2B) with AUCs of 0.775 and 0.789, respectively. ROC curves for ferritin and lymphocytes were displayed properly in the same plot when depicted by Analyze-it ® (Fig. Two biomarkers (ferritin and lymphocytes) were chosen for their diverse direction of association with mortality (a higher value of ferritin is associated with mortality, while a lower value of lymphocytes is associated with mortality). To further clarify this issue to the reader, a dataset on evaluation of biomarkers for predicting in-hospital mortality for the COVID-19 patients (online resource 1) were used to depict ROC curves by three statistical software Analyze-it ® (Analyse-it Software Ltd, Leeds, UK), Medcalc ® (MedCalc Software Ltd, Ostend, Belgium), and IBM SPSS Statistics ®. Using the above formula, the correct AUCs for the duration of hospitalization and duration of symptoms are 0.654 and 0.667, respectively. For procalcitonin, CRP, and FiO 2 a higher value is associated with mortality, while for the duration of hospitalization and duration of symptoms, a lower value is associated with mortality resulting in their misleadingly small AUCs. plotted five ROC curves with inverse direction of association with mortality. Thus, more than one ROC curve in a graph can be displayed only if the direction of association is similar. Unlike other statistical software that give emphasis to ROC analysis (e.g., Analyze-it ® and Medcalc ®), the IBM SPSS statistics ® did not automatically detect the test direction and the user has to choose which direction is positive (Fig. When a smaller test result indicates more positive test, the algorithm simply estimates AUC ( ώ) using the formula ( ώ = 1 – ω.). The IBM SPSS statistics ® ROC algorithm computes AUC ( ω) based on the default assumption that a larger test result indicates more positive test. Besides its robust set of statistical functions, it has a user-friendly interface. IBM SPSS Statistics ® is one of the most popular statistical analysis software.
0 kommentar(er)
