Background

Physical activity (PA) levels are reported to decrease during pregnancy and postpartum [1]. Western European women in Norway spend more time in moderate-to-vigorous physical activity (MVPA) in pregnancy than ethnic minority women, and by 3 months postpartum, their MVPA had increased significantly contrasted with women with ethnic South Asian background (182 MVPA minutes/week (bouts ≥ 10 MVPA min)) [2]. Increased PA level reduces the risk of adverse health outcomes (e.g., gestational diabetes, postpartum depression, cardio-vascular disease) [3, 4], which are more pronounced in ethnic minority groups [5,6,7,8]. Thus, it is important to understand factors that influence MVPA in ethnic minority women to inform health promotion.

Women’s PA behavior is influenced by a number of factors and their interactions [9] such as ethnicity [10, 11], previous PA level [1], socio-economic position [12], children [10], health status [13], PA friendliness of the neighborhood [2], cultural norms and low level of health literacy [14,15,16] and national/local political context [17, 18]. Further, the migration context, including acculturation, is an additional PA determinant for ethnic minorities [9]. Psychological acculturation has been defined as the result of intra-individual change processes resulting when person moving into a new cultural environment [19]. Acculturation may involve shifts in attitudes, behavior and social factors, thus, a growing body of research on PA in ethnic minority groups centers on acculturation [20].

Epidemiological studies often use simple acculturation proxies as it is not feasible to employ comprehensive acculturation instruments. According to Doucerain and co-workers [19] studies that employ simple proxies should avoid referring to these as measures of acculturation, but refer to specific domains of acculturation, also supported by Bornstein [21]. The current study is based on these recommendations.

Two commonly studied domains of acculturation are language and social contact with individuals with ethnic majority background [22]. Skills in the majority language may enhance ethnic minority women’s health literacy and ability to find, understand and use health information, and make informed health choices [23]. Social contact with individuals with ethnic majority background may positively influence PA, given that such contact provides health information and social support for PA [24, 25], and given that ethnic majority individuals have adopted a higher PA level (descriptive norms) [26, 27].

There are few studies of the association between acculturation and PA during pregnancy and postpartum. In studies of pregnant Latina women in US, a high level of acculturation (measured as preference of English language) was associated with meeting PA guidelines [28] and participation in sports/exercise [29]. In contrast, Latinas with a low level of acculturation (preference of Spanish language) reported more frequent household/caregiving PA [29]. Studies based on objective measures of MVPA from the postpartum period show that Latina women in the US who preferred Spanish language were more physically active [30]. In contrast, for Hispanic women in the US, no association was observed between an overall acculturation-score, based on media language preference (Spanish or English) and social contact with the majority population (Americans), and objectively recorded MVPA [31]. Cross-sectional studies of non-pregnant Latino/Hispanic US resident women of the association between acculturation and self-reported PA [32] and objectively recorded PA [30, 33,34,35,36,37] yielded inconsistent findings.

We are not aware of studies of the association between acculturation and PA in pregnancy or early postpartum in Europe, but there are cross-sectional studies in non-pregnant women showing a consistent positive association between various acculturation proxies and self-reported leisure-time PA [38, 39], but not with PA-intensity level [40, 41].

Our aims were to investigate if the two specific acculturation domains (language skills and social contact with members of majority group) were associated with PA level in ethnic minority women. Our research questions were [1] is social contact with the majority population or [2] Norwegian language skills associated with total MVPA min/day, during pregnancy and postpartum among ethnic minority women [3]? Are the associations modified by stage of pregnancy or early postpartum period? We hypothesized that [1] social contact with ethnic majority women and [2] level of Norwegian language skills were positively associated with MVPA during pregnancy/postpartum.

Methods

Study Design, Population/Setting and Data Collection

Data was obtained from the population-based STORK-G cohort study of pregnant women from Oslo, Norway. Pregnant women attending three child health clinics for antenatal care were recruited between May 2008 and May 2010. Trained midwives conducted face-to face interviews at the child health clinics at visit 1 (mean gestational week 15, early pregnancy), visit 2 (mean gestational week 28, late pregnancy), and visit 3 (mean 14 weeks postpartum). MVPA was objectively recorded immediately following each visit. Information material and questionnaires were translated to eight languages: Arabic, English, Sorani, Somali, Tamil, Turkish, Urdu, and Vietnamese, covering the largest ethnic groups in Oslo [42]. In addition, a professional translator assisted during interviews if required. Inclusion criteria were [1] living in one of the three collaborating city districts, [2] planned to give birth at collaborating hospitals, [3] in gestational week ≤ 20, [4] not suffering from diseases necessitating intensive hospital follow-up during pregnancy, [5] not already included with a pregnancy lasting ≥ 22 weeks, and [6] able to communicate in Norwegian or any of the other eight languages. Study methods are described in detail elsewhere [42]. Written informed consent was obtained from all participants ahead of the study. The Regional Committee for Medical and Health Research Ethics for South Eastern Norway (ref: 2007/894) and the Norwegian Data Inspectorate approved the study protocol [42]. Only study participants who were born abroad, or born in Norway with two immigrant parents, were eligible in the current study.

Primary Outcome: Moderate to Vigorous Physical Activity

Total MVPA (min/day) was recorded by the SenseWear™ Pro3 Armband [43] (SWA) at visit 1–3 [44,45,46,47]. At each visit, women were asked to wear the SWA across the right triceps brachii 24 h per day over 4–7 days and remove it only for water activities. Raw data was integrated into 60-seconds epochs using the manufacturer’s software [48]. The summed value of 1-min epochs was used to estimate metabolic equivalents (METs). MVPA were defined as minute epochs ≥ 3METs. One valid SWA day was defined as ≥ 19.2 h/day of wear time, and SWA data from each visit was valid given ≥ 2 valid SWA days.

Main Exposures

Contact with Ethnic Norwegians

Data on cross-cultural social contact was collected at visit 1 and operationalized by two items that measured frequency of visits by ethnic Norwegians during the last year (item 1) or frequency of help from ethnic Norwegians (item 2). The response alternatives were never, seldom, weekly, and daily. Women who reported “never” on both items were categorized as having no contact, while women who reported seldom, weekly or daily on at least one item were categorized as having contact with ethnic Norwegians.

Self-Reported Norwegian Language Skills

Self-reported Norwegian language skills at visit 1 was rated as poor, not very good, fair, good and very good. A binary variable reflecting low skill level (poor and not very good) and a high skill level (fair, good and very good) was used in the analyses.

Confounders and Other Baseline Characteristics

Confounders were age, body mass index (BMI) kg/m2, education, and ethnicity (Supplementary Fig. 1). Age was treated as a continuous outcome. BMI was obtained from Tanita-BC 418 MA (Tanita, Tokyo, Japan). Highest educational level categories were primary school or less (≤ 10 years), high school (10–12 years) and university or college. Ethnicity was defined by the participant’s country of birth, or that of her mother if born outside Europe or North America [42]. Ethnic categories were South Asian origin, Middle Eastern origin and other.

Statistical Methods

Descriptive Analyses

Descriptive data are presented as mean, standard deviation (SD), or frequencies and proportions, for participants with and without valid PA data across time-points (Table 1).

Table 1 Characteristic of the cohort with and without valid physical activity (PA) data at three timepoints
Full size table

Analyses of the Associations

In separate statistical models we analyzed the associations between each main exposure and MVPA. We used mixed effects linear regression analyses with random intercepts. Level 1 data consisted of repeated measures of MVPA (visits 1–3), nested within women at level 2. Level 2 was treated as random effects in the analyses. In model 1, we employed the visit number (visit 1–3) to model time, but to control for variation in gestational and postpartum week of recording PA, we controlled for gestational/postpartum week centered at the mean week at each visit. We controlled for SWA wear-time mean-centered at each visit and included an interaction term between visit and main exposure to investigate time-varying association with MVPA. In model 2, we additionally adjusted for the confounders age, BMI, educational level, and ethnicity. To assess effect-modification, we explored separate demographic variables in two-way interaction terms with acculturation: (a) ethnicity, (b) parity (binary categories: nullipara and uni-/multipara), and (c) migration status (binary categories born abroad and born in Norway with two immigrant parents).

Few women had missing data on the two main exposures (2.1–5.3%). The sample percentage with missing PA data was 26.5%, 38.4% and 65.3% at visit 1, 2 and 3, respectively (Table 1). Reasons for missing PA data were not attending study visit, not accepting to wear SWA, or < 2 days wear time. Given the proportion of missing data, and to strengthen the plausibility of the missing-at-random assumption, we performed multiple imputation by chained equations [49]. We used predictive mean matching and accounted for dependencies between repeated measures of MVPA using the MICE package in RStudio version 1.1.419 [50]. We generated 50 imputed datasets with 100 burn-in iterations. We performed all other analyses in STATA 15, including multiple imputation analyses (MIA) using the STATA command mi estimate to obtain pooled estimates based on mixed effects regression analyses across the 50 imputed datasets [50]. We present estimates from MIA in the paper, while estimates from complete-case-analyses (CCA) are presented in text if diverting from the MIA (49).

In Supplementary Tables 1, complete CCA results are presented. Sensitivity analyses were performed to assess confounding effects of occupational status, previous PA level, self-reported health, and immigration status in separate models to avoid overadjustment. The results from sensitivity analyses were consistent with main results (data not shown).

Results

Sample Characteristics

The study sample consisted of 487 ethnic minority women included at visit 1 (Fig. 1).

Fig. 1
figure 1

Flowchart of study sample and drop-out

Full size image

The mean (SD) age was 29.1 (5.0) years and BMI was 25.4 (5.0) kg/m2. 90% were born outside Norway, and 22% needed a professional translator at study visits. The mean (SD) week for SWA monitoring was gestational week 15.6 (3.5) at visit 1 and gestational week 28.3 (1.5) at visit 2, and postpartum week 13.9 (2.5) at visit 3. The mean (SD) SWA wear-time ranged from 23.3 to 23.5 (0.5-0.7) hours across visits. There were minor differences between women with and without valid PA data at visit 1, but the latter had higher BMI and a larger proportion had ethnic background from South Asia (Table 1).

Descriptive Analyses of MVPA During Pregnancy and Postpartum

Mean (SD) MVPA min/day was 78.0 (62.4) in early pregnancy, 63.6 (51.1) min/day in late pregnancy and 80.7 (63.9) min/day at postpartum. MVPA declined from early pregnancy to late pregnancy for women with low and high levels of contact with ethnic Norwegians and Norwegian language skills. MVPA increased postpartum across sub-groups, except for women without contact with ethnic Norwegians, where the mean value did not change (Table 2).

Table 2 Descriptive analyses of moderate to vigorous physical activity min/day by levels of main exposures
Full size table

Associations Between Specific Domains of Acculturation and MVPA

In pregnancy, there were small and non-significant differences in MVPA between women with and without contact with ethnic Norwegians and Norwegian language skills (Table 3, Model 2). Estimates obtained from CCA and MIA agreed. In postpartum, there were non-significant MVPA differences between women with low versus high Norwegian language skills in MIA and CCA. Women with social contact with ethnic Norwegians accumulated an additional 17 MVPA min/day (95% CI: − 0.60, 34.54, p = 0.058) in MIA (Table 3). The difference was borderline significant in MIA, and statistically significant in CCA, which indicated women with social contact with ethnic Norwegians accumulated an additional 27 MVPA min/day (95% CI: 8.60, 44.54, p = 0.004) (Supplementary Table 1) compared with women without contact. Parity, ethnicity, and migration status did not modify the associations between the two acculturation domains and MVPA in the MIA (results not presented).

Table 3 Association between main exposures and moderate-to-vigorous physical activity. (Mixed effect linear regression analyses of imputed data)
Full size table

Discussion

To the best of our knowledge, this is the first study of associations between specific domains of acculturation and objectively recorded MVPA in pregnancy and postpartum in a cohort of ethnic minority women in Europe. Contact with ethnic Norwegians and Norwegian language skills were not associated with MVPA in pregnancy. At postpartum, there is some support for claiming that women with contact with ethnic Norwegians are more physically active than those with no contact.

Associations in Pregnancy

Previous studies have indicated an association between preference for the majority language and self-reported PA in pregnancy among immigrants [28, 29, 51]. A plausible reason that the current study indicates no difference in total MVPA in pregnancy is that negative and positive associations with different types of PA (e.g., household and recreational domains) demonstrated elsewhere [29], are “evened out”. Positive associations between acculturation and self-reported exercise demonstrated in previous studies [28, 51] are prone to social desirability bias, and may explain why the current analyses of objectively recorded MVPA did not replicate previous findings. The conflicting findings may also result from analyses of different domains of acculturation, different national contexts, and different migrant populations [52,53,54].

The absence of associations between social contact/language skills and MVPA during pregnancy, imply that pregnancy-specific barriers common among ethnic minority women such as cultural norms (e.g., discouraging physical activity in pregnancy, obligation to perform domestic task), safety concern for the fetus (e.g., lack of information), and health issues (e.g., nausea, muscle pain, anthropometric changes) [14, 55], reduce the impact of other factors determining PA levels.

Associations in Early Postpartum

The CCA and MIA indicate that ethnic minority women having cross-cultural accumulate 27 and 17 MVPA min/day more than women with no cross-cultural contact, although borderline significant in MIA. Postpartum MVPA duration obtained objectively, was notably greater in Mexian-American women with limited cross-cultural social contact in childhood compared to their highly exposed counterparts [30]. The different operationalization of cross-cultural contact in the two studies may partly explain the conflicting results. Another reason may be that the Norwegian context differs from the US context; the maternity leave period and the social arena created by the public maternity care services by offering pram walk groups in Norway may promote PA and explain the difference with respect to US-based studies [30] Another study of Hispanic women in the US indicated no association between acculturation and MVPA [31]. However, acculturation was measured using a multidimensional acculturation scale with acculturation expressed as a single score [31], which may camouflage different associations between separate acculturation domains and MVPA [19] investigated in the current study.

The positive association observed postpartum between cross-cultural social contact and MVPA may be mediated through different mechanisms: As Western women are generally more active [2], social contact with Western women may give access to health information concerning PA, while descriptive norms and social support for doing PA may also promote PA in ethnic minority women [26, 27]. A possible explanation for the positive influence of cross-cultural contact on MVPA in postpartum may be that the mechanisms of social contact are more pronounced at postpartum since a clear shift in PA engagement among Western women occur at this timepoint [2]. Further, it may also suggest sensitivity to situation-specific demands (e.g., cultural norms) revealing dynamic adaptability within specific domains [14, 21]: In the postpartum period, cultural norms for PA may be more aligned between the new culture and the inherited culture, leading to a greater psychological impact of cross-cultural contact and increased PA levels, than during pregnancy [14].

Although, our study revealed that cross-cultural contact may influence postpartum PA, promotion of PA among ethnic minority women in pregnancy and postpartum must be informed by evidence of multiple PA determinants and their interaction sensitive to ethnic groups [10, 14, 56,57,58,59,60,61]. Future intervention studies are warranted to provide stronger evidence of the magnitude of the effect of promoting cross-cultural contact in antenatal health care services.

Strengths and Limitations

This study has several strengths. The models are based on objectively recorded MVPA at multiple time points in pregnancy and postpartum, thus, MVPA estimates are not prone to recall [62] or social desirability bias [63]. Use of interpreters and translated study material promoted inclusion of women with poor Norwegian language skills and strengthened the sample’s representativeness [9, 64], and enhanced the study’s potential to investigate if associations differ by ethnic groups and immigration status. Analyses of specific acculturation domains acknowledge that different domains may influence PA differently [19, 21].

The study has some limitations. The specific measures of acculturation domains are not validated. The study does not capture how individuals balance heritage and host culture within the domain [21, 22]. Another limitation is the large proportion of participants with missing PA data. To mitigate potential bias, we performed multiple imputation, and, for transparency purposes, we have presented the CCA in supplementary material [49, 65]. Small group sizes in some subgroups reduces the statistical power. We included women with two valid SWA days, which is below the recommended cut-off (≥ 3 days) [63]. This may have negatively influenced the reliability of the MVPA estimates [65]. Additionally, wearing an activity monitor may have influenced the participants’ PA behavior.

Conclusion

Cross-cultural contact or skills in the majority population’s language were not associated with MVPA during pregnancy. However, the results indicated that having cross-cultural contact may have a positive influence on MVPA in early postpartum.