Cutoff Value of Phase Angle by Bioelectrical Impedance Analysis at Admission as a Prognostic Factor in Patients with Acute Heart Failure

This study evaluates the prognostic utility of phase angle in acute heart failure by identifying an optimal cutoff value and evaluating model performance through multi-variable configuration with restricted splines. Recent developments in this field include the use of bioelectrical impedance vector analysis, or BIVA, for precise assessment of hydration status and cell membrane integrity alongside multi-variable survival models, incorporated by electrical impedance-derived biomarkers to improve early identification of high-risk patients.

Current challenges include ensuring accurate, reliable BIVA measurements over hospitalization, controlling confounding factors in diverse patient populations, and securing an adequate sample size for accurate estimations in prognostic modeling. We address the gap in the lack of precise cutoff values for phase angle as a prognostic marker and the need for reliable and accurate statistical methods to provide accurate predictions of mortality and rehospitalization. This protocol uses non-invasive, quick, and bedside method, applying Cox regression to identify risk factors and to graphically display adjusted threshold for precise risk prediction in acute heart failure.

[Instructor] After installing R and R Studio on the computer, open R Studio. Click on the file tab, go to new file, and select R script to display the script file in the upper left corner above the console tab. In the new script file, install the required survival, MASS, and pspline packages. Load the packages by typing and executing the library functions for survival, MASS, and pspline. Import the dataset by specifying the name and path of the data file in the command and assign it to an object called dataset.

Determine the Cox proportional hazards regression model, adjusted for diabetes mellitus, systolic blood pressure, serum sodium, and phase angle using the breslow method and calculate the 95% confidence intervals for the model coefficients. Then, calculate the Akaike information criterion, the Bayesian information criterion, and the C statistic, along with its 95% confidence interval. For the graphical representation, create the survival object using the surv function with survival days and outcome as inputs. Fit the Cox proportional hazards regression model using the survival object and adjust for diabetes mellitus, systolic blood pressure, serum sodium, and a penalized spline of phase angle with two degrees of freedom. Then, predict the values for the fitted spline and their standard errors using the model. Finally, plot the fitted model on a graph using the predicted values to visualize the adjusted hazard ratio for phase angle. Add the corresponding lines above and below the fitted curve to represent the upper and lower bounds of the prediction interval. Identify the cutoff point where the adjusted hazard ratio exceeds one, indicating that lower phase angle values are associated with an increased risk of in-hospital mortality and 90-day readmission or death. This table summarizes the prognostic value of various clinical and bioimpedance parameters using multi-variable Cox regression models, with outcomes consisting of in-hospital mortality and 90-day readmission or death. In model one for adverse events, a lower phase angle was independently associated with increased risk. In model two, which included different variables, a lower phase angle remained a significant predictor of adverse outcomes. The C statistic for model one indicated moderate predictive capacity for adverse events. In the in-hospital mortality model, a lower phase angle was a highly significant predictor of death. Lower systolic blood pressure was also significantly associated with in-hospital mortality. The in-hospital mortality model had the highest discriminatory ability among all models.