Search

RU-2861556-C1 - METHOD FOR DIFFERENTIAL DIAGNOSIS OF COMMUNITY-ACQUIRED BACTERIAL PNEUMONIA AND VIRAL LUNG INJURY

RU2861556C1RU 2861556 C1RU2861556 C1RU 2861556C1RU-2861556-C1

Abstract

FIELD: medicine. SUBSTANCE: invention relates to pulmonology and intensive care, and can be used for differential diagnosis of the aetiology of lung injury. Within the first 24 hours of hospitalisation, the concentration of total protein in the blood, heart rate (HR), the presence of a decreased level of consciousness, the need for vasopressor administration, and the ratio of the absolute number of neutrophils to lymphocytes are determined. Based on the obtained indicators, the probability of bacterial pneumonia (P) is calculated using logistic regression according to the original mathematical formula. If the P value is 45% or more, a high probability of pneumonia of bacterial aetiology is determined. EFFECT: increasing the accuracy of differential diagnosis of bacterial lung injury for timely decision-making on the prescription of antibacterial therapy and patient routing by using a set of the most significant clinical and diagnostic indicators available at an early stage of hospitalisation. 1 cl, 1 dwg, 1 ex

Inventors

  • Rachina Svetlana Aleksandrovna
  • Strelkova Daria Aleksandrovna
  • Vlasenko Anna Egorovna
  • Kupriushina Olga Aleksandrovna
  • Avdeev Sergei Nikolaevich

Dates

Publication Date
20260505
Application Date
20250908

Claims (10)

  1. A method for determining the probability of bacterial pneumonia, which consists of determining the concentration of total protein, heart rate (HR), decreased level of consciousness, neutrophil/lymphocyte index, and the need for vasopressors in the first 24 hours of hospitalization, calculates the probability of bacterial pneumonia using the formula:
  2. where P is the probability that pneumonia has a bacterial etiology;
  3. Z=1.16–0.046×OB+0.783×BS+0.0132×HR+1.584×VAS+0.0308×NkL;
  4. Z is the coefficient of the logistic regression equation;
  5. OB – concentration of total protein in the blood, g/l;
  6. BS - decreased level of consciousness: 1 - yes, 0 - no;
  7. HR - heart rate;
  8. VAZ - need for vasopressors in the first 24 hours of hospitalization: 1 - yes, 0 - no;
  9. NkL - the ratio of the absolute number of neutrophils to lymphocytes;
  10. at P equal to or greater than 45%, a high probability of pneumonia of bacterial etiology is determined.

Description

The invention relates to medicine and can find application in general therapy, pulmonology, and infectious pathology. It can be used for the differential diagnosis of community-acquired bacterial pneumonia (CBP) and viral lung disease in adults. Early etiological verification of lung damage is unavailable in most institutions; therefore, methods based on a combined assessment of clinical, laboratory, and instrumental data are becoming more important. The developed invention can be used in therapeutic and intensive care units of Russian hospitals prior to the etiological diagnosis of infectious lung disease and can be useful both for patient routing and for deciding on the prescription of antibacterial drugs. For the purpose of early differential diagnosis of viral and community-acquired pneumonia in children, a logistic regression model was developed based on clinical (age, broncho-obstructive syndrome) and laboratory data (absolute neutrophil count, relative band neutrophil count, platelet distribution width by volume) [Kozyrev E.A., Grigoriev S.G., Babachenko I.V., Orlov A.V., Martens E.A., Nikitina E.V., Aleksandrova E.V., Marchenko N.V., Novokshonov D.Yu., Orlova E.D. Differential diagnosis of viral and bacterial community-acquired pneumonia in children using a logistic regression model. Journal of Infectology. 2023; 15(1):25-35. https://doi.org/10.22625/2072-6732-2023-15-1-25-35]. A nomogram for predicting severe adenovirus pneumonia in children was developed. [Zhang J, Xu C, Yan S, Zhang X, Zhao D and Liu F (2023) A nomogram for predicting severe adenovirus pneumonia in children. Front. Pediatr. 11:1122589. doi: 10.3389/fped.2023.1122589]. The disadvantage of models developed on the pediatric population is that they are not applicable to the adult population. Logistic regression, SVM, neural networks, random forests, and gradient boosting can be used for modeling. All of the above methods, except logistic regression, are black-box methods, making their interpretation significantly difficult. Furthermore, these are ML methods and require large amounts of data. Therefore, logistic regression in this case is the method that best suits the initial data (binary output variable, variable types at the input, limited data volume) and the stated goals and objectives of the study (forecast and interpretation of results). The technical result of the proposed invention is an increase in the accuracy of determining the probability that pneumonia has a bacterial etiology based on clinical diagnostic indicators. The technical result of the claimed invention is achieved through a method for differential diagnosis of bacterial pneumonia and viral lung damage, which consists of determining a number of factors: a decrease in total protein, an increased heart rate, a decreased level of consciousness, an increased index neutrophils/lymphocytes, the need for vasopressors in the first 24 hours of hospitalization with calculation of the probability of bacterial pneumonia using the formula: P is the probability that pneumonia has a bacterial etiology, Z = , Z is the coefficient of the logistic regression equation calculated on a sample with missing values replaced by medians, OB - concentration of total protein in the blood, LS - decreased level of consciousness (1 - yes, 0 - no); HR - heart rate, VAZ - taking vasopressors on the first day of hospitalization (1 - yes, 0 - no), NkL is the ratio of the absolute number of neutrophils to lymphocytes, and at P equal to or greater than 45% we determine a high probability that pneumonia has a bacterial etiology. To use the created model, a nomogram was developed (Fig. 1). On the nomogram, each risk factor is assigned a score, determined using the "Score" scale. A perpendicular line is drawn upward from the specified factor until it intersects with the "Score" scale. The intersection is the score corresponding to that risk factor. The scores for all factors are summed. The probability of PFS (as a fraction of 1), corresponding to the resulting score, is determined as follows: a point corresponding to the resulting score is found on the "Score" scale. A perpendicular line is drawn downward from this point until it intersects with the "Risk" axis. This intersection is the predicted probability. To develop the proposed method, we included 300 patients with viral lung disease and 100 patients with CAP in a retrospective case-control study. At the first stage of the analysis, we identified factors that were most significantly associated with the likelihood of CAP compared with viral lung disease: decreased total protein, unilateral infiltration according to CT or chest radiography, increased heart rate, decreased level of consciousness, increased neutrophil/lymphocyte index, need for vasopressors in the first 24 hours of hospitalization, increased urea levels and lower incidence of proteinuria, higher frequency of chills and lower incidence of general weakness. Logistic regression (the me