Development and initial validation of a multivariable predictive Early Adversity Scale for Schizophrenia (EAS-Sz) using register data to quantify environmental risk for adult schizophrenia diagnosis after childhood exposure to adversity

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)


BACKGROUND: Additional to a child's genetic inheritance, environmental exposures are associated with schizophrenia. Many are broadly described as childhood adversity; modelling the combined impact of these is complex. We aimed to develop and validate a scale on childhood adversity, independent of genetic and other environmental liabilities, for use in schizophrenia risk analysis models, using data from cross-linked electronic health and social services registers.

METHOD: A cohort of N = 428 970 Western Australian children born 1980-2001 was partitioned into three samples: scale development sample ( N = 171 588), and two scale validation samples (each N = 128 691). Measures of adversity were defined before a child's 10th birthday from five domains: discontinuity in parenting, family functioning, family structure, area-level socioeconomic/demographic environment and family-level sociodemographic status. Using Cox proportional hazards modelling of follow-up time from 10th birthday to schizophrenia diagnosis or censorship, weighted combinations of measures were firstly developed into scales for each domain, then combined into a final global scale. Discrimination and calibration performance were validated using Harrell's C and graphical assessment respectively.

RESULTS: A weighted combination of 42 measures of childhood adversity was derived from the development sample. Independent application to identical measures in validation samples produced Harrell's Concordance statistics of 0.656 and 0.624. Average predicted time to diagnosis curves corresponded with 95% CI limits of observed Kaplan-Meier curves in five prognostic categories.

CONCLUSIONS: Our Early Adversity Scale for Schizophrenia (EAS-Sz), the first using routinely collected register data, predicts schizophrenia diagnosis above chance, and has potential to help untangle contributions of genetic and environmental liability to schizophrenia risk.

Original languageEnglish
Pages (from-to)4990-5000
Number of pages11
JournalPsychological Medicine
Issue number11
Early online date12 Jul 2022
Publication statusPublished - 12 Aug 2023

Cite this