WS More or Less: Is Suicide Seasonal?

Episode Summary

Executive Summary: The episode examines myths and evidence about seasonal patterns in suicide, finding that in the UK risk is highest on 1 January and low on Christmas Day, while older studies often showed a spring peak. It then explores why prediction remains difficult, noting that traditional risk factors barely beat chance, though machine-learning models may improve accuracy if validated in real-world clinical use.

Main Topics: Seasonal patterns in suicide: The program contrasts long-held beliefs about a spring/summer peak with newer UK data suggesting January, especially New Year’s Day, is the highest-risk period. Christmas and New Year myths: Listener concern about Christmas and January prompts a review showing December 25 has lower risk, while January 1 is markedly higher. Why seasonal effects might exist: Researchers discuss speculative explanations such as the 'broken promise effect' and serotonin changes, but emphasize that evidence is weak or absent. Limits of suicide risk prediction: A psychologist explains that decades of research on risk factors only predicts suicidal thoughts and behavior slightly better than chance. Machine learning and future prediction: New algorithms using many variables simultaneously show promise for improving prediction, but have not yet been tested in clinical trials. Prevention and support: The episode ends by stressing practical help-seeking, encouraging listeners to direct worried individuals toward GPs, mental health professionals, and support services.

Key Arguments: Older research across multiple countries found a robust late-spring/early-summer peak in suicide, but UK data suggest the pattern may be changing. In the UK study cited, December had lower suicide risk than January and April, and 1 January was the single riskiest day. The high risk at the start of the year is not likely to be a recording artefact from date-entry errors. Theories for seasonal spikes, including the 'broken promise effect' and serotonin changes, remain speculative and unsupported by strong evidence. Traditional predictors like severe depression are poor at forecasting suicide because most depressed people do not die by suicide. Machine-learning approaches can combine hundreds of factors and may improve prediction substantially, but their real-world usefulness is unproven without prospective trials. The most evidence-based response for concern about someone is to encourage professional help and social support rather than rely on prediction alone.

Data Points: Annual global deaths by suicide: 800,000 - Estimated by the World Health Organisation at the start of the program Highest monthly risk in UK study: January - Month with the highest suicide risk, though not significantly higher than spring Lowest monthly risk in UK study: December - About 10% lower suicide risk than January and April Highest-risk day: 1 January - In the UK study, the day with the highest suicide risk Risk increase on 1 January: About 40% higher - Compared with the rest of the year Risk on 25 December: About 26% lower - Compared with the rest of the year Chance of predicting suicidal thoughts/behaviors with traditional models: Marginally better than chance - Jessica Ribeiro’s meta-analysis across 50 years of research Chance level AUC: 0.5 - Benchmark used to describe random prediction Machine-learning AUC range: 0.8 to 0.92 - Reported prediction performance of risk algorithms Perfect prediction AUC: 1.0 - Theoretical upper bound for the statistic Lifetime suicide probability among severely depressed individuals: 2% - Used to show why depression alone is a weak predictor General U.S. lifetime suicide probability: 1.6% - Compared with the rate among severely depressed individuals

Pivotal Quotes: "the month with the highest suicide risk was January, but it wasn't significantly higher than spring" — Nav Kapoor: Summarizing the UK seasonal findings "Our results were a bit sobering in that our ability to predict suicidal thoughts and behaviours was only marginally better than chance, so almost coin flip levels." — Jessica Ribeiro: Describing the limitations of traditional suicide risk prediction "The best evidence is that people seek and signpost to appropriate help." — Nav Kapoor: Final advice on prevention and intervention

Implications: Listeners should not assume Christmas is the main danger period; New Year’s Day may be more concerning. More importantly, suicide prediction is still limited, so prevention should focus on timely support, professional help, and better-validated tools.