Mean and standard deviations were calculated for continuous variables, and frequencies and percentages for categorical variables. Age of onset, co-morbidity, impairment, clinical severity and treatment contact of those with bipolar disorder were compared between the ethnic groups using analysis of variance (ANOVA) and Chi-square test. A series of multiple logistic regression models were used to generate odds ratios (ORs) and 95% confidence intervals using DSM-IV lifetime BPD, BP-I and BP-II as the main outcome variables and other DSM-IV mental disorders and chronic medical disorders as predictor variables adjusted for age, gender and ethnicity. We also examined the association between BPD with socio-demographic variables including age, gender, ethnicity, education, employment, income and marital status as predictor variables. We checked the fitness of the final design-based logistic regression models using Archer–Lemeshow goodness-of-fit test (Archer and Lemeshow, 2006). Age of onset distributions were estimated separately across birth of cohorts
using life-table method implemented in SAS version 9.2 (Heeringa et al., 2000). Standard errors (SE) and significance tests were estimated using the Taylor series linearisation method. Multivariate significance was evaluated using Wald w2 tests based on design corrected coefficient variance–covariance matrices. Statitical significance was evaluated at the o0.05 level using two sided tests. All statistical analyses were carried out using the
Statistical Analysis Software (SAS) System version 9.2 (Cary, NC)