Adjusted relative risk estimates of maternal autoimmune disease in relation to pregnancy.

Vaginal delivery and CS models were adjusted for age, calendar year, infant sex and place of birth. Abortion models were adjusted for age, calendar year and infant place of birth. We had no information about fetal sex or number of babies in abortions. Repeating vaginal delivery and CS models for singletons only did not change the estimates. Negative binomial regression models showed that the Poisson models were not over dispersed.