274Early adversity scale for schizophrenia (EAS-SZ) constructed and validated using linked register data
Abstract Background A range of adversities have been implicated as risks for schizophrenia. Adversities often cluster, with synergistic impact, which may vary by age of exposure. We expand on current understanding, and propose a method for ranking combinations of adversities associated with risk of schizophrenia, to derive a risk prediction measure. Methods We used prospectively collected data for 430,000 children born 1980–2001 in Western Australia, and their parents. Follow up continued until 2015 using linked State registers, identifying 1,620 children with schizophrenia. Five domains of adversity exposure to age 10 were considered. Using Cox modeling of a 40% training data subset, we categorised adversity exposure by associated rates of schizophrenia. Firstly, for each domain separately, numerous constructs of adversity exposure were screened for independent association with schizophrenia. Those with p < =0.2 were combined, using augmented backwards elimination, to define a minimal domain set of jointly associated constructs. Combination was summarised as the linear predictor corresponding to the optimum Cox model. Secondly, these domain summaries were combined with each other to form a global Cox model, predictive of association with schizophrenia. Harrell’s Concordance was calculated using a 30% assessment sample which did not overlap with our training sample. Prognostic categories were defined and tested. Results Harrell’s Concordance was 0.655. Dose response was observed. Conclusions Our scale combines many adversity measures into a single construct. It displays dose response and predicts association considerably above chance. Key messages EAS-SZ allows a range of adversity exposure profiles to be ranked according to association with schizophrenia.