Objective: To measure inter- and intra-observer reliability in the amniotic fluid index (AFI).
Methods: To measure inter-observer reliability, three observers measured the AFI at random in 32 patients without knowing each other's results. To determine intra-observer reliability, each investigator made repeated measurements on ten other patients. Patients at 26-41 weeks' gestation who were referred for an antenatal ultrasound examination were asked to participate. Intra-observer reliability and interobserver reliability were expressed as intra-class correlation coefficients. Inter-observer agreement was also expressed as the kappa statistic after dichotomizing the AFI (less than 8 cm, 8 cm or more).
Results: The intra-class correlation coefficient for intraobserver reliability exceeded 0.8, and for inter-observer reliability it was 0.67 (95% confidence interval [CI] 0.52- 0.79), with an overall kappa of 0.44 (95% CI 0.22-0.65). The percentage disagreement between pairs of observers reached 22-28%.
Conclusion: Measurements of AFI have good intraobserver and moderate inter-observer agreement. In clinical use, an AFI level near a cutoff value should be interpreted with caution.
© 1994 The American College of Obstetricians and Gynecologists