Explainable artificial intelligence models for predicting risk of suicide using health administrative data in Quebec.

Suicide is a complex, multidimensional event, and a significant challenge for prevention globally. Artificial intelligence (AI) and machine learning (ML) have emerged to harness large-scale datasets to enhance risk detection. In order to trust and act upon the predictions made with ML, more intuitive user interfaces must be validated. Thus, Interpretable AI is one of the crucial directions which could allow policy and decision makers to make reasonable and data-driven decisions that can ultimately lead to better mental health services planning and suicide prevention. This research aimed to develop sex-specific ML models for predicting the population risk of suicide and to interpret the models. Data were from the Quebec Integrated Chronic Disease Surveillance System (QICDSS), covering up to 98% of the population in the province of Quebec and containing data for over 20,000 suicides between 2002 and 2019. We employed a case-control study design. Individuals were considered cases if they were aged 15+ and had died from suicide between January 1st, 2002, and December 31st, 2019 (n = 18339). Controls were a random sample of 1% of the Quebec population aged 15+ of each year, who were alive on December 31st of each year, from 2002 to 2019 (n = 1,307,370). We included 103 features, including individual, programmatic, systemic, and community factors, measured up to five years prior to the suicide events. We trained and then validated the sex-specific predictive risk model using supervised ML algorithms, including Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost) and Multilayer perceptron (MLP). We computed operating characteristics, including sensitivity, specificity, and Positive Predictive Value (PPV). We then generated receiver operating characteristic (ROC) curves to predict suicides and calibration measures. For interpretability, Shapley Additive Explanations (SHAP) was used with the global explanation to determine how much the input features contribute to the models' output and the largest absolute coefficients. The best sensitivity was 0.38 with logistic regression for males and 0.47 with MLP for females; the XGBoost Classifier with 0.25 for males and 0.19 for females had the best precision (PPV). This study demonstrated the useful potential of explainable AI models as tools for decision-making and population-level suicide prevention actions. The ML models included individual, programmatic, systemic, and community levels variables available routinely to decision makers and planners in a public managed care system. Caution shall be exercised in the interpretation of variables associated in a predictive model since they are not causal, and other designs are required to establish the value of individual treatments. The next steps are to produce an intuitive user interface for decision makers, planners and other stakeholders like clinicians or representatives of families and people with live experience of suicidal behaviors or death by suicide. For example, how variations in the quality of local area primary care programs for depression or substance use disorders or increased in regional mental health and addiction budgets would lower suicide rates.

A case-control study on predicting population risk of suicide using health administrative data: a research protocol.

INTRODUCTION: Suicide has a complex aetiology and is a result of the interaction among the risk and protective factors at the individual, healthcare system and population levels. Therefore, policy and decision makers and mental health service planners can play an important role in suicide prevention. Although a number of suicide risk predictive tools have been developed, these tools were designed to be used by clinicians for assessing individual risk of suicide. There have been no risk predictive models to be used by policy and decision makers for predicting population risk of suicide at the national, provincial and regional levels. This paper aimed to describe the rationale and methodology for developing risk predictive models for population risk of suicide. METHODS AND ANALYSIS: A case-control study design will be used to develop sex-specific risk predictive models for population risk of suicide, using statistical regression and machine learning techniques. Routinely collected health administrative data in Quebec, Canada, and community-level social deprivation and marginalisation data will be used. The developed models will be transformed into the models that can be readily used by policy and decision makers. Two rounds of qualitative interviews with end-users and other stakeholders were proposed to understand their views about the developed models and potential systematic, social and ethical issues for implementation; the first round of qualitative interviews has been completed. We included 9440 suicide cases (7234 males and 2206 females) and 661 780 controls for model development. Three hundred and forty-seven variables at individual, healthcare system and community levels have been identified and will be included in least absolute shrinkage and selection operator regression for feature selection. ETHICS AND DISSEMINATION: This study is approved by the Health Research Ethnics Committee of Dalhousie University, Canada. This study takes an integrated knowledge translation approach, involving knowledge users from the beginning of the process.

[Health Service Use in 24 Months Prior to Suicide in Quebecers with and without Diagnosed Mental and Substance Use Disorders]. / Consultations médicales et types de services de santé utilisés dans les deux années précédant le suicide auprès des Québécois diagnostiqués avec et sans troubles mentaux et troubles avec utilisation de substances.

Objectives Population based studies on linked health administrative databases (HADBs) characterizing those who die by suicide and their use of health services are rare. The objectives of this study were to describe the use of health services among people who died by suicide in Quebec, with and without previously receiving a mental disorder (MD) and dependencies (D) (MD/D) diagnosis, and (2) living Quebecers diagnosed with MD/D. Methods This study is based on an analysis of data from the Integrated Chronic Disease Surveillance System of Quebec (SISMACQ), which combines databases on outpatient medical and emergency services and hospitalizations. The population of the study consists of Quebecers aged 15 years and over and insured between April 1, 1996 and March 31, 2013 under the Régie de l'assurance maladie du Québec (RAMQ). For the purposes of this study, the 7 years preceding suicide were examined in the HADBDS for the three following cohort groups: (i) cases of suicide with a diagnosed MD/D in the 7 years preceding the date of death; (ii) cases of suicide without a MD/D diagnosis in the 7 years preceding the date of death; and (iii) a control group of living persons at the time of death of the suicide case (5 controls, 1 case) with a MD/D diagnosis within 7 years matched by region, sex and age group of the case. Results The results show that cases of suicide without a MD/D (about 25% of suicide cases) were less likely to have consulted than those with a MD/D. Suicide cases with and without a MD/D were less likely to be heavy users of ambulatory health services (≥ 4 visits) than matched living controls. They were also more likely to consult for a physical disorder alone and less likely to consult for mental health reasons. Compared to cases of suicide with a MD/D, suicide cases without MD/D were less likely to be hospitalized and more likely to have visited only an emergency room. Suicide cases diagnosed with a MD/D were more likely to be hospitalized and use emergency services alone than the other two groups. Matched living controls with a MD/D were more likely to use outpatient services alone. Conclusions These results should be compared with those emerging from systematic suicide case audits. These show a prevalence of mental disorders of 90%, especially depression, personality disorders and substance use disorders. They also show deficits in the recognition and treatment of mental disorders, which would correspond to the 25% of cases of suicide in the HADBs not diagnosed in the last 7 years. Future studies should include multivariate analyses to better elucidate health service use trajectories and patient vulnerability profiles.

Out-of-pocket costs in the year after early breast cancer among Canadian women and spouses.

BACKGROUND: We lack comprehensive information about the extent of out-of-pocket costs after diagnosis of early breast cancer and their effects on the family's financial situation. METHODS: This longitudinal study assessed out-of-pocket costs and wage losses during the first year after diagnosis of early breast cancer among Canadian women and spouses. Out-of-pocket costs for treatments and follow-up, consultations with other practitioners, home help, clothing, and natural health products were estimated, with information collected from telephone interviews. Generalized linear models were used to identify women at risk of having higher costs and the effects of out-of-pocket costs on perceptions of the family's financial situation. RESULTS: Overall, 829 women (participation, 86.2%) and 391 spouses participated. Women's median net out-of-pocket costs during the year after diagnosis were $1002 (2003 Canadian dollars; mean = $1365; SD = $1238), and 74.4% of these costs resulted from treatments and follow-up. Spouses' median costs were $111 (mean = $234; SD = $320), or 9% of couples' total expenses. In multivariable analyses, the percentage of women with out-of-pocket costs of $1773 or more (upper quartile) was statistically significantly associated with higher education, working at diagnosis, living more than 50 km from the hospital where surgery was performed, and having two and three different types of adjuvant treatment (all 2-sided P values ≤ .01). However, when considered simultaneously with wage losses, out-of-pocket costs were not associated with perceived deterioration in the family's financial situation; rather, wage losses were the driving factor. CONCLUSIONS: Overall, out-of-pocket costs from breast cancer for the year after diagnosis are probably not unmanageable for most women. However, some women were at higher risk of experiencing financial burden resulting from these costs.

Out-of-pocket costs for accessing adjuvant radiotherapy among Canadian women with breast cancer.

PURPOSE: Patients with breast cancer incur out-of-pocket costs when they receive adjuvant radiation. These treatments are administered in dedicated centers on a daily basis over 4 to 5 weeks. We assessed the extent of out-of-pocket costs to access radiotherapy and identified women at risk of experiencing higher costs. PATIENTS AND METHODS: This prospective study was conducted among 800 women from eight Quebec hospitals (participation, 86%), of whom 693 women received adjuvant radiotherapy. Costs to access treatment (transportation, parking, lodging, and meals) were estimated on the basis of information collected by telephone interviews. Generalized linear models were used to identify women at risk of having higher costs. RESULTS: Mean and median total costs (including financial aid received) to access radiotherapy were 445 Canadian dollars (Can$; standard deviation, Can$407) and Can$311, respectively. Women who lived at home but far from the radiotherapy center had the highest risk of experiencing weekly costs of at least Can$122 (highest quartile; prevalence ratio [PR], 5.8; 95% CI, 4.7 to 7.1) followed by women who lived away from home and farther from the center (PR, 4.5; 95% CI, 3.4 to 6.0) and by women who lived away from home and closer to the center (PR, 2.14; 95% CI, 1.3 to 3.5). When women stayed at a cancer lodge during treatment, costs were reduced by 80%. CONCLUSION: Considered in and of themselves, out-of-pocket costs for access to radiotherapy appeared relatively modest. Financial assistance initiatives for women who lived away from home during treatments contributed to a substantial reduction of access costs. These efforts should be maintained, but financial assistance should also target patients with cancer who live at home during treatments but have to travel relatively long distances on a daily basis.

Psychological distress and physical health in the year after diagnosis of DCIS or invasive breast cancer.

Ductal carcinoma in situ (DCIS) has an excellent prognosis, but its management can resemble that of early invasive breast cancer. We compared aspects of quality of life of women with DCIS to that of women with invasive disease during the first year after treatment initiation. Participants came from consecutive series of women with newly diagnosed, non-metastatic breast cancer treated in eight Quebec hospitals in 2003. Psychological distress and health-related quality of life were measured using the Psychiatric Symptom Index (PSI) and the SF-12 mental and physical component scales (MCS, PCS). Data were obtained 1, 6, and 12 months after the start of treatment. We used generalized linear models to compare mean scores and explored the possible clinical significance of between-group differences with effect size (ES). Participation and retention among eligible women were high, 86 and 97%, respectively. Among the 800 women who completed all interviews, 13.4% (n = 107) had DCIS and 86.6% (693) invasive disease. No statistically significant between-group differences were found at 1, 6, or 12 months in psychological state (PSI and MCS: P values from 0.065 to 0.904; ES from -0.01 to -0.21). Women with DCIS reported significantly higher levels of physical health, particularly when compared at 1 month to women with invasive disease who had chemotherapy (P value < 0.0001; ES = 0.82). Measured in symptoms of psychological distress, the better prognosis or less aggressive management of DCIS does not offset the general psychological effects of a cancer diagnosis to any great degree.