Establishment Of A Predictive Model And Prognostic Analysis Of Sepsis-induced Cardiac Dysfunction Based On Left Ventricular Global Longitudinal Strain

Background Sepsis-induced cardiac dysfunction significantly increases the risk of death of septic patients.Early detecting of sepsis induces cardiac dysfunction,and adjusting fluid therapy and vasoactive drugs according to the patient’s cardiac function can improve the patient’s prognosis.Early detection of sepsis-induced cardiac dysfunction has great significance to septic patients.However,there were few studies on the risk factors and predictive models of sepsis-induced cardiac dysfunction.Objective Finding the risk factors of cardiac dysfunction in patients with sepsis,establish a predictive model,and analyzing the prognosis of patients with sepsis-induced cardiac dysfunction.Method Selected patients with sepsis and septic shock in the Department of Intensive Care Medicine（ICU）,Northern People’s Hospital between September 2019 and January2021,and patients with previous cardiac dysfunction were excluded.Included patients were measured left ventricular global longitudinal strain（LV GLS）by echocardiography within 72 hours,and the patients were divided into cardiac dysfunction group（LV GLS>-17%）and normal cardiac function group（LV GLS≤-17%）according to the ultrasound results.Difference of the general indicators such as gender,age,body mass index（BMI）,septic shock,hypertension,diabetes,chronic renal insufficiency,chronic respiratory failure and source of infection of the two groups of patients were analyzed.Difference of clinical indicators such as respiratory rate,blood pressure,heart rate,body temperature,central venous pressure（CVP）,fluid treatment,vasopressor dosing intensity（VDI）and inotropic were analyzed.Difference of organ function score such as Acute Physiology and Chronic Health Evaluation（APACHE II）,Sequential Organ Failure Assessment（SOFA）were analyzed.Difference of laboratory indicators white blood cells,hemoglobin,platelets,albumin,bilirubin,creatinine,procalcitonin（PCT）,high-sensitivity troponin I（Hs-Tn I）,N-terminal pro-brain natriuretic peptide（N-terminal pro-brain natriuretic peptide（NT-pro BNP）,p H lactic acid（Lactate,Lac）and Venous-to-arterial carbon dioxide partial pressure difference[P（v-a）CO₂]≥6mm Hg were analyzed.Difference of ejection fraction（EF）between the two groups was analyzed.Find the best cut-off value according to the receiver characteristic curve（ROC）,and establish a logistics regression model to analyze the relevant risk factors according to the cut-off value.Hosme-Lemeshow was used to test the goodness of fit of the model rows,ROC was used for discrimination test,bootstrap self-sampling method was used to internally verify the model,and a nomogram was drawn.Simultaneously analyzed the difference of ICU mortality,hospital mortality,tracheal intubation rate,Acute Kidney Injury（AKI）,fibrillation incidence,continuous renal replacement therapy（CRRT）frequency,hospital stay,ICU stay of patients in the normal cardiac function group and the cardiac dysfunction group.Then drawn the Kaplan-Meier survival curve and used the Log-rank test to calculate the Hazard Ratio（HR）.Result 1）.A total of 136 patients were enrolled in the study,12 patients were excluded,124 cases included,73 patients in the normal cardiac function group and 51patients in the cardiac dysfunction group.2).There was no statistical difference in general indicators between the two groups in gender,age,BMI,the proportion of patients with septic shock,and previous diseases（p>0.05）.The clinical indicators of the two groups of patients:respiratory rate,blood pressure,heart rate,body temperature,CVP and fluid treatment amount were not statistically different（p>0.05）;Organ function scores of the two groups:APACHE II scores were not statistically different.There was no statistical difference in laboratory indicators such as white blood cells,hemoglobin,platelets,albumin,bilirubin,creatinine and P（v-a）CO₂>6mm Hg,LVEF of two groups of patients（p>0.05）.3).The cardiac dysfunction group had a higher PCT,Hs-Tn I,NT-pro BNP,Lac,SOFA,VDI than the normal cardiac function group（p<0.05）.Among them,PCT=40 ng/m L,Hs-Tn I=0.131 ng/m L,NT-pro BNP=3270 pg/m L,Lac=4.2 mmol/L,SOFA=11,VDI=57μg/min are the best cut-off values for diagnosing sepsis-induced cardiac dysfunction.The cardiac dysfunction group patients had more rate of PCT≥40 ng/ml,rate of Hs-Tn I≥0.131 ng/ml,rate of NT-pro BNP≥3270 pg/ml,rate of Lac≥4.2 mmol/L,and rate of SOFA≥11 than the normal cardiac function group patients（p<0.05）.4).The model was ln[p/（1-p）]=-2.18+[1.90 Hs-Tn I（<0.131 ng/m L vs.≥0.131 ng/ml,<0.131 ng/m L=0,≥0.131 ng/m L）]+[1.12 PCT（<40 ng/m L vs.≥40 ng/m L,<40ng/m L=0,≥40 ng/m L=1）]+[1.03 Lac（<4.2 mmol/L vs.≥4.2 mmol/L,<4.2 mmol/L=0,≥4.2 mmol/L=1）]+[0.98 NT-pro BNP（<3270 pg/m L vs.≥3270 pg/m L,<3270pg/m L=1）],and the Hosme-Lemeshow test and ROC curve were performed and they showed that the goodness of fit（p=0.719）and the discriminability of the model（Area Under Roc Curve,AUC=0.838）were good.5).A nomogram is drawn for 4 independent risk factors.The scores of the 4 factors are as follows:Hs-Tn I≥0.131 ng/m L 100 points;PCT≥40 ng/m L 59 points;Lac≥4.2mmol/L 54 points;NT-pro BNP≥3270 pg/m L 52 points.6).Internal verification using bootstrap self-sampling method found that the model fits well,with a good discriminability[C-index=0.822（95%CI=0.750～0.894）].7).The prognostic analysis found that the cardiac dysfunction group had higher ICU mortality,hospital mortality,atrial fibrillation incidence（p<0.05）and death risk than the normal cardiac function group.（HR=3.104,95%CI 1.617 to 5.957,p<0.001）.There was no statistical difference in the tracheal intubation rate,AKI rate,CRRT rate,length of hospital stay,and length of stay in ICU between the two groups（p>0.05）.Conclusion Hs-Tn I≥0.131 ng/m L,PCT≥40 ng/m L,NT-pro BNP≥3270 pg/m L,Lac≥4.2 mmol/L were independent risk factors for sepsis-induced cardiac dysfunction,and its diagnostic efficacy in constructing a sepsis-induced cardiac dysfunction prediction model was acceptable.Patients with sepsis-induced cardiac dysfunction had a higher risk of death,atrial fibrillation,and hospitalization than patients with normal cardiac function.In the next step,a larger sample size study was needed to verify the model externally.