Taxometric results showed very strong evidence that serious suicide risk is a discrete category (e.g., like presence or absence of strep throat) as opposed to dimensional (e.g., like the amount of back pain). In fact, the robust nature of our taxometric results consistently surpassed the most stringent criteria used to establish a categorical solution.
What differentiates people who die by suicide from other people? This is a key question for any clinician or professional who would like to be able to predict which people are most at risk for death by suicide. In particular, suicide researchers would like to know whether suicidality a categorical trait—that is, people either are seriously suicidal, or they aren’t—or whether it is more of a continuous trait, where people can fall anywhere on a spectrum from not suicidal at all to highly suicidal.
Now two researchers funded by the Military Suicide Research Consortium, Jill Holm-Denoma at the University of Denver and Tracy Witte of Auburn University, have offered the most definitive answer yet to this important question. By analyzing a trove of data collected by various other researchers funded by the Consortium, Holm-Denoma and Witte were able to conclude that suicidality is best thought of as a categorical trait—that is, it is most accurate to assume that a person either is or isn’t seriously suicidal and then try to figure out which of the two it is, rather than trying to place a person on a spectrum from highly suicidal to not suicidal at all. This understanding has important implications both for the assessment of suicidality and for its treatment.
Historically, Holm-Denoma explains, many suicide specialists have assumed that suicidality is a categorical trait much like schizophrenia. Thus it made sense to try to group people into risk categories: those at high risk of attempting suicide versus those at low risk, or perhaps high, medium, and low risk. But there have been various problems associated with such categorizations, Holm-Denoma says, and it has proved difficult to come up with clear, reliable divisions, so it remained possible that suicidality is more a continuous trait than a categorical one.
To get a more definitive answer, Holm-Denoma and Witte analyzed data from 1,773 men and women, most of them with a military connection, from 16 different studies funded by the Military Suicide Research Consortium. Because all of these study participants had completed the Common Data Elements assessment, a 57-item questionnaire assessing suicidality and related factors, Holm-Denoma and Witte could look for suicide-relevant patterns among this large group of subjects, many of whom had participated in the Consortium studies because they had suicidal thoughts or had attempted suicide, or both.
To find out from these subjects whether suicidality is a categorical trait or a continuous train, Holm-Denoma and Witte used a technique called taxometric analysis. Holm-Denoma explains the technique with a simple example: If you took data on height and baldness from a large group of adults, you would discover that the two were related—that the bald individuals were likely to be taller than average. But the two traits are correlated only because men are both most likely to be bald and are taller on average than women, and if you restrict your analysis to only men or to only women, the correlation between height and baldness disappears. This indicates that the population can be divided into two natural groupings, or taxa—a taxon for men and a taxon for women. It is possible to apply this sort of analysis to any population in order to discover whether a taxon exists. “It’s a series of statistical techniques,” Holm-Denoma explains. “You look at convergence across a variety of techniques to tell if there is a taxon present.”
However, it is possible to miss a taxon if the sample is not drawn up the right way—for example, if there are not enough examples of the relevant phenomenon in the sample. This is why it was so important that Holm-Denoma and Witte were able to get common data from the nearly 1,800 subjects from the 16 different Consortium studies. The large number of subjects provided them with enough statistical power to determine with relative certainty whether suicidality is categorical or dimensional.
What they found, using standard taxometric techniques such as MAMBAC and MAXCOV, was that suicidality is clearly categorical nature—that is, that people who score high on measures of suicide risk are indeed categorically different from those with low scores. Thus it makes sense for clinicians to ask which category a patient belongs to.
Furthermore, Holm-Denoma says, their analysis suggests that the high-risk and low-risk categories are doing more than simply classifying subjects based on their levels of emotional distress; instead, the subjects in the high-risk category score higher on measures that specifically point to an increased risk of suicidality. A key limitation of this study was that the researchers were not able to examine directly whether people in the high-risk category were more likely than those in the low-risk category to attempt suicide in the future. This is something that will need to be tested in subsequent research studies, Holm-Denoma says.
These results have a number of important implications for clinicians. For one thing, the analysis indicates that it should be possible to accurately place patients in a high- or low-risk group using fewer items on an assessment, thus making it possible to assess patients more quickly without losing any validity. The analysis also suggests which specific measures of suicidality would make sense to include on a brief assessment; however, Holm-Denoma says, it will require a future study to test such an assessment and make sure that it actually does place people accurately into high-risk and low-risk groups.
In addition to the implications for assessing suicide risk, Holm-Denoma and Witte’s results may also help improve suicide intervention efforts. For example, some of the most powerful and effective suicide interventions require a great deal of time and other resources. Since such resources are limited, it is important to be able to identify those patients in the high-risk category and devote those resources to them.
Finally, the results of the study have implications for future suicide research. Since people in the high-risk group are categorically different from those in the low-risk group, it makes sense to focus suicide research on those who fall in the high-risk category. Yet, much current suicide research is done on subjects—such as college students who volunteer for a study—who are at low risk of suicide. This is not the best approach, Holm-Denoma says, for such things as identifying risk factors among those people who are actually most likely to attempt suicide.
In short, Holm-Denoma and Witte’s research tells us that the people most likely to attempt suicide are indeed a distinct category of people, and future clinical work with and research on suicidal patients should take that into account.