Piecewise linear regression model of hemostasis dynamics in donors and patients with schizophrenia

Abstract

This work is to develop and use a new approach to the analysis of differences between the velocity profiles of registered light signals in groups of patients with schizophrenia and healthy donors. The present study involved 62 patients (all female) with schizophrenia or schizoaffective disorders in the acute period (observational study — 2016—2017). 44 patients were diagnosed with paranoid schizophrenia with an attack-progredient type of course (F20.01 according to ICD-10) or with a continuous type of course (F20.02). The fibrinodynamics test (FD) was performed on the T-2 thrombodynamics device (Hemacore LLC, Moscow), which makes it possible to monitor the processes of coagulation and fibrinolysis in the cuvette channels filled with fresh blood plasma. The result of the test are the brightness profiles of the clot. In the proposed approach, each profile is characterized by a vector of distances between it and other analyzed profiles. At that the distances between the profiles are calculated as the sum of the modules of differences at different points of the observation interval. This method makes it possible to overcome a certain loss of information that occurs when using a standard method based on the calculation of several parameters that characterize the analyzed velocity profiles. The method includes the construction of all kinds of statistically significant piecewise linear regression models that link the distances to two profiles. The velocity profiles corresponding to the points lying on opposite sides of the break point of such models have significantly different shapes. In practice, it turns out that for many piecewise linear regression models, the break points also significantly separate the velocity profiles for the compared groups of donors and patients with schizophrenia. Statistically significant deviations at the level p≤0.01 exist for 16.1% of 3143 statistically significant piecewise linear models. The developed method makes it possible to visually establish the relationship between the shape of the velocity profile and the presence of the disease.