Causal evaluation of ML-enabled interventions in nursing and midwifery

Source article: Machine Learning in Causal Effect Estimation and Causal Evaluation of ML-Enabled Interventions in Nursing-Related Research: A Scoping Review

Machine learning (ML) informs nursing practice, but prediction cannot establish intervention or workflow effects. Aim To map ML use within causal effect estimation and causal evaluations of ML-enabled interventions in nursing and midwifery. Eight electronic sources were searched for reports published January 2015-June 2026, followed by backward citation searching. Eligibility required a causal estimand, corresponding identification strategy, substantive ML role, and nursing or midwifery relevance. Two reviewers…

By the Truvace deskOctober 14, 20262 min read1 sourcePeer-reviewedTRV-2026-1339· cite & verify

Trace impact readingContested

P 68

Both sides are scored from claims and sources, not community votes.

G 70

In brief

The review searched eight electronic sources for reports from January 2015 to June 2026 and included fifteen studies in seventeen reports. Eligible work needed a causal estimand, an identification strategy, a substantive ML role, and relevance to nursing or midwifery.

The findings matter because prediction accuracy alone cannot show whether a tool changes care or patient outcomes. The evidence is still thin and uneven, so it remains uncertain how reliably ML-enabled nursing workflows improve outcomes; the authors call for better alignment of methods and workflows.

Main points

Fifteen studies (17 reports) met the inclusion criteria after independent screening by two reviewers.
Five studies addressed acute deterioration or sepsis; others covered care prioritization, falls, post-discharge care, and nurse- or midwife-led services.
Five studies used ML within causal effect estimation to examine longitudinal and spillover effects, treatment-effect heterogeneity, and targeting strategies.
Ten studies evaluated ML-enabled interventions or workflows, with favorable effects reported alongside null or imprecise controlled estimates.

The gain

Among ML-enabled nursing interventions or workflows evaluated causally, studies reported favorable effects on mortality, care escalation, message-review time, and nursing processes.

The problem

Several controlled estimates for ML-enabled nursing interventions were null or imprecise, and the evidence base is small and methodologically uneven.

The rundown

The review set out to map how machine learning is used in causal effect estimation and in causal evaluations of ML-enabled interventions in nursing and midwifery. It started from the premise that ML informs nursing practice but prediction cannot establish intervention or workflow effects.

Of the included studies, five applied ML inside causal estimation methods, looking at longitudinal and spillover effects, heterogeneity of treatment effects, and targeting strategies. Ten evaluated ML-enabled interventions or workflows, covering outcomes such as mortality, care escalation, message-review time, and nursing processes.

The authors recommend that future studies align estimands, identification strategies, analytic methods, and workflows, and evaluate model-plus-workflow effects on nursing processes and patient outcomes.

What this doesn’t fix

The evidence base is small and methodologically uneven, and several controlled estimates were null or imprecise, so the favorable results are not conclusive.