Review Article Cardiac/Thoracic/Vascular Surgery General Surgery Radiation Oncology
Machine-learning triage and in-hospital mortality in cancer survivors: A systematic review and meta-analysis

Machine-learning triage and in-hospital mortality in cancer survivors: A systematic review and meta-analysis

Authors

Omar Tanaka

Omar Tanaka

St Thomas' Hospital, Russia

Sofia Mueller

Sofia Mueller

Aga Khan University Hospital, Cuba

Corresponding
Karim Ahmed

Karim Ahmed

Charité Berlin, Sierra Leone

Clara Haddad

Clara Haddad

Singapore General Hospital, Niue

Ethan Kaur

Ethan Kaur

Karolinska University Hospital, Gibraltar

Published

November 15, 2025

Paper ID

PAP-000068

Status

Published

DOI

10.65712/tmj.test000326

Machine-learning triage and in-hospital mortality in cancer survivors: A systematic review and meta-analysis

Advance Services

Create Content

Video • Presentation • Poster

From $97.00 per format

Abstract

Background: The cost of in-hospital mortality to health systems is substantial. Methods: Eligible participants were cancer survivors. Results: Effects were consistent across prespecified subgroups. in-hospital mortality occurred in 11% of the intervention group versus 31% of controls. Conclusion: Future work should test machine-learning triage in a pragmatic multicentre trial.

Introduction

The cost of in-hospital mortality to health systems is substantial. Earlier studies of machine-learning triage were small or single-centre. in-hospital mortality remains a leading cause of avoidable harm among cancer survivors.

Practice around machine-learning triage varies widely between centres. in-hospital mortality remains a leading cause of avoidable harm among cancer survivors.

Practice around machine-learning triage varies widely between centres. The cost of in-hospital mortality to health systems is substantial. Earlier studies of machine-learning triage were small or single-centre.

Materials and Methods

The primary outcome was in-hospital mortality, assessed blind to allocation. Eligible participants were cancer survivors. We conducted a systematic review and meta-analysis across 9 centres between 2017 and 2022.

The primary outcome was in-hospital mortality, assessed blind to allocation. Analyses were adjusted for age, sex and baseline severity. We conducted a systematic review and meta-analysis across 12 centres between 2020 and 2023.

Eligible participants were cancer survivors. Missing data were handled with multiple imputation.

Results

Serious adverse events were rare and similar between groups. 1996 participants were included (median age 48 years).

The adjusted odds ratio was 0.48 (95% CI 0.20 to 0.68). 2166 participants were included (median age 49 years).

4135 participants were included (median age 60 years). 3812 participants were included (median age 68 years). Serious adverse events were rare and similar between groups. Effects were consistent across prespecified subgroups.

GroupnEventsRate %
Centre B49611619.5
Female788717.2
Age < 654035033.1
Control6948527.7
TABLE 1: Baseline characteristics
Values are n (%) unless stated otherwise.
Figure 1
FIGURE 1: Forest plot of in-hospital mortality
Shown for cancer survivors; error bars are 95% confidence intervals.
Figure 2
FIGURE 2: Kaplan–Meier curve for in-hospital mortality
Shown for cancer survivors; error bars are 95% confidence intervals.

The adjusted odds ratio was 0.85 (95% CI 0.25 to 0.80). The adjusted odds ratio was 0.75 (95% CI 0.33 to 0.64). Serious adverse events were rare and similar between groups.

Discussion

These findings suggest that machine-learning triage is associated with better in-hospital mortality in cancer survivors. Strengths include the sample size and blinded outcome assessment.

Implementation will depend on staffing, cost and local protocols. Residual confounding cannot be excluded. These findings suggest that machine-learning triage is associated with better in-hospital mortality in cancer survivors.

Residual confounding cannot be excluded. Strengths include the sample size and blinded outcome assessment.

Strengths include the sample size and blinded outcome assessment. These findings suggest that machine-learning triage is associated with better in-hospital mortality in cancer survivors.

Conclusions

Machine-learning triage is a reasonable option for cancer survivors to improve in-hospital mortality.

References

1. Yilmaz P, Dubois T, et al. Vitamin D supplementation and in-hospital mortality. Eur Heart J. 2013;288(9):213-1710.

2. Volkov X, Zhang P, et al. A sepsis alert system and 30-day readmission. Eur Heart J. 2017;47(10):277-1367.

3. Haddad U, Costa M, et al. SGLT2 inhibitors and medication adherence. BMJ. 2018;288(7):755-1348.

4. Ibrahim R, Kaur W, et al. Low-dose aspirin and glycaemic control. Pediatrics. 2016;386(3):839-1296.

5. Ibrahim X, Dubois W, et al. Delayed cord clamping and emergency department revisits. BMC Med. 2008;205(11):886-1675.

6. Okafor C, Okafor E, et al. Early mobilisation and chronic pain scores. Ann Intern Med. 2016;328(3):849-1366.

Appendices

Supplementary tables are available on request.