Tappin, D., Mackay, D., Reynolds, L. et al. Minimizing sample bias due to stigmatized behaviours: the representativeness of participants in a cohort study of alcohol in pregnancy. BMC Med Res Methodol 22, 138 (2022). https://doi.org/10.1186/s12874-022-01629-2
Stigmatized behaviours are often underreported, especially in pregnancy, making them challenging to address. The Alcohol and Child Development Study (ACDS) seeks to inform prevention of foetal alcohol harm, linking self-report as well as a maternal blood alcohol biomarker with child developmental outcomes. Samples were requested using passive, generic consent. The success of this approach at minimizing bias is presented comparing characteristics of women who provided samples to those who did not.
All pregnant women in the study city were sent a Patient Information Sheet (PIS) with their first NHS obstetric appointment letter. The PIS informed them that the NHS would like to take an extra blood sample for research purposes, unless they opted out. Neither the women nor the midwives were informed that the samples might be tested for an alcohol biomarker. This paper examines the extent to which women who provided the extra sample were representative of women where no sample was provided, in terms of routinely collected information: age; body mass index; area-based deprivation; previous pregnancies, abortions and caesarians; smoking status and carbon monoxide level; self-reported alcohol use, gestation and birth weight of their baby. Chi-square and Mann-Whitney U tests were used to compare groups.
3436 (85%) of the 4049 pregnant women who attended their appointment provided the extra sample. Women who did not were significantly younger (p < 0.001), more materially deprived (p < 0.001), and less likely to be considered for intervention based on self-reported alcohol use (p < 0.001). There were no significant differences between the two groups on other routine data.
The use of passive consent without disclosure of the specific research focus resulted in a high level of sample provision. There was no evidence that study blinding was breached, and women who provided a sample were more likely to report alcohol consumption. Passive consent to draw additional blood for research purposes at routine antenatal venipuncture reduced sampling bias compared to asking women to give blood for an alcohol study. This methodology may be useful for other stigmatised behaviours.