Suicide remains a common cause of death, and a meta-analysis by Franklin and colleagues argued that physicians are still quite limited in their ability to predict the risk for suicide. Their study, which was published in the February 2017 issue of Psychological Bulletin, assessed 365 studies of risk factors for suicide.
The main finding of the meta-analysis was that no prediction tool for suicide was highly sensitive and specific. Prediction with these instruments was only slightly better than that related to chance alone. Moreover, there was no trend toward improved performance of predictive abilities over time, and risk factors remained fairly stagnant over the 50 years of included research. Studies with longer periods of follow-up were not superior to shorter studies in terms of finding meaningful risk factors for suicide.
One limitation of previous research is that many studies assessed mental health at a single point in time. The current research by Melhem and colleagues assessed multiple clinical predictors of suicide attempt among young people over time, with an emphasis on how the trajectories of these variables affect the risk for suicide attempt.
Study Synopsis and Perspective
A new assessment tool that detects severity and fluctuation of depressive symptoms accurately predicts the risk for suicidality in high-risk individuals, a new study suggests.
Investigators followed more than 600 young adults with a mean age of 23.8±8.5 years who were at elevated risk for suicidal behavior because their parents had been diagnosed with mood disorders.
Parents and children were followed over a 12-year period and periodically evaluated through standard assessments.
The most accurate predictor of suicidal behavior was having severe depressive symptoms that fluctuated over time.
The investigators combined this finding with other relevant factors, such as younger age, mood disorders, childhood abuse, and personal and family history of suicide attempts to create a Prediction Risk Score (PRS).
They found that having a score of ≥3 of these risk factors indicated a higher risk for suicidal behavior, with 87% sensitivity, which is far superior to currently available models.
"We recommend that clinicians pay particular attention in their assessment of depression to the severity of both current and past depression and the variability in these symptoms," lead author Nadine M. Melhem, PhD, associate professor of psychiatry, University of Pittsburgh School of Medicine, told Medscape Medical News.
"Our PRS is a valuable addition to predict suicide attempt in high-risk individuals and uses variables that are already collected as part of the medical history assessment in clinical settings," she said.
Lack of Predictability "Unacceptable"
"The prediction of suicidal behavior continues to be among the most challenging tasks in psychiatry," Dr Melhem observed.
"A recent meta-analysis of the past 50 years of research showed that the prediction of suicidal ideation, attempt, and suicide deaths was only slightly better than chance, which is unacceptable for a life-threatening outcome like suicidal behavior," she said.
"Here, we focused on symptoms and their variability over time because, while psychiatric diagnoses are important predictors of suicidal behavior, diagnoses and stable or trait-like predictors are of limited value because the risk of suicidal behavior varies during the course of psychiatric illness," she reported.
The researchers "examined the trajectories of impulsivity, aggression, impulsive aggression, depression symptoms, irritability, and hopelessness in a longitudinal study of offspring of parents with mood disorders."
They examined whether changes in these measures over time could predict suicidal attempts and time to onset of the attempt, above and beyond any psychiatric diagnoses and other predictors.
In addition, they computed a risk score according to their models and assessed its performance.
The study included 663 offspring of 318 parents (referred to as probands) with a lifetime history of mood disorders; more than half had a lifetime history of an actual suicide attempt.
Recruitment of participants took place between July 15, 1997 and September 6, 2005, and participants were followed through January 21, 2014.
"Suicide attempt" in offspring was defined in both "broad" and "narrow" terms.
"Broad" included an actual attempt or a suicide-related behaviour whereas "narrow" included only actual attempts.
A 10-fold cross-validation model was used for statistical analysis; machine-learning regression techniques were used to determine a subset of predictors with the strongest effects.
Probands and offspring were interviewed at baseline, and then at annual follow-ups for a 12-year period using a variety of established scales.
Monitor Severity, Variability
The number of male and female offspring were approximately even (47.7% female), with a mean age of 23.8±8.5 years; most participants (70.3%) were white.
Participants were followed for a median (range) of 8.1 (95% CI: 1, 15.4) years.
Among the offspring, 71 (10.7%) had suicide attempts over the course of the study. Of these, 51 were first-time attempts (incidence rate=8.4%).
Among individuals who attempted suicide, the mean number of suicide attempts and actual attempts were 1.2±0.6 and 1.3±0.7, respectively.
The mean lethality (for actual attempts only) was 1.7±2, "which corresponded to physical damage for which medical attention was needed," the investigators wrote.
The median time from the last assessment point to suicide attempt was 45 (95% CI: 1, 126) weeks and 30 (95% CI: 4, 126) weeks for actual attempt.
The researchers identified a 2-class model for each of the factors they identified and found that class 2 consistently showed higher mean scores and variability vs class 1 in measures of hopelessness, impulsivity, aggression, and irritability.
For depression, they identified a 3-class model, in which class 3 showed the highest mean scores and variability (2.08±0.75 vs 0.59±0.42 vs −0.44±0.28 [class 1]; F2644=1216.7; P <.001; Cohen d class 3 vs 2=2.97; Cohen d class 3 vs 1=7.31; Cohen d class 2 vs 1=3.17).
Participants with suicide attempt were more likely than participants without attempt to belong to class 3 depression symptoms, with higher mean and variability (16 [22.9%] vs 27 [4.7%]; P <.001; Cohen d =0.72), the authors reported.
They were also more likely to belong to class 2 for impulsivity, aggression, impulsive aggression, and irritability.
The only statistically significant trajectory that predicted increased risk for suicide attempt in the offspring turned out to be class 3 depression symptoms with higher mean and variability (model 1 odds ratio [OR]=6.53 [95% CI: 2.53,16.87]; t =3.88; P
Class 3 depression symptoms remained associated with an almost 8-fold increased risk for suicide attempt (OR=7.69 [95% CI: 2.37, 24.90]; t =3.40; P =.001) when the researchers included the trajectories for all measures, even after controlling for demographics and proband history of actual attempt.
In fact, class 3 depression symptoms were the only statistically significant trajectory to predict suicide attempt and remained so, even after the researchers controlled for additional clinical characteristics (OR=4.72 [95% CI: 1.47, 15.21]; t =2.60; P =.01).
"We recommend that clinicians monitor and treat depression symptoms over time to reduce their severity and fluctuation in high-risk young adults to reduce their risk for suicide attempt," Dr Melhem said.
New Tool in the Toolbox
In addition to class 3 depression symptoms, the only ones that remained statistically significant in predicting the narrow and index attempt were:
- Younger age (≤30 years) (OR=0.82 [95% CI: 0.74, 0.9])
- Lifetime history of unipolar and bipolar disorder (OR=4.71 [95% CI: 1.63, 13.58] and OR=3.4 [95% CI: 0.96, 12.04], respectively)
- History of childhood abuse (OR=2.98 [95% CI: 1.4, 6.38])
- Proband actual attempt (OR=2.24 [95% CI: 1.06, 4.75])
The researchers computed a risk score using these statistically significant predictors.
A risk score of ≥3 resulted in the highest sensitivity (87.3%) and moderate specificity (63%; area under the curve=0.80) for suicide attempt.
Although the positive predictive value was low, it increased as the prevalence of attempts increased.
"The PRS is not only a valuable addition to the physician's toolkit in predicting suicide risk, but it can also be done at little cost since the information is already being collected as part of standard evaluations," Dr Melhem commented.
Greater Predictive Accuracy
Commenting on the study for Medscape Medical News, Beth Salcedo, MD, board president of the Anxiety and Depression Association of America, said that it "underscores the need for more and better treatments for depression," since we need to be "confident that the treatments we have cannot only get people better but also keep people better and control symptoms over time."
Dr Salcedo, who is also the medical director of the Ross Center, a large outpatient practice with offices in Washington, DC and New York, New York, noted that it "has been long known that depression is a predictor of suicidal ideation and attempts, and the more severe the symptoms, the more concern one should have about that risk."
"It would be great if this new data could lead to a more accurate way of assessing an individual's risk, as this is something we are not at all able to do in current clinical practice," said Dr Salcedo, who was not associated with the current study.
Dr Melhem agreed, stating, "We are working on further improving the accuracy of this prediction model by adding objective biological markers."
The study was funded by the National Institute of Mental Health. Melhem reports receiving research support from the National Institute of Mental Health, Brain and Behavior Research Foundation, and the American Foundation for Suicide Prevention. The other authors' disclosures are listed on the original paper. Salcedo has disclosed no relevant financial relationships.
- Researchers enrolled 663 offspring of 318 parents diagnosed with mood disorders treated in inpatient units in New York and Pennsylvania. More than half of parents had a history of suicide attempt. Offspring were recruited between 1997 and 2005 and followed into 2014. They were screened annually with a battery of psychiatric tests, including measurements of suicidality.
- Suicide attempt was defined broadly as a true attempt or aborted effort or suicidal ideation that prompted emergency referral. Researchers also performed an analysis limited to participants with a true suicide attempt.
- The main study analysis focused on longitudinal clinical features that might influence the risk for suicide attempt. Researchers also evaluated other variables traditionally associated with the risk for suicide.
- 47.7% of the 663 participants were female, and the mean age of participants at the time of censored observations was 23.8±8.5 years; 70.3% of the cohort was white.
- The median follow-up time was 8.1 years, and 10.7% had a suicide attempt during this time. The mean number of suicide attempts per participant who had attempted suicide was 1.2±0.6.
- More severe and variable symptoms of depression were the only clinical characteristics significantly associated with the risk for suicide attempt over time. Compared with participants without depression, the OR for suicide attempt among participants with severe depression was 4.72 (95% CI: 1.47, 15.21). This risk was similar for both the broad and narrow definitions of suicide attempt.
- The finding of severe depression was associated with a sensitivity of 87.3% and specificity of 63% for suicide attempt.
- Impulsivity, impulsive aggression, and irritability over time were not significantly linked with the risk for suicide attempt.
- Age ≤30 years was predictive against suicide attempt (OR=0.82 [95% CI: 0.74, 0.90]). Lifetime histories of unipolar and bipolar disorders were associated with ORs for suicide attempt of 4.71 (95% CI: 1.63, 13.58) and 3.4 (95% CI: 0.96, 12.04).
- A history of child abuse was associated with an OR for suicide attempt of 2.98 (95% CI: 1.4, 6.38), and the respective OR for suicide attempt by a parent was 2.24 (95% CI: 1.06, 4.75).
- The principal finding of a previous meta-analysis by Franklin and colleagues was that no prediction tool for suicide was highly sensitive and specific. Moreover, there was no trend toward improved performance of predictive abilities over time, and risk factors for suicide remained fairly stagnant over the 50 years of included research. Studies with longer periods of follow-up were not superior to shorter studies in terms of finding meaningful risk factors for suicide.
- In the current study by Melhem and colleagues, more severe and variable symptoms of depression were the only clinical characteristic significantly associated with the risk for suicide attempt over time.
- Implications for the Healthcare Team: The healthcare team should closely track symptoms of depression among young people. More severe symptoms may be a sign of suicidality.