Ultrasound fetal biometry: number needed to audit

Fetal measurement audit is important in order to improve and maintain accuracy. The aim of this study was to model the potential outcomes of audits of different numbers of images and measurements, and explore the implications with respect to the improvement and maintenance of standards.

The outcomes of image and estimated fetal weight audits of 3, 10, 25 and 100 patients were modelled using statistical techniques, namely the binomial distribution for images scored as pass/fail and sample size methods for fetal weight estimates. Scored images were modelled for pass rates of 45%, 60%, 75% and 90%. The difference between mean fetal weight errors that would be significant (p < 0.05) was calculated.

The 95% confidence interval on modelled pass rates ranged from ±32% to ±56% for 3 images, ±18% to ±31% for 10 images, ±11% to ±19% for 25 images and ±6% to ±10% for 100 images. The differences between mean fetal weight errors that would be significant were 16.8% (3 samples), 9.2% (10 samples), 5.8% (25 samples) and 2.9% (100 samples).

The results of small audits lack precision; if not interpreted carefully, they may result in inappropriate targeting of interventions. Guidance is provided on careful interpretation and management of results to ensure that remedial training and support is appropriately targeted.