Case Report Allergy/Immunology Cardiology
Effect of machine-learning triage on falls among patients on haemodialysis

Effect of machine-learning triage on falls among patients on haemodialysis

Authors

Mateo Lindqvist

Mateo Lindqvist

Toronto General Hospital, Palestinian Territory Occupied

James Farouk

James Farouk

Karolinska University Hospital, Qatar

Corresponding
Clara Ahmed

Clara Ahmed

Aga Khan University Hospital, Aruba

Sarah Rahman

Sarah Rahman

American University of Beirut Medical Center, Central African Republic

Layla Tanaka

Layla Tanaka

Cleveland Clinic Abu Dhabi, Reunion

Published

December 30, 2023

Paper ID

PAP-000039

Status

Published

DOI

10.65712/tmj.test000297

Effect of machine-learning triage on falls among patients on haemodialysis

Advance Services

Create Content

Video • Presentation • Poster

From $97.00 per format

Abstract

Background: Practice around machine-learning triage varies widely between centres. Methods: Analyses were adjusted for age, sex and baseline severity. Results: falls occurred in 20% of the intervention group versus 24% of controls. Serious adverse events were rare and similar between groups. Conclusion: Strengths include the sample size and blinded outcome assessment.

Introduction

Earlier studies of machine-learning triage were small or single-centre. falls remains a leading cause of avoidable harm among patients on haemodialysis. The cost of falls to health systems is substantial.

falls remains a leading cause of avoidable harm among patients on haemodialysis. Practice around machine-learning triage varies widely between centres. The cost of falls 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.

The cost of falls to health systems is substantial. Guidelines make only a weak recommendation on machine-learning triage.

Discussion

These findings suggest that machine-learning triage is associated with better falls in patients on haemodialysis. Strengths include the sample size and blinded outcome assessment. Residual confounding cannot be excluded.

Implementation will depend on staffing, cost and local protocols. Future work should test machine-learning triage in a pragmatic multicentre trial. Residual confounding cannot be excluded.

Implementation will depend on staffing, cost and local protocols. Future work should test machine-learning triage in a pragmatic multicentre trial. Strengths include the sample size and blinded outcome assessment.

Conclusions

Machine-learning triage should be considered for patients on haemodialysis to improve falls.

References

1. Tanaka C, Moreau G, et al. SGLT2 inhibitors and chronic pain scores. BMC Med. 2016;224(2):430-1200.

2. Rahman A, Haddad J, et al. Continuous glucose monitoring and 30-day readmission. BMC Med. 2008;297(8):864-1763.

3. Kim N, Herrera W, et al. High-flow nasal oxygen and chronic pain scores. BMC Med. 2020;53(10):66-1440.

4. Dubois V, Patel W, et al. Telemedicine follow-up and falls. Diabetes Care. 2008;141(3):214-1412.

5. Nasser X, Osei R, et al. Machine-learning triage and glycaemic control. Lancet. 2018;314(11):689-1304.

6. Mensah W, Schmidt I, et al. SGLT2 inhibitors and surgical site infection. Diabetes Care. 2008;287(4):324-1386.

7. Schmidt C, Mueller O, et al. Early enteral nutrition and patient satisfaction. JAMA. 2008;255(9):717-1648.

Appendices

The statistical analysis plan is in the online supplement.

Materials and Methods

Eligible participants were patients on haemodialysis. We conducted a pilot feasibility study across 12 centres between 2017 and 2024. Analyses were adjusted for age, sex and baseline severity.

We conducted a pilot feasibility study across 7 centres between 2015 and 2025. Missing data were handled with multiple imputation. Eligible participants were patients on haemodialysis. The primary outcome was falls, assessed blind to allocation.

Results

1466 participants were included (median age 82 years). Serious adverse events were rare and similar between groups.

The adjusted odds ratio was 0.78 (95% CI 0.32 to 0.87). Serious adverse events were rare and similar between groups. Effects were consistent across prespecified subgroups.

Effects were consistent across prespecified subgroups. The adjusted odds ratio was 0.48 (95% CI 0.45 to 0.94).

GroupnEvents
Control782105
Female71427
Age < 6516184
Control6742
Centre A34586
Centre A33445
Centre A28519
TABLE 1: Subgroup analysis
Values are n (%) unless stated otherwise.
Figure 1
FIGURE 1: Kaplan–Meier curve for falls
Shown for patients on haemodialysis; error bars are 95% confidence intervals.

falls occurred in 22% of the intervention group versus 29% of controls. The adjusted odds ratio was 0.84 (95% CI 0.46 to 0.62).