AI-driven Clinical Decision Support for Perioperative Blood Orders (SPATH)

Intelligent Clinical Decision Support for Preoperative Blood Management: A Cluster-Randomized Clinical Trial

20 million patients have surgery in the United States every year, with approximately 1 million of those patients requiring life-saving blood transfusion. Presurgical preparation for transfusion is important to allow for safe and timely transfusion during surgery; however, excessive preparation is unfortunately common, costly, and contributes to blood waste. This study aims to evaluate an intelligent clinical decision support system that helps clinicians prepare blood for patients who are likely to need it, while avoiding excessive preparation for patients who don't, potentially improving patient safety while reducing blood waste and healthcare costs.

Study Overview

• Status: Recruiting
• Conditions: Surgery
• Intervention / Treatment: Other: Usual care; Other: S-PATH clinical decision support system

• Study Type: Interventional
• Enrollment (Estimated): 50
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Sunny S Lou, MD, PhD; Phone Number: 314-362-5000; Email: slou@wustl.edu

Study Locations

• United States
  • Missouri
    • St Louis, Missouri, United States, 63110
      • Recruiting
      • Washington University / Barnes Jewish Hospital
      • Contact: Sunny S Lou, MD, PhD; Phone Number: 314-362-5000; Email: slou@wustl.edu

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

Description

Clinician Level Exclusion Criteria:

• Clinician (resident physician or advanced practice provider) who works at a preoperative assessment clinic

Clinician Level Exclusion Criteria:

• None

Patient Inclusion Criteria:

• Scheduled for surgery in one of the main operating room areas (non-remote) at Barnes Jewish Hospital
• Evaluated in-person at one of the preoperative assessment clinics affiliated with BJC Healthcare
• Have a valid S-PATH model prediction prior to their preoperative assessment clinic visit

Patient Exclusion Criteria:

• Pregnant
• Presence or history of red cell alloantibodies

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Screening
• Allocation: Randomized
• Interventional Model: Crossover Assignment
• Masking: None (Open Label)

Number of Arms

2

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Active Comparator: Usual care; Usual care for determining presurgical blood orders, including use of the institutional Maximum Surgical Blood Ordering Schedule (MSBOS)	Other: Usual care; Including use of the conventional Maximum Surgical Blood Ordering Schedule (MSBOS)
Experimental: S-PATH; Access to the S-PATH clinical decision support system	Other: S-PATH clinical decision support system; Access to the S-PATH electronic health record (EHR)-integrated clinical decision support system

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Frequency of patients with a type and screen order placed during the preoperative clinic assessment visit	This is a binary outcome at the patient / surgical case level, which will be aggregated across all patients in the study to produce an overall frequency. Placement of a type and screen order during the preoperative clinic assessment visit will be evaluated at the patient / surgical case level. This includes orders placed and collected during the preoperative clinic assessment visit, as well as orders signed during the preoperative clinic assessment visit or subsequent follow up care and held to be drawn on the day of surgery.	Decision made during the preoperative assessment clinic visit

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Frequency of a valid type and screen order at the start of surgery	This is a binary outcome at the patient / surgical case level, which will be aggregated across all patients in the study to produce an overall frequency. Valid type and screen order at the start of surgery (defined as resulted prior to 1h after anesthesia start). The 1 hour buffer is used to account for orders that were drawn prior to surgery but may not have resulted by the time the patient enters the operating room. This will include orders placed by the clinicians working in preoperative assessment clinic and any orders that may be placed by the day of surgery anesthesia or surgical teams. This will not include type and screen orders that have expired by the start of surgery. This is a secondary efficacy outcome that reflects type and screen ordering decisions by all members of the patient's care team.	Start of surgery (1 hour after Anesthesia Start)
Frequency of red cell transfusion during surgery	This is a binary outcome at the patient / surgical case level, which will be aggregated across all patients in the study to produce an overall frequency. Administration of allogeneic packed red blood cells during surgery will be retrieved from the electronic health record. Documentation of transfusion is mandatory and the electronic health record scan is used to confirm the correct patient. This is a secondary efficacy outcome, as lack of a type and screen may prevent discretionary intraoperative transfusion.	During surgery
Frequency of emergency release blood use during surgery	This is a binary outcome at the patient / surgical case level, which will be aggregated across all patients in the study to produce an overall frequency. Administration of emergency release allogeneic packed red blood cells will be retrieved from the electronic health record. These are either documented as volumes under the MTP tab or individually scanned as uncrossmatched red cells. This is the primary safety outcome of the study as emergency release blood may be administered if red cell transfusion is urgently indicated but a type and screen is not available. The frequency of this outcome will be reported overall across both groups, and stratified by whether the patient had an active type and screen at the start of surgery.	During surgery
Frequency of red cell transfusion during surgery without an active type and screen at the start of surgery	This is a binary outcome at the patient / surgical case level, which will be aggregated across all patients in the study to produce an overall frequency. This is the secondary safety outcome intended to capture false negative results, i.e., patients who required transfusion during surgery but did not have a preoperative type and screen. This outcome will be collected electronically from the EHR and will capture patients for whom the type and screen was resulted after the start of surgery (i.e., after 1h after anesthesia start), suggesting it was collected after the patient entered the operating room	During surgery
Frequency of transfusion reaction	This is a binary outcome at the patient / surgical case level, which will be aggregated across all patients in the study to produce an overall frequency. Transfusion reactions are reported by clinical teams to the blood bank, and the transfusion medicine service investigates and classifies each transfusion reaction. These investigation reports will be transmitted to the study team for adverse event reporting and outcome collection. This is also a secondary safety outcome for the study. Transfusion reactions will be reported stratified by category, as hemolytic transfusion reactions are the most relevant to the trial.	From time of surgery to hospital discharge or 30 days after surgery

Other Outcome Measures

Outcome Measure	Measure Description	Time Frame
Viewing S-PATH predictions (Implementation outcome)	Clinician-level measure of the frequency of viewing S-PATH predictions, retrieved from EHR audit log data	Through study completion, an average of 1 year
Acceptance of S-PATH recommendation (Implementation outcome)	Clinician-level measure of the frequency of the clinician's blood order during the preoperative clinic visit (1' outcome) matching S-PATH's recommendation	Through study completion, an average of 1 year
Cost assessment	Estimate of the cost savings associated with S-PATH use	Through study completion, an average of 1 year

Collaborators and Investigators

• Sponsor: Washington University School of Medicine
• Collaborators: National Heart, Lung, and Blood Institute (NHLBI)

Publications and helpful links

General Publications

• Yang P, Zijlstra EP, Hall BL, Gregory SH, Jackups R Jr, Li J, Abraham J, Lou SS. Challenges in reliable preoperative blood ordering: A qualitative interview study. Transfusion. 2024 Oct;64(10):1889-1898. doi: 10.1111/trf.18012. Epub 2024 Sep 16.
• Lou SS, Liu H, Lu C, Wildes TS, Hall BL, Kannampallil T. Personalized Surgical Transfusion Risk Prediction Using Machine Learning to Guide Preoperative Type and Screen Orders. Anesthesiology. 2022 Jul 1;137(1):55-66. doi: 10.1097/ALN.0000000000004139.
• Lou SS, Liu Y, Cohen ME, Ko CY, Hall BL, Kannampallil T. National Multi-Institutional Validation of a Surgical Transfusion Risk Prediction Model. J Am Coll Surg. 2024 Jan 1;238(1):99-105. doi: 10.1097/XCS.0000000000000874. Epub 2023 Sep 22.
• Lou SS, Kumar S, Goss CW, Avidan MS, Kheterpal S, Kannampallil T; Multicenter Perioperative Outcomes Group. Multicenter Validation of a Machine Learning Model for Surgical Transfusion Risk at 45 US Hospitals. JAMA Netw Open. 2025 Jun 2;8(6):e2517760. doi: 10.1001/jamanetworkopen.2025.17760.

Study record dates

Study Major Dates

• Study Start (Actual): October 28, 2025
• Primary Completion (Estimated): December 1, 2026
• Study Completion (Estimated): January 1, 2027

Study Registration Dates

• First Submitted: October 28, 2025
• First Submitted That Met QC Criteria: October 29, 2025
• First Posted (Actual): November 3, 2025

Study Record Updates

• Last Update Posted (Actual): February 27, 2026
• Last Update Submitted That Met QC Criteria: February 23, 2026
• Last Verified: February 1, 2026

More Information

Terms related to this study

• Keywords: anesthesia; machine learning; surgery; artificial intelligence; clinical decision support; perioperative blood management; pretransfusion testing
• Other Study ID Numbers: 202506199; K23HL166880 (U.S. NIH Grant/Contract)

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES

IPD Plan Description

For each clinician, the final dataset will include self-reported demographic information, exit interview transcripts, and behavioral and implementation outcomes relating to usage of the proposed clinical decision support (CDS) intervention.

For each patient, the final dataset will include clinical and safety outcomes for the study.

Trial materials, including consent forms, protocols, and code book for qualitative analysis will also be shared.

To anonymize the data, all clinicians and patients will be assigned random identifiers to replace clinician names and patient medical record numbers. No dates for surgical encounters will be retained. All other incidental identifiers (such as mentions of care team member names, encounter numbers, or hospital units) will be stripped.

• IPD Sharing Time Frame: Data will be made available within 12 months of the grant period ending or on publication of a manuscript using the data, whichever comes first. Data will be retained in the WashU Libraries Open Scholarship Digital Research Materials Repository for at least 10 years.
• IPD Sharing Access Criteria: The proposed clinical trial will involve a near-complete sampling of clinicians who work at the host institution's preoperative assessment clinic. Even with removal of identifiers, we believe it would be difficult to protect the identities of clinician participants given the type of demographic data collected and the very restricted setting in which recruitment will occur. In addition, even if patient data is anonymized, it may be possible to reidentify the patients using metadata alone. Therefore, access controls will be required prior to sharing of the all data, including patient and clinician data. Specifically, data will be shared only under a data-sharing agreement according to institutional guidelines that provides for: (1) a commitment to using the data only for research purposes and not to identify any individual participant; (2) a commitment to securing the data using appropriate computer technology; and (3) a commitment to destroying the data after analyses are complete.
• IPD Sharing Supporting Information Type: STUDY_PROTOCOL; SAP; ICF; ANALYTIC_CODE; CSR

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No