Implementing Machine Learning Models for Suicide Risk Prediction in Clinical Practice: Focus Group Study With Hospital Providers

Adolescent, Parent, and Provider Perceptions of a Predictive Algorithm to Identify Adolescent Suicide Risk in Primary Care

OBJECTIVE

To understand adolescent, parent, and provider perceptions of a machine learning algorithm for detecting adolescent suicide risk prior to its implementation primary care.

METHODS

We conducted semi-structured, qualitative interviews with adolescents (n = 9), parents (n = 12), and providers (n = 10; mixture of behavioral health and primary care providers) across two major health systems. Interviews were audio recorded and transcribed with analyses supported by use of NVivo. A codebook was developed combining codes derived inductively from interview transcripts and deductively from implementation science frameworks for content analysis.

RESULTS

Reactions to the algorithm were mixed. While many participants expressed privacy concerns, they believed the algorithm could be clinically useful for identifying adolescents at risk for suicide and facilitating follow-up. Parents' past experiences with their adolescents' suicidal thoughts and behaviors contributed to their openness to the algorithm. Results also aligned with several key Consolidated Framework for Implementation Research domains. For example, providers mentioned barriers inherent to the primary care setting such as time and resource constraints likely to impact algorithm implementation. Participants also cited a climate of mistrust of science and health care as potential barriers.

CONCLUSIONS

Findings shed light on factors that warrant consideration to promote successful implementation of suicide predictive algorithms in pediatric primary care. By attending to perspectives of potential end users prior to the development and testing of the algorithm, we can ensure that the risk prediction methods will be well-suited to the providers who would be interacting with them and the families who could benefit.

Developing a clinical decision support system software prototype that assists in the management of patients with self-harm in the emergency department: protocol of the PERMANENS project

BACKGROUND

Self-harm presents a significant public health challenge. Emergency departments (EDs) are crucial healthcare settings in managing self-harm, but clinician uncertainty in risk assessment may contribute to ineffective care. Clinical Decision Support Systems (CDSSs) show promise in enhancing care processes, but their effective implementation in self-harm management remains unexplored.

METHODS

PERMANENS comprises a combination of methodologies and study designs aimed at developing a CDSS prototype that assists clinicians in the personalized assessment and management of ED patients presenting with self-harm. Ensemble prediction models will be constructed by applying machine learning techniques on electronic registry data from four sites, i.e., Catalonia (Spain), Ireland, Norway, and Sweden. These models will predict key adverse outcomes including self-harm repetition, suicide, premature death, and lack of post-discharge care. Available registry data include routinely collected electronic health record data, mortality data, and administrative data, and will be harmonized using the OMOP Common Data Model, ensuring consistency in terminologies, vocabularies and coding schemes. A clinical knowledge base of effective suicide prevention interventions will be developed rooted in a systematic review of clinical practice guidelines, including quality assessment of guidelines using the AGREE II tool. The CDSS software prototype will include a backend that integrates the prediction models and the clinical knowledge base to enable accurate patient risk stratification and subsequent intervention allocation. The CDSS frontend will enable personalized risk assessment and will provide tailored treatment plans, following a tiered evidence-based approach. Implementation research will ensure the CDSS' practical functionality and feasibility, and will include periodic meetings with user-advisory groups, mixed-methods research to identify currently unmet needs in self-harm risk assessment, and small-scale usability testing of the CDSS prototype software.

DISCUSSION

Through the development of the proposed CDSS software prototype, PERMANENS aims to standardize care, enhance clinician confidence, improve patient satisfaction, and increase treatment compliance. The routine integration of CDSS for self-harm risk assessment within healthcare systems holds significant potential in effectively reducing suicide mortality rates by facilitating personalized and timely delivery of effective interventions on a large scale for individuals at risk of suicide.

Impact of Automated Prognostication on Traumatic Brain Injury Care: A Focus Group Study

BACKGROUND

Prognosticating outcomes for traumatic brain injury (TBI) patients is challenging due to the required specialized skills and variability among clinicians. Recent attempts to standardize TBI prognosis have leveraged machine learning (ML) methodologies. This study evaluates the necessity and influence of ML-assisted TBI prognostication through healthcare professionals' perspectives via focus group discussions.

METHODS

Two virtual focus groups included ten key TBI care stakeholders (one neurosurgeon, two emergency clinicians, one internist, two radiologists, one registered nurse, two researchers in ML and healthcare and one patient representative). They answered six open-ended questions about their perceptions and potential ML use in TBI prognostication. Transcribed focus group discussions were thematically analyzed using qualitative data analysis software.

RESULTS

The study captured diverse perceptions and interests in TBI prognostication across clinical specialties. Notably, certain clinicians who currently do not prognosticate expressed an interest in doing so independently provided they had access to ML support. Concerns included ML's accuracy and the need for proficient ML researchers in clinical settings. The consensus suggested using ML as a secondary consultation tool and promoting collaboration with internal or external research resources. Participants believed ML prognostication could enhance disposition planning and standardize care regardless of clinician expertise or injury severity. There was no evidence of perceived bias or interference during the discussions.

CONCLUSION

Our findings revealed an overall positive attitude toward ML-based prognostication. Despite raising multiple concerns, the focus group discussions were particularly valuable in underscoring the potential of ML in democratizing and standardizing TBI prognosis practices.

Factors influencing clinician and patient interaction with machine learning-based risk prediction models: a systematic review

Machine learning (ML)-based risk prediction models hold the potential to support the health-care setting in several ways; however, use of such models is scarce. We aimed to review health-care professional (HCP) and patient perceptions of ML risk prediction models in published literature, to inform future risk prediction model development. Following database and citation searches, we identified 41 articles suitable for inclusion. Article quality varied with qualitative studies performing strongest. Overall, perceptions of ML risk prediction models were positive. HCPs and patients considered that models have the potential to add benefit in the health-care setting. However, reservations remain; for example, concerns regarding data quality for model development and fears of unintended consequences following ML model use. We identified that public views regarding these models might be more negative than HCPs and that concerns (eg, extra demands on workload) were not always borne out in practice. Conclusions are tempered by the low number of patient and public studies, the absence of participant ethnic diversity, and variation in article quality. We identified gaps in knowledge (particularly views from under-represented groups) and optimum methods for model explanation and alerts, which require future research.

External Validation and Updating of a Statistical Civilian-Based Suicide Risk Model in US Naval Primary Care

Importance

Suicide remains an ongoing concern in the US military. Statistical models have not been broadly disseminated for US Navy service members.

Objective

To externally validate and update a statistical suicide risk model initially developed in a civilian setting with an emphasis on primary care.

Design, Setting, and Participants

This retrospective cohort study used data collected from 2007 through 2017 among active-duty US Navy service members. The external civilian model was applied to every visit at Naval Medical Center Portsmouth (NMCP), its NMCP Naval Branch Health Clinics (NBHCs), and TRICARE Prime Clinics (TPCs) that fall within the NMCP area. The model was retrained and recalibrated using visits to NBHCs and TPCs and updated using Department of Defense (DoD)-specific billing codes and demographic characteristics, including expanded race and ethnicity categories. Domain and temporal analyses were performed with bootstrap validation. Data analysis was performed from September 2020 to December 2022.

Exposure

Visit to US NMCP.

Main Outcomes and Measures

Recorded suicidal behavior on the day of or within 30 days of a visit. Performance was assessed using area under the receiver operating curve (AUROC), area under the precision recall curve (AUPRC), Brier score, and Spiegelhalter z-test statistic.

Results

Of the 260 583 service members, 6529 (2.5%) had a recorded suicidal behavior, 206 412 (79.2%) were male; 104 835 (40.2%) were aged 20 to 24 years; and 9458 (3.6%) were Asian, 56 715 (21.8%) were Black or African American, and 158 277 (60.7%) were White. Applying the civilian-trained model resulted in an AUROC of 0.77 (95% CI, 0.74-0.79) and an AUPRC of 0.004 (95% CI, 0.003-0.005) at NBHCs with poor calibration (Spiegelhalter P < .001). Retraining the algorithm improved AUROC to 0.92 (95% CI, 0.91-0.93) and AUPRC to 0.66 (95% CI, 0.63-0.68). Number needed to screen in the top risk tiers was 366 for the external model and 200 for the retrained model; the lower number indicates better performance. Domain validation showed AUROC of 0.90 (95% CI, 0.90-0.91) and AUPRC of 0.01 (95% CI, 0.01-0.01), and temporal validation showed AUROC of 0.75 (95% CI, 0.72-0.78) and AUPRC of 0.003 (95% CI, 0.003-0.005).

Conclusions and Relevance

In this cohort study of active-duty Navy service members, a civilian suicide attempt risk model was externally validated. Retraining and updating with DoD-specific variables improved performance. Domain and temporal validation results were similar to external validation, suggesting that implementing an external model in US Navy primary care clinics may bypass the need for costly internal development and expedite the automation of suicide prevention in these clinics.

The use of advanced technology and statistical methods to predict and prevent suicide

In the past decade, two themes have emerged across suicide research. First, according to meta-analyses, the ability to predict and prevent suicidal thoughts and behaviours is weaker than would be expected for the size of the field. Second, review and commentary papers propose that technological and statistical methods (such as smartphones, wearables, digital phenotyping and machine learning) might become solutions to this problem. In this Review, we aim to strike a balance between the pessimistic picture presented by these meta-analyses and the optimistic picture presented by review and commentary papers about the promise of advanced technological and statistical methods to improve the ability to understand, predict and prevent suicide. We divide our discussion into two broad categories. First, we discuss the research aimed at assessment, with the goal of better understanding or more accurately predicting suicidal thoughts and behaviours. Second, we discuss the literature that focuses on prevention of suicidal thoughts and behaviours. Ecological momentary assessment, wearables and other technological and statistical advances hold great promise for predicting and preventing suicide, but there is much yet to do.

Complex modeling with detailed temporal predictors does not improve health records-based suicide risk prediction

Suicide risk prediction models can identify individuals for targeted intervention. Discussions of transparency, explainability, and transportability in machine learning presume complex prediction models with many variables outperform simpler models. We compared random forest, artificial neural network, and ensemble models with 1500 temporally defined predictors to logistic regression models. Data from 25,800,888 mental health visits made by 3,081,420 individuals in 7 health systems were used to train and evaluate suicidal behavior prediction models. Model performance was compared across several measures. All models performed well (area under the receiver operating curve [AUC]: 0.794-0.858). Ensemble models performed best, but improvements over a regression model with 100 predictors were minimal (AUC improvements: 0.006-0.020). Results are consistent across performance metrics and subgroups defined by race, ethnicity, and sex. Our results suggest simpler parametric models, which are easier to implement as part of routine clinical practice, perform comparably to more complex machine learning methods.

Association of HLA class II alleles with suicidal behavior in a Transylvanian population

Background: Suicide is a complex phenomenon determined by the interaction of various risk factors. The Major Histocompatibility Complex is the most polymorphic gene cluster of the entire human genome, being linked to both the regulation of the immune system and various psychiatric diseases. The aim of this study was to identify HLA-DQB1 and DRB1 alleles and genotypes susceptible to influence suicidal behavior.

Methods: We explored the association of HLA-DQB1 alleles with the suicidal behavior on a sample of 427 individuals (including 110 suicide attempters) from Transylvania, as well as the association of HLA-DRB1 alleles with the suicidal behavior on a sample of 271 individuals (including 50 suicide attempters), using the single specific primer-PCR (SSP-PCR) technique.

Results: We found that the HLA-DQB1*02, *03 and *06 alleles, the DQB1*02/*03, DQB1*02/*06, DRB1*12/*15 and DRB1*07/*13 genotypes, as well as the DQB1*06~DRB1*07 and DQB1*02~DRB1*13 haplotypes, were more frequent in suicide attempters. In contrast, the HLA-DQB1*04 and DQB1*13 alleles, the DQB1*02/*05 and DQB1*03/*05 genotypes and the DQB1*03~DRB1*13 haplotype were less frequent in the case group.

Conclusion: HLA-DQB1*02, *03 and *06 alleles and the DQB1*02/*03 and *02/*06 genotypes are susceptible to favor a suicide behavior, while the HLA-DQB1*04 and *13 alleles and the DQB1*02/*05 and *03/*05 genotypes were protective against such behavior. A similar analysis regarding the HLA-DRB1 alleles detected a possible risk for suicidal behavior among individuals possessing either the DRB1*12/*15 or the DRB1*07/*13 genotypes. DQB1*06~DRB1*07 and DQB1*02~DRB1*13 haplotypes were found susceptible to favor a suicidal behavior, while DQB1*03~DRB1*13 exhibited a protective influence.