Bias from Differential Exposure Measurement Error in a Study of Flight Attendants.

BACKGROUND: Self-reported occupational exposures are often used in epidemiological studies when actual exposure measurements are unavailable, which could cause measurement error and bias study results. This study provides a numeric example of this potential bias. METHODS: A study of block hours and preterm birth was used as an illustrative example. This study included 577 flight attendants, ages 18-45 yr, who gave birth to a term (37 or greater gestational weeks) or preterm (20-36 gestational weeks) infant between 1992 and 1996. Flight attendants self-reported the number of block hours flown during the first trimester of pregnancy; the number of block hours flown during the first trimester of pregnancy was also calculated from airline records. No adjustment for confounding was performed for this illustrative example. RESULTS: Although flight attendants having term and preterm births self-reported similar hours worked during the first trimester (median 213 vs. 215 block hours), airline records showed that flight attendants having term births worked more hours than those having preterm births (median 146 vs. 104 block hours). Using self-reported block hours, there was no association between block hours and preterm birth; when using airline records, an inverse association was observed. DISCUSSION: In this example, differential measurement error from use of self-reported block hours obscured an inverse association apparent when using airline records, demonstrating the importance of accurate exposure assessment for identifying occupational risk factors for health outcomes.

Duke Scholars

Author Candice Yang Johnson Family Medicine and Community Health