Table 1 Details of included studied, systematic review mixed-method in order to identify most common suicide risk factors and ML outcomes
Authors, published year | Country | Study design | Sample size | Participants | Most important suicide risk factors | The best ML algorithms | Outcome |
|---|---|---|---|---|---|---|---|
(Alexopoulos et al., 2021) [59] | USA | randomized clinical trial | 249 | Older adults (aged 60 or above) diagnosed with major depression | Hopelessness, Neuroticism, And Low General Self-Efficacy, Avoidance/Rumination, Activation, Disability [12-Item World Health Organization Disability Assessment Schedule II (WHODAS-II), NEO Neuroticism Subscale, [Beck Hopelessness Scale (BHS), [Subscale Of The Behavioral Activation For Depression Scale (BADS) | least absolute shrinkage and selection operator (LASSO) | AUC = 0.735 |
(Bai et al., 2021) [24] | China | cohort | 159 | Individuals diagnosed with major depressive disorder who experience suicidal ideation, compared to healthy participants | Total Cholesterol (TC), Triglyceride (TG), HDLC, Low Density Lipoprotein Cholesterol (LDLC), APOA1, Apo Lipoprotein B (APOB), Lipoprotein A (LPA), Hypersensitive C-Reactive Protein (CRP), TRSF, Homocysteine (HCY), And AAT, BMI, Age, Sex, Education, Marital Status, Family History Of Mental Illness, Smoke, Hamilton Rating Scale For Depression (HDRS Scores), Beck Scale For Suicide Ideation-Chinese Version (BSI-CV Scores) | Random Forest | AUC = 0.928 |
(Balbuena et al.,2022) [60] | Norway, Saskatoon | retrospective cohort | 185,889 | General population | Income, Age, Sex, Marital Status, Month Of Birth, Living With A Spouse, Education, Social Deprivation, Mood Symptoms, Nervous Or Worried, Anxiety, Irritable, Felt Confident And Calm, Felt Happy And Optimistic, Felt Down, Depression, Felt Lonely, Smoke, Alcohol, Physical Activity, Exposure To Smoke-Filled Rooms, And Exposure To Secondary Smoke As A Child, Substance Abuse, Triglyceride, HDL-Cholesterol, Glucose, Total Cholesterol, Waist-Hip Ratio, Blood Pressure Medications, BMI, Having An Injury Requiring Hospitalization, Eating Disorder, Personality Disorder, Schizophrenia, Mania, ADHD, Repeated Episodes Of Self-Harm (RESH) | COX (male) Random Forest (female) | AUC (male): 0.57, AUC (female): 0.50 |
(Ballester et al.2021) [61] | Brazil | Prospective cohort study | 1560 | General population | Sex, Socioeconomic Status, Age, Education, Quality Of Life, Mental Disorder Symptoms | gradient tree boosting | True positive rate = 0.79, False positive rate = 0.28 |
(Barak-Corren et al., 2017) [62] | USA | retrospective cohort | 1,724,785 | Inpatient and outpatient | Mental Disorders, Personality Disorders, Schizophrenia, Psychotic Disorders, Open Wounds, Superficial Injury, Sex, Age, Race, Ethnicity, Veteran Status, Marital Status, Substance Abuse, Alcohol | Naive Bayesian classifier | AUC = 0.77, Sensitivity = 0.54, Specificity = 90% |
(Bhak et al., 2019) [63] | South Korea | cohort | 182 | The study included individuals who had attempted suicide, patients diagnosed with major depressive disorder, and a group of healthy controls | Age, Sex, Family History Of Depression, Family History Of Suicide Attempt, Depression, Antidepressant, Hamilton Rating Scale For Depression–17, SSI Scale For Suicidal Ideation | random forest classifier | Accuracy = 92.6% in distinguishing suicide attempters from major depressive disorder patients 87.3% in distinguishing major depressive disorder patients from controls 86.7% in distinguishing suicide attempters from controls Average accuracy = 88.87 |
(Cho et al., 2020) [64] | South Korea | cohort | 372,813 | Individuals who are subscribers of health insurance or recipients of medical aid and have been referred for a check-up | Y-Glutamyl Trans peptidase, BMI, Demographics: Age, Exercise, Alcohol | random forest | AUC during follow-up = 0.849, AUC after follow-up = 0.818, Average AUC = 0.8335, positive predictive value during follow-up = 0.007, positive predictive value after follow-up = 0.003, Average positive predictive value = 0.005, negative predictive value during and after follow-up = 0.999, Average negative predictive value = 0.999, False negative rate during follow-up = 0.183, False negative rate after follow-up = 0.341, Average False negative rate = 0.262, False positive rate during follow-up = 0.247, False positive rate after follow-up = 0.212, Average False positive rate = 0.2295 , accuracy during follow-up = 0.754, accuracy after follow-up = 0.788, Average accuracy = 0.771, Sensitivity during follow-up = 0.817, Sensitivity after follow-up = 0.657, Average sensitivity = 0.737, Specificity during follow-up = 0.754, Specificity after follow-up = 0.788 Average Specificity = 0.771 |
(Cho et al., 2021) [42] | South Korea | cohort | 48,047 | Individuals who underwent health screening | Sex, Insurance, Age, Depression, Antidepressants, Aspartate Transaminase, Waist Circumference, Sleeping Pills, BMI, Benzodiazepine, Glucose, Cholesterol, Lipoprotein (HDL), Hemoglobin, Triglyceride, Systolic Blood Pressure, Low-Density Lipoprotein (LDL), Creatinine, Y-Glutamyl Transpeptidase | random forest | AUC = 0.818, positive predictive value = 0.007, negative predictive value = 0.999, Accuracy = 0.832, Sensitivity = 0.600, Specificity = 0.833 |
(Choi et al., 2021) [50] | South Korea | longitudinal study | 78,796 | General population | Income, Age, Drinking Age, Average Sleep Time Per Day, Education, Home Ownership, Economic Activity, Uncomfortable Experience In Past 2 Weeks, Sex, Depression, Anxiety, Athletic Ability, Pain/Discomfort | XGBoost classifier | AUC suicidal ideation = 0.950 AUC suicide attempts = 0.990 Average AUC = 0/97, Precision suicidal ideation = 0.874, Precision suicide attempts = 0.977, Average precision = 0.9255 , Recall suicidal ideation = 0.893, Recall suicide attempts = 0.986 Average recall = 0/9395, Accuracy suicidal ideation = 0.893, Accuracy suicide attempts = 0.986, Average accuracy = 0.9395 |
(Kyu Sung Choi et al., 2021) [65] | South Korea | longitudinal study | 31,720 | General population | Age, Sex, Anxiety, Resilience, Self-Esteem, Depression, Neuropsychiatric Disorder, Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), State-Trait Anxiety Inventory-State Anxiety (STAI-S, Or STAI-X1), Resilience Appraisal Scale (RAS), Rosenberg Self-Esteem Scale (RSES), Mini International Neuropsychiatric Interview (MINI) | deep graph neural network model (GIN) | AUC = 0.878, Sensitivity = 76.3, Specificity = 83.4 |
(Choi et al., 2018) [46] | South Korea | cohort | 819,951 | Korean who are covered by Medicare and Medicaid insurance inpatient or outpatient | Disability, Endocrine, Nutritional And Metabolic Diseases, Mental And Behavioral Disorders, Diseases Of The Nervous System, Diseases Of The Eye And Adnexa, Diseases Of The Ear And Mastoid Process, Diseases Of The Respiratory System, Diseases Of The Genitourinary System, Injury, Poisoning, Factors Influencing Health Status, Sex, Age, Type Of Insurance, Household Income | Cox regression | AUC = 0.688 |
(R. Yates Coley et al., 2021) [25] | USA | diagnostic/prognostic study | 13,980,570 | Outpatient appointments with a mental health professional | Bipolar, Schizophrenia, Mental Disorder Symptoms, Encounters In Past 5 Y With Mental Health Diagnosis, Depression, Anxiety, Race, Sex, Age, Medicaid Insurance, Alcohol, Substance Abuse, Patient Health Questionnaire 9 (PHQ-9 Score) | Logistic regression with LASSO | AUC = 0.822 |
(Edgcomb et al., 2021) [43] | USA | longitudinal cohort study | 15,644 | Adults with serious mental illnesses are at high risk for both medical illness and serious suicide attempts and self-harm | Medical Comorbidity, Age, Number of Medical Hospitalizations In The Previous Year, Alcohol, Bipolar, In-Hospital Mortality Score, Walraven Score | Classification and Regression Tree algorithm (CART) | AUC = 0.86 |
(Edgcomb et al., 2021) [31] | USA | retrospective cohort study | 142,476 | women with depression, bipolar disorder, schizophrenia or schizoaffective disorder | Medical Comorbidity, History of Pregnancy-Related Mental Illness, Age, Van Walraven Score | Classification and Regression Tree algorithm (CART) | AUC repository1 = 0.73, AUC repository2 = 0.71, Average AUC = 0.72 , Accuracy repository1 and repository2 = 0.84, Sensitivity repository1 = 73.4, Sensitivity repository2 = 83.3, Average Sensitivity = 78.35 , Specificity repository1 = 84.1, Specificity repository2 = 82.2, Average Specificity = 83.15 |
(Etter et al., 2017) [66] | USA | Prospective cohort study | 2134 | General population | Sex, Race, Age, Substance Abuse | Logistic regression | AUC = 0.84 |
(Ge et al., 2020) [67] | China | retrospective study | 1916 | patients with major depressive disorder | Age, Sex, Marital Status, Vocational Status, Ethnicity, Depression, Anxiety, Free Triiodothyronine (FT3), Free Thyroxine (FT4), Thyroxine, Triiodothyronine, Thyrotropin Stimulating Hormone (TSH), Hamilton Depression Scale (HAMD), Hamilton Anxiety Rating Scale (HAMA) | neural network | AUC = 0.76, Accuracy = 70.08, Sensitivity = 70.68, Specificity = 67.09 |
(Hill et al., 2019) [68] | USA | longitudinal study | 4834 | General population | HIV/STD/Sex-Related, Depression, Risky Behavior, Relationship, Friends Have Succeeded in Committing Suicide In The Past 12 Months, Resident Mother, Race, Substance Abuse | A Classification Tree Approach (CTA) | Accuracy Wave 1 = 85.1, Accuracy Wave 2 = 71.7, Average accuracy = 78.4, Sensitivity Wave 1 = 69.8, Sensitivity Wave 2 = 90.6, Average sensitivity = 80.2, Specificity Wave 1 = 85.7, Specificity Wave 2 = 70.9, Average specificity = 78.3 |
(Kim et al., 2021) [69] | South Korea | retrospective analysis | 7824 | Students who have been referred for a health check-up | Family Problems, Interpersonal Passivity, Aberrant Experiences, Social Avoidance, Shyness, Disaffiliativeness, Juvenile Conduct Problems, Aggression, Activation, Hopelessness, Helplessness, Self-Doubt, Inefficacy, Stress/Worry, Anxiety, Anger Proneness, Behavior-Restricting Fears, Multiple Specific Fears, Malaise, Demoralization, Low Positive Emotions, Cynicism, Antisocial Behavior, Ideas Of Persecution, Dysfunctional Negative Emotions, Hypomanic Activation, Emotional/Internalizing Dysfunction, Thought Dysfunction, Behavioral/Externalizing Dysfunction, Substance Abuse, Gastro-Intestinal Complaints, Head Pain Complaints, Neurological Complaints, Cognitive Complaints, Somatic Complaints, Suicidal/Death Ideation (SUI) Scale, PSY-5 (Personality Psychopathology Five) Scales, AES Aesthetic-Literary Interests, MEC Mechanical-Physical Interest | Random forest | AUC suicide ideation = 0.844, AUC suicide attempt = 0.851, Average AUC = 0.8475 , Precision suicide ideation = 0.920, Precision suicide attempt = 0.953, Average precision = 0.9365 , Recall suicide ideation = 0.929, Recall suicide attempt = 0.950, Average recall = 0.9395 , Accuracy suicide ideation = 92.9, Accuracy suicide attempt = 95 , Average accuracy = 93.95 , |
(Haroz et al., 2022) [70] | USA | retrospective cohort | 13,420 | Pediatric patients | Sex, Age, Race, Anxiety, Depression, Alcohol, Obesity, Tobacco, ADHD, Brain Injury, Psychotic Disorders, Anemia, Bipolar, Migraine, Autism, Diabetes, Drug, Asthma, Learning Disorder, Intellectual Disability, PTSD, Epilepsy, Cancer, Hemiplegia, Cerebrovascular Disease, Other Developmental Disorder, Renal Disease, Mild Liver Disease, Rheumatoid Disease, Heart Failure, Vascular Disease, Metastatic Tumor, AIDS, Severe Liver Disease, Myocardial Infraction, Leukemia And Lymphoma, Personality Disorder, Charlson Comorbidity Index | Binary Logistic Regression model | AUC Model 2: ASQ + HER = 0.843, positive predictive value Model 2: ASQ + HER = 0.103, False negative rate Model 2: ASQ + HER = 0.183, Accuracy Model 2: ASQ + HER = 0.815, Sensitivity Model 2: ASQ + HER = 0.817 |
(Melhem et al., 2019) [71] | USA | longitudinal study | 663 | offspring of parents (or probands) with mood disorders | Bipolar, Posttraumatic Stress, Hyperactivity, Childhood Abuse, Psychotherapy, Antidepressants, Nonpsychotropic Medications, Hopelessness, Impulsivity, Impulsive Aggression, Irritability, Mood Disorders, Depression, Anxiety, Aggression, Sex, Age, Race, Income, Alcohol, Substance Abuse | logistic regression | AUC = 0.80, Sensitivity = 87.3, Specificity = 63 |
(Machado et al., 2020) [72] | USA | Prospective cohort study | 77,746 | General population | BMI, Psychotic Disorders, Personality Disorders, Mood Disorder, Anxiety, Age, Sex, Race, Marital Status, Education, Income, Substance Abuse | Elastic net regularization | AUC distinguished individuals who had attempted suicide from those who had not = 0.89, AUC participants with lifetime major depressive episodes = 0.89, Average AUC = 0.89, positive predictive value distinguished individuals who had attempted suicide from those who had not = 0.0455, positive predictive value participants with lifetime major depressive episodes = 0.1048, Average positive predictive value = 0.07515 , negative predictive value distinguished individuals who had attempted suicide from those who had not = 0.9980, negative predictive value participants with lifetime major depressive episodes = 0.9944, Average negative predictive value = 0.9962 , Sensitivity distinguished individuals who had attempted suicide from those who had not = 0.7451, Sensitivity participants with lifetime major depressive episodes = 0.7742, Average sensitivity = 0.75965 , Specificity distinguished individuals who had attempted suicide from those who had not = 0.8922, Specificity participants with lifetime major depressive episodes = 0.8586, Average specificity = 0.8754 |
(Miranda et al., 2022) [73] | USA | retrospective study | 38,807 | posttraumatic stress disorder (PTSD) patients, a population with a high risk of suicide | Glucose, Chloride, Red Cell Distribution Width, Mean Corpuscular Hemoglobin, Hemoglobin, Hematocrit, Mean Corpuscular Volume, White Blood Cell, Anion Gap, Blood-Urine, ABS Neutrophils, Glucose-Urine, Glucose (Bedside Test), ABS Basophils, Red Blood Cell, Potassium, International Normalized Ratio, Ionized Calcium ISTAT, Ionized Calcium, Bacteria, Base Excess, Total Protein, Calcium, Red Blood Cells-Urine, Sodium, Prothrombin Time, Mean Platelet Volume, Platelets | deep-learning (DRNN) | AUC = 0.932 |
(Zheng et al., 2020) [74] | USA | cohort | 236,347 | Patients who visited three specified health centers | Age, Sex, Family Problems, Unemployment, Anxiety, Depression, Mood Disorder, Anxiety, Substance Abuse, Alcohol, Tobacco, Arthritis, Back Pain, Vehicle Accidents, Head Pain Complaints, PTSD, Esophageal Disorders, Hypertension, Esophageal Reflux, Serotonin Reuptake Inhibitor, Physical Restraints, Educational Circumstances Problems, Injury, Poisoning, Psychosocial Issues, Drug Use, Asthma, Schizophrenia, Joint Pain, Nervous System Disorders, Neuropathy, Bipolar, Developmental Disorders, Fibromyalgia, Impulse Control Disorders, Personality Disorders, Endocrine Disorders, Forearm, Migraines, Borderline Personality Disorder | Deep Learning model | AUC 1-year suicide attempt = 0.792, AUC retrospective and prospective cohorts = 0.769, Average AUC = 0.7805 |
(Zhu et al., 2020) [75] | China | cohort | 169 | inpatients (Patients With Bipolar II Disorder) | Age, Education, Sex, Marital Status, Family History Of Mental Disorder, Family History Of Suicide, Occupation, Seasonal Characteristics, Depression, Hypomania, Substance Abuse, Course Of Disease, Combined Somatic Disorder, Psychotic Characteristics, Hamilton Depression Rating Scale (Scores Of HAMD-17), Hamilton Depression Rating Scale (Scores Of HAMD-16), Hamilton Anxiety Rating Scale (Scores Of HAMA), Nurses’ Global Assessment Of Suicide Risk (NGASR) | support vector machines (SVMs) | Accuracy = 0.84 |
(Setoyama et al., 2016) [76] | Japan | cohort | 339 | psychiatric patients | Blood Plasma: 3-Hydroxybutyrate (3HB), Betaine, Citrate, Creatinine, And Gamma-Aminobutyric Acid (GABA), Hamilton Rating Scale For Depression (HAMD)-17, Patient Health Questionnaire (PHQ)-9 | logistic regression | AUC = 0.75 |
(Van Mens et al., 2020) [77] | Scotland | longitudinal study | 3508 | General population | Burdensomeness, Mental Disorder Symptoms, Impulsivity, Defeat, Entrapment, Internal Entrapment, Interpersonal Needs, Depression, External-Entrapment, Stress, Wellbeing, Twarted-Belonginess, Optimism, Perfectionism, Resilience, Interpersonal Needs, Lack of Social Support, Reengagement, Acquired Capability, Goal Disengagement, Age, Sex | random forest | AUC = 0.83, positive predictive value = 0.52 |
(Su et al., 2020) [78] | USA | retrospective study | 41,721 | pediatric population who visited specified health center | Sex, Age, Race, Ethnicity, Depression, Anxiety, Mood Disorder, Problems Related To Lifestyle, Emotional State, ADHD, Conduct Disorder, Bipolar, Personal History Of Certain Other Disease, Joint Disorder, Sertraline, Hydroxyzine Pamoate, Ethinyl Estradiol, Lorazepam, Escitalopram, Citalopram, Fluoxetine, Cholecalciferol, Hydroxine Hcl, Venlafaxine, Guanfacine, Trazodone, Prozac, Risperidone, Zoloft, Aripiprazole, Quetiapine, Culture Urine, Acetaminophen Level, Urinalysis With Microscopic, POC Pregnancy Urine, MDMA, POC Urinalysis Dipstick, Urinalysis With Microscopic, POC Pregnancy Urine, Salicylate Level, GC/Chlamydia, Oxycodone Urine | logistic regression | AUC = 0.86 |
(Forkmann et al., 2020) [79] | Germany | longitudinal prospective study | 308 | The study focused on participants over the age of 18 who were admitted to a psychiatric hospital due to severe suicidal ideation or a suicide attempt within two weeks of admission | Burdensomeness, Belongingness, Depression, Pessimistic And Hopeless Cognitions, Fearlessness About Death, Pain Tolerance, Capability For Suicide, Self-Injurious Thoughts And Behaviors Interview (SITBI), German Capability For Suicide Questionnaire (GCSQ), Interpersonal Needs Questionnaire (INQ), Rasch-Based Depression Screening (DESC-I), Beck Hopelessness Scale (BHS), Beck Scale For Suicide Ideation (BSS) (Interpersonal Theory Of Suicide, Pos. 4–8) | logistic regression | AUC = 0.729 |
(Adams et al., 2021) [80] | Denmark | Case–cohort study | 15,953 | Case: all hospital outpatient clinic visits were reported to the Danish National Patient Registry Control: all persons born or legally residing in Denmark on 1 January 1995 | Sedative, Hypnotic, Anxiolytic, Age, Marital Status, Income, Immigrant, Alcohol, Opioid, Cannabis, Cocaine, Hallucinogen, Nicotine, Other Psychoactive Substance | random forests | AUC men = 0.77, AUC women = 0.86, Average AUC = 0.815 |
(Barak-Corren et al., 2020) [81] | USA | case control | 3,714,105 | Case: Cases were defined as individuals having at least 1 captured suicide attempts, Control: candidate that are likely to capture suicide attempts | Sex, Race, Ethnicity, Poisoning By Unspecified Drug Or Medicinal Substance, Drug Withdrawal Syndrome, Cocaine Dependence, Unspecified Use, Drug-Induced Organic Affective Syndrome, Borderline Personality, Unspecified Use, Unspecified Neurotic Disorder, Unspecified Use, Multiple And Unspecified Open Wound Of Upper Limb, Without Mention Of Complication, Unspecified Use, Open Wound Of Upper Arm, Without Mention Of Complication, Unspecified Use, Open Wound Of Hand Except Fingers Alone With Tendon Involvement, Assault By Cutting And Piercing Instrument, Bipolar Affective Disorder, Unspecified Degree, Unspecified Personality Disorder, Bipolar Affective Disorder, Mixed, Mixed, Or Unspecified Drug Abuse, Acute Alcoholic Intoxication In Alcoholism (Alcohol), Unspecified Drinking Behavior, Cocaine Abuse, Opioid Type Dependence, Unspecified Drug Dependence, Depression, Antipsychotics, Antidepressants, And Mood Stabilizers (Eg, Lithium), Antiretroviral Medications Used To Treat HIV/AIDS (Eg, Ritonavir) Were Associated With An Increased Risk For A Suicide Attempt. Nicotine Replacement Therapy Also Consistently Emerged As A Leading Predictor Across The Nonpediatric Hospitals, Toxicology Tests, Toxicology Screening Tests, Choriogonadotropin-Beta, Levels Of Valproic Acid, BCH, Lithium Level | naive Bayes classifiers | AUC = 0.73 |
(Barros et al., 2017) [82] | USA | case control | 707 | Case: mental health patients with suicide risk, Control: mental health patients without suicide risk | Bipolar, Mixed Episode, Adjustment Disorder, Dysthymia, Depression, Anxiety, Age, Sex, Marital Status, Parental Status, Education, Occupation | support vector machine | Accuracy = 0.78, Sensitivity = 0.77, Specificity = 0.79 |
(Berkelmans et al., 2021) [53] | Netherland | case control | 602,270 | Case: The people who died by suicide, Control: people who did not die due to suicide | Age, Sex, Income, Education, Immigration Background, Urban city, Place In Household, Healthcare Costs, Province, Social Benefits, Mental Disorder Symptoms, Physical Health Problems | logistic regression | AUC = 0.77 |
(Delgado-Gomez et al., 2016) [83] | Spain | case control | 902 | Case 1: first-time suicide attempters, Case 2: psychiatric inpatients without current or past history of suicidal behavior, Control: healthy controls | Sex, Age, Marital Status, Education, Employment Status, Suicidal Risk, Personality and Life Events Items and The Corresponding Samejima’s Parameters (PLE Scale Item) | Decision tree | Precision = 0.89, Sensitivity = 0.89, Specificity = 0.89 |
(D. Delgado-Gomez et al., 2011) [84] | Spain | case control | 879 | control: healthy blood donors, case: psychiatric inpatients | Barrat’s Impulsiveness Scale Version 11 (BIS-11), International Personality Disorder Evaluation Screening Questionnaire (IPDESQ) | Support vector machine | Accuracy = 80.26 |
(D. Delgado-Gomez et al., 2012) [85] | Spain | case control | 883 | Case: Adults with suicide attempt, Control: adults without suicide attempt | Age, Holmes–Rahe Social Readjustment Rating Scale (SRRS Questionnaire), International Personality Disorder Examination Screening Questionnaire (IPDE-SQ Questionnaire) | multivariate classification techniques | Accuracy = 0.79, Sensitivity = 78.30, Specificity = 0.91 |
(DelPozo-Banos et al., 2018) [86] | United Kingdom | Case–cohort study | 54,684 | Case: those who died by suicide, Control: | Prescription Of Psychotropic, Depression, Anxiety, Self-Harm, Age, Sex, Alcohol, Drug Abuse | ANN | Sensitivity = 64.57, Specificity = 81.86 |
(Gradus et al., 2021) [87] | Denmark | Case–cohort study | 288,157 | Case: Cases were all persons who made a nonfatal suicide attempt, Control: 5% random sample of the population at risk | Psychiatric Medications, Poisoning, Stress, Substance Abuse | Random forest | AUC (male): 0.89, AUC (female): 0.91, Average AUC: 0/9 |
(Jiang et al., 2021) [30] | Denmark | Case–cohort study | 14,737 | Cases: all persons who died by suicide and had an incident depression diagnosis, Control: 5% random sample of all individuals in Denmark | Mental Disorders, Somatic Disorders, Surgeries, Medications, Age, Marital Status, Immigrant, Family Suicide Death History, Employment, Income | Decision tree and random forest | AUC decision tree (male): 0.71, AUC random forest (female): 0.75, Average AUC: 0.73 |
(Harman et al., 2021) [88] | USA | case control | 11,369 | Case: individuals with suicidal ideation and/or attempt, Control: healthy controls | Impulsivity, Prodromal Psychosis, Loneliness, Worthlessness, Behavioral Problems | random forest | AUC: 0.77 |
(Navarro et al., 2021) [89] | Canada | prognostic study | 1623 | General population | Mother and Father Education and Mother and Father Age, Socioeconomic Status, Neighborhood Characteristics, Parents’ Psychological State, Parents’ Antisocial Behaviors, Parenting Practices, Sex, Prematurity | Random forest | AUC (male): 0.62, AUC (female): 0.72, Average AUC: 0.67 , positive predictive value (male): 0.62, positive predictive value (female): 0.60, Average positive predictive value: 0.61 , negative predictive value (male): 0.71, negative predictive value (female): 0.75, Average negative predictive value: 0.73 , Sensitivity (male): 0.32, Sensitivity (female): 0.50, Average sensitivity: 0.41, Specificity (male): 0.82, Specificity (female): 0.76, Average Specificity: 0.79 |
(McKernan et al., 2019) [90] | USA | case control | 8879 | Case: fibromyalgia patients with suicide ideation and attempt, Control: general population | Sex, Race, Age, Comorbid Medical Illness, Drug Dependence, Inpatient Utilization, Mental Disorder Symptoms, Obesity, Polysomatic Complaints | bootstrapped L-1 penalized regression | AUC suicide ideation: 0.8, AUC suicide attempt: 0.82, Average AUC: 0.81, Recall suicide ideation: 0.14, Recall suicide attempt: 0.08, Average recall: 0.11 |