Factors associated with cervical cancer screening participation among immigrants of Russian, Somali and Kurdish origin: a population-based study in Finland

Our data came from the Migrant Health and Well-being Survey, 2010-2012, of the Finnish National Institute for Health and Welfare (THL. The sample included 3000 immigrants, 1000 from each of the three countries of origin, i.e. those from Russia or the Soviet Union, Somalia, and Kurds from Iraq or Iran [29].The three major migrant groups were selected to represent different geographical areas. Russians were the largest, Somalis were the fourth largest, and Kurds were the sixth largest groups of the migrants in Finland listed by foreign language. The latter are Iraqi and Iranian refugees who have been among the largest groups of quota refugees accepted into Finland in recent years [30]. The Somali group were mainly refugees and Muslims. The sample was stratified by municipality and ethnic groups and randomly drawn from the Finnish National Population Registry. The inclusion criteria were native language, the above-mentioned countries of origin, age 18-64, having lived in Finland for at least one year, and living in one of the six cities with high proportions of immigrants: Helsinki, Espoo, Vantaa, Turku, Tampere, and Vaasa. For one part of the study protocol, the participation rates were as follows: Russians 70%, Somalis 51%, and Kurds, 61%. We obtained responses to the question about cervical cancer screening (Pap test) participation in the previous five years from 620 women (291 Russians, 132 Somalis, 197 Kurds) in the age group 25-60, the target age group for the Finnish national screening program.

Trained bilingual interviewers conducted structured face-to-face interviews, using the native languages of the participants. These interviews covered several topics on health, well-being and health service utilization. The dependent variable was self-reported participation in screening in the previous five years in Finland; answers were recorded as ‘yes’ or ‘no,’ without distinguishing whether such tests were opportunistic or taken as part of an organized screening program. Based on previous studies on potential factors associated with cervical cancer screening [9‐12, 17, 21, 31], we divided the explanatory variables into four main categories. They were socio-demographic factors and variables related to immigration, health services, health status and women’s health (Fig. 1).

Statistical analyses used the StataSE/13 software package. First, basic descriptive information, frequencies, and distributions, were explored by country of origin. We detected several interactions (p < 0.15); the interaction of country of origin with the length of stay in Finland (p = 0.0326), literacy (p = 0.0268), and marital status (p = 0.1054). We conducted further analyses separately for each country of origin. We produced age-standardized proportions (%) using predictive margins [32]. For the continuous variables, we calculated means and standard deviations (SDs) among those who responded to the question on screening participation in the previous five years (n = 620). Next, we analyzed age-adjusted bivariate associations between the screening participation in the previous five years and potential explanatory variables by country of origin. Logistic regression analyses provided Odds Ratios (ORs), 95% Confidence Intervals (CIs), and p-values.

Based on the age-adjusted results, we created four models for each immigrant group, considering factors from the previous stage, using alpha level p < 0.15 as the threshold. We then grouped these models; for instance, for the model representing socio-demographic factors, we selected the most significant variables for each immigrant group. We did the same for the models representing factors related to immigration, health services use, and health status. Thus, in the analyses, variables within one variable group (for example, socio-demographic) were adjusted for age plus each other. To create a final parsimonious and robust multiple-adjusted model, from the results of the previous analyses, we chose only those factors with a statistical significance level of p < 0.15. Thus, all the variables included in the final model were simultaneously adjusted for each other. To avoid multi-correlation problems, before creating any multiple-adjusted models, we checked correlations within potential explanatory variables. If there were high (r > 0.60) correlations, we included only one of those highly correlating variables simultaneously in the same model.

All our analyses involved Finite Population Correction (FPC) and the effects of non-response were statistically corrected with Inverse Probability Weighting (IPW) [33]. Potential predictors of non-response were the migrant group, age group, sex, city, and marital status. Using the Bayesian Information Criterion (BIC), the best-fitting model was chosen, and it included all the main effects of the covariates. Interactions between variables little improved the predicting power of the model. Afterwards the model to IPW was applied; commonly utilized in correcting the effects of non-response. To obtain representative aggregated results across cities, we further calibrated the IPW weights with respect to the population sizes of the strata.

Table 1 illustrates the main characteristics of the study participants. The mean age of Russians was the highest (42); their employment outside the home was relatively high (63%), and so was their Finnish or Swedish language proficiency (90%). However, a relatively low proportion were married or living in common-law partnership (63%). The mean age of migration was 30; about 82% of them lived in the metropolitan area Helsinki, Espoo, and Vantaa, and 63% had lived in Finland for ten years or longer. They reported similar rates of discrimination experiences (40%) as Kurds (39%) but higher rates compared with Somalis (17%).

Among Somalis, the mean age was 39; the mean number of family members were 5.4 (compared with 3.8 for Kurds and 2.5 for Russians). The mean age of migration was 28; about 90% of them lived in the metropolitan area, while some 70% had lived in Finland for ten years or longer. Almost all of them reported Islam as their main religion (99%), 75% Kurds also reported Islam, while 91% Russians were Christian. Somalis had relatively good self-rated health status (79%), compared with 58% in Kurds and 64% in Russians. Only 14% Somalis had high school education, compared with 87% Russians and 42% Kurds.

Kurdish women were the youngest (average 38), and the mean age of migration was 27. Of them, 55% lived in metropolitan area, while 62% had lived in Finland for ten years or longer. This group had almost the same proportion of those having given birth at least once (91%, compared with 90% Somalis). We had 34% Kurds and 22% Somalis reporting problems due to FGM. Questions on FGM were targeted at those with evidence of prevalence in their country of origin based on UNICEF Statistics [34], hence, Russians were excluded.

Table 1 displays the weighted age-adjusted proportions of screening participation by country of origin. Significant differences existed between these countries (p < 0.001). The participation rate was highest among Russians [73.9% (95% CI = 68.1-79.7)], followed by Kurds [61.3% (55.0-67.7)]; Somalis had the lowest [34.7% (26.4-43.0)].

Table 2 illustrates the age-adjusted bivariate associations between potential factors and participation in screening in the previous five years by country of origin. When adjusted for age, the significant factor for higher screening participation common to all groups was having had at least one general health exam (p-values ranging from 0.025 to <0.001) or gynecological check-up (p < 0.001) in the previous five years.

Table 3 displays the results of the first multiple logistic regression models; the estimates within each of the four variable groups are adjusted for age plus each other. Here, the only significant common factor was having had a gynecological check-up in the previous five years (p < 0.001).

Table 4 illustrates the results of the final multiple logistic regression models; the estimates are adjusted simultaneously for all other variables in each model. Among Russians, in addition to age, significant factors (p < 0.05) associated with higher likelihood of screening participation were literacy in Finnish/Swedish (p = 0.003) and gynecological health check-ups in the previous five years (p < 0.001). Discrimination experiences (p = 0.085) and higher age of immigration (p = 0.104) approached statistical significance.

Among Somalis, the significant factor associated with higher likelihood of screening participation was having had at least one gynecological check-up in the previous five years (p < 0.001); living outside of metropolitan area significantly decreased the likelihood of participation (p = 0.035).

Among Kurds, factors increasing the likelihood of screening participation were as follows: high school education (p = 0.014), being employed (p = 0.007), having given birth at least once (p = 0.014), and having had a gynecological check-up in the previous five years (p < 0.001).

The strengths of this study are that it has a population-based design and satisfactory response rates among Russian and Kurdish participants. To promote willingness to respond to gender-specific issues, the interviewers were mostly female. Bilingual interviewers spoke the same language as the participants, thus promoting understanding about the study purpose and specific interview questions [29]. The study has some limitations. Screening participation in the previous five years was self-reported; as such, some recall bias [45, 46] or reporting errors with participants unfamiliar with screening might exist. Further, some interviews had male interviewers when a female interviewer was unavailable; this may have caused some reporting bias or item-specific non-response, especially in questions related to reproductive health. However, the interviewers were trained to specify unfamiliar terms where necessary.

The variation in screening participation among the immigrant groups might be partially due to factors such as cultural beliefs and norms, which were not included in the data collection for this study. A qualitative research with an ethnographic approach [47, 48] might be successful in studying such factors, especially among Somali women. An ethnographic method would provide a deeper understanding of Somali women’s participation barriers and thus an opportunity to support their screening participation.

These limitations notwithstanding, this study has succeeded, for the first time in Finland, in identifying some enabling factors and barriers associated with screening participation among immigrant groups. This study demonstrates that, although Finland, like some other European countries [11, 20, 25, 26] offers ‘cost-free’ screening targeting all eligible women, disparities exist in using this service among different immigrant groups. These results are worth considering when planning interventions to enhance screening participation among immigrants.