Computerised adaptive testing across the paranoia continuum

Background To drive improvement in clinical services, an important innovation will be to regularly assess patients’ psychotic experiences in order to guide, monitor and, when needed, alter treatment provision. The great heterogeneity in presentations of psychosis means that a comprehensive assessment battery is impractical. A plausible solution is computerised adaptive testing (CAT), which uses real-time computation to present the most informative questions to an individual. Fewer questions are needed to reach similar precision as a full questionnaire.Objective We tested the potential of a CAT for paranoia to halve the number of items that need to be presented.Methods We used the established item response theory psychometric properties of the 10-item Revised Green et al Paranoid Thoughts Scale (Persecution) to run CAT simulations in four datasets in which participants had completed the full scale: a representative survey of 10 382 UK adults; a clinical trial with 319 patients with psychosis; a cohort study of 836 National Health Service (NHS) male patients with psychosis; and a clinical trial with 89 patients with persecutory delusions. The CAT algorithm used the graded response model and the test was concluded when the SE of estimation dropped below 0.3 or five items had been answered.Findings On average, the CAT administered 4.2, 4.0, 4.2 and 4.0 items to each person in the four datasets. The correlations between the CAT score and the full-scale paranoia score were 0.95, 0.94, 0.94 and 0.87. Minimal systematic error in paranoia estimation occurred (mean bias scores=−0.01, –0.06, −0.07 to –0.10). Estimation was the least precise for people at the boundary of normal and elevated levels of paranoia.Conclusions In datasets with people across the whole paranoia continuum, accurate estimates of paranoia can be provided by a CAT with fewer than half the items of the full scale. Tailored testing may work well with people with psychosis.Clinical implications CAT may be a way to implement informative measurement-based care in psychosis services.