MODEL SPECIFICATIONS. Comedian racial group membership (Asian = 0, White = 1) was specified as the level-1 predictor; joke funniness, offensiveness, interpersonal, and intergroup anxiety were specified as level-1 mediators; and stereotype agreement was specified as a level-1 outcome variable. This model also estimates the direct effect between the predictor (comedian racial group membership) and the outcome variable (stereotype agreement).
MODEL SPECIFICATIONS. Comedian racial group membership (Asian = 0, White = 1) was specified as the level-1 predictor; joke funniness, offensiveness, interpersonal, intergroup anxiety, and stereotype agreement were all specified as the level-1 outcome variables. To model the high skewness of the two anxiety variables, this model assumed that these two variables were from the Gamma distribution/family. The rest outcome variables were assumed to be from the Gaussian (normal) distribution/family. The identity link function was used for all the estimations, including the Gamma distributed anxiety variables. Thus the interpretations for all the parameter estimates are the same as interpreting regular linear regression estimates (Fox & Xxxxxxxx, 2011). For the current analysis, there was no data transformation involved, or the observed responses were directly modeled/predicted. Therefore, the magnitudes (or the effect sizes) of the parameter estimates should be judged on a 1-7 scale, given all the responses were measured using 7-point Likert-type items.
MODEL SPECIFICATIONS. The original transdiagnostic, clinically-based, individualized risk calculator was developed using a retrospective cohort study leveraging EHRs of the SLaM boroughs of Lambeth and Southwark, firstly validated in the SLaM boroughs of Croydon and Lewisham7 and secondly validated in C&I11 in the UK. In summary, a Cox model was used to predict the hazard ratio of developing any psychotic disorder over time (see Supplementary Methods 2-2 for definition) as primary outcome of interest. The predictors included age (at the time of the index diagnosis), gender, age*gender, self-assigned ethnicity, and cluster index diagnosis (ICD-10 diagnostic spectra: acute and transient psychotic disorders (ATPD), bipolar mood disorders, nonbipolar mood disorders, anxiety disorders, personality disorders, developmental disorders, childhood/adolescence onset disorders, physiological syndromes, mental retardation). Self-assigned ethnicity and index diagnoses were operationalized as indicated in Supplementary Tables 2-4 & 2-5. A weighted sum of covariates with the model weights from the Cox model resulted in the Prognostic Index (PI). From this, the risk of the individual developing a psychotic disorder within a time period (between one and six years) could be calculated.14 Since this model was originally developed on a retrospective cohort,7 it excluded cases with an onset of psychosis within the first three months to minimize the short-term diagnostic instability of baseline ICD-10 index diagnoses. However, during the subsequent implementation study12,15 an updated version of the model was adapted for prospective use (i.e. not excluding transitions occurring in the first three months), demonstrating similar prognostic performance (Supplementary Table 2-6). Furthermore, a lookback period was additionally used in the current study (Supplementary Methods 3), to minimize the risk of misclassification of index diagnosis date. The specifications of the present model are fully detailed in Supplementary Table 2-7. A main difference compared to the SLaM dataset was that there were no patient-level ethnicity data in Commercial. To mitigate this issue, aggregate ethnicity coefficients were generated for patients who had Metropolitan Statistical Area (MSA) and state-level ethnicity data using Integrated Public Use Microdata Series (IPUMS) census data (xxx.xxxxx.xxx). The geographical information from IPUMS were matched with the geographical data available for each patient in the study populat...
