La catatonie progressive par accès périodiques ou « catatonie périodique » de Karl Leonhard 2e partie. Diagnostic probabiliste par Bayes-CPAP, opérationnalisation et preuve de concept

Abstract

In this second part of the study, we propose an implementation of Karl Leonhard's phenotype of periodic catatonia for ICD-10/11-trained psychiatrists. If the goal is to identify the same subgroup of patients within the schizophrenia spectrum disorders (SSD), we will refer to it as “progressive periodic catatonia” or PPC to emphasize the difference with the classical diagnostic procedure which only makes sense in the context of systems neuropsychiatry. This attempt differs from the previous ones in that we use criteria accessible to practitioners trained in ICD-10/11 and by the replacement of the Boolean polythetic logic by a more integrative Bayesian logic. The principle is simple and consists of using the characteristics whose probability of occurrence is conditioned by the PPC vs non-PPC diagnosis (nPPC). If neither of them is discriminating in isolation, the combination of their likelihood ratios allows us, from a determined prevalence of PPC, to estimate an a posteriori probability of diagnosis. The procedure, named “Bayes-PPC”, requires the evaluation of nine criteria: two concern the evolutionary course, six the residual syndrome (two psychomotor distortions, two negative symptoms, and the absence of two positive symptoms), and the family aggregation. A feasibility study was conducted on 38 PPC and 21 nPPC individuals. Starting from the SSD populations or restricting to SSD with deficit schizophrenia (80% of the population), Bayes-PPC allows for a correct classification of 69 and 67% of the patients respectively. For a posteriori probability higher than 95%, the sensitivity for PPC is 82 and 77% respectively, and the specificity is 100% in all cases. For a posteriori probability less than 5%, the sensitivity for nPPC is 48 and 47% respectively and the specificity is 100% in both cases. These results are encouraging and must now be confirmed in a study involving a larger population and by researchers not trained in the neuropsychiatric diagnostic procedure. They nevertheless suggest the possibility of allowing these same untrained psychiatrists to study the PPC phenotype.