Exploring the factor structure of the Health of the Nation Outcomes Scale (HoNOS) in a sample of patients with schizophrenia, schizotypal and delusional disorders.

<bold>Background: </bold>The Health of the Nation Outcomes Scale (HoNOS) is recommended for use by the English National Service Framework for Mental Health and by the working group on outcome indicators for severe mental illnesses to the Department of Health. It was developed to measure the health and social functioning of people with severe mental illness. Since the development of the HoNOS many have debated its latent structure. This paper examines the latent structure of the HoNOS using current factor analysis techniques.<bold>Method: </bold>HoNOS data for 12,910 patients with ICD10 diagnoses F20 to F29 at a UK National Health Service Mental Health Trust were analysed using exploratory, confirmatory and bifactor analysis for categorical data. The fit of models was assessed using relative and absolute fit indices.<bold>Results: </bold>Exploratory followed by confirmatory factor analysis identified a four factor solution which fit the data better than existing models. The corresponding bifactor factor solution identified three robust factors and one weak factor after accounting for a general factor. The factor loadings on the general factor were not appreciably different when compared to a unidimensional factor solution indicating the existence of a common trait.<bold>Conclusion: </bold>Existing models proposed in the literature did not fit well in our data. Factor analysis identified a new four factor solution. These factors showed clinical relevance according to published literature. The bifactor model demonstrated that there is not much loss of information when the HoNOS is used as a unidimensional construct. Further studies should explore this structure in larger samples and in alternative sample populations. A bifactor approach may have implications for how the HoNOS is used in practice, since there is ongoing debate on whether HoNOS item scores should be aggregated for interpretation. [ABSTRACT FROM AUTHOR]