Implementation of Suicide Risk Models in Health Systems

Evaluating Effectiveness and Implementation of a Risk Model for Suicide Prevention Across Health Systems

The goal of this clinical trial is to evaluate a suicide risk model in patients receiving behavioral health care treatment. The main question it aims to answer is: Does the implementation of the suicide risk model reduce suicide attempts? Researchers will compare the outcomes of patients identified by the model to those in a usual care group.

Suicide is a major public health concern in the United States; nearly 50,000 individuals die by suicide annually and almost 1.5 million attempt suicide. To date, identification of individuals at risk for suicide has relied on suicide risk screening practices, including using a variety of self-reported instruments. However, sensitivity of these measures is only moderate; more precise tools for identifying patients at risk for suicide are needed. Suicide risk models, developed by our team, incorporate health records data and historical self-report screening questionnaire responses to improve accuracy of risk prediction. Our models have outperformed traditional clinical screening and similar risk models for adults receiving care in outpatient mental health specialty settings. However, while statistically accurate, they have not been evaluated in real world care; whether the models actually increase identification or result in patients receiving more suicide prevention services, fewer crisis services, or making fewer suicide attempts is unknown. There is substantial clinical interest in implementing suicide risk models but little scientific evidence about the effectiveness of these models in real-world settings compared to standard screening practices alone. Additionally, there is almost no guidance for their implementation in healthcare. The proposed project leverages the NIMH-funded Mental Health Research Network (MHRN), a collaboration of large health systems with established clinical data infrastructure to support multi-site studies. MHRN members Henry Ford Health, Kaiser Permanente Northwest, and HealthPartners will participate in this project and collectively serve >170,000 behavioral health patients per year. The patient populations are diverse, including thousands of individuals with Medicaid and Medicare. Each of these systems has implemented a suicide prevention care model in their behavioral health departments, including robust suicide risk screening and assessment processes. However, none of these systems has implemented a suicide risk identification model. The proposed project includes a pragmatic trial approach with randomization of behavioral health clinics across the three participating health systems. It is innovative because it seeks to implement an MHRN suicide risk model (intervention) into each system's existing suicide prevention care model (usual care) to increase the reach and effectiveness of the suicide prevention care models. Sites will receive implementation planning support based on stakeholder feedback from preliminary studies and deliverables include an implementation planning tool kit to facilitate spread. This high-impact study has important clinical implications as health systems consider whether it makes sense to enhance their existing suicide prevention care models with a suicide risk model. It is timely because many health systems are advancing toward suicide risk model implementation without evidence to support this innovation.

Clinics are the unit of randomization, all behavioral health clinics will be randomized, over the course of three waves, to cross over from usual care to intervention (implementation of the suicide risk model).

The suicide attempt risk model uses documented histories of medical and psychiatric diagnoses, medications, and health service utilization to predict risk of a suicide attempt in the 90 days following an outpatient visit in behavioral health clinics.

The number and proportion of visits followed by any suicide attempt (ICD-10 diagnosis codes) occurring within 90 days of an index visit.

The number and proportion of visits identified by PHQ9 item 9 or the risk model or both, stratified by race/ethnicity, where the denominator is the number of visits in the study period.

The number and proportion of visits with a completed risk assessment (C-SSRS), stratified by race/ethnicity, where the denominator is the number of visits in the study period.

The number and proportion of visits with a documented safety plan, lethal means counseling, or caring contacts subsequent to the index encounter, where the denominator is the number of visits in the study period.

The number and proportion of visits with any contact with behavioral health within 14 days of the index encounter, where the denominator is the number of visits in the study period.

Inclusion Criteria: 18+ years old 1+ visit to a behavioral health clinic at participating sites Exclusion Criteria: None

We will make our documentation, research methods and protocol, data collection tools, and a de-identified dataset of data that underlie results in publications freely available, upon request, to interested researchers beginning 6 months after publication of the main analyses. Materials will be shared through the MHRN website or a secure file transfer. Creation of a deidentified dataset for sharing may include redaction of some information to prevent re-identification or because the data is proprietary. The de-identified dataset will be available for non-commercial research use to external investigators via a data-sharing agreement and under the auspices of the Site-PIs. Users must agree to the conditions of use governing access to the data. The study team will be available for support. Information related to errors in the data, future releases, and publication lists will also be shared with users.

Materials will be shared, upon request, to interested researchers beginning 6 months after publication of the main analyses for up to one year.

Materials will be shared with interested researchers through the MHRN website, data may be shared for secondary analyses through a secure file transfer site.

Ahmedani BK, Simon GE, Stewart C, Beck A, Waitzfelder BE, Rossom R, Lynch F, Owen-Smith A, Hunkeler EM, Whiteside U, Operskalski BH, Coffey MJ, Solberg LI. Health care contacts in the year before suicide death. J Gen Intern Med. 2014 Jun;29(6):870-7. doi: 10.1007/s11606-014-2767-3. Epub 2014 Feb 25.

Simon GE, Johnson E, Lawrence JM, Rossom RC, Ahmedani B, Lynch FL, Beck A, Waitzfelder B, Ziebell R, Penfold RB, Shortreed SM. Predicting Suicide Attempts and Suicide Deaths Following Outpatient Visits Using Electronic Health Records. Am J Psychiatry. 2018 Oct 1;175(10):951-960. doi: 10.1176/appi.ajp.2018.17101167. Epub 2018 May 24.

Hedegaard H, Curtin SC, Warner M. Increase in Suicide Mortality in the United States, 1999-2018. NCHS Data Brief. 2020 Apr;(362):1-8.

Yarborough BJH, Ahmedani BK, Boggs JM, Beck A, Coleman KJ, Sterling S, Schoenbaum M, Goldstein-Grumet J, Simon GE. Challenges of Population-based Measurement of Suicide Prevention Activities Across Multiple Health Systems. EGEMS (Wash DC). 2019 Apr 12;7(1):13. doi: 10.5334/egems.277.

Rossom RC, Richards JE, Sterling S, Ahmedani B, Boggs JM, Yarborough BJH, Beck A, Lloyd K, Frank C, Liu V, Clinch SB, Patke LD, Simon GE. Connecting Research and Practice: Implementation of Suicide Prevention Strategies in Learning Health Care Systems. Psychiatr Serv. 2022 Feb 1;73(2):219-222. doi: 10.1176/appi.ps.202000596. Epub 2021 Jun 30.

Simon GE, Shortreed SM, Johnson E, Rossom RC, Lynch FL, Ziebell R, Penfold ARB. What health records data are required for accurate prediction of suicidal behavior? J Am Med Inform Assoc. 2019 Dec 1;26(12):1458-1465. doi: 10.1093/jamia/ocz136.

Simon GE, Rutter CM, Peterson D, Oliver M, Whiteside U, Operskalski B, Ludman EJ. Does response on the PHQ-9 Depression Questionnaire predict subsequent suicide attempt or suicide death? Psychiatr Serv. 2013 Dec 1;64(12):1195-202. doi: 10.1176/appi.ps.201200587.

Yarborough BJH, Stumbo SP. Patient perspectives on acceptability of, and implementation preferences for, use of electronic health records and machine learning to identify suicide risk. Gen Hosp Psychiatry. 2021 May-Jun;70:31-37. doi: 10.1016/j.genhosppsych.2021.02.008. Epub 2021 Mar 4.

Coleman KJ, Stewart CC, Bruschke C, et al. Identifying people at risk for suicide: Implementation of screening for the Zero Suicide Initiative in large health systems. Advances in Psychiatry and Behavioral Health. 2021;1(1):67-76.

National Action Alliance for Suicide Prevention. A prioritized research agenda for suicide prevention: An action plan to save lives. Rockville, MD. 2014.

https://www.jointcommission.org/-/media/tjc/documents/resources/patient-safety-topics/suicide-prevention/general_suicide_reduction_tools.pdf