Identifying People at Risk for Suicide: Implementation of Screening for the Zero Suicide Initiative in Large Health Systems

Introduction

A recent report from the National Center for Health Statistics at the Centers for Disease Control and Prevention found that death from suicide increased by 35% across U.S. populations from 1999 to 2018.¹ It is estimated that for every suicide death there are 25 attempts,² suggesting multiple opportunities for prevention. Increasing suicide rates have prompted many national organizations to improve identification and intervention strategies for those at highest risk. For example, the Joint Commission released an official National Patient Safety Goal on suicide prevention to be implemented in all hospital settings by July 2020.³ In response to multiple agency recommendations regarding suicide prevention, the Suicide Prevention Resource Center was created to address suicide prevention throughout the U.S.⁴ As part of their efforts, in partnership with other collaborating national advocacy groups, they created the Zero Suicide Framework specific to health care settings.⁵

The Zero Suicide Framework is a flexible set of evidence-based clinical practices and implementation strategies encompassing seven domains (lead, train, identify, engage, treat, transition, improve) that are designed to mitigate suicide risk, enhance protective factors, and close gaps in health care that leave at-risk patients vulnerable.⁵ The domains are meant to be used as needed, depending upon the resources of any one health care system, and are designed as a set of recommendations to improve the quality of suicide prevention efforts. One of the key clinical practices of this Framework is “identification” which includes two components: screening for suicidal ideation and assessment of suicide risk including intent and plans. Without identifying those at risk for suicide, there is no opportunity to enact the clinical practices of the framework: engage, treat, and transition.

Unfortunately, with few exceptions, screening for mental health conditions in general including suicidal ideation are not standard parts of care in U.S. health settings. Screening rates for depression in U.S. primary care settings are universally low, hovering at less than 5% of patients, generally^6,7 and rates of systematic screening specifically for suicide are even lower. A few health systems, most notably the Veteran’s Health Administration (VHA), the Department of Defense (DOD), and Henry Ford Health System (HFHS) have implemented robust and systematic suicide screening and risk assessment programs.^8–12 As a result, the VHA has achieved a suicide risk screening rate of 94.5%,⁹ and the U.S. Air Force significantly reduced suicide rates among its members.¹⁰ The “perfect depression care” program at HFHS^11,12 has resulted in some of the lowest rates of suicide in the nation and has been used as a model program for the Zero Suicide Framework.

Identifying people at risk for suicide depends in part on having measures and methods that are accurate determinants of risk. Clinical leaders are often concerned that asking about suicidal ideation will require significant resources to manage. One approach to address this concern is to use a staged screening and risk assessment approach as recommended for the identify domain of the Zero Suicide Framework: the first stage screens for suicidal ideation and the second stage determines the level to which the risk is imminent by assessing intent and plans. This staged approach helps to deploy resources to more accurately match level of risk with the appropriate evidence-based approach.¹³

The first stage can be done with specialized screening instruments such as The Ask Suicide-Screening Questions (ASQ);¹⁴ however most large health care systems are already using tools for depression screening/symptom severity assessment, such as the Patient Health Questionnaire (PHQ-9)^15,16 which includes a single question asking about the frequency in the last two weeks of a patient having “thoughts that you would be better off dead, or of hurting yourself in some way.” In this way health care systems can meet their national quality metrics for depression screening and care¹⁷ while also identifying patients at risk for suicide. The second stage entails assessing suicide risk severity and intent so that appropriate treatment plans can be enacted. For this stage, the Joint Commission³ has recommended many processes and measures including the Columbia-Suicide Severity Rating Scale (C-SSRS).¹⁸ The C-SSRS can be used for both screening and risk assessment (both steps in identify) but takes longer than the PHQ9 and providers must be trained to use the assessment portion of the instrument.

In addition to considering the characteristics of and the evidence base for screening and assessment tools, the feasibility of administration and the potential to integrate tools into the current workflows of health care systems are critical to address. One important consideration is how broadly to screen.^4,19,20 Some health care systems are taking a broad approach and screening everyone in a population regardless of known risk. For example, the national quality metrics for depression screening and care¹⁷ have a universal requirement for annual screening of all patients 12 years and older and many systems are using the PHQ9 to meet this requirement. Because it can be difficult to get “buy-in” to screen broadly (especially initially), many systems choose to begin with selective screening of groups known to be at higher risk for suicide. For instance, they may screen all persons with a depression diagnosis, as that is a group known to have higher suicide rate than those without this diagnosis.¹ This approach is consistent with the minimum requirement from the Joint Commission’s mandate for screening and assessment.³

A second major feasibility consideration is the setting in which screening will occur.^19,20 Most experts recommend screening in specialty behavioral health care where a more extensive process can be used to both screen and assess level of risk.³ This approach is likely to screen those most at risk and comprehensively address both elements of the identify clinical practice of the Zero Suicide Framework. Unfortunately, it is also likely to miss patients, as many people who die by suicide are not receiving specialty mental health care at that time or have not been diagnosed with any mental health conditions.²¹ For this reason, other settings have been proposed for suicide screening such as the emergency department, hospitals, and primary care.³ However, these settings have a number of barriers to screening for mental health conditions and suicide risk including providers who are not well trained or comfortable managing patients with suicide risk especially if assessment must be done in addition to screening, and the time and resources it takes to screen and assess risk for suicide.

The current study presents the rationale and results for the implementation of the screening clinical practice in the identify domain of the Zero Suicide Framework over a period of 2 – 8 years in five regions of a large integrated health care system, serving over 10 million members. Screening involved a two-stage approach with the PHQ9 and the C-SSRS. Findings reflect a learning health care system partnership which paired clinical leadership with embedded research scientists to work towards the long-term goal of “zero suicides”.²¹

Methods

Learning Healthcare System Partnership

Kaiser Permanente (KP) is a large, integrated health care delivery system, composed of eight healthcare regions across the U.S., which provides care and insurance and oversees hospitals and medical offices for a defined membership population. Each region is autonomous with respect to policies and operations; however, they all work together under the direction of a national entity, the Program Office, located in Oakland, CA. Many national clinical initiatives implemented across KP regions are supported by the Care Management Institute²³ which resides in the Program Office. Operational leaders and research scientists from each KP region are chosen to represent a topic area for these national initiatives (e.g. suicide risk screening clinical workflow developers) and are responsible for enacting the goals of the larger group in their respective regions.

Implementation of the Zero Suicide Framework at KP, led by the Care Management Institute, first began in 2014, and then gained traction in 2016. Traction came with a growing awareness of the Zero Suicide Framework and an interest in the significant outcomes in reducing suicide mortality demonstrated by other organizations in the health care industry.^9–12 One KP region began assessing risk for suicide in the second quarter (Q2) 2012, serving as an early adoption mentor for the other regions who followed in 2017 and 2018. To provide a coordinated approach to implementation, an interregional Suicide Prevention Learning Collaborative workgroup was formed which included mental health leaders and champions from each KP region, risk management and patient safety partners, researchers, key external collaborators and field experts, and KP members with lived experience of suicide themselves or in family members.

A two-step screening process, informed by research generated by KP researchers,^15,16,24 and the Joint Commission recommendations³ was adopted to identify individuals at risk of suicide attempt. Initially the 9^th question from the PHQ9 was completed and then was followed by the C-SSRS. In addition, it was decided that the initiative would target behavioral health and addiction medicine providers because of the expertise in these departments for assessing the level of risk and to provide safety and treatment planning. It also was decided to use the PHQ9 and C-SSRS with children as young as 10 years old given that suicidal ideation could not be reliably detected with commonly available instruments in younger children.²⁵ The screening practices for children and adolescents also included other measures such as the PHQ for adolescents (PHQA) and the PROMIS measures for depression followed by the C-SSRS, and in some cases a single step C-SSRS was implemented.

The actual workflows used to administer the PHQ9 and subsequent C-SSRS varied across KP regions. For the most part mental health specialty providers were already using the PHQ9 to assess depression symptoms, including suicidal ideation because of the national quality metrics for depression screening and care.¹⁷ However, only one region systematically used the C-SSRS to conduct the second screening for suicide risk severity before implementation of the Zero Suicide Framework. In order to facilitate the systematic use of the C-SSRS, the Care Management Institute led the creation of a version of the C-SSRS for the electronic medical record platforms used in each region. They also developed a standard training for staff on the background and administration of the C-SSRS in order to increase comfort level with the tool.

Systematic measurement of the efforts to identify patients at risk for suicide was critical to the success of adopting the Zero Suicide Framework. The Care Management Institute created a standardized set of metrics and processes to support cross-regional implementation of screening practices. The processes for obtaining and standardizing the data were built upon the work previously done by research scientists who were part of the KP regional implementation teams.^22,26 A formal agreement was created among the KP regions participating in the Zero Suicide Framework that stipulated what data would be abstracted, how and when it would be processed and shared with the Care Management Institute, in what format it would be shared, and with which stakeholders. Funding for the measurement process was provided by the National Institutes of Mental Health, the Care Management Institute, and the local KP regions.

Settings and Population

Five regions of KP were included in this study across four states: Washington, Oregon, California, and Colorado. All eight regions agreed to consider implementation of the Zero Suicide Initiative, however, only five were regularly reporting on their efforts at the time of this analysis (please see reasons in the discussion). To describe the systems in which screening was implemented, patients with continuous membership in these regions in 2019 were included (n = 9,948,080). All regions were participating in the Mental Health Research Network (MHRN), a nationwide consortium of public-domain research centers based in large, not-for-profit health care systems in the U.S.²⁷ The institutional review boards for human subjects in each region approved all study procedures. Informed consent was not required due to the retrospective nature and minimal risks of the study.

Data and Sources

All screening and membership data were abstracted from electronic medical record sources and other secure databases used to collect patient-reported outcomes in the departments of psychiatry and addiction medicine. Specifically, all responses to the PHQ9 and C-SSRS were obtained between 04/01/2012 and 10/31/2019 and demographics including age, gender, race/ethnicity, insurance status, and neighborhood census-tract derived education were used from 2019. These data are organized in a virtual data warehouse (VDW) for all systems to facilitate population-based research.²⁶ Protected health information remains at each health care system, but sites apply common data definitions and formats to ensure equivalent deidentified data for analysis. Only frequencies are shared between institutions for analyses.

Analyses

Screening and membership data are presented as descriptive statistics. Means and standard deviations are used to summarize continuous variables and frequencies and percentages are used for categorical variables. The rate of C-SSRS administration was the key metric used to assess how each KP region implemented the screening clinical practice in the Zero Suicide Framework identify domain. The rate was defined as a C-SSRS completed within 2 business days of a positive 9^th item (at least 1 on a scale of 0-3) of the PHQ9/PHQA among patients 10 years and older seen in specialty behavioral health or addiction medicine departments.

Results

Descriptive statistics for the membership in 2019 across all five participating KP regions (n = 9,948,080) are provided in Table 1. The majority of KP members were between ages 20 to 64 years old (61.5%), female (51.8%), White (53.3% both Hispanic and non-Hispanic), had commercial insurance (70.7%), and had higher census-track estimated neighborhood education (56.7%). Table 2 presents demographics for the denominator for the C-SSRS administration rate in 2019 (positive response on the 9^th item of the PHQ9; n = 70,036). Most patients reporting positive suicidal ideation were adults 18 to 64 years old (82.7%), women (65.1%), had commercial insurance (66.0%), and were non-Hispanic White (57.3%).

Table 1.

Descriptive characteristics for patients who were continuously enrolled in one of five Kaiser Permanente regions during 2019 (n = 9,948,080).

Characteristic	n	%
Age (years)
≤ 19	2,192,469	22.0%
20 - 39	2,699,562	27.1%
40 – 64	3,423,963	34.4%
≥ 65	1,632,086	16.4%
Sex
Female	5,152,675	51.8%
Male	4,794,707	48.2%
Other/Unknown	698	0.0%
Race
White	5,301,995	53.3%
Asian	1,413,685	14.2%
Black	710,677	7.1%
Hawaiian/Pacific Islander	94,883	1.0%
American Indian/Alaskan Native	66,838	0.7%
Multiple Race/Other	74,509	0.7%
Unknown	2,285,499	23%
Hispanic Ethnicity (% Yes)	2,717,901	27.3%
Insurance
Commercial	7,035,047	70.7%
Medicare	1,679,763	16.9%
Medicaid	813,214	8.2%
Private	392,885	3.9%
Other	27,171	0.3%
Higher Neighborhood Education (% Yes)	5,640,561	56.7%

Table 2.

Descriptive statistics for all Patient Health Questionnaires (PHQ9) completed in psychiatry and addiction medicine in 2019 (n = 70,036) that indicated* a patient needed further assessment with a Columbia Suicide Severity Rating Scale (C-SSRS). Data are presented in aggregate for 5 Kaiser Permanente (KP) regions. This indication was a response of “some/most/almost all days in the last two weeks” to the 9^th item of the PHQ9, “How often in the last two weeks have you had thoughts that you would be better off dead, or of hurting yourself in some way?” Data are presented as sample size and percentage. The rates of C-SSRS administration by KP region are shown in Figure 1.

	n	%
Age (years)
10 to 17	4,940	7.1%
18 to 64	57,890	82.7%
≥ 65	7,206	10.3%
Sex
Female	45,569	65.1%
Male	24,436	34.9%
Insurance
Commercial	46,218	66.0%
Medicare	10,029	14.3%
Medicaid	9,206	13.1%
Private	4,437	6.3%
Other	181	0.3%
Race/Ethnicity
Non-Hispanic White	40,137	57.3%
Hispanic	15,230	21.7%
Non-Hispanic Black	5,159	7.4%
Asian	4,813	6.9%
Hawaiian/Pacific Islander	660	0.9%
American Indian/Alaskan Native	769	1.1%
Multiple Race/Other	476	0.7%
Unknown	2,783	4.0%

Figure 1 provides the proportion of C-SSRS screenings completed when the PHQ9 item 9 score was at least 1 over time starting in 2012 in each of the five KP regions. The pattern of C-SSRS screening rates varied greatly among regions. One region (A) was an early adopter and served as a guide for the processes implemented in other regions. Two regions started administration of the C-SSRS in 2017 (C,D) with one achieving 100% administration rates by the end of 2017 (C) and one achieving 85% by December 2018 (D). Finally, the last 2 of the 5 regions to implement the C-SSRS (B,E) began in January 2018 (B) and January 2019 (E) and were 85% (B) and 69% (E) respectively by October 2019.

Figure 1.

Rates of Columbia Suicide Severity Rating Scale (C-SSRS) administration in psychiatry and addiction medicine over time (April 2012 – October 2019) for each Kaiser Permanente region (A – E). Rates are calculated for all Patient Health Questionnaires (PHQ9) with a positive response (some/most/almost all days in the last two weeks) to the 9^th item of the PHQ9: “How often in the last two weeks have you had thoughts that you would be better off dead, or of hurting yourself in some way?”.

Discussion

A learning health care system approach²² to implementing the screening clinical practice in the Zero Suicide Framework identify domain, including strong collaborative partnerships between health care system leaders, providers and researchers, significantly increased screening rates from 0% to 82% across 5 regions of KP serving nearly 10 million patients in a period of less than three years. One region was able to complete the C-SSRS in 100% of patients with an indication of suicidal ideation as measured by the 9^th item of the PHQ9 after a six-month implementation period.

Many factors could account for the differences between KP regions in rates of C-SSRS administration when indicated. One is the comparative size and complexity of the regions. The range in annual membership among the 5 KP regions was just over 600,000 to 4.7 million, nearly 1,300 to 9,400 physicians, and some regions served entire states. Although each region is part of the same company, each region’s organizational culture, structure, labor management partnership structure and function, and competing health system priorities are very different. Another was the implementation of the C-SSRS within the electronic medical record. Although this was created by the Care Management Institute for all the regions, local processes were still necessary to activate the C-SSRS for use by providers in each region. Depending upon the local teams’ relationships with their health information technology personnel this took several months. Another was the creation of specific clinical workflows, acceptable to all parties involved, that would clearly state what was to be done and who would do it when the screening indicated risk for suicide. Many of the regions would not implement screening, even if they had the technology to do so, without a clear plan for providing an evidence-based intervention for these patients.

Considerations for Implementation

We believe that there are six essential components necessary to successfully implement the screening clinical practice in the Zero Suicide Framework identify domain for specialty behavioral health and addiction medicine. The first is to obtain “buy-in” from the clinical systems in which screening will take place. One important component of this process is understanding the investments that clinical systems have already made in obtaining patient reported outcomes related to suicide prevention such as the PHQ9 for depression screening. The learning health care system approach²² is particularly important to this step as it embeds researchers in clinical teams which can then help those teams present evidence that will convince executive leaders in these systems to adopt effective practices and implementation strategies. This first step is also critical to the “lead” domain in the Zero Suicide Framework.⁵

The second is an assembly of all key decision-makers in a workgroup that is coordinated by staff dedicated to the initiative. Key decision-makers should include health technology experts, workflow consultants, clinical experts on screening for and treatment of suicidal ideation such as providers and research scientists, administrators who oversee and train staff, clinical informatics for data reporting, implementation scientists, and people with lived experience who have experience with health care system workflows regarding screening for suicide. The third is the use of validated instruments for screening.³ Careful consideration should be made for choosing evidence-based instruments that already align with the health care systems’ needs for patient-reported outcome quality metrics¹⁷ such as the PHQ9 which is recommended for depression screening and follow-up care.

The fourth is the automation, whenever possible, of the screening algorithm and collection of the patient reported outcomes. The PHQ9 and C-SSRS can be incorporated into web-based electronic medical record platforms that communicate with patients and store information in the medical record for clinicians to review. The fifth is the creation of clear workflows that designate responsibility for various clinical responses, depending upon the outcome of the scoring algorithm. For example, when a patient indicates clear intent for suicide, to whom does that person speak, what is said in the conversation, when must this conversation take place, and how is this process documented? Even though we only describe the results of the KP efforts to implement the screening clinical practice of the identify domain, in reality, the other domains (lead, train, engage, treat, improve) must also be in place for screening to succeed.

And finally, the sixth recommendation is to plan for performance feedback involving a standardized set of metrics created with benchmarks for screening that can be distributed to all clinical partners involved in the work. A strategy for mitigating poor performance should also be clearly articulated including who will be responsible for the remediation and how it will be done. This will also address the “improve” domain of the Zero Suicide Framework.

Limitations

One of the major limitations of the study is that data were only presented for specialty behavioral health settings. Most experts³ recommend screening in specialty behavioral health care where a more extensive process can be used to both screen and assess level of risk. Unfortunately, it is also likely to miss patients, as many people who die by suicide are not receiving specialty mental health care at that time.^21,24 There are well-documented racial and ethnic disparities in specialty mental health treatment initiation,^28–32 and implementation of screening in non-specialty care settings such as primary care might be better at identifying risk for some groups.^24,33 Rural residents, too, are more likely to have mental health-related office visits with a primary care provider than with a psychiatrist.³³ Because of these important issues efforts are underway to use the learnings from the roll out of screening in specialty mental health and addiction medicine to screen in KP emergency departments and primary care settings. Recent research highlights a successful approach for increasing culturally and linguistically appropriate depression screening which could be applied to screening and assessment for suicide risk.³⁴

Another limitation of the study was that other health care organizations were not included, such as Federally Qualified Health Centers, solo practioner medical offices, or other health care delivery models. An advantage of KP is that it serves a racially and ethnically diverse population, with some regions being majority Hispanic/Non-Hispanic Black. The insurance coverage of the membership is also very diverse including high deductible plans and limited access to specialty mental health care. We believe that as long as any health care setting can meet the requirements for electronic medical records as outlined in the Affordable Care Act,³⁵ and they have an infrastructure for creating national quality metrics such as those required by the National Center for Quality Assurance,¹⁷ the processes we report in this study could be implemented across a variety of settings.

Future Directions

Expanding suicide screening and assessment within health care settings promises the possibility of saving lives and of greatly reducing the burden of suicide within health care system populations. Some recent studies suggest that health care systems could improve the efficiency of screening and target screening to those most in need. For instance, several research teams have developed suicide identification algorithms that use data from electronic health records, pharmacy data and other health system information to estimate which people are at greater risk of suicide.³⁶ This approach shows promise, but there are a number of unanswered questions about how to integrate this approach into health care systems. Other studies have explored whether genetic markers could identify persons at risk,³⁷ but to date there have not been any strong markers that have emerged.

The changes in care with the advent of COVID-19 have dramatically increased the use of virtual visits and many health care systems plan to continue this expansion of telehealth. Several studies have tested the use of computerized or remote suicide screening.^38–40 There is some suggestion that this approach could successfully identify persons at risk, but these methods have primarily been used in University settings. Moreover, many tools that have been shown to be effective in clinical settings haven’t been tested in virtual delivery. Finally, little is known about how well persons identified through such methods could be effectively linked to needed care. We will likely see continuing expansion of these approaches in the future and additional research to help ensure that they are used effectively.

Key Points:

The following six actions are recommended to implement wide-scale screening for the risk for suicide:

• Health System “Buy-In”. Align screening for suicide risk with efforts clinical systems have already made in obtaining patient reported outcomes to assist executive leadership achieve goals in which they are already invested.

• Cross-disciplinary Collaboration. Assemble key decision-makers in a workgroup that is coordinated by staff dedicated to the initiative.

• Evidence-based Screening Tools. Use validated instruments recommended by the Joint Commission.

• Electronic Health Record-embedded Clinical Tools. Automate any screening algorithm and collection of patient-reported outcomes.

• Well-Defined Clinical Workflows. Define clear workflows that outline who is responsible for what aspects of the process of screening, depending upon patient response.

• Performance Feedback. Create and maintain a regular set of metrics with benchmarks for screening that can be distributed to all clinical partners involved in the work.

Synopsis:

The current study presents the rationale and results for the implementation of the “identify” step of the Zero Suicide Framework in 5 regions of Kaiser Permanente serving over 10 million members. Suicide risk screening rates increased from 0% to 82% in a period of less than three years. Findings reflect a learning healthcare system partnership which paired clinical leadership with embedded research scientists to work towards the long-term goal of zero suicides.