Optimizing Suicide Research

A recent article reviews challenges in suicide research and proposes ways to address them.

The authors discuss multiple sources of bias that may limit the usefulness and validity of findings in current suicide research. These include publication bias, in which negative findings are less likely to be published, skewing the body of research in favor of those that do get published. They also note a bias toward publishing studies that demonstrate statistical significance without considering effect sizes (i.e., the strength of findings). This can result in studies of risk factors that are not robust enough to advance research and prevention practices.

In addition to these current challenges, the authors warn against a practice called “p-hacking,” in which researchers develop a hypothesis after running multiple exploratory analyses in search of significant findings, instead of setting out to confirm a hypothesis. While this practice is common in clinical research, it is unclear how widespread it is in suicide-related studies.

Other challenges discussed include low power (i.e., low chance of finding a real statistically significant effect) due to small sample sizes (i.e., small number of people studied). The authors also argue that definitions of constructs such as suicide risk, intent, and attempts vary across studies, which limits the replication of positive results and makes it difficult to interpret the body of research.

To help identify and address these biases and limitations in suicide research, the authors advocate for more “open science,” such as:

• Providing free and open access to research products such as conference papers, book chapters, and dissertations. Research could be archived in online repositories that allow for a more extensive review of research protocols and analyses by peers prior to publication. A preregistration process like that used by the National Institutes of Health could require that researchers specify their analysis plans upfront, ensuring that exploratory and confirmatory analyses are differentiated from one another. 
• Describing detailed methodology in research manuscripts, including issues of effect size and power.
• Encouraging the use of standard data collection instruments to facilitate the aggregation of findings across studies. This could provide sample sizes large enough to power the statistical models needed to test hypotheses.
• Performing preliminary analyses of data during data collection, which could allow researchers to decide whether to stop or continue their efforts.
• Replicating studies that form the basis for key suicide prevention theories to ensure that findings are consistent and robust, which could also help with determining which findings are valid.

Carpenter, T. P., & Law, K. C. (2021). Optimizing the scientific study of suicide with open and transparent research practices. Suicide and Life-Threatening Behavior, 51, 36–46.