Assessing Diagnostic Classification in an Emergency Department: Implications for Daily Time Series Studies of Air Pollution

Seven independent assessments of diagnosis were obtained for 92 records of nontrauma emergency department visits in Saint John, New Brunswick, Canada, in 1994. The hospital database was 1.18 times as likely (p < 0.05) as six external physician raters to classify visits as cardiorespiratory, which was consistent for high- and low-pollution days, ^K was 0.70 (95 percent confidence interval (Cl) 0.68–0.73). ^K, s were: asthma, 0.69 (95% Cl 0.64–0.73); chronic obstructive pulmonary disease, 0.78 (95% Cl 0.74–0.83); respiratory infections, 0.53 (95% Cl 0.49–0.57); cardiac, 0.84 (95% Cl 0.79–0.88); and other, 0.66 (95% Cl 0.62–0.71). Substantial or better interobserver agreement was seen, respiratory infections notwithstanding, and there was no evidence of diagnostic bias in relation to daily air pollution level. Am J Epidemiol 1998; 148: 666–70.