A World Psychiatry umbrella review synthesised 69 meta-analyses covering 970 studies and 221 putative risk or protective factors for psychotic disorders, grading each association by strength and estimating the share of cases each modifiable factor could in principle account for.

Six factors reached the top tier of evidence, including low birthweight (under 2,000 g, odds ratio 1.77) and young paternal age. Among potentially modifiable factors, the largest population attributable fractions were childhood adversity (about 41%), Toxoplasma gondii seropositivity (about 20%), daily or near-daily cannabis use (about 13%) and the clinical high-risk state (about 13%). Several ethnicity and migration findings carried convincing evidence, reflecting social adversity rather than biology.

For clinic, this is a map for counselling and case-finding rather than a screening tool. It supports asking about childhood adversity and heavy cannabis use, taking the clinical high-risk state seriously, and remembering that much psychosis risk is social and potentially modifiable. These are associations, so they guide where to look, not what causes an individual's illness.

Umbrella review of 69 meta-analyses, 970 studies and 221 risk or protective factors for psychosis.
Top-tier evidence included low birthweight (under 2,000 g, odds ratio 1.77) and young paternal age.
Largest modifiable population attributable fractions: childhood adversity about 41%, Toxoplasma about 20%, heavy cannabis about 13%.
Ask about childhood adversity and near-daily cannabis use, and treat the clinical high-risk state seriously.
Several ethnicity and migration associations reflect social adversity, not biology.

Why it matters

It separates the many weak associations around psychosis from the few strong, modifiable ones worth acting on in prevention and counselling.

The statistics, in plain English

A population attributable fraction estimates how many cases might be avoided if a factor were removed, but it assumes the association is causal, which these observational data cannot prove; higher evidence classes mean more consistent, larger pooled associations.