Fig. 5

An example of application of Eq. 2 to algorithms from validation studies. Positive likelihood ratio (DLR⁺) is plotted against the disease prevalence of the validation study. Each line is drawn corresponding to the reported PPV value for each algorithm. Seven studies (eight algorithms) were selected from Fig. 2A of the systematic review by McCormick et al. [17]. Ordering of the algorithms in the legend corresponds to the order of lines in the graph. Included algorithms and the reported PPVs are: Merry 2009 (0.9688), Kiyota 2004p (primary diagnosis) (0.9411), Ainla 2006 (0.933), Kiyota 2004s (primary or secondary diagnosis) (0.9245), Barchielli 2010 (0.8602), Hammar 2001 (0.8583), Heckbert 2004 (0.8302), and Varas-Lorenzo 2008 (0.7202). The gray band indicates disease prevalence between 0.1 and 0.3. Horizontal blue solid lines indicate DLR⁺ values that are consistent with all eight algorithms; blue dotted lines indicate DLR⁺ values that are consistent with the “median” algorithm