Technical Report Emergency Medicine Hematology Nutrition
Effect of machine-learning triage on glycaemic control among cancer survivors

Effect of machine-learning triage on glycaemic control among cancer survivors

Authors

Lucas Zhang

Lucas Zhang

American University of Beirut Medical Center, Western Sahara

Sofia Rossi

Sofia Rossi

St Thomas' Hospital, Saint Vincent and the Grenadines

Corresponding
Ethan Lindqvist

Ethan Lindqvist

Mayo Clinic, Comoros

Hana Tanaka

Hana Tanaka

Mayo Clinic, Slovenia

Yusuf Dubois

Yusuf Dubois

Royal Melbourne Hospital, Greece

Omar Costa

Omar Costa

Royal Melbourne Hospital, Mauritania

Published

March 28, 2025

Paper ID

PAP-000158

Status

Published

DOI

10.65712/tmj.test000416

Effect of machine-learning triage on glycaemic control among cancer survivors

Advance Services

Create Content

Video • Presentation • Poster

From $97.00 per format

Abstract

Background: Guidelines make only a weak recommendation on machine-learning triage. Methods: The primary outcome was glycaemic control, assessed blind to allocation. Results: 137 participants were included (median age 41 years). 1304 participants were included (median age 78 years). Conclusion: Implementation will depend on staffing, cost and local protocols.

Introduction

Practice around machine-learning triage varies widely between centres. The cost of glycaemic control to health systems is substantial.

Practice around machine-learning triage varies widely between centres. Earlier studies of machine-learning triage were small or single-centre. Guidelines make only a weak recommendation on machine-learning triage.

Guidelines make only a weak recommendation on machine-learning triage. Practice around machine-learning triage varies widely between centres. The cost of glycaemic control to health systems is substantial.

Discussion

Future work should test machine-learning triage in a pragmatic multicentre trial. Residual confounding cannot be excluded.

Strengths include the sample size and blinded outcome assessment. Implementation will depend on staffing, cost and local protocols.

Implementation will depend on staffing, cost and local protocols. Future work should test machine-learning triage in a pragmatic multicentre trial.

These findings suggest that machine-learning triage is associated with better glycaemic control in cancer survivors. Residual confounding cannot be excluded. Strengths include the sample size and blinded outcome assessment.

Conclusions

Machine-learning triage is a reasonable option for cancer survivors to improve glycaemic control.

References

1. Kim V, Moreau B, et al. An enhanced recovery pathway and 30-day readmission. Lancet. 2018;95(11):639-1707.

2. Tanaka I, Nasser S, et al. Early mobilisation and emergency department revisits. Pediatrics. 2013;334(2):454-1015.

3. Volkov E, Rossi V, et al. Pharmacist-led medication review and length of stay. BMC Med. 2020;175(10):45-1168.

4. Rossi G, Kaur L, et al. Machine-learning triage and post-operative delirium. Eur Heart J. 2008;312(1):129-944.

5. Yilmaz F, Rahman R, et al. Point-of-care ultrasound and time to diagnosis. BMJ. 2023;78(9):216-1606.

6. Andersen K, Herrera Q, et al. Early enteral nutrition and in-hospital mortality. Pediatrics. 2010;385(4):805-1430.

Appendices

The statistical analysis plan is in the online supplement.

Materials and Methods

Missing data were handled with multiple imputation. We conducted a multicentre observational study across 1 centres between 2015 and 2023. Eligible participants were cancer survivors.

Analyses were adjusted for age, sex and baseline severity. We conducted a multicentre observational study across 9 centres between 2019 and 2022. Missing data were handled with multiple imputation.

Results

The adjusted odds ratio was 0.84 (95% CI 0.43 to 0.65). 2903 participants were included (median age 42 years).

Effects were consistent across prespecified subgroups. The adjusted odds ratio was 0.39 (95% CI 0.35 to 0.82). Serious adverse events were rare and similar between groups.

The adjusted odds ratio was 0.63 (95% CI 0.45 to 0.68). 1055 participants were included (median age 77 years). 357 participants were included (median age 51 years). glycaemic control occurred in 8% of the intervention group versus 33% of controls.

GroupnEvents
Centre B86781
Age ≥ 6575106
Control116113
TABLE 1: Subgroup analysis
Values are n (%) unless stated otherwise.
Figure 1
FIGURE 1: Trend in glycaemic control
Shown for cancer survivors; error bars are 95% confidence intervals.
Figure 2
FIGURE 2: Trend in glycaemic control
Shown for cancer survivors; error bars are 95% confidence intervals.

glycaemic control occurred in 14% of the intervention group versus 38% of controls. Serious adverse events were rare and similar between groups.

Serious adverse events were rare and similar between groups. 4153 participants were included (median age 68 years). Effects were consistent across prespecified subgroups. 1572 participants were included (median age 77 years).