Cancer survival prognosis guides treatment decisions; AUC 0.87 slightly better than oncologist panel in one regional cohort; no broader problem scale stated.