Machine Learning Assisted Differentiation of Low Acuity Patients at Dispatch: The MADLAD Randomi

31 Mar 20264 min read0 viewsJournal Feed

GIST

by Douglas Nils Spangler, Simon Morelli, David Smekal, Lennart Edmark, Hans Blomberg Background Resource Constrained Situations (RCS) at Emergency Medical Dispatch centers where there are more patients requiring an ambulance than there are available ambulances are common. Machine Learning (ML) techniques offer a promising but largely untested approach to assessing relative risks among these patients.

The study aims to establish whether the provision of ML-based risk scores predicting patient outcomes improves the ability of dispatchers to identify patients at high risk for deterioration in RCS and dispatch the first available ambulance to them. Methods and findings We performed a parallel-group, randomized trial of adult patients assessed by a dispatch nurse at two study sites in Sweden as requiring a low-priority ambulance response in RCS.

Patients were randomized 1:1 to be prioritized with the aid of an ML-based risk assessment tool, or per current clinical practice. The primary outcome was defined in terms of whether the first available ambulance was sent to the patient with the highest National Early Warning Score (NEWS 2) based on subsequently collected vital signs.

Clinical Editorial

Dispatch Context and Study Design

This randomized parallel-group trial evaluated whether an ML-based risk assessment tool used by dispatch nurses could improve identification of high-risk patients among those deemed low-priority for ambulance response.
Conducted at two Swedish study sites, enrolling adult patients assessed as requiring a low-priority response in resource-constrained emergency dispatch (RCS) settings.
Participants were randomized 1:1 to receive ML-assisted prioritization versus standard clinical practice.
Primary outcome: whether the first available ambulance was sent to the patient with the highest NEWS 2 score based on vital signs collected after dispatch.

Major Signal and Statistical Ambiguities

Among 1,245 RCS encounters, the ML-assisted group achieved correct prioritization for the primary outcome in 68.3% of cases, compared with 62.5% in the control group.
Reported odds ratio for correct prioritization: 1.28 (95% CI 1.00–1.63), p = 0.047.
The study notes the primary outcome was underpowered due to smaller-than-expected sample size and is limited to low-priority responses in two Swedish regions.

Context within Evidence Landscape

The trial addresses resource-constrained dispatch environments where timely differentiation between high- and low-risk patients is critical.
Findings suggest ML-based decision support can influence dispatcher decisions and potentially enhance rapid risk stratification at the point of dispatch.
Generalizability is limited by the focus on low-priority cases, the regional setting, and sample size constraints.

Limitations and Open Questions

The study was underpowered for its primary outcome and restricted to low-priority patients in two regions, limiting broader applicability.
It remains unclear how ML performance translates to higher-priority scenarios or different emergency systems.
Further research is needed to assess performance in larger, more diverse cohorts, across varying priority levels, and in other settings.

Operational Implications and Next Steps

If replicated, ML-assisted prioritization could be incorporated to support dispatch decisions and potentially improve rapid identification of deteriorating patients.
Future work should explore scalability, calibration across regions, and impact across the full spectrum of priority levels.

Safety and Trial Integrity

The trial is registered (ClinicalTrials.gov NCT04757194) and reports pre-specified outcomes with statistical testing, though limitations regarding power and generalizability are acknowledged.