Clinical Decision Support Guided by Risk Models in Suicide Screening: A Randomized Clinical Trial

Academic Background

Suicide prevention is a crucial topic in global public health, particularly in healthcare environments where identifying and intervening in suicide risks are top priorities. Traditional methods for identifying suicide risks rely mainly on patient self-reports, feedback from support networks, or face-to-face screening. However, these methods have certain limitations, such as patients being unwilling to disclose suicidal thoughts or limited healthcare resources hindering comprehensive screening of all patients. In recent years, with the development of big data and artificial intelligence technologies, statistical model-based tools for assessing suicide risk have been introduced into clinical practice to assist clinicians. However, the effectiveness of such tools within clinical decision support systems (CDSs) has not been sufficiently validated.

This study aims to evaluate the effectiveness of a risk model-based CDS in assessing suicide risk, particularly comparing “interruptive” and “noninterruptive” CDS designs in clinical practice. Interruptive CDS uses pop-up windows to remind clinicians to conduct further assessments of suicide risk, whereas noninterruptive CDS displays static icons or visual cues to present risk information. The study seeks to verify, through a randomized clinical trial (RCT), which CDS design is more effective, thereby providing scientific evidence for future suicide prevention strategies.

Source of the Paper

This paper, written by Colin G. Walsh et al., represents a collaboration among researchers from institutions such as Vanderbilt University Medical Center. The article was published on January 3, 2025, in the journal JAMA Network Open and is titled “Risk Model–Guided Clinical Decision Support for Suicide Screening: A Randomized Clinical Trial.” The study received funding from institutions such as the National Institute of Mental Health (NIMH) in the United States.

Study Design and Methods

Study Overview

This research is a comparative effectiveness randomized clinical trial aimed at evaluating the application of CDS systems guided by risk models in suicide risk assessment. The study was conducted from August 17, 2022, to February 16, 2023, at Vanderbilt University Medical Center’s neurology outpatient clinics. The participants were patients undergoing routine care. The primary objective was to compare the effectiveness of interruptive versus noninterruptive CDS systems in encouraging clinicians to conduct in-person suicide risk assessments.

1. Participants and Randomization

The study included 561 patients involving 596 clinical encounters. Patients were randomized at a 1:1 ratio into interruptive or noninterruptive CDS groups via the electronic health record (EHR) system during appointments. The average age of the participants was 59.3 years, and 52% of them were women.

2. Intervention Design

• Interruptive CDS: When a patient’s suicide risk score exceeded the predefined threshold, the system displayed a pop-up reminder alongside a patient summary panel icon. Clinicians could dismiss the pop-up while the icon remained visible.
• Noninterruptive CDS: The system displayed an “Elevated Suicide Risk Score” icon in the patient summary panel without triggering a pop-up window. Hovering over the icon provided further details, similar to the interruptive system.

3. Data Collection and Analysis

The primary outcome was the proportion of clinical encounters where clinicians decided to conduct face-to-face suicide risk assessments. Secondary outcomes included the occurrence of suicidal ideation, suicide attempts, and baseline screening rates from the prior year. Manual medical record review was conducted to determine whether suicide risk assessments were documented.

Main Results

1. Primary Outcome

In the interruptive CDS group, 42% of encounters (¹²¹⁄₂₈₉) led to clinicians deciding to assess suicide risk, compared to only 4% of encounters (¹²⁄₃₀₇) in the noninterruptive CDS group. After adjusting for the clustering effect among individual clinicians, the interruptive CDS increased the likelihood of in-person assessment by 17.7 times compared to the noninterruptive system (95% CI: 6.42–48.79; p < .001). Screening rates in the interruptive group were significantly higher than the baseline screening rate of 8% observed in the prior year.

2. Secondary Outcomes

Although the noninterruptive CDS group achieved a higher proportion of documentation among those who chose to screen (92%, ¹¹⁄₁₂ encounters) compared to the interruptive group (52%, ⁶³⁄₁₂₁ encounters), the absolute documentation rate was higher in the interruptive group (22%, ⁶³⁄₂₈₉) than in the noninterruptive group (4%, ¹¹⁄₃₀₇; p < .001). During the trial, neither group reported any instances of suicidal ideation or suicide attempts.

Conclusion

This study demonstrates that interruptive CDS significantly outperforms noninterruptive CDS in prompting clinicians to initiate face-to-face suicide risk assessments. The findings support the use of risk-model-guided CDS, particularly in healthcare settings where universal screening is not currently practiced. Further research is needed to validate the long-term impact of such CDS systems on reducing suicide attempts and behaviors.

Research Highlights

1. Innovation: This study is the first randomized clinical trial to evaluate the effectiveness of risk-model-guided CDS systems in suicide prevention, filling a critical gap in the field.
2. Practicality: The results show that interruptive CDS can significantly increase the proportion of clinicians conducting in-person suicide risk assessments, offering a practical tool for clinical settings.
3. Scientific Value: The study provides actionable evidence to inform future suicide prevention strategies and offers valuable insights into integrating artificial intelligence technologies into clinical workflows.

Research Implications and Value

The scientific value of this study lies in its rigorous evaluation of the effectiveness of risk-model-guided CDS systems in suicide prevention through a randomized clinical trial. Interruptive CDS was shown to significantly increase the likelihood of clinicians conducting suicide risk assessments, particularly in settings lacking universal screening protocols. These findings provide critical evidence supporting the integration of artificial intelligence technologies into clinical workflows for suicide prevention.

Additionally, the study highlights key challenges in CDS design, such as the potential for alert fatigue with interruptive systems despite their effectiveness. Future research should focus on optimizing CDS designs to balance improved screening rates with minimal disruption to clinical workflows.

This research provides significant contributions to the field of suicide prevention and offers valuable recommendations for clinical practices and public health policy-making.