Why Are Suicidal Thoughts and Actions Difficult to Explain and Predict?

Distinguish outcomes, compare explanations, and ask what prediction and prevention evidence can establish.

Level: Intermediate
Reading load: 9 core readings in 5 units
Prerequisites: Some familiarity with psychological research and the difference between association and causation. What Is a Mental Disorder? and How Do Psychological Disorders Develop and Persist? provide useful preparation. Statistical terms are introduced in the guides.
Last reviewed: September 2026

Introduction

Explanation, prediction, and prevention meet in this field without becoming the same achievement. A factor associated with suicidal thinking may reveal little about later action; a model may separate groups while estimating probabilities poorly; an intervention may reduce events despite incomplete forecasting or mechanistic knowledge. Understanding the evidence begins by giving each of these questions its own outcome, population, and timescale.

This path begins with what researchers measure, then compares explanations and prospective evidence. It moves from model evaluation to changes within a person over hours, and ends with school and post-detention prevention trials. Throughout, distinguish suicidal thoughts, attempts, deaths and nonsuicidal self-injury. They overlap empirically without forming one inevitable progression.

The aim is to evaluate a claim by specifying its outcome, population and time horizon. Follow the reasoning through the statistical detail: what comparison was made, what was observed, and what would still have to be shown? The longer model review may take more than one sitting.

Unit 1 of 5

What exactly is being explained or predicted?

Before comparing theories, establish what counts as the outcome. Millner and colleagues expose the consequences of brief labels; Nock then offers a functional account of nonsuicidal self-injury. Classification, function and intent each require evidence.

Reading
1 of 9
Measuring Suicidal Behaviors
Millner, A. J., Lee, M. D., & Nock, M. K. (2015). Single-item measurement of suicidal behaviors: Validity and consequences of misclassification. PLOS ONE, 10(10), Article e0141606. https://doi.org/10.1371/journal.pone.0141606
Methodological contribution · Moderate · Full work: 17 pages
Follow the movement from a survey label to a classified account.
Why you’re reading it. Establish why the denominator and measurement process matter before evaluating risk factors.
A question to consider. If two studies use the same outcome label but classify different experiences together, what should be checked before combining their results?
Reading
2 of 9
Why Do People Hurt Themselves?
Nock, M. K. (2009). Why do people hurt themselves? New insights into the nature and functions of self-injury. Current Directions in Psychological Science, 18(2), 78–83. https://doi.org/10.1111/j.1467-8721.2009.01613.x
Foundational argument · Moderate · Full work: 6 pages
Separate a proposed maintaining function from the definition of the behavior.
Why you’re reading it. Introduce nonsuicidal self-injury as an outcome requiring explanation in its own right.
A question to consider. How could two behaviors share a proposed function while differing in intent, and what evidence would prevent you from simply assuming that function?

Unit 2 of 5

What could explain a transition from thoughts to action?

Theories can agree that thoughts and actions require different explanations while disagreeing about the processes involved. Compare those proposals, then examine a prospective cohort. Temporal order improves the question but does not remove confounding or turn associations into a forecasting tool.

Reading
3 of 9
Ideation-to-Action Theories of Suicide
Klonsky, E. D., Saffer, B. Y., & Bryan, C. J. (2018). Ideation-to-action theories of suicide: A conceptual and empirical update. Current Opinion in Psychology, 22, 38–43. https://doi.org/10.1016/j.copsyc.2017.07.020
Major synthesis · Foundational argument · Moderate · Full work: 6 pages
Compare the theories’ proposed pathways and the time scales on which they should operate.
Why you’re reading it. Turn the ideation-to-action distinction into competing, testable explanations.
A question to consider. Choose two accounts and describe a finding that would favor one over the other without merely renaming the observed outcome.
Reading
4 of 9
Predicting Future Attempts in an Adolescent Cohort
Mars, B., Heron, J., Klonsky, E. D., Moran, P., O'Connor, R. C., Tilling, K., Wilkinson, P., & Gunnell, D. (2019). Predictors of future suicide attempt among adolescents with suicidal thoughts or non-suicidal self-harm: A population-based birth cohort study. The Lancet Psychiatry, 6(4), 327–337. https://doi.org/10.1016/s2215-0366(19)30030-6
Landmark empirical study · Moderate · Full work: 11 pages
Track who enters each baseline group and who is observed at follow-up.
Why you’re reading it. Examine an actual prospective comparison after the theoretical proposals.
A question to consider. Which part of a theory could this five-year design test, and which claim about an immediate transition would remain unresolved?

Unit 3 of 5

What would count as a useful prediction?

A significant association, good ranking and accurate probabilities are different accomplishments. Begin with a critique of model evaluation, then inspect a systematic appraisal. These papers share an author and a methodological perspective; treat their arguments and disclosed model-development involvement explicitly.

Reading
5 of 9
How Should Suicide Prediction Models Be Evaluated?
Whiting, D., & Fazel, S. (2019). How accurate are suicide risk prediction models? Asking the right questions for clinical practice. Evidence Based Mental Health, 22(3), 125–128. https://doi.org/10.1136/ebmental-2019-300102
Critical reappraisal · Methodological contribution · Moderate · Full work: 4 pages
Keep the denominator of each performance measure visible.
Why you’re reading it. Learn why usefulness requires a specified comparison and evidence about consequences.
A question to consider. In an invented dataset with 100 events among 10,000 observations, a model identifies 80 events but also flags 990 non-events. What fraction of its positive predictions is correct, and what does that number leave unanswered?
Reading
6 of 9
Prediction Models for Self-Harm and Suicide
Seyedsalehi, A., Bailey, J., Ogonah, M. G. T., Fanshawe, T. R., & Fazel, S. (2025). Prediction models for self-harm and suicide: A systematic review and critical appraisal. BMC Medicine, 23(1), Article 549. https://doi.org/10.1186/s12916-025-04367-6
Major synthesis · Contemporary update · Methodological contribution · Challenging · Full work: 21 pages
Follow one performance claim from the reported metric through validation and risk-of-bias appraisal.
Why you’re reading it. Apply the earlier evaluation questions to a larger body of model research.
A question to consider. A model ranks outcomes well in a new dataset. What calibration and impact evidence would you still request before accepting a claim of practical benefit?

Unit 4 of 5

What changes within a person over hours?

The earlier cohort asks about years, but many proposed processes could operate much faster. Repeated reports make fluctuations visible. This unit uses one paper with two samples to examine the extra information gained—and the gap between observing change and predicting action.

Reading
7 of 9
Real-Time Fluctuations in Suicidal Ideation
Kleiman, E. M., Turner, B. J., Fedor, S., Beale, E. E., Huffman, J. C., & Nock, M. K. (2017). Examination of real-time fluctuations in suicidal ideation and its risk factors: Results from two ecological momentary assessment studies. Journal of Abnormal Psychology, 126(6), 726–738. https://doi.org/10.1037/abn0000273
Landmark empirical study · Methodological contribution · Moderate · Full work: 13 pages
Distinguish simultaneous covariation from prediction of the next report after accounting for the current one.
Why you’re reading it. Examine why the temporal resolution of evidence must match the explanatory question.
A question to consider. How could a factor distinguish people over years yet fail to anticipate changes over hours? Give a measurement explanation and a substantive explanation.

Unit 5 of 5

What can prevention evidence establish?

Prevention need not wait for a certain forecast about each person. Compare a school-level trial with an individually randomized trial after detention. Revisit outcome definitions, populations and time horizons, then ask whether an effect identifies a particular mechanism.

Reading
8 of 9
School-Based Prevention: The SEYLE Trial
Wasserman, D., Hoven, C. W., Wasserman, C., Wall, M., Eisenberg, R., Hadlaczky, G., Kelleher, I., Sarchiapone, M., Apter, A., Balazs, J., Bobes, J., Brunner, R., Corcoran, P., Cosman, D., Guillemin, F., Haring, C., Iosue, M., Kaess, M., Kahn, J.-P., … Carli, V. (2015). School-based suicide prevention programmes: The SEYLE cluster-randomised, controlled trial. The Lancet, 385(9977), 1536–1544. https://doi.org/10.1016/s0140-6736(14)61213-7
Landmark empirical study · Moderate · Full work: 9 pages
Compare the offered intervention, the analytic sample and the outcomes actually recorded.
Why you’re reading it. Evaluate a population approach alongside the earlier concerns about individual forecasting.
A question to consider. Using Millner’s measurement argument, what could a programme-related change in reporting mean for interpretation of a self-reported outcome?
Reading
9 of 9
Prevention After Pretrial Jail Detention
Weinstock, L. M., Jones, R. N., Miller, T. R., Brown, G. K., Arias, S. A., Graves, H. R., Miller, I. W., Cerbo, L., Rexroth, J., Fitting, H., Russell, D., Kubiak, S., Stein, M. D., Matkovic, C., Yen, S., Gaudiano, B. A., & Johnson, J. E. (2025). Safety planning vs standard care for suicide prevention after pretrial jail detention: A randomized clinical trial. JAMA Network Open, 8(11), Article e2543156. https://doi.org/10.1001/jamanetworkopen.2025.43156
Landmark empirical study · Contemporary update · Moderate · Full work: 14 pages
Keep composite events, repeat attempts and deaths separate when interpreting the findings.
Why you’re reading it. Bring the path together: a causal comparison can support an intervention without settling its mechanism.
A question to consider. How does this trial’s evidence differ from SEYLE’s in population, randomization and outcome—and which explanatory question remains open in both?

Scope and omissions

This intermediate path examines research questions, not personal risk assessment or treatment delivery. It does not survey every theory, diagnosis, prevention programme or social determinant. Nonsuicidal self-injury and broader self-harm categories are distinguished rather than treated as interchangeable.

Much of the selected evidence comes from Europe and North America, with important recruitment and follow-up restrictions. The school and detention settings expose different institutional contexts, but neither stands for all communities. Forecasting models, functions and group associations must not become deterministic descriptions of individuals. Detailed psychotherapy, crisis-service organization and population policy require additional paths.

Synthesis

Precision about the target makes explanations and forecasts more informative. Measurement can blur differences, a theory of thinking may leave action unexplained, and prospective association remains short of a complete causal account. Model performance changes with the metric, setting, and time horizon. Repeated observations reveal movement that longer intervals miss, reminding us that a group pattern and an individual course are different objects of inquiry.

The prevention trials add a further distinction. Evidence that an intervention changes a specified outcome does not depend on perfect individual prediction, and it does not by itself identify the active mechanism. Counts, first occurrences and deaths must retain their separate meanings.

Choose a research question and define its outcome, population and time window. Compare two possible explanations and an observation that could distinguish them. Identify a variable that might improve prediction without being a cause. Then propose an ethical prevention comparison: what would a beneficial result establish, what would remain uncertain, and whose experience would the study leave out?

For a broader foundation for these distinctions, revisit How Do Psychological Disorders Develop and Persist?. To connect the argument to measurement across psychology, continue with How Do We Know What a Psychological Test Measures?.