Additional file 2.
Figure showing performances of the exposure scoring system in predicting Chikungunya virus infection in the development population (n = 1,863) in the SEROCHIK study, 2006. The exposure score was built up with six covariates and two derivatives (eight components) including residence area (north, west, south, east), housing type (collective/individual), a first interaction term between residence area and housing type, deciles of altitude, household size (1, 2 to 4, ≥ 5 persons), history of recent Chikungunya-related picture in the neighbours (no/yes/don't know), a second interaction term between the household size and history of Chikungunya, and last, the answer to the question: 'Is Chikungunya virus a mosquito-borne virus (no/yes)?'. It was developed from a population of 2,101 eligible adult individuals (≥ 15 years of age) enrolled in the SEROCHIK survey. After elimination of 238 individuals (11.3%) due to missing data, the score displayed a range of 320 eigen values in 1,863 individuals according to a continuous multimodal distribution. The discrimination (or the ability to distinguish infected from uninfected individuals) and calibration (or the adequation between predicted and observed infections over a range of probabilities) performances of the exposure scoring system were considered both satisfactory in the development population (receiver operator characteristic area or Az index: 0.70, 95% CI 0.67, 0.72; goodness of fit F-adjusted test, P = 0.840).
Format: JPEG Size: 36KB Download file
Gérardin et al. Arthritis Research & Therapy 2013 15:R9 doi:10.1186/ar4137