CoChat

SmartAlert - Implementing Machine Learning-Driven Clinical Decision Support for Inpatient Laboratory Utilization Reduction

Open access

The DOI matches its Crossref record; 19 of the 19 references it cites were found in a registry or an index; 10 of the 14 sources its claims cite were checked for support.

Abstract

Repetitive laboratory testing that is unlikely to yield clinically useful information is a common practice that burdens patients and increases health care costs. Education and feedback interventions have limited success, while general test ordering restrictions and electronic alerts impede appropriate clinical care. We introduce and evaluate SmartAlert, a machine learning-driven clinical decision support (CDS) system integrated into the electronic health record that predicts stable laboratory results to reduce unnecessary repeat testing. This Case Study describes the implementation process, challenges, and lessons learned from deploying SmartAlert targeting complete blood count (CBC) utilization in a randomized controlled pilot across 9270 admissions in eight acute care units across two hospitals between August 15, 2024, and March 15, 2025. The results show a significant decrease in the number of CBC results within 52 hours of SmartAlert display (1.54 vs. 1.82; P<0.01) without adverse effect on secondary safety outcomes, representing a 15% relative reduction in repetitive testing. Implementation lessons learned include interpretation of probabilistic model predictions in clinical contexts, stakeholder engagement to define acceptable model behavior, governance processes for deploying a complex model in a clinical environment, user interface design considerations, alignment with clinical operational priorities, and the value of qualitative feedback from end users. In conclusion, a machine learning-driven CDS system backed by a deliberate implementation and governance process can provide precision guidance on inpatient laboratory testing to safely reduce unnecessary repetitive testing. (Funded by the Agency for Science, Technology, and Research and others.).

Overview

CoChat Summary

TL;DR

SmartAlert, an EHR-integrated machine learning clinical decision support tool, was deployed in a randomized pilot to reduce repetitive inpatient complete blood count (CBC) testing. It significantly reduced CBC results within 52 hours of alert display (1.54 vs. 1.82; P<0.01) without adverse effects on secondary safety outcomes.

Key findings

  • In a 7-month randomized pilot across 9270 admissions in eight acute care units at two hospitals, the treatment group had 1.54 CBC results within 52 hours of SmartAlert display versus 1.82 in the control group (P<0.01), a 15.4% relative reduction.
  • A significant reduction in CBC results was also seen 28 hours after an alert; no significant differences were detected in ICU transfers within 52 hours, encounter-level ICU transfer rate, length of stay, readmission rate, or mortality.
  • Of 486 displayed SmartAlerts, 30.1% resulted in order adjustments, 49.6% were overridden, and 20.3% were dismissed; among overrides, “medical necessity” (40.9%) and “more review needed” (38.7%) were most common.
  • The model’s positive predictive value remained near the 90% goal in silent prospective evaluation (88.1%, 95.4%, and 93% for white blood cell counts, hemoglobin, and platelet counts, respectively); retrospective recall was 29.3%, 59.5%, and 100%, respectively.

Methods

  • Randomized controlled pilot conducted August 15, 2024, to March 15, 2025, in eight acute care inpatient units across two hospitals; each admission was randomly assigned via an Epic SmartDataElement to treatment (alert displayed when workflow criteria met) or control (alert hidden), excluding primary bone marrow transplant and hematology admissions.
  • Primary outcome was mean number of CBC results within 52 hours of the alert trigger; secondary outcomes included CBC results within 28 hours, ICU transfers within 52 hours, encounter-level ICU transfer rate, length of stay, readmission rate, and encounter-level mortality.
  • Poisson regression compared mean CBC counts; Mann–Whitney U test compared length of stay; Fisher’s exact test compared ICU transfers, readmission, and mortality; 95% confidence intervals were obtained through bootstrapping. Governance included CDS committee review, user codesign interviews with 18 clinicians, silent prospective evaluation, and a FURM assessment.

Limitations

  • The paper states the work was conducted at a single academic medical center with established governance committees that may not exist elsewhere, though the architecture and governance process are described as replicable.
  • The paper states the pilot focused on general medical and surgical settings, and additional evaluation would be necessary in other settings such as intensive care units and hematology wards.
  • Evident from the text: the intervention targeted only CBCs, alerts were displayed only between 7 a.m. and 6 p.m., and data retrieval delays led to batched pre-calculation rather than fully real-time prediction, which may limit immediate generalizability to other tests or workflows.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1056/aics2501086

Cites

19 of 19 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

5 Valid · 2 Not valid · 0 Not checked · 22 Own results and interpretations

2 more claims

Create a free account to see all 29 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 19
  • Probably not a substantial claim. 11

Create a free account to see these claims.

Citations

Cites

Research works printed but not linked: 1

  1. Committee on the Learning Health Care System in America, Institute of Medicine. Best care at lower cost: the path to continuously learning health care in America. (Smith M, Saunders R, Stuckhardt L, McGinnis JM, eds.). Washington, DC: National Academies Press (US), 2013 (http://www.ncbi.nlm.nih.gov/books/NBK207225/).

Metadata

Authors
Liang, A. S., Amrollahi, F., Jiang, Y., Corbin, C. K., Kim, G. Y. E., Mui, D., Crowell, T., Acharya, A., Mony, S., Punnathanam, S., McKeown, J., Smith, M., Lin, S., Milstein, A., Schulman, K., Hom, J., Pfeffer, M. A., Pham, T. D., Svec, D., Chu, W., Shieh, L., Sharp, C., Ma, S. P., & Chen, J. H.
Affiliations
  • Stanford Medicine (US)
  • Stanford University (US)
  • SmarterDx, New York, NY, USA
  • Johns Hopkins University (US)
  • Stanford Health Care (US)
  • Technology and Digital Solutions, Stanford Health Care, Palo Alto, CA
Venue
NEJM AI
Published
2026-06-25
Volume
3
Issue
7
ISSN
2836-9386 (ISSN-L)
OpenAlex
W7165639712
Cited by
0
References
18
FWCI
0.00
Type
journal article
Language
en
Open access
Locations
  • NEJM AI (publishedVersion)
  • PubMed (publishedVersion)
  • PubMed Central (submittedVersion)
Funders
  • Agency for Science, Technology and Research
  • National Institutes of Health (UM1TR004921)
  • Stanford Bio-X
  • National Institute of Allergy and Infectious Diseases (R01AI178121)
  • National Institute of Neurological Disorders and Stroke (U01NS134358)
  • National Center for Advancing Translational Sciences
Topics
  • Health Sciences › Health Professions › Health Information Management › Electronic Health Records Systems
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
Keywords
clinical decision support, laboratory test utilization, complete blood count, electronic health record alerts, machine learning prediction, inpatient laboratory testing, implementation lessons learned, model governance, stakeholder engagement, user interface design, health care costs

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.