OBJECTIVE: Identifying and explaining sources of heterogeneity in schizophrenia would help elucidate its aetiology and course. In this paper we examine heterogeneity in terms of age-at-first-admission, diagnosis and gender by decomposing a large dataset using mixture modelling. METHOD: Using the Queensland Mental Health Statistics System, we first extracted age-at-first-admission data for schizophrenia (ICD8/9 295) to represent a 'narrow' definition of schizophrenia (N= 7651); we then added paraphrenia (297) and other non-organic psychoses (298) for a 'broad' definition (N= 10 199). Mixture models were fitted to these narrowly and broadly defined distributions for both males and females. RESULTS: For narrowly defined schizophrenia a three-component model best fitted both male and female distributions. While the mean ages of these components were very similar for both males and females, the ratios of males to females crossed from an excess of males in the 'youngest' component to an excess of females in the 'oldest' component. When using the broad definition, four components best fitted the underlying distributions. While the first three were similar to those found for narrowly defined schizophrenia, the additional fourth component reverted to a male excess; however, the mean age for males was 10 years younger than for females. CONCLUSION: Our findings suggest that subtypes based on age-at-first-admission can be identified, although the number identified depends on how inclusively schizophrenia is defined. While there appear to be the same number with similar mean ages for both genders, there are differences in the proportions of males to females. Further work to fully characterize their nature is warranted.
OBJECTIVE: Identifying and explaining sources of heterogeneity in schizophrenia would help elucidate its aetiology and course. In this paper we examine heterogeneity in terms of age-at-first-admission, diagnosis and gender by decomposing a large dataset using mixture modelling. METHOD: Using the Queensland Mental Health Statistics System, we first extracted age-at-first-admission data for schizophrenia (ICD8/9 295) to represent a 'narrow' definition of schizophrenia (N= 7651); we then added paraphrenia (297) and other non-organic psychoses (298) for a 'broad' definition (N= 10 199). Mixture models were fitted to these narrowly and broadly defined distributions for both males and females. RESULTS: For narrowly defined schizophrenia a three-component model best fitted both male and female distributions. While the mean ages of these components were very similar for both males and females, the ratios of males to females crossed from an excess of males in the 'youngest' component to an excess of females in the 'oldest' component. When using the broad definition, four components best fitted the underlying distributions. While the first three were similar to those found for narrowly defined schizophrenia, the additional fourth component reverted to a male excess; however, the mean age for males was 10 years younger than for females. CONCLUSION: Our findings suggest that subtypes based on age-at-first-admission can be identified, although the number identified depends on how inclusively schizophrenia is defined. While there appear to be the same number with similar mean ages for both genders, there are differences in the proportions of males to females. Further work to fully characterize their nature is warranted.