Protective subgroups for suicidality: Decision tree analysis of protective and risk factors in a representative sample of Australian adults

  • Jennifer Ma (Speaker)

Activity: Talk or presentationOral presentation


Background: Protective factors are thought to lower suicide risk or help a person defend against suicidal thoughts. However, it is currently unclear how protective factors for suicide interact with established risk factors to disrupt the developmental pathways from suicidal thoughts to behaviours. This study used Decision Tree (DT) analysis to identify the most relevant protective and risk factors, and their relative importance, in predicting suicidality in the general population.
Methods: Online quota sampling (age, gender, location) was used to recruit 1,187 Australian adults (54.3% women; M age = 46.76 years, SD = 18.44) from market research panels to participate in an anonymous online survey. Sociodemographic information and a range of empirically and theoretically supported protective (i.e., self-esteem, resilience, emotional regulation, social support, meaning in life, service utilisation) and risk factors for suicide (i.e., mental and physical health, stressful and traumatic life events, alcohol use, insomnia, hopelessness, defeat and entrapment, interpersonal risk, intrusive future imagery, exposure to suicide, acquired capability) were measured in relation to past month suicide ideation and attempt. A DT model with 41 protective and risk factors was built using the Chi-Square Automatic Interaction Detector algorithm.
Results: Two thirds (65.5%) of participants reported no suicidality, 16% ideation only (SI), and 18% both ideation and attempts (SI and SA). The DT model identified 18 different protective and risk profiles. Perceived burden (PB) was the strongest predictor of past month suicidality. Low levels of PB (≤ 25th quartile) correctly classified 90.1% of those with no suicidality. In this low PB subgroup, mental health diagnosis, self-esteem, and intrusive future imagery did not confer additional risk of SI or SA, regardless of level. Low access to means (≤ 25th quartile) was protective against suicidality at the highest levels of PB (≥75th quartile).
Discussion: Findings suggest that when protective and risk factors are analysed together, risk factors are statistically identified as more important for the prediction of suicidality. Out of 41 protective and risk factors measured, low PB – comprising low feelings of self-hatred and liability to others – was found to be most protective against suicidality. The 18 different subgroups identified in the study highlight the complexity of interrelationships between suicide protective and risk factors. Further examination of these relationships, beyond their independent associations with suicide-related outcomes, could aid in the identification of risk profiles for suicidal behaviour.
Period20 Jul 2023
Event titleCanberra Health Annual Research Meeting (CHARM) 2023
Event typeConference
LocationCanberra, Australia, Australian Capital TerritoryShow on map
Degree of RecognitionNational