Figure 5. Identification of cut-off values for the scoring model. (A) Receiver operating characteristics (ROC) curve was applied in order to identify the numeric scoring value which would correspond to a high likelihood of fulfilling all target values for continuous glucose monitoring (CGM) data. Based on the data the cut-off limit was set to a scoring value of ≥80 which is visualized by the color green. This resulted in an accuracy of 93.4% when compared with a ‘true’ target fulfillment as previously described. The specificity was 97.1%, which means that very few CGM periods would be falsely classified as green. The sensitivity was 61.4%, showing that the score is a stricter measure of glucose control than commonly used AGP metrics. (B) Distributions of commonly used AGP metrics and score. The green segment represents the highest scoring segment in which the likelihood of fulfilling the treatment targets is high. The dotted lines indicate the currently applied target value for each of the AGP metrics. Patients with a score <80 are represented by gray for visualization purposes in order to highlight the relationship to target values in the green segment. The graph in the lower right corner displays the distribution of score for the whole cohort grouped into five color-coded (in intervals of 20) segments. Dark red and red represent the two lowest segments of score and are hence furthest from achieving the treatment guidelines for CGM data. The majority of the patients are found in the orange and yellow segments (ie, with a score between 40 and 79). AGP, ambulatory glucose profile; AUC, area under the curve; CV%, coefficient of variation; FPR, false positive rate.