Predicting Suicide Action: Secondary Analysis of ADD Health Data

Background

According to the CDC the incidence of suicide has increased 24% since 1999, most significantly for girls aged 10-14 (CDC, 2015). Despite comprehensive research strategies and significant resources, over 47,770 people every year die by suicide. This translates to 117 deaths per day; additionally, someone attempts to take their own life every 30 seconds (Drapeau & McIntosh, 2015). Research on suicide includes risk and protective factors, childhood antecedents, and predictors of suicide behavior; however, research has not improved our ability to differentiate between those who only think about suicide from those who actually attempt to take their own life. The purpose of this study was to analyze commonly known risk factors of suicidality in a predictive model to assess whether or not it is possible to distinguish between those who ideate about and those who attempt suicide.

Methods

Using the National Longitudinal Study of Adolescent to Adult Health wave IV data, a binomial logistic regression was performed to assess the following research question: What factors most strongly predict whether or not an individual will engage in suicidal action compared to only ideating about suicide? Predictors included gender, mental health symptoms (e.g., anxiety, isolation, overwhelmed), and relationship with parents. Results are presented at α=.10 with 95% confidence intervals.

Results

Results of the analysis indicate that the model was a better fit to the data than the null model (x²₍₁₉₎ =28.34; p<.1). Pseudo R² indicates that the model is useful in explaining group membership (Cox & Snell=.13; Nagelkerke=.23). The Hosmer and Lemeshow goodness of fit test indicates a good fit between the model and the data (x²₍₈₎=12.42; p=.13). Parameter estimates showed that feeling overwhelmed (OR=;.52;95% CI:.29,.92), are unable to shake off the ‘blues’ (OR=2.54;95% CI:.1.35,4.77), having trouble staying asleep (OR=1.36;95% CI:.95,1.94), and having a family member or  friend attempt to kill themselves (OR=2.78;95% CI:.87,8.91), are significantly predictors of attempting suicide. The results indicate that the odds of making a suicide attempt, compared to persons just thinking (ideating) about suicide, are 2.78 times higher for those with family/friends who had previously attempted suicide; 2.54 times higher for those who felt like they were unable to shake off the ‘blues’ in the 7 days prior their interview; 1.36 times higher for those who are often unable to stay asleep; and .521 higher for those who felt overwhelmed in the 7 days prior to their interview. The overall success rate of the model to predict suicide action is 86.7%.  

Implications

Being able to differentiate between those who think about suicide from those who attempt would directly influence practice in regards to the ways in which individuals are approached, assessed, and subsequently referred to for services in the midst of a suicidal crisis.