Original Article Symptoms associated with psychosis risk in an adolescent birth cohort: improving questionnaire utility with a multidimensional approach Sebastian Therman, 1 Markus Heinimaa, 3 Jouko Miettunen, 4 Matti Joukamaa, 6,7 Irma Moilanen, 5 Pirjo Mäki 4 and Juha Veijola 4,2 1 Department of Mental Health and Substance Abuse Services, National Institute for Health and Welfare, and 2 Academy of Finland, Helsinki, and 3 Department of Psychiatry, University of Turku, Turku, and 4 Departments of Psychiatry and 5 Child Psychiatry Institute of Clinical Medicine, University of Oulu and Oulu University Hospital, Oulu, and 6 Social Psychiatry Unit, Tampere School of Public Health, University of Tampere, and 7 Department of Psychiatry, Tampere University Hospital, Tampere, Finland Corresponding author: Mr Sebastian Therman, Department of Mental Health and Substance Abuse Services, National Institute for Health and Welfare, Lintulahdenkuja 4, P.O. Box 30, FI 00271 Helsinki, Finland. Email: sebastian.therman@thl.fi Received 7 September 2009; accepted 11 Feburary 2011 Abstract Aim: Specialized self-report ques- tionnaires have been developed for detection of symptoms indicative of psychosis risk. The identification of at-risk individuals is typically based on sum scores, which assume equal severity and discriminability of all symptoms, and a single dimension of illness. Our aim was to test whether separable dimensions of risk could be identified in the general population. Methods: We explored the latent structure of one such questionnaire using full-information item factor analysis, deriving exploratory models from the PROD-Screen questionnaire responses of the adolescent general population based on the Northern Finland 1986 Birth Cohort (n = 6611). Results: A three-dimensional factor structure of positive, negative and general symptoms emerged. The factor structure, the appropriateness of the statistical model and the appli- cation of the results to the detection of heightened psychosis risk are discussed. Conclusions: In explicitly taking into account the multidimensionality and varying symptom severity of the included items, the current model provides an improvement in questionnaire-based assessment of psychosis risk. Key words: adolescent, full-information item factor analysis, prodro- mal, psychotic-like, questionnaire. INTRODUCTION Several self-report questionnaires have been devel- oped for initial screening of symptoms indicative of imminent psychosis, among them are theYouth Psy- chosis at Risk Questionnaire, 1 the Prodromal Ques- tionnaire 2 and the PROD-Screen. 3 Their utility looks promising, as the concordant validity of these pro- dromal questionnaires can be quite good when comparing risk-group classifications with those generated by interviews. 1 Furthermore, although the focus of these ‘prodromal’ questionnaires is on predicting short-term psychosis risk, these traits and symptoms are clinically relevant in themselves, and assessing them is important regardless of later exacerbation. As the above-mentioned questionnaires arise from the prodromal research tradition, they primarily address recent-onset psychiatric symp- toms. In contrast, psychosis proneness and schizo- typy questionnaires tend to measure stable and subclinical traits. Yet the content is very similar, and the extensive psychosis proneness literature can, therefore, inform the analysis of clinical high- risk symptoms. In particular, there is a consensus regarding the factor structure of psychosis prone- ness and schizotypy questionnaires: separable positive, negative and disorganization latent dimen- sions have been found; 4–6 and some aspects of psychosis proneness are more predictive of later psychosis than others. 7 It would, therefore, seem worthwhile to investigate also the dimensionality of questionnaires probing ‘prodromal’ experiences. Currently the responses to ‘prodromal’ question- naires are typically evaluated by counting symp- tomatic responses and using cut-off scores for these Early Intervention in Psychiatry 2011; 5: 343–348 doi:10.1111/j.1751-7893.2011.00290.x First Impact Factor released in June 2010 and now listed in MEDLINE! © 2011 Blackwell Publishing Asia Pty Ltd 343