Search

BACKGROUND Machine learning is a promising tool in the area of suicide prevention due to its ability to combine the effects of multiple risk factors and complex interactions. The power of machine learning has led to an influx of studies on suicide prediction, as well as a few recent reviews. However, the existing systematic reviews did not differentiate between data sources, and have largely failed to summarize the most important risk factors. OBJECTIVE The objective of our study was to assess the degree to which machine learning techniques capture incidents of suicidal thoughts and behaviors, and ascertain the most important predictors of suicidal thoughts and behaviors identified via machine learning. METHODS A systematic literature search of PubMed,‎ Medline, Embase, PsycINFO, Web of Science,‎ Cumulative Index to Nursing and Allied Health Literature‎ (CINAHL)‎, and Allied and Complementary Medicine Database (AMED) to identify all studies that have used machine learning to predict suicidal thoughts and behaviors using administrative or survey data‎ was performed. The search was conducted for articles published between January 1, 2019 and May 11, 2022. In addition, all articles identified in three recently published systematic reviews (the last of which included studies up until January 1, 2019) were retained, if they met our inclusion criteria. To evaluate the predictive power of machine learning methods in predicting suicidal thoughts and behaviors box plots were used to summarize the distribution of the area under the receiver operating characteristic curve (AUC) values by suicide outcome (i.e., suicidal thoughts, suicide attempt, and death by suicide, as well as all suicide outcomes combined) and machine learning method. Mean AUCs with 95% confidence intervals (CIs) were computed for each suicide outcome by study design, study data source, total sample size, and sample size of cases, and machine learning methods employed. The most important risk factors identified were summarized. RESULTS The search strategy identified a total of 2,200 unique records, of which 104 articles met the inclusion criteria. Machine learning algorithms achieved good prediction of suicidal thoughts and behaviors (i.e., an AUC between 0.80 and 0.89); however, their predictive power appears to differ across suicide outcomes. The AUCs for boosting algorithms achieved good prediction of suicidal thoughts, death by suicide, and all suicide outcomes combined, while neural network algorithm achieved good prediction of suicide attempt. The risk factors for suicidal thoughts and behaviors differed depending on the data source and the population under study. CONCLUSIONS The predictive utility of machine learning for suicidal thoughts and behaviors is largely dependent on the approach used. The findings of the current review should prove helpful in preparing future machine learning models. CLINICALTRIAL N/A

Background Mental health problems and substance use co-morbidities during and after the COVID-19 pandemic are a public health priority. Identifying individuals at high-risk of developing mental health problems and potential sequela can inform mitigating strategies. We aimed to identify distinct groups of individuals (i.e., latent classes) based on patterns of self-reported mental health symptoms and investigate their associations with alcohol and cannabis use. Methods We used data from six successive waves of a web-based cross-sectional survey of adults aged 18 years and older living in Canada (6,021 participants). We applied latent class analysis to three domains of self-reported mental health most likely linked to effects of the pandemic: anxiety, depression, and loneliness. Logistic regression was used to characterize latent class membership, estimate the association of class membership with alcohol and cannabis use, and perform sex-based analyses. Results We identified two distinct classes: (1) individuals with low scores on all three mental health indicators (no/low-symptoms) and (2) those reporting high scores across the three measures (high-symptoms). Between 73.9 and 77.1% of participants were in the no/low-symptoms class and 22.9–26.1% of participants were in the high-symptom class. We consistently found across all six waves that individuals at greater risk of being in the high-symptom class were more likely to report worrying about getting COVID-19 with adjusted odds ratios (aORs) between 1.72 (95%CI:1.17–2.51) and 3.51 (95%CI:2.20–5.60). Those aged 60 + were less likely to be in this group with aORs (95%CI) between 0.26 (0.15–0.44) and 0.48 (0.29–0.77) across waves. We also found some factors associated with class membership varied at different time points. Individuals in the high-symptom class were more likely to use cannabis at least once a week (aOR = 2.28, 95%CI:1.92–2.70), drink alcohol heavily (aOR = 1.71, 95%CI:1.49–1.96); and increase the use of cannabis (aOR = 3.50, 95%CI:2.80–4.37) and alcohol (aOR = 2.37, 95%CI:2.06–2.74) during the pandemic. Women in the high-symptom class had lower odds of drinking more alcohol during the pandemic than men. Conclusions We identified the determinants of experiencing high anxiety, depression, and loneliness symptoms and found a significant association with alcohol and cannabis consumption. This suggests that initiatives and supports are needed to address mental health and substance use multi-morbidities.

BACKGROUND The COVID-19 global pandemic has increased the burden of mental illness on Canadian adults. However, the complex combination of demographic, economic, and lifestyle factors and perceived health risks contributing to patterns of anxiety and depression has not been explored. OBJECTIVE The aim of this study is to harness flexible machine learning methods to identify constellations of factors related to symptoms of mental illness and to understand their changes over time during the COVID-19 pandemic. METHODS Cross-sectional samples of Canadian adults (aged ≥18 years) completed web-based surveys in 6 waves from May to December 2020 (N=6021), and quota sampling strategies were used to match the English-speaking Canadian population in age, gender, and region. The surveys measured anxiety and depression symptoms, sociodemographic characteristics, substance use, and perceived COVID-19 risks and worries. First, principal component analysis was used to condense highly comorbid anxiety and depression symptoms into a single data-driven measure of emotional distress. Second, eXtreme Gradient Boosting (XGBoost), a machine learning algorithm that can model nonlinear and interactive relationships, was used to regress this measure on all included explanatory variables. Variable importance and effects across time were explored using SHapley Additive exPlanations (SHAP). RESULTS Principal component analysis of responses to 9 anxiety and depression questions on an ordinal scale revealed a primary latent factor, termed “emotional distress,” that explained 76% of the variation in all 9 measures. Our XGBoost model explained a substantial proportion of variance in emotional distress (r2=0.39). The 3 most important items predicting elevated emotional distress were increased worries about finances (SHAP=0.17), worries about getting COVID-19 (SHAP=0.17), and younger age (SHAP=0.13). Hopefulness was associated with emotional distress and moderated the impacts of several other factors. Predicted emotional distress exhibited a nonlinear pattern over time, with the highest predicted symptoms in May and November and the lowest in June. CONCLUSIONS Our results highlight factors that may exacerbate emotional distress during the current pandemic and possible future pandemics, including a role of hopefulness in moderating distressing effects of other factors. The pandemic disproportionately affected emotional distress among younger adults and those economically impacted.

Treating mental illnesses in primary care is increasingly emphasized to improve access to mental health services. Although family physicians (FPs) or general practitioners are in an ideal position to provide the bulk of mental health care, it is unclear how best to remunerate FPs for the adequate provision of mental health services. We examined the quantity of mental health services provided in Ontario's blended fee-for-service and blended capitation models. We evaluated the impact of FPs switching from blended fee-for-service to blended capitation on the provision of mental health services in primary care and emergency department using longitudinal health administrative data from 2007 to 2016. We accounted for the differences between those who switched to blended capitation and non-switchers in the baseline using propensity score weighted fixed-effects regressions to compare remuneration models. We found that switching from blended fee-for-service to blended capitation was associated with a 14% decrease (95% CI 12-14%) in the number of mental health services and an 18% decrease (95% CI 15-20%) in the corresponding value of services. This result was driven by the decrease in services during regular-hours. During after-hours, the number of services increased by 20% (95% CI 10-32%) and the corresponding value increased by 35% (95% CI 17-54%). Switching was associated with a 4% (95% CI 1-8%) decrease in emergency department visits for mental health reasons. Blended capitation reduced provision of mental health services without increasing emergency department visits, suggesting potential efficiency gain in the blended capitation model in Ontario.

Background The COVID-19 global pandemic has increased the burden of mental illness on Canadian adults. However, the complex combination of demographic, economic, and lifestyle factors and perceived health risks contributing to patterns of anxiety and depression has not been explored. Objective The aim of this study is to harness flexible machine learning methods to identify constellations of factors related to symptoms of mental illness and to understand their changes over time during the COVID-19 pandemic. Methods Cross-sectional samples of Canadian adults (aged ≥18 years) completed web-based surveys in 6 waves from May to December 2020 (N=6021), and quota sampling strategies were used to match the English-speaking Canadian population in age, gender, and region. The surveys measured anxiety and depression symptoms, sociodemographic characteristics, substance use, and perceived COVID-19 risks and worries. First, principal component analysis was used to condense highly comorbid anxiety and depression symptoms into a single data-driven measure of emotional distress. Second, eXtreme Gradient Boosting (XGBoost), a machine learning algorithm that can model nonlinear and interactive relationships, was used to regress this measure on all included explanatory variables. Variable importance and effects across time were explored using SHapley Additive exPlanations (SHAP). Results Principal component analysis of responses to 9 anxiety and depression questions on an ordinal scale revealed a primary latent factor, termed “emotional distress,” that explained 76% of the variation in all 9 measures. Our XGBoost model explained a substantial proportion of variance in emotional distress (r2=0.39). The 3 most important items predicting elevated emotional distress were increased worries about finances (SHAP=0.17), worries about getting COVID-19 (SHAP=0.17), and younger age (SHAP=0.13). Hopefulness was associated with emotional distress and moderated the impacts of several other factors. Predicted emotional distress exhibited a nonlinear pattern over time, with the highest predicted symptoms in May and November and the lowest in June. Conclusions Our results highlight factors that may exacerbate emotional distress during the current pandemic and possible future pandemics, including a role of hopefulness in moderating distressing effects of other factors. The pandemic disproportionately affected emotional distress among younger adults and those economically impacted.

In Canada's most populous province, Ontario, family physicians may choose between the blended fee-for-service (Family Health Group [FHG]) and blended capitation (Family Health Organization [FHO] payment models). Both models incentivize physicians to provide after-hours (AH) and comprehensive care, but FHO physicians receive a capitation payment per enrolled patient adjusted for age and sex, plus a reduced fee-for-service while FHG physicians are paid by fee-for-service. We develop a theoretical model of physician labor supply with multitasking to predict their behavior under FHG and FHO, and estimable equations are derived to test the predictions empirically. Using health administrative data from 2006 to 2014 and a two-stage estimation strategy, we study the impact of switching from FHG to FHO on the production of a capitated basket of services, after-hours services and nonincentivized services. Our results reveal that switching from the FHG to FHO reduces the production of capitated services to enrolled patients and services to nonenrolled patients by 15% and 5% per annum and increases the production of after-hours and nonincentivized services by 8% and 15% per annum.

Psychiatric hospitalizations could be reduced if mental illnesses were detected and treated earlier in the primary care setting, leading to the World Health Organization recommendation that mental health services be integrated into primary care. The mental health services provided in primary care settings may vary based on how physicians are incentivized. Little is known about the link between physician remuneration and psychiatric hospitalizations. We contribute to this literature by studying the relationship between physician remuneration and psychiatric hospitalizations in Canada's most populous province, Ontario. Specifically, we study family physicians (FPs) who switched from blended fee-for-service (FFS) to blended capitation remuneration model, relative to those who remained in the blended FFS model, on psychiatric hospitalizations. Outcomes included psychiatric hospitalizations by enrolled patients and the proportion of hospitalized patients who had a follow-up visit with the FP within 14 days of discharge. We used longitudinal health administrative data from a cohort of practicing physicians from 2006 through 2016. Because physicians practicing in these two models are likely to be different, we employed inverse probability weighting based on estimated propensity scores to ensure that switchers and non-switchers were comparable at the baseline. Using inverse probability weighted fixed-effects regressions controlling for relevant confounders, we found that switching from blended FFS to blended capitation was associated with a 6.2% decrease in the number of psychiatric hospitalizations and a 4.7% decrease in the number of patients with a psychiatric hospitalization. No significant effect of remuneration on follow-up visits within 14 days of discharge was observed. Our results suggest that the blended capitation model is associated with fewer psychiatric hospitalizations relative to blended FFS.

Team-based primary care offers a wide range of health services to patients by using interdisciplinary health care providers committed to delivering comprehensive, coordinated and high-quality care through team collaboration. Ontario's Family Health Team (FHT), the largest team-based practice model in Canada, was introduced to improve access to and effectiveness of primary health care services, and was available primarily for physicians paid under blended capitation models (Family Health Organizations and Family Health Networks). Using health administrative data on physicians practicing under blended capitation models in Ontario between 2006 and 2015, we study the impact of switching from non-FHT to FHTs on the production of capitated comprehensive care services, after-hours services, non-incentivized services, and services provided to non-enrolled patients by family physicians. We find that when in FHTs, physicians increase the production of total services and non-incentivized services by 26% and 5% per annum and reduce capitated comprehensive care services by 3.2% per annum. When in FHTs, physicians also see and enroll more patients relative to those practicing in non-FHTs. We find evidence of improved access to physician's services under team-based primary care, but switching to FHTs has no effect on the production of after-hours services and services provided to non-enrolled patients.

This paper explores the use of fuzzy regression-discontinuity design in the context where multiple treatments are applied at the threshold. The identification result shows that, under a very strong assumption of equality of treatment probability changes at the cutoff point, a difference in fuzzy discontinuity identify a treatment effect of interest. Using the data from the National Health Interview Survey (NHIS), we apply this identification strategy to evaluate the causal effect of the Affordable Care Act (ACA) on health care access and utilization of old Americans. We find results suggesting that the implementation of the Affordable Care Act has led to an increase in the hospitalization rate of elderly American -- 5% more hospitalization. It has caused a minor increase of cost-related direct barrier to access to care -- 3.6% increase in the probability of delaying care for cost reasons. The ACA has also exacerbated cost-related barriers to follow-up and continuity care -- 7% more elderly couldn't afford prescriptions, 7% more couldn't see a specialist and, 5.5% more couldn't afford a follow-up visit - as result of ACA.

We measure the response of physicians to monetary incentives using matched administrative and time-use data on specialists from Quebec (Canada). These physicians were paid fee-for-service contracts and supplied a number of different services. Our sample covers a period during which the Quebec government changed the prices paid for clinical services. We apply these data to a multitasking model of physician labour supply, measuring two distinct responses. The first is the labour-supply response of physicians to broad-based fee increases. The second is the response to changes in the relative prices of individual services. Our results confirm that physicians respond to incentives in predictable ways. The own-price substitution effects of a relative price change are both economically and statistically significant. Income effects are present, but are overridden when prices are increased for individual services. They are more prominent in the presence of broad-based fee increases. In such cases, the income effect empirically dominates the substitution effect, which leads physicians to reduce their supply of clinical services.