RU2806726C1 - Method for predicting probability of death in patients with covid-19 hospitalized on basis of laboratory and instrumental studies - Google Patents

Abstract

FIELD: medicine.

SUBSTANCE: invention relates to the field of medicine, namely, to internal and infectious diseases, and can be used to predict the onset of death in patients with COVID-19. Risk factors obtained during hospitalization are determined. The ECG reveals the presence of: extrasystole, intraventricular blockades, atrioventricular block, atrial fibrillation, intraventricular conduction, pacemaker rhythm. According to the results of computed tomography of the chest organs: 0 points - if there is no lesion, 1 point - a lesion of up to 25% of lung volume, 2 points - a lesion of 26 to 50% of lung volume, 3 points - a lesion of 51 to 75% of lung volume, 4 points - a lesion of more than 76% of lung volume. The results of pulse oximetry are evaluated as 0 points - above 95, 1 point - from 90 to 94, 2 points - from 75 to 89, 3 - below 75. According to the blood test, the level of proBNP to IL is determined as 0 points - the norm, 1 point - above the norm, the level of glucose, alkaline phosphatase, ESR, potassium, troponin and calcium, the number of leukocytes, neutrophils and monocytes. Identify the gender and age of the patient. The probability of predicting the result (PPR) is calculated according to the stated formula. If the PPR value is greater than 0.47, a high probability of death is predicted.

EFFECT: method makes it possible to predict the probability of death in COVID-19 patients hospitalized in a hospital.

1 cl, 1 dwg, 4 ex

Description

COVID-19 (abbreviation for CoronaVirus Disease 2019) is an acute respiratory infection caused by the SARS-CoV-2 virus.

• U07.1 COVID-19, virus identified

Novelty:

On March 11, 2020, the World Health Organization (WHO) announced the development of the COVID-19 pandemic. The emergence of a new infectious agent has posed challenges for healthcare professionals to quickly diagnose the disease and provide timely medical care, especially to the most vulnerable groups of patients, in order to prevent adverse outcomes.

Currently, indicators such as the level of D-dimer, interleukin 6 (IL-6), albumin, lactate dehydrogenase, fibrinogen, ferritin, C-reactive protein, leukocyte count, lymphocytes, and platelets in the blood of sick people. Based on these indicators, various countries are attempting to create various prognostic models and scales for assessing the risk of certain events - hospitalization, transfer to intensive care, or death from COVID-19. As the fight against the COVID-19 pandemic continues, the search for ways to optimize the management of patients with COVID-19 to improve the effectiveness of therapy remains relevant.

The purpose of the invention is to develop a method for predicting the probability of death in patients aged 18 years or more hospitalized with COVID-19.

Description

The invention relates to the field of medicine, namely to internal and infectious diseases, and can be used to predict unfavorable outcomes in patients hospitalized with a diagnosis of COVID-19.

The patented method for predicting the likelihood of death in patients with COVID-19 includes determining risk factors for the development of an unfavorable outcome of the disease during hospitalization of the patient in a hospital based on the patient’s gender and age indicators and the results obtained from instrumental and laboratory studies.

Calculation of the prognostic coefficient of VPR (probability of predicting the result - hospital death in patients with COVID-19) is carried out using the formula:

VPR=1/(1+e ^-x ), where:

e is the base of the natural logarithm;

x=2.569*(ECG ES)+2.145*(ECG EX)+1.502*(ECG IV b)+1.368*(CT OGK)+0.414*(ECG AB-b)+0.404*(proBNP)+0.359* (gender)+0.302*(Saturation)+0.287*(ECG AF)+0.193*(glucose)+0.165*(leukocytes)+0.039*(age)+0.009*(ALP) - 0.018*(ESR) - 0.048*( neutrophils) -0.198*(monocytes) - 0.292*(potassium) - 0.485*(ECG n/g) - 4.148*(troponin) -4.973*(calcium),

and with a value greater than 0.47, a high probability of death is predicted.

Possible prediction of the outcome will reduce the risk of death through preventive measures and rational allocation of medical resources for the treatment of patients with COVID-19.

The method is carried out as follows:

1. A mathematical model was constructed to predict the probability of death. The machine learning algorithm Multifactorial Logistic Regression (MLR) was used as an algorithm to build the model. The quality of the constructed model was assessed using ROC-AUC (Fig. 1), accuracy, specificity and sensitivity. Statistical analysis of the results was carried out using the Python programming language (Python 3.9). For calculations, built-in functions from the Statsmodels, Sklearn and Scipy modules were used.

2. A formula was created (PRP - probability of predicting the result) based on laboratory and instrumental data obtained from the analysis of inpatient records of hospitalized patients with COVID-19 in the infectious diseases departments of the Federal State Budgetary Institution "Central Clinical Hospital with a Polyclinic" of the Department of the Russian Federation in the period from March 2020 to February 2022 .

3. Next, using the resulting VPR formula, the prognostic probability of death in patients hospitalized with COVID-19 is calculated.

4. If the obtained VPR result is less than 0.47, then death is unlikely; greater than or equal to 0.47, then death is highly probable.

The method for predicting the probability of death is carried out by taking into account data:

1. Instrumental research:

1.1. Presence on ECG:

• extrasystoles (ECG ES) (0 - no, 1 - yes),

• pacemaker rhythm (ECG EX) (0 - no, 1 - yes)

• intraventricular block (ECG intraventricular block) (0 - no, 1 - yes)

• atrioventricular block (ECG AV block) (0 - no, 1 - yes)

• atrial fibrillation (ECG AF) (0 - no, 1 - yes)

• disturbance of intraventricular conduction (ECG n i/g p) (0 - no, 1 - yes)

1.2. Results of chest computed tomography (CT CT):

0 - no defeat,

1 - CT1 (damage to 25% of the lung tissue volume),

2 - CT2 (damage from 26 to 50% of the lung tissue volume),

3 - CT3 (damage from 51 to 75% of the lung tissue volume),

4 - CT4 (damage to more than 76% of the lung tissue volume).

1.3. Pulse oximetry results (Saturation):

0 - above 95,

1 - from 90 to 94,

2 - from 75 to 89,

3 - below 75.

2. Laboratory blood test results:

2.1. proBNP level to IL

0 - normal (up to 125 pg/ml),

1 - above normal.

2.2. glucose level (mmol/l)

2.3. leukocyte count (10^9 /l)

2.4. alkaline phosphatase (ALP) level (IU/l)

2.5. ESR level (mm/hour)

2.6. number of neutrophils (10^9 /l)

2.7. monocyte count (10^9 /l)

2.8. potassium level (mmol/l)

2.9. troponin level (µg/l)

2.10. calcium level (mmol/l)

3. Gender and age characteristics of the patient:

3.1. Floor

0 - male,

1 - female.

3.2. Age (full years)

Based on the level of the analyzed indicators, the diagnostic coefficient “Probability of death” is calculated - PPR (probability of predicting the result).

This invention provides the ability to predict mortality after diagnosis of COVID-19.

The mathematical method we have developed for predicting the probability of death in patients with COVID-19 allows us to make a statistically sound prediction of the probability of death depending on the data of instrumental and laboratory studies, as well as the gender and age of the patient.

The use of the proposed method in clinical practice will make it possible to choose the optimal personalized approach to the treatment of patients with COVID-19, which will avoid the development of situations that significantly increase the frequency of complications, adverse outcomes and high material costs.

The accuracy of predicting the probability of death when using this method is confirmed by the accuracy values - 86.6% [84.2; 89.0]%, specificity - 86.8% [83.5; 90.2]%, sensitivity - 86.4% [82.9; 89.9]% on the training set and accuracy - 83.9% [77.0; 89.7]%, specificity - 85.7% [77.1; 93.1]%, sensitivity - 80.6% [68.6; 91.7]%) - on the test sample.

Additionally, the accuracy of the “Probability of Death” prediction model was calculated using the ROC AUC method (Fig. 1). The accuracy of the ROC AUC method is 92.0% [90.1; 94.1]% on training and 89.3% [82.3; 95.1]% on the test sample. The resulting prognostic model is of very good quality (80-90%), because ROC-AUC is 89.3% on the test set.

This mathematical model was used in 435 patients diagnosed with COVID-19, hospitalized at the Federal State Budgetary Institution "Central Clinical Hospital with a Clinic" of the Presidential Administration of the Russian Federation, of which 329 patients recovered, 106 patients died. Death was declared based on the protocol for ascertaining the death of a person.

The invention is illustrated by a detailed description, tables, clinical examples and an illustration depicting ROC-AUC and a method for finding the “cut point” between sensitivity and specificity in ROC analysis.

Clinical examples

Example #1

Patient B., 92 years old, was hospitalized in the infectious diseases department of the Central Clinical Hospital with a Polyclinic on October 11, 2021 with complaints of general weakness, increased body temperature to 38°C, sore throat, disturbances of smell and taste.

From the medical history: on 10/08/2021, the patient developed a sore throat and difficulty in nasal breathing. From 10/09/2021, a low-grade fever was noted; a PCR swab was taken from the nasopharynx for coronavirus infection (the result was positive). On 10/11/2021, the patient was hospitalized at the Federal State Budgetary Institution “Central Clinical Hospital with a Polyclinic” of the Administration of the Russian Federation for further examination and treatment.

On examination: general condition of moderate severity due to intoxication-asthenic syndrome. Position active. The skin and visible mucous membranes are clean. BMI 23.88. Body temperature 38°C. Breathing through the nose is difficult. NPV 18 per minute. SpO2 98%.

Breathing is harsh on auscultation. No adverse breath sounds are heard. On auscultation of the heart, heart sounds are muffled and there are no murmurs. Heart rate 84 per minute. Blood pressure 117/68 mmHg. The abdomen is soft and painless on palpation in all parts. The lower edge of the liver does not protrude from under the costal arch. The stool is regular and formed. Urination is frequent, in small portions. There are no meningeal signs. The function of the cranial nerves is not impaired.

Concomitant pathology: IHD: atherosclerotic cardiosclerosis. Atherosclerosis of the carotid arteries. Hypertension, stage 2, stage 3, risk 3 CVD; History of syncope (vasovagal syncope). Diffuse multinodular toxic goiter, grade 2-3. Drug euthyroidism. CVB. Encephalopathy. Hyperurecemia.

According to CT of the chest cavity dated 10/12/2021: the CT picture corresponds to right-sided polysegmental pneumonia with a high probability of a viral nature (including COVID-19), with an approximate volume of lung tissue damage <25% (no more than 5%) - CT1. Minimal effusion in the left pleural cavity. Hypostatic changes in the lower lobes of the lungs.

On the ECG dated 10/12/21: the rhythm is correct, sinus. Heart rate 59/min. EOS is deviated to the left. Violation of intraventricular conduction along the anterior left branch and right bundle branch.

Patient indicators:

The “Probability of Death” (PPD) was calculated using the formula:

x=2.569*(ECG ES)+2.145*(ECG EX)+1.502*(ECG IV b)+1.368*(CT OGK)+0.414*(ECG AB-b)+0.404*(proBNP)+0.359* (gender)+0.302*(Saturation)+0.287*(ECG AF)+0.193*(glucose)+0.165*(leukocytes)+0.039*(age)+0.009*(ALP) - 0.018*(ESR) - 0.048*( neutrophils) -0.198*(monocytes) - 0.292*(potassium) - 0.485*(ECG i.v. p) - 4.148*(troponin) -4.973*(calcium)=

=2.569*(0)+2.145*(0)+1.502*(1)+1.368*(1)+0.414*(0)+0.404*(0)+0.359*(1)+0.302*(0)+0.287 *(0)+0.193*(6.3)+0.165*(2.8)+0.039*(92)+0.009*(120) - 0.018*(5) - 0.048*(1.96) - 0.198*( 0.34) - 0.292*(3.3) - 0.485*(0) -4.148*(0.001)-4.973*(1.2)

VPR=1/(1+e ^-x )=0.87

A value of 0.87>0.47 means that this patient has a high probability of death.

After calculation, the outcome of the disease was clarified: on October 31, 2021, against the background of increasing respiratory and cardiovascular failure, at 07:00 the patient was declared biologically dead. Causes of death ARF (RDS).

Thus, the occurrence of death in this patient was accurately predicted by the formula.

Example No. 2

Patient P., 62 years old, was hospitalized in the infectious diseases department of the Central Clinical Hospital with a Polyclinic on November 27, 2020 with complaints of general weakness, shortness of breath during exercise, a feeling of heaviness in the chest, an increase in body temperature to 39.0°C, sweating, decreased appetite, loose stools from 26.11.2020.

From the anamnesis: the patient considers herself sick on the 7th day, when weakness appeared and a low-grade fever was noted. On November 23, 2020, a PCR swab was taken from the nasopharynx for coronavirus infection (the result was positive). Since November 26, 2020, the patient has noted a deterioration in her general health and the appearance of loose stools. She was hospitalized at the Federal State Budgetary Institution "Central Clinical Hospital with a Clinic" of the Department of the Russian Federation for further examination and treatment.

On examination: the general condition is serious. Position active. Skin and visible mucous membranes: normal color, clean. Nutrition: excessive. Pastiness, swelling: no. Body temperature 38°C. Breathing through the nose is free. NPV 18 per minute. SpO2 70%.

Breathing is harsh on auscultation. No adverse breath sounds are heard. On auscultation of the heart, heart sounds are muffled and there are no murmurs. Heart rate 80 per minute. Blood pressure 125/70 mm. Hg The abdomen is soft and painless on palpation in all parts. The lower edge of the liver does not protrude from under the costal arch. The stool is regular, since November 26. liquid. Urination is normal. There are no meningeal signs. The function of the cranial nerves is not impaired.

Concomitant pathology: IHD: atherosclerotic cardiosclerosis. Nutritional obesity.

According to a CT scan of the chest cavity dated November 27, 2020: the CT picture corresponds to bilateral polysegmental pneumonia, probably of a viral nature (including COVID-19), with an approximate volume of lung tissue damage on the right up to 69%, on the left - up to 78, which corresponds to the degree lesions CT4. Quantitative hilar lymphadenopathy.

On the ECG dated 10/12/21: the rhythm is correct, sinus. Heart rate 59/min. EOS is deviated to the left. Violation of intraventricular conduction along the anterior left branch and right bundle branch.

Patient indicators:

The “Probability of Death” (PPD) was calculated using the formula:

x=2.569*(ECG ES)+2.145*(ECG EX)+1.502*(ECG IV b)+1.368*(CT OGK)+0.414*(ECG AB-b)+0.404*(proBNP)+0.359* (gender)+0.302*(Saturation)+0.287*(ECG AF)+0.193*(glucose)+0.165*(leukocytes)+0.039*(age)+0.009*(ALP) - 0.018*(ESR) - 0.048*( neutrophils) -0.198*(monocytes) - 0.292*(potassium) - 0.485*(ECG i.v. p) - 4.148*(troponin) -4.973*(calcium)=

=2.569*(0)+2.145*(0)+1.502*(0)+1.368*(4)+0.414*(0)+0.404*(0)+0.359*(0)+0.302*(2)+0.287 *(0)+0.193*(4.8)+0.165*(3.7)+0.039*(62)+0.009*(0) - 0.018*(18) - 0.048*(3.22) - 0.198*( 0.15) - 0.292*(4.5) - 0.485*(0) -4.148*(0.008)-4.973*(1.2)

VPR=1/(1+e ^-x )=0.91

A value of 0.91>0.47 means that this patient has a high probability of death.

After calculation, the outcome of the disease was clarified: on December 19, 2020, against the background of increasing respiratory and cardiovascular failure, at 18:10 the patient was declared biologically dead. Causes of death ARF (RDS).

Thus, the occurrence of death in this patient was accurately predicted by the formula.

Example No. 3

Patient P., 59 years old, was hospitalized in the infectious diseases department of the Central Clinical Hospital with a Polyclinic on October 27, 2020, with complaints of general weakness, dry cough, and fever up to 38°C.

From the medical history: the patient has been in self-isolation since October 13, 2020 after contact with a patient with COVID-19. On 10/13/20, a PCR swab was taken from the nasopharynx for coronavirus infection (the result was positive). On October 16, 2020, the patient’s body temperature increased for the first time to 38.2°C, a dry cough and mushy stool appeared. Outpatient treatment was prescribed, but on October 27, 2020, the patient was hospitalized at the Federal State Budgetary Institution “Central Clinical Hospital with a Polyclinic” of the Department of the Russian Federation for further examination and treatment due to persistent fever.

On examination: general condition of moderate severity due to intoxication-asthenic syndrome. Position active. The skin and visible mucous membranes are clean. BMI 27.0. Body temperature 38.7 C. Breathing through the nose is free. NPV 18 per minute. SpO2 95%.

Breathing is harsh on auscultation. No adverse breath sounds are heard. On auscultation of the heart, heart sounds are muffled and there are no murmurs. Heart rate 76 per minute. Blood pressure 145/97 mmHg. The abdomen is soft and painless on palpation in all parts. The lower edge of the liver does not protrude from under the costal arch. The stool is regular and formed. Urination is normal. There are no meningeal signs. The function of the cranial nerves is not impaired.

From concomitant pathology: Hypertension, stage 2, stage 3, risk of cardiovascular disease 3;

According to CT of the chest cavity dated October 22, 2020: the CT picture corresponds to right-sided polysegmental pneumonia with a high probability of a viral nature (including COVID-19), with an approximate volume of lung tissue damage of 9% on the right and 28% on the left, which corresponds to CT2 .

On the ECG dated 10.22.20: the rhythm is correct, sinus. Heart rate 100/min. EOS is deviated to the left. Sinus tachycardia. Frequent supraventricular (abberant) extrasystole of the trigeminy type. Diffuse changes in the LV myocardium with signs of insufficiency of blood supply to the posterolateral wall of the LV.

Outcome of the disease: recovery. As a result of the therapy, the patient's condition improved, her temperature steadily returned to normal, and there was no respiratory failure. Repeated throat/nose swabs did not detect SARS-CoV-2 RNA.

Patient indicators:

The “Probability of Death” (PPD) was calculated using the formula:

x=2.569*(ECG ES)+2.145*(ECG EX)+1.502*(ECG IV b)+1.368*(CT OGK)+0.414*(ECG AB-b)+0.404*(proBNP)+0.359* (gender)+0.302*(Saturation)+0.287*(ECG AF)+0.193*(glucose)+0.165*(leukocytes)+0.039*(age)+0.009*(ALP) - 0.018*(ESR) - 0.048*( neutrophils) -0.198*(monocytes) - 0.292*(potassium) - 0.485*(ECG i.v. p) - 4.148*(troponin) -4.973*(calcium)=

=2.569*(0)+2.145*(0)+1.502*(0)+1.368*(2)+0.414*(0)+0.404*(0)+0.359*(1)+0.302*(0)+0.287 *(0)+0.193*(6.95)+0.165*(6)+0.039*(59)+0.009*(0) - 0.018*(27) - 0.048*(5.08) - 0.198*(0, 5) - 0.292*(3.6) - 0.485*(1) -4.148*(0.001)-4.973*(1.2)

VPR=1/(1+e ^-x )=0.38

A value of 0.38≤0.47 means that death is unlikely for this patient. After the calculation, the outcome of the disease was specified: recovery. Thus, the outcome of the disease in this patient was accurately predicted by the formula.

Example No. 4

Patient A., 35 years old, was hospitalized in the infectious diseases department of the Central Clinical Hospital with a Polyclinic on April 15, 2020, with complaints of general weakness, dry cough, increased body temperature to 37.5°C, a feeling of heaviness in the chest, headache, sore throat.

From the anamnesis: on April 10, 2020, the patient’s body temperature first increased to 37.5 C, weakness, rhinorrhea, headache, and sore throat appeared. A PCR swab for coronavirus infection was taken from the nasopharynx on April 10, 2020 (the result was positive). Outpatient treatment was prescribed, but on April 15, 2020, the patient was hospitalized at the Federal State Budgetary Institution “Central Clinical Hospital with a Polyclinic” of the Administration of the Russian Federation for further examination and treatment due to persistent fever.

On examination: general condition of moderate severity due to intoxication-asthenic syndrome. Position active. The skin and visible mucous membranes are clean. Body temperature 37.2 C. Breathing through the nose is free. NPV 18 per minute. SpO2 98%.

Breathing is harsh on auscultation. No adverse breath sounds are heard. On auscultation of the heart, heart sounds are muffled and there are no murmurs. Heart rate 80 per minute. Blood pressure 110/70 mm. Hg The abdomen is soft and painless on palpation in all parts. The lower edge of the liver does not protrude from under the costal arch. The stool is regular and formed. Urination is normal. There are no meningeal signs. The function of the cranial nerves is not impaired.

Concomitant pathology: Chronic gastritis, in remission. Iron deficiency anemia, moderate severity. Endometriosis.

According to CT of the chest organs dated 04/16/2020: no convincing CT data for the presence of inflammatory changes in the lungs were obtained in this study, which corresponds to CT 0.

On the ECG dated April 15, 2020: the rhythm is correct, sinus. Heart rate 68/min. EOS normal position. There is no evidence for diffuse changes or acute focal pathology.

Outcome of the disease: recovery. As a result of the therapy, the patient's condition improved, her temperature steadily returned to normal, and there was no respiratory failure. Repeated throat/nose swabs did not detect SARS-CoV-2 RNA.

Patient indicators:

The “Probability of Death” (PPD) was calculated using the formula:

x=2.569*(ECG ES)+2.145*(ECG EX)+1.502*(ECG IV b)+1.368*(CT OGK)+0.414*(ECG AB-b)+0.404*(proBNP)+0.359* (gender)+0.302*(Saturation)+0.287*(ECG AF)+0.193*(glucose)+0.165*(leukocytes)+0.039*(age)+0.009*(ALP) - 0.018*(ESR) - 0.048*( neutrophils) -0.198*(monocytes) - 0.292*(potassium) - 0.485*(ECG n/g) - 4.148*(troponin) -4.973 *(calcium)=

=2.569*(0)+2.145*(0)+1.502*(0)+1.368*(0)+0.414*(0)+0.404*(0)+0.359*(0)+0.302*(0)+0.287 *(0)+0.193*(4.9)+0.165*(7.4)+0.039*(35)+0.009*(0) - 0.018*(8) - 0.048*(5.03) - 0.198*( 0.44) - 0.292*(4) - 0.485*(0) -4.148*(0.001)-4.973*(1.2)

VPR=1/(1+e ^-x )=0.02

A value of 0.02≤0.47 means that death is unlikely for this patient. After the calculation, the outcome of the disease was specified: recovery. Thus, the outcome of the disease in this patient was accurately predicted by the formula.

Claims (6)

A method for predicting the likelihood of death in patients with COVID-19 hospitalized in a hospital based on laboratory and instrumental studies includes determining risk factors obtained during the patient’s hospitalization in a hospital based on the results of instrumental studies: assessment of electrocardiography (ECG) data - the presence of extrasystoles on the ECG (ECG ES), intraventricular block (ECG i/g b), atrioventricular block (ECG AV-b), atrial fibrillation (ECG AF), intraventricular conduction (ECG n i/g p), and pacemaker rhythm (ECG EX), Moreover, the absence of each of these factors is assessed as 0 points, and the presence - 1 point; assessment of the results of computed tomography of the chest organs (CT OGK): as 0 points - if there is no lesion, 1 point - damage to 25% of the lung volume, 2 points - damage from 26 to 50% of the lung volume, 3 points - damage from 51 to 75 % lung volume, 4 points - damage to more than 76% of lung volume; assessment of pulse oximetry results (Saturation) as 0 points - above 95, 1 point - from 90 to 94, 2 points - from 75 to 89 3 - below 75; analyze the indicators obtained during a laboratory blood test: the level of proBNP to IL as 0 points - the norm is up to 125 pg/ml, 1 point - above the norm, and also determine the numerical indicators of the level of glucose, alkaline phosphatase (ALP), ESR, potassium, troponin and calcium, number of leukocytes, neutrophils and monocytes; assess the gender-age characteristics of the patient: gender - man - 0 points, woman - 1 point, enter the numerical value of age, then calculate the predictive coefficient of the VPR - the probability of predicting the result using the formula: VPR = 1/(1+е- ^х ), Where: VPR - probability of predicting the outcome - hospital death in patients with COVID-19, e is the base of the natural logarithm; x=2.569*(ECG ES)+2.145*(ECG EX)+1.502*(ECG IV b)+1.368*(CT OGK)+0.414*(ECG AB-b)+0.404*(proBNP)+0.359* (gender)+0.302*(Saturation)+0.287*(ECG AF)+0.193*(Glucose)+0.165*(leukocytes)+0.039*(age)+0.009*(ALP) - 0.018*(ESR) - 0.048*( neutrophils) -0.198*(monocytes) - 0.292*(potassium) - 0.485*(ECG n/g) - 4.148*(troponin) -4.973*(calcium), and if the value is greater than 0.47, a high probability of death is predicted.

Images