Figure 6

Outbreak timing as a function of network size for all ICD-9 codes. Pearson’s r for the existing sites as well as for sites generated using the two MCM variants, K-median model, and randomly selected sites for all 30 ICD-9 codes. Recall that Pearson’s r is used to quantify the quality of outbreak timing detection for a surveillance network. At n=19, the average Pearson’s r for the existing sites is 0.98 (±0.0002), while it is 0.99 (±0.0002) for MCM-C, MCM-NC, and K-median, respectively, and 0.97 (±0.02) for random. All algorithmic methods offer site placements that perform extremely well, achieving at least 0.67 correlation for networks with as few as one site. It only requires two sites selected using either MCM variant to achieve correlation of at least 0.9, and only 11 sites are needed to match the performance of the existing network.