Table 3 Methodology and results of drug selection studies included in a literature review of drug sales data analysis for surveillance of infectious diseases
Author | Disease | Method | Results of the algorithm evaluation |
|---|---|---|---|
Pelat et al. [33] 2010 | GI | Hierarchical clustering procedure ,CUSUM | Identification of 4 therapeutic classes relevant to gastroenteritis outbreak detection. Detection performance of a multiple voter algorithm: sensibility 100%, specificity 95%, timeliness 1.7 weeks. |
Cami et al. [31] 2009 | ILI | Aggregate mining algorithm | Identification of product categories with outbreak detection performance superior to predefined categories and more strongly correlated with the disease data. |
Wallstrom et al. [34] 2007 | ILI | Unsupervised time-series clustering algorithm | Distinction between OTC products for allergy and OTC products for influenza symptoms |
Li et al. [32] 2005 | ILI/GI | Canonical correlation analysis | Identification of eight diagnoses that have strong association with electrolyte sales (r = 0.96) |
Magruder et al. [21] 2004 | ILI | Unsupervised stepwise clustering algorithm | Identification of 16 OTC product groups with similar historical trends |