Changing Influence of Testing Methodologies on Legionellosis Diagnosis, Ontario, 1978–2006. Observed (dots) and expected (curves) probabilities of test positivity by methodology in 1401 cases of legionellosis diagnosed in Ontario, Canada from 1978 to 2006. In bottom right graph, solid line and white dots represent seroconversion based on paired serological testing, while broken line and gray dots represent diagnosis based on single high-titre positive serology. Expected probabilities are generated using logistic regression models, with probability of test positivity regressed against year and (in the case of culture and single high-titre immunofluorescent antibody) year-squared. Urine antigen testing was introduced at the laboratory in 1984, and consequently no specimen was positive by this method prior to that date.