Missing data were imputed using multiple imputations by full
Missing data were imputed using multiple imputations by fully conditional specification. Fifty complete datasets were created. To make the MAR assumption more plausible, every previously described variable was used for the imputation model, [23,24] including the outcome. In addition, waist size was introduced as a continuous auxiliary variable. Since the outcome variable contained missing values, sensitivity analyses were conducted, using the Multiple Imputations then Deletion method that enabled to take this missingness into account in the imputation model but not in the analyses . Information on the last CCS date was missing for 12.2% of the sample, and Thapsigargin status for 2.0%. Overall, there were fewer than 5% missing data for all the explanatory variables, with the exception of BMI (7.0%), gynaecological follow-up (6.4%), alcohol consumption (7.0%) and depressive disorder (5.6%). SAS 9.4® was used to perform the analyses.
Results The distribution and the CCS rates for every women’s characteristic are presented in Table 1. Naturalized women and women of foreign nationality accounted for 4.7% and 3.8% of the 31,024 women in our sample, respectively. Overweight and obese women accounted for 22.9% and 11.3% of the sample, respectively. French-born women (main group) had the highest screening rate (87.1%), followed by French-born women with at least one foreign parent (85.7%), naturalized women (82.0%) and women of foreign nationality (73.8%). The CCS rate increased from 77.5% among obese women to 85.0% among overweight women, 88.1% among normal-weight women and 88.6% among underweight women. Overweight and obesity were more frequent among the naturalized women and women of foreign nationality than among French-born women (Table 2). After adjusting for the covariates, the CCS rates were similar among all French-born women, regardless of their parents’ origin. Women of foreign nationality had an 11% lower CCS rate (95% CI: 8%–14%) than women in the main group. Naturalized women had a 2% lower screening rate, but this result did not achieve statistical significance (Table 3). After adjusting for the covariates, BMI was strongly associated with CCS (underweight PR = 1.02 [95% CI: 0.99–1.04]; overweight 0.98 [0.97-0.99]; obese 0.91 [0.90-0.93]) but adjustment for BMI did not modify the estimates of the association between migration status and CCS (Table 3). One of the underlying hypotheses of mediation model was not verified, therefore BMI was not a mediator of the association between migration origin and CCS. The interaction term between migration status and BMI did not reach statistical significance (p = 0.236), but it was significant when both BMI and migration status were considered in two groups (obese/non-obese and French/foreign nationality, p = 0.014). When stratifying by BMI category, after adjustment for the covariates, non-naturalized immigrants showed an 11% (7%–14%) lower CCS rate than French-born with two French parents when normal weight, a 9% (2%–16%) lower CCS rate when overweight, and an 18% (5%–30%) lower CCS rate when obese. In contrast, the CCS rates among French-born women with parents of foreign origin were similar to those among the women in the main group in all BMI categories. No clear pattern was observed for naturalized immigrant women, and the estimates did not achieve statistical significance.
Discussion We observed lower CCS rate among women of foreign origin, which is consistent with the literature, and we found that the more belonging to the host country was important, the more the CCS rate was closer to what it was in the main group [6,7,9,10,36]. Studies have reported differences in CCS rates according to the women’s country of birth. A Spanish study reported that the lower CCS rates observed were limited to women born in low-income countries , but a US study found more contrasted results. They observed lower CCS rates among all women born abroad, but only women from Asia and India stood out with much lower CCS rates . This heterogeneity in CCS rate among women of foreign origin was observed in our data, where CCS rates were lower among naturalized and foreign women of African origin than of another origin.