The retrospective study was approved by the Ethics Committee of Fujian Provincial Hospital (Ethics Code: K2020-05-014) and Fujian Medical University Union Hospital (Ethics Code: 2019KJCX006) from January 2013 to December 2019. Transthoracic echocardiography was performed within 24 hours of admission on all participants enrolled in the study. The study involved a total of 494 patients. The study included adult patients (> 18 years of age) diagnosed with sepsis based on Sepsis-3 (Third International Consensus Definitions for Sepsis): (1) the presence of infection confirmed by microbiological culture or clinical diagnosis and (2) Sequential Organ Failure Assessment (SOFA) score ≥ 2. Exclusion criteria included age less than 18 years old, hemodialysis or end-stage renal disease, acute coronary syndrome, post-renal causes of renal injury, valvular heart disease, heart failure, cardiopulmonary resuscitation before ICU admission, active malignancy, survival time less than 24 hours, pregnant women, and poor quality echocardiographic images. The grouping was determined using a pre-seeded random number generator in R software (version 4.1.0). After random assignment, patients were split into two cohorts: training (n = 371) and validation (n = 123) according to a 3:1 ratio.
Demographic and clinical data, physiological parameters, transthoracic echocardiographic parameters, the laboratory data including leukocyte counts, hemoglobin, platelet counts, serum creatinine, urea nitrogen, creatine kinase isoenzyme, and albumin measured within the first 24 hours of ICU admission, site of infection and type of microorganism, chronic conditions, comorbidities, and the use of invasive mechanical ventilation and continuous renal replacement therapy (CRRT). According to the Acute Physiology and Chronic Health Evaluation (APACHE) II score and Sequential Organ Failure Assessment (SOFA) score, the severity of illness and organ failure was evaluated on the first day of ICU admission.
Transthoracic echocardiography was performed exclusively by sonographers. Parameters were obtained from long-axis and short-axis parasternal views; apical four-chamber, two-chamber, and long-axis views; and subcostal views using two-dimensional Doppler echocardiography. Data were collected on early transmitral flow velocity (E), late diastolic velocity of mitral inflow (A), early diastolic mitral annulus velocity (e′), E/A ratio, and E/e′ ratio. The patients were classified according to American Society of Echocardiography (ASE 2009) guidelines  and the simplified definition based on the echocardiographic parameters.
Definitions and outcomes
As defined by Kidney Disease Improving Global Outcomes (KDGIO), AKI is a condition associated with renal decline. The level of anemia is assessed by the reference standard set by the World Health Organization (WHO). Based on the simplified definition of diastolic function in sepsis and septic shock, we classified cardiac functions into the following two categories: normal cardiac function and unnormal cardiac function which includes systolic dysfunction (ejection fraction of <50%), and or diastolic dysfunction. The diastolic dysfunction was classified into three grades (grades I, II, and III) as suggested by Lanspa et al.
To assess the distribution of variables, Shapiro-Wilk tests were used. Continuous parametric data were expressed as mean±standard deviation (SD), while nonparametric distributions were expressed as median (interquartile range). Numbers (percentages) were used to express categorical data. Unpaired Student t-tests were used to compare parametric continuous variables and Mann-Whitney U tests to compare non-parametric continuous variables. An analysis of categorical variables was conducted using the chi-squared test. In the training cohort, univariate and multivariate factors were analyzed using logistic regressions. A multivariate logistic regression based on forward stepwise selection was conducted on all variables with P<0.05 in the univariate logistic analyses. Nomogram predicting the hospitalized mortality was determined using the independently selected significant variables. The C-index was used to measure the discrimination ability of the nomogram. A C-index of 0.5 showed no discrimination, while a C-index of 1.0 implied good discrimination. Additionally, we used a calibration curve to determine the relationship between observed frequency and assumed probability, with a 1000-bootstrapped sample of the training cohort. We evaluated the accuracy of the model by measuring the area under the receiver-operating characteristic (ROC) curve (AUC). The nomogram was assessed in the validation cohort to determine if it was stable and general. Furthermore, we evaluated the clinical utility of the final nomogram with a decision curve analysis by calculating the net benefit at various threshold probabilities.