Dr. Gregory Brown discussed overall needs in the field of measurement and assessment of suicidal risk. In the 1999 Surgeon General's “A Call to Action to Prevent Suicide,” one of the points made was the need to advance the science of suicide prevention. Dr. Brown stressed the importance of suicide prevention programs, including evaluation. This requires developing reliable and valid measures of suicide ideation and suicidal behavior.

The effectiveness of prevention programs to prevent suicide is unknown. Suicidal people are routinely and purposefully excluded from the vast majority of randomized clinical trials for psychiatric disorders. Further confounding the issue, most studies do not use screening measures for suicidal ideation, but frequently use a single item from the Hamilton Depression Scale. Only a small number of these studies report a change in suicidality, but without proper measurement the meaning is unclear.

Suicide ideation is an independent predictor of suicide risk above and beyond affective illness.

Gregory Brown

The field of measurement would benefit from two changes: the use of screening measures with appropriate psychometric properties in clinical studies, and the adoption of common, operationalized nomenclature.

Dr. Brown briefly reviewed the existing measures. These include quick screening measures that have been used in community surveys, clinical trials, and primary care settings. One of the most widely used and best evaluated measures is the Scale for Suicide Ideation (SSI). It is a 19-item scale, available as interview, self-report, and computer-administered. If a person endorses an item indicating intent to commit suicide, then the rest of the scale is administered. It has been standardized on both inpatient and outpatient psychiatric samples. It has also been used in emergency rooms, primary care settings, jails, and in college student samples.

Dr. Brown described a study he did with Dr. Aaron Beck involving almost 7,000 patients. This was a prospective study with up to 20-year follow-up with psychiatric outpatients. Patients received standardized, structured interviews and standardized assessment measures. These data were matched to the National Death Index, and death certificates were obtained for those who had died. Through this process, 49 suicide cases were identified. The average length of follow-up was ten years, and the average length of time to death was approximately 4.3 years from the baseline interview. Patients who scored above 3 on the SSI were about 6.5 times more likely to commit suicide than patients who scored below this cut-off. The scores on several other scales were also elevated: the Beck Hopelessness Scale (about 4.4 risk ratio), Beck Depression Inventory, Beck Anxiety Scale, and the Hamilton Rating Scale for Depression. Psychiatric hospitalization, suicide attempts, bipolar disorder, major depression, and suicide ideation were significant risk factors. Analyses further revealed that “recurrent depression is much more predictive of completed suicide than single episodic depression, underscoring] the chronicity of the disorder may very well be a key risk factor for suicide.”

Trying to predict suicide, a binary event occurring at a low base-rate, is very difficult, as further discussed by Drs. Robert Gibbons and Ming Tsuang. The best approach has been to use the Poisson distribution which is the limiting form of the binomial. This method allows statistical prediction of an event, but it does not allow exact prediction. Another problem is the underestimation of suicide by using only death certificates as Dr. Tsuang pointed out. He noted that family members sometimes present suicide notes, even though the death certificate states cause of death as “unknown” or “accidental.”

Dr. Brown summarized by making several recommendations. (1) There should be more consistent use of measures across studies to make comparisons possible. It is important to be able to examine findings across studies since suicide is a low base rate event. (2) Multivariate models are preferable for predicting risk; they have greater predictive value, as compared to single measures. (3) Measures need to be developed for the elderly, minorities, males, and females, since there is some evidence that suicide ideation and behavior varies across groups. (4) Refine and research a universally adopted nomenclature. (5) Use complete, standardized suicide assessment measures in clinical trials, not single items, and not the sole use of non-suicidal measures. (6) Develop greater interest among clinical researchers for working with suicidal patients. (7) Include evaluation components in all suicide programs.