Machine-Learning Prediction and Reducing Overdoses With EHR Nudges

NCT06806163 | Recruiting DMC

• 2 other identifiers: STUDY22040068R01DA044985-04
• Study Type: interventional
• Target: 1,350
• Locations: 1 country
• Sites: 1

Brief Summary

The goal of this cluster randomized clinical trial is to test a clinician-targeted behavioral nudge intervention in the Electronic Health Record (EHR) for patients who are identified by a machine-learning based risk prediction model as having an elevated risk for an opioid overdose. The clinical trial will evaluate the effectiveness of providing a flag in the EHR to identify individuals at elevated risk with and without behavioral nudges/best practice alerts (BPAs) as compared to usual care by primary care clinicians. The primary goals of the study are to improve opioid prescribing safety and reduce overdose risk.

Conditions

Opioid Overdose; Opioid Use; Opioid Use Disorder; Opioids

Keywords

Machine learning; Risk prediction; Behavioral nudge; Best Practice Alert; Artificial intelligence; Primary care; Electronic Health Record Platforms; Opioid Overdose; Opioids

Outcome Measures

Primary Outcomes (1)

• Prescribing Practices Composite Score

  This outcome measures a composite of three prescribing practices associated with a reduced risk of opioid overdose. Each of the three is assigned a value of one point, with the composite score ranging from 0 to 3: 1. Naloxone prescription: Evidence of a prescription for naloxone. 2. Opioid dosage \< 50 morphine milligram equivalents (MME) per day: No prescriptions exceeding 50 MME/day during the measurement period. 3. No opioid and benzodiazepine overlap: No concurrent prescriptions for opioids and benzodiazepines. The composite score will be treated as a 3-point ordinal measure reflecting adherence to these prescribing practices.

  Assessed at 4 months following study enrollment (i.e., at 4 months after the first encounter in the study period. An encounter refers to the 1st primary care visit for a patient enrolled in the study.)

Secondary Outcomes (7)

• Prescribing Practices Composite Score--6 Month Measure

  Assessed at 6 months following study enrollment (i.e., at 6 months after the first encounter in the study period. An encounter refers to the 1st primary care visit for a patient enrolled in the study.)

• Active Naloxone Prescription

  Assessed at 4 & 6 months after study enrollment by reviewing data from 12 months preceding index date (i.e., at 4 & 6 months after enrollment). An active naloxone prescription is recorded if one exists at any point during the year before the index date.

• Average Daily Opioid Dosage > 50 MME

  Assessed at 4 and 6 months after study enrollment, based on the average daily MME calculated over the 7 days preceding the index date (i.e., 4 and 6 months after enrollment).

• Overlapping Opioid Benzodiazepine Prescriptions

  Assessed at 4 and 6 months after study enrollment, based on overlap occurring on the index date (i.e., 4 and 6 months after enrollment) or within the 28 days preceding the index date.

• Overlapping Opioid Benzodiazepine Prescriptions Where Average Daily Opioid MME > 50

  Assessed at 4 and 6 months after study enrollment, based on overlap occurring on the index date (i.e., 4 and 6 months after enrollment) or within the 28 days preceding the index date.

Study Arms (3)

Usual Care

ACTIVE COMPARATOR

Patients in the practices randomized to the Usual Care arm will receive standard care practice without change.

Behavioral: Usual Care

EHR-Embedded Elevated-Risk Flag

EXPERIMENTAL

An elevated-risk flag will be embedded in the EHR and prominently displayed in the chart during encounters for patients identified as having elevated-risk for opioid overdose.

Behavioral: EHR-Embedded Elevated-Risk Flag

EHR-Embedded Elevated-Risk Flag with Behavioral Nudges

EXPERIMENTAL

An elevated-risk flag will be embedded in the EHR and prominently displayed in the chart during encounters for patients identified as having elevated-risk for opioid overdose. This flag will be combined with a set of best practice alerts/behavioral nudges that will trigger when certain conditions are met during encounters with elevated-risk patients.

Behavioral: EHR-Embedded Elevated-Risk Flag with Behavioral Nudges

Interventions

EHR-Embedded Elevated-Risk Flag BEHAVIORAL

Clinicians seeing patients at elevated predicted risk will see a flag on the EHR 'storyboard' during in person or telephone encounters indicating the patient is at elevated predicted risk of opioid overdose. The clinician will have the option of including this information into their decision-making process when providing care. There will be no best practice alerts/behavioral nudges in this arm.

EHR-Embedded Elevated-Risk Flag

Usual Care BEHAVIORAL

Patients in the practices randomized to the Usual Care arm will receive standard care practice without change.

Usual Care

EHR-Embedded Elevated-Risk Flag with Behavioral Nudges BEHAVIORAL

Clinicians seeing patients at elevated predicted risk for opioid overdose will see a flag on the EHR storyboard indicating that the patient is at elevated predicted risk. Clinicians will also receive up to 4 best practice alerts/behavioral nudges during an in-person or telephone primary care encounter with elevated risk patients when certain requirements are met: 1) if the patient does not have an active naloxone prescription on their medication list, the clinicians will receive an active choice alert during any medication ordering to encourage naloxone prescription; 2) if the patient's opioid dosage is \>50 MME, OR they are ordered a new opioid prescription, OR they have an overlapping opioid and benzodiazepine prescription order, the clinicians will receive an accountable justification alert when the relevant order is entered.

Also known as: EHR-Embedded Elevated-Risk Flag with Best Practice Alerts (BPAs)

EHR-Embedded Elevated-Risk Flag with Behavioral Nudges

Eligibility Criteria

• Age: 18 Years+
• Sex: all
• Healthy Volunteers: No
• Age Groups: Adult (18-64), Older Adult (65+)

You may qualify if:

• Received an opioid prescription within the past year
• Age 18 years or older at the time of the opioid prescription
• At least one visit to an internal medicine or family care practice within the past year

You may not qualify if:

• Diagnosis of malignant cancer within the past year
• Enrollment in hospice care

Sponsors & Collaborators

• National Institute on Drug Abuse (NIDA): collaborator
• University of Pittsburgh: lead

Study Sites (1)

University of Pittsburgh

Pittsburgh, Pennsylvania, 15213, United States

RECRUITING

Related Publications (9)

• Lo-Ciganic WH, Huang JL, Zhang HH, Weiss JC, Wu Y, Kwoh CK, Donohue JM, Cochran G, Gordon AJ, Malone DC, Kuza CC, Gellad WF. Evaluation of Machine-Learning Algorithms for Predicting Opioid Overdose Risk Among Medicare Beneficiaries With Opioid Prescriptions. JAMA Netw Open. 2019 Mar 1;2(3):e190968. doi: 10.1001/jamanetworkopen.2019.0968.

  PMID: 30901048 BACKGROUND

• Lo-Ciganic WH, Huang JL, Zhang HH, Weiss JC, Kwoh CK, Donohue JM, Gordon AJ, Cochran G, Malone DC, Kuza CC, Gellad WF. Using machine learning to predict risk of incident opioid use disorder among fee-for-service Medicare beneficiaries: A prognostic study. PLoS One. 2020 Jul 17;15(7):e0235981. doi: 10.1371/journal.pone.0235981. eCollection 2020.

  PMID: 32678860 BACKGROUND

• Lo-Ciganic WH, Donohue JM, Hulsey EG, Barnes S, Li Y, Kuza CC, Yang Q, Buchanich J, Huang JL, Mair C, Wilson DL, Gellad WF. Integrating human services and criminal justice data with claims data to predict risk of opioid overdose among Medicaid beneficiaries: A machine-learning approach. PLoS One. 2021 Mar 18;16(3):e0248360. doi: 10.1371/journal.pone.0248360. eCollection 2021.

  PMID: 33735222 BACKGROUND

• Lo-Ciganic WH, Donohue JM, Yang Q, Huang JL, Chang CY, Weiss JC, Guo J, Zhang HH, Cochran G, Gordon AJ, Malone DC, Kwoh CK, Wilson DL, Kuza CC, Gellad WF. Developing and validating a machine-learning algorithm to predict opioid overdose in Medicaid beneficiaries in two US states: a prognostic modelling study. Lancet Digit Health. 2022 Jun;4(6):e455-e465. doi: 10.1016/S2589-7500(22)00062-0.

  PMID: 35623798 BACKGROUND

• Guo J, Gellad WF, Yang Q, Weiss JC, Donohue JM, Cochran G, Gordon AJ, Malone DC, Kwoh CK, Kuza CC, Wilson DL, Lo-Ciganic WH. Changes in predicted opioid overdose risk over time in a state Medicaid program: a group-based trajectory modeling analysis. Addiction. 2022 Aug;117(8):2254-2263. doi: 10.1111/add.15878. Epub 2022 Apr 3.

  PMID: 35315173 BACKGROUND

• Hulsey E, Hershey TB, Parker LS, Kuza C, Fedro-Byrom S, Gellad WF. Overdose Risk Prediction Algorithms: The Need for a Comprehensive Legal Framework. Health Affairs Forefront. 2022 November 22. doi: 10.1377/forefront.20221118.549875.

  BACKGROUND

• Gellad WF, Yang Q, Adamson KM, Kuza CC, Buchanich JM, Bolton AL, Murzynski SM, Goetz CT, Washington T, Lann MF, Chang CH, Suda KJ, Tang L. Development and validation of an overdose risk prediction tool using prescription drug monitoring program data. Drug Alcohol Depend. 2023 May 1;246:109856. doi: 10.1016/j.drugalcdep.2023.109856. Epub 2023 Mar 27.

  PMID: 37001323 BACKGROUND

• Nguyen K, Wilson DL, Diiulio J, Hall B, Militello L, Gellad WF, Harle CA, Lewis M, Schmidt S, Rosenberg EI, Nelson D, He X, Wu Y, Bian J, Staras SAS, Gordon AJ, Cochran J, Kuza C, Yang S, Lo-Ciganic W. Design and development of a machine-learning-driven opioid overdose risk prediction tool integrated in electronic health records in primary care settings. Bioelectron Med. 2024 Oct 18;10(1):24. doi: 10.1186/s42234-024-00156-3.

  PMID: 39420438 BACKGROUND

• Militello LG, Diiulio J, Wilson DL, Nguyen KA, Harle CA, Gellad W, Lo-Ciganic WH. Using human factors methods to mitigate bias in artificial intelligence-based clinical decision support. J Am Med Inform Assoc. 2025 Feb 1;32(2):398-403. doi: 10.1093/jamia/ocae291.

  PMID: 39569464 BACKGROUND

Related Links

• Center for Pharmaceutical Policy \& Prescribing website

MeSH Terms

Conditions

Opiate Overdose; Opioid-Related Disorders

Condition Hierarchy (Ancestors)

Drug Overdose; Prescription Drug Misuse; Drug Misuse; Substance-Related Disorders; Chemically-Induced Disorders; Narcotic-Related Disorders; Mental Disorders

Study Officials

• Walid F Gellad, MD, MPH

  University of Pittsburgh Center for Pharmaceutical Policy and Prescribing

  PRINCIPAL INVESTIGATOR

Central Study Contacts

Lead Research Program Coordinator, CP3

CONTACT

(412) 692-4889; cp3@pitt.edu

Study Design

• Study Type: interventional
• Phase: not applicable
• Allocation: RANDOMIZED
• Masking: SINGLE
• Who Masked: INVESTIGATOR
• Purpose: HEALTH SERVICES RESEARCH
• Intervention Model: PARALLEL
• Sponsor Type: OTHER
• Responsible Party: PRINCIPAL INVESTIGATOR
• PI Title: Professor of Medicine

Model Details: This interventional study will be a cluster randomized trial across UPMC primary care practices. Practices will be randomized into one of 3 clinician-targeted intervention groups: 1. Usual care 2. Elevated-risk flag only 3. Elevated-risk flag combined with behavioral nudge alerts. The electronic health record (EHR) based intervention will be applied in the participating practices for clinicians whose patients have been identified as elevated-risk through the risk prediction algorithm.

Study Record Dates

• First Submitted: January 31, 2025
• First Posted: February 3, 2025
• Study Start: March 10, 2025
• Primary Completion (Estimated): May 31, 2026
• Study Completion (Estimated): May 31, 2026
• Last Updated: March 20, 2026

Record last verified: 2025-10

Locations

US (1)