Socioeconomic position and lower dietary moderation among Chinese immigrant women in the United States

Abstract

Objective

To examine associations of education and occupation, as indicators of socioeconomic position (SEP), with dietary intake and diet quality in a sample of Chinese immigrant women.

Design

Cross-sectional. Data collection included four days of dietary recalls and information on education and current occupation for participants and their spouses.

Setting

Philadelphia, PA, USA.

Subjects

423 Chinese immigrant women recruited 10/05-4/08.

Results

In multivariate models, both higher education level and occupation category were significantly associated with higher energy density and intake of energy and sugar. Education was additionally associated with intake of sugar-sweetened beverages (p=0.01) and lower dietary moderation (p=0.01). With joint categorization based on both education and occupation, we observed significant trends indicating higher energy density (p=0.004) and higher intake of energy (p=0.001) and sugar (p=0.04), but less dietary moderation (p=0.02) with higher SEP.

Conclusions

In this sample of US Chinese immigrants, higher SEP as indicated by education level and occupation category was associated with differences in dietary intake, and with less dietary moderation. While higher SEP is typically linked to healthier diet in higher income nations, in these immigrants the association of SEP with diet follows the pattern of their country of origin – a lower-income country undergoing the nutrition transition.

INTRODUCTION

Lower socioeconomic position (SEP), quantified using such indicators as level of education, occupational status, and income, is associated with poorer diet and health in higher-income countries, including the United States (US) ^1–8. However, these associations may not be valid across all subgroups. Despite their lower SEP, for example, low-acculturation Asian immigrant samples appear to have better diet and health than their more acculturated, higher SEP peers ⁹ and the mainstream white population ^{10, 11}.

Asian immigrants’ transition to increased risk for overweight and chronic disease following migration to the US ^12–21 might be attributed in part to dietary acculturation, or the adoption of the dietary practices of the host population ²². Few studies, however, have examined the association of SEP with dietary intake in US immigrants. The association of SEP with dietary intake among immigrants may parallel the nutrition transition observed in developing countries, in which economic development and higher SEP is associated with less healthy lifestyle and increased risk for chronic disease ^{8, 23}. The objective of this analysis was to describe associations of diet and diet quality with level of education and occupational category as indicators of SEP in a sample of Chinese immigrant women.

METHODS

Study sample

Between October 1, 2005, and April 30, 2008, we recruited 436 healthy, premenopausal women from community organizations and contacts into a study of diet and mammographic breast density. Eligibility criteria included Chinese heritage, migration from Asia ≤20 years ago, and being of mammography screening age. Exclusion criteria were: postmenopausal status (no menstruation in the past year); history of breast augmentation/reduction, prophylactic mastectomy, or any cancer except non-melanoma skin cancer; current pregnancy; current breastfeeding or breastfeeding within last 9 months; or symptoms of new breast problem, such as palpable lump, skin changes, or nipple discharge. Participants received $20 as reimbursement for their time. The study was approved by the Fox Chase Cancer Center Institutional Review Board.

Data collection

Interviewers conducted detailed health interviews that elicited information on various health behaviors, health and reproductive history, and sociodemographic characteristics including level of education, current and usual occupation of the participant and of her spouse, if applicable, and level of acculturation.

Participants were asked about the highest level of education they completed, and responses were collapsed into three categories: (1) 0–8 years, (2) 9–12 years, with or without completion of high school or vocational / technical school, and (3) at least some college, university, or graduate school. Each participant was also asked to select a category for her own and for her husband’s current occupation from among the following: not employed; farmer/farm worker; machine/vehicle operator; craftsworker; service worker; clerical worker; sales worker; manager/administrator; or professional/technical. These were collapsed into three categories: (1) not employed, farmer/farm worker, machine or vehicle operator, craftsworker, or service worker, (2) clerical or sales worker, and (3) manager/administrator or professional/technical, roughly corresponding to categories used in other research ^{24, 25}. In analyses based on husband’s occupation, the participant’s own occupation was used (n=44) if she was unmarried or her husband was not employed. Results were similar but more pronounced when analyses were based on the participant’s husband’s rather than her own occupation; thus, analyses on occupation category were based primarily on husband’s occupation. For 2% (n=8) of the sample, no occupation was specified for either the participant or her spouse because both the participant and her spouse were unemployed, or because the participant was unemployed and unmarried.

Acculturation was measured using an abridged version of the General Ethnicity Questionnaire - American version (GEQA) ²⁶. The original, 37-item GEQA scale, developed for immigrant and American-born Chinese college students, was found to have good validity and reliability in Chinese immigrants in previous studies ^{26, 27}. For the current study we dropped items that showed little variability in response during pilot testing among a sample of middle-aged Chinese women, most of whom migrated to the US in adulthood (e.g., ‘I was raised in a way that was American’, ‘How much do you speak English at school?’). The remaining eleven items dealt with exposure to or familiarity with American people, culture, and activities (e.g., ‘Now, I am exposed to American culture,’ ‘I go to places where people are American,’ ‘I celebrate American holidays’) and showed high internal reliability (alpha=0.91).

Trained interviewers followed a standardized protocol for conducting two 48-hour dietary recall interviews about two weeks apart for each participant, with the mean over the four days used in analysis, and for entering responses into the Nutrition Data System for Research (NDS-R, Nutrition Coordinating Center, University of Minnesota). Foods not included in the NDS-R database were added by creating recipes for new mixed dishes, or by the Nutrition Coordinating Center, which bases nutrient values on information from food manufacturers, foreign food composition tables, the scientific literature, and other available databases. In addition to providing estimates of nutrient intake, the NDS-R assigns each food item to one of 166 possible food subgroups and estimates serving counts for each food subgroup. Food items are counted at the whole food level when appropriate (e.g., bread, apple pie, French fries), or at the component/ingredient level (e.g., lasagna, soup, fruit salad, sandwiches) to capture intake of ingredients. Food subgroup definitions and serving sizes were based primarily on recommendations from the 2005 Dietary Guidelines for Americans 2005 ²⁸ and the Food Guide Pyramid ²⁹. Food and Drug Administration serving sizes ³⁰ were used for foods not included among current recommendations, such as cookies and fruit drinks.

Statistical analyses

Of 436 women enrolled in the study, three women subsequently did not complete baseline questionnaires, and 10 were excluded for not having completed dietary interviews (n=8) or questions on occupation (n=1) or country of birth, an adjustment variable in multivariate regression analyses (n=1), leaving a sample of 423 women. We observed no statistically significant differences in age, level of education, length of US residence, or GEQ-A score between the sample of 423 women and the 10 women with information on these factors but excluded from analysis.

We used the Diet Quality Index-International (DQI-I) ³¹ to examine four components of diet quality: (1) variety within protein sources and across food groups; (2) adequacy of intake of vegetables, fruits, grains, fiber, protein, iron, calcium, and vitamin C; (3) moderation of intake of total fat, saturated fat, cholesterol, sodium, and empty calorie foods; and (4) overall balance with respect to macronutrients and fatty acid composition ³¹. Details on the components of the DQI-I are given in the Appendix. Other dietary outcome variables of interest were selected for their direct relevance to diet quality and its components. These included energy density, percent of energy from fat, carbohydrates, and protein, and intake of total energy, cholesterol, dietary fiber, sodium, sugar, and sugar-sweetened beverages.

The primary predictors of interest were level of education (three categories) and occupation category (three categories). From these variables, we also created three categories jointly classifying women according to both education and occupation: (1) least educated women in the lowest occupational category, (2) most educated women in the highest occupational category, and (3) an intermediate category including all other women. We used linear regression analyses to examine associations of education and occupation as predictors of dietary intake.

The balance component of the DQI-I could not be treated as a continuous outcome variable. Therefore, to use a consistent approach across all DQI-I components, we created categories of approximate tertiles, then conducted logistic regression analyses for polychotomous outcomes using proportional odds models. Estimates from these models can be interpreted as the log odds of falling into a higher vs. lower category ³². The score test ³² was used to test the assumption of proportional odds – i.e., that the cumulative logits are equal. Variables for which the assumption of proportional odds was violated (dietary adequacy, DQI-I score) were dichotomized at the median and analyzed in logistic regression for binary outcomes. Sugar-sweetened beverages were consumed by less than half of the study sample and so was also analyzed as a dichotomous outcome variable (any vs. no consumption) in logistic regression analyses for binary outcomes.

Models for energy intake, DQI-I score, and DQI-I components adjusted for age (continuous years), country of birth (China or elsewhere), marital status (married or not), and GEQA score (continuous). All other models additionally adjusted for energy intake (continuous kcal). Models were first run for education and occupation separately, then for joint categories of education and occupation. Tests for trend were conducted using a variable representing ordinal values for education, occupation, or their jointly classified category. All statistical analyses were conducted using SAS (version 9.1.3, 2005, SAS Institute, Cary, NC).

RESULTS

Among the 423 women in the sample, mean (SD) age was 43.9 (4.5) years, with a range of 35–56 years (Table 1). The women had lived in the US for a mean (SD) of 7.6 (4.8) years and migrated to the US at a mean (SD) age of 36.4 (6.5) years. Almost all (97%) were born in China, and most (70%) spoke no English at home. Mean (SD) GEQA score was 2.1 (0.7), out of a possible range of 1 (least acculturated) to 5 (most acculturated). Half of the women (50%) reported up to eight years of education, while 36% reported at least nine years of education up to technical or vocational school, and 15% had at least some college education. Most women (82%) were classed in the farm-, machine-, crafts-, or service worker category, 7% in the clerical or sales worker category, and 11% in the managers/administrators or professional/technical workers category. Women with a higher level of education or in a higher occupation category had higher mean GEQA score, were more likely to speak English at home, and were less likely to be married. Education and occupation categories were also strongly interrelated, with 63% of women with at least some college education holding an administrator / professional position, and 86% of women in administrator / professional positions having at least some college education.

Table 1.

Sociodemographic variables by education and occupation category in a sample of US Chinese immigrant women (n=423).

		Education categorya			Occupation categoryb
	All women (N=423)	1 (n=206)	2 (n=147)	3 (n=70)	1 (n=338)	2 (n=34)	3 (n=51)
Mean (SD) age (years)	43.9 (4.5)	44.1 (4.7)	43.8 (4.3)	43.6 (4.4)	43.8 (4.5)	44.9 (4.5)	44.1 (4.6)
Born in China (%)	97	98	96	97	97	94	100
Level of education (%)
0–8 years	49	--	--	--	57	35	4
9–12 years	35				38	44	10
at least some college	17c				6c	21	86
Occupational category (%)
Machine operator, farm, craft, or service worker, or not employed	80	93	86	27	--	--	--
Clerical or sales worker	8	6	10	10
Manager, administrator, or professional	12	1	3c	63
Not married (%)	7	4	5	20	7	3	14
Mean (SD) age at migration (years)	36.4 (6.5)	36.6 (6.3)	36.5 (6.6)	35.6 (6.8)	36.3 (6.5)	37.1 (7.0)	36.3 (6.4)
Mean (SD) length of US residence (years)	7.5 (4.8)	7.5 (4.6)	7.3 (4.7)	8.0 (5.4)	7.5 (4.7)	7.8 (5.1)	7.8 (5.2)
Mean (SD) acculturation score	2.1 (0.7)	1.9 (0.6)	2.1 (0.7)	2.9 (0.6)	2.0 (0.7)	2.4 (0.6)	2.9 (0.6)
Speak no English at home (%)	70	86	65	33	75	74	33
Mean (SD) intake per day
Energy (kcal)	1358 (363)	1308 (365)	1391 (343)	1435 (382)	1332 (354)	1438 (354)	1476 (406)
Energy density (kcal/g)	0.80 (0.24)	0.75 (0.25)	0.80 (0.21)	0.97 (0.23)	0.77 (0.23)	0.83 (0.20)	1.00 (0.24)
Total fat (% kcal)	24.5 (6.1)	23.5 (5.9)	24.3 (5.9)	27.7 (6.1)	23.9 (6.0)	26.2 (5.9)	27.1 (6.2)
Total sugar (g)	40.0 (22.6)	36.2 (20.7)	38.0 (19.9)	55.2 (26.9)	37.2 (20.6)	40.8 (20.9)	58.0 (27.9)
% consuming sugar-sweetened beverages	10	6	8	22	7	17	19
Mean (SD) diet quality scores
Variety	15.6 (3.4)
Adequacy	30.4 (4.6)	30.3 (4.5)	30.6 (4.6)	30.3 (4.9)	30.3 (4.5)	30.7 (4.8)	30.7 (5.1)
Moderation	22.4 (5.2)	23.4 (5.0)	21.6 (5.3)	21.3 (5.3)	22.7 (5.2)	21.3 (5.1)	21.6 (5.3)
Balance	1.87 (2.00)	1.80 (1.90)	2.03 (2.17)	1.77 (1.91)	1.85 (2.00)	1.65 (1.67)	2.20 (2.20)
DQI-I score	70.4 (8.5)	70.9 (7.9)	69.8 (8.7)	69.7 (9.5)	70.3 (8.3)	69.8 (7.7)	70.9 (10.0)

^aEducation category 1 = 0–8 years of school, 2 = 9–12 years, 3 = at least some college.

^bOccupation category 1 = machine operator, farm, craft, or service worker, or not employed, 2 = clerical or sales worker, 3 = manager, administrator, or professional.

^cProportions do not add to 100% due to rounding.

In linear regression models (Table 2), both higher education level and higher occupation category were associated with higher energy density, and higher intake of energy and sugar. More education was also significantly associated with a higher percent of energy from fat and, in logistic regression analyses, with greater likelihood of consuming sugar-sweetened beverages (odds ratio (OR)=3.9 (95% confidence interval (CI) 1.5–10.3) for highest vs. lowest education category, trend p=0.01). In logistic regression analyses with diet quality as the outcome variable (Table 3), higher education level was additionally associated with lower dietary moderation. We observed no associations for either education or occupation category with percent of energy from saturated fat, carbohydrates, and protein, with intake of cholesterol, dietary fiber, and sodium (not shown), or with other measures of diet quality.

Table 2.

Adjusted beta estimates ^a and p-values for trend for education and occupation categories in multivariate linear regression models (n=423).

	Education categoryb			Occupation categoryc
	1 (n=206)	2 (n=147)	3 (n=70)	1 (n=338)	2 (n=34)	3 (n=51)
Energy	ref	85.8 (0.03)	137.0 (0.01)	ref	107.5 (0.11)	153.2 (0.01)
Trend pd		0.002			0.002
Energy density	ref	0.01 (0.54)	0.12 (0.0002)	ref	−0.01 (0.88)	0.12 (0.0004)
Trend p		0.002			0.001
Total fat (%)	ref	0.3 (0.64)	2.3 (0.01)	ref	1.2 (0.27)	1.3 (0.20)
Trend p		0.03			0.14
Total sugar	ref	−1.6 (0.44)	9.4 (0.002)	ref	−1.4 (0.69)	11.8 (0.0002)
Trend p		0.04			0.001

^aAdjusted for age, marital status, birthplace, and score on General Ethnicity Questionnaire – American version. All models except for energy intake were also adjusted for energy intake.

^bEducation category 1 = 0–8 years of school, 2 = 9–12 years, 3 = at least some college.

^cOccupation category 1 = machine operator, farm, craft, or service worker, or not employed, 2 = clerical or sales worker, 3 = manager, administrator, or professional.

^dP-values for trend estimated by including education or occupation category as an ordinal variable in the linear regression model.

Table 3.

Adjusted odds ratios^a and corresponding 95% confidence intervals for being in higher category for Diet Quality Index-International (DQI-I) scores, estimated from logistic regression (n=423).

	Education categoryb			Occupation categoryc
	1 (n=206)	2 (n=147)	3 (n=70)	1 (n=338)	2 (n=34)	3 (n=51)
Variety d	1.0	1.0 (0.6–1.5)	1.3 (0.7–2.3)	1.0	1.4 (0.7–2.8)	1.4 (0.8–2.5)
Trend p e		0.44			0.47
Adequacy f	1.0	1.3 (0.9–2.1)	0.8 (0.5–1.6)	1.0	1.2 (0.6–2.4)	1.1 (0.6–2.1)
Trend p		0.87			0.74
Moderation d	1.0	0.6 (0.4–0.8)	0.6 (0.3–1.1)	1.0	0.6 (0.3–1.1)	0.8 (0.4–1.5)
Trend p		0.01			0.39
Balance d	1.0	1.2 (0.8–1.8)	1.1 (0.6–2.0)	1.0	1.0 (0.5–1.9)	1.5 (0.8–2.8)
Trend p		0.49			0.81
DQI-I score f	1.0	0.8 (0.5–1.3)	0.8 (0.4–1.4)	1.0	1.1 (0.5–2.2)	1.2 (0.6–2.3)
Trend p		0.31			0.40

^aAdjusted for age, marital status, birthplace, and score on General Ethnicity Questionnaire – American version.

^bEducation category 1 = 0–8 years of school, 2 = 9–12 years, 3 = at least some college.

^cOccupation category 1 = machine operator, farm, craft, or service worker, or not employed, 2 = clerical or sales worker, 3 = manager, administrator, or professional.

^dModeled in logistic regression for polychotomous outcome, with outcome categorized into tertiles.

^eP-values for trend estimated by including education or occupation category as an ordinal variable in the logistic regression model.

^fModeled in logistic regression for binary outcome (above vs. below median).

When women were jointly classified according to both education and occupation categories (Table 4), those in the highest education/occupation category consumed an average of 197.3 kcal and 12.2 g of sugar per day more than women in the lowest education/occupation category. They also had significantly higher dietary energy density. A trend toward lower dietary moderation (p=0.02) was also evident, although dietary moderation did not decrease monotonically with SEP category.

Table 4.

Adjusted estimates^a and p-values for trend for jointly categorized education and occupation in multivariate linear or logistic regression models (N=423).

	Joint categories of education and occupation b
	1 (n=192)	2 (n=187)	3 (n=44)	Trend p-value
	Beta estimatesc (p-values)
Energy (kcal)	ref	92.7 (0.01)	197.3 (0.004)	0.001
Energy density (kcal/g)	ref	0.01 (0.56)	0.15 (0.0001)	0.004
Total fat (% kcal)	ref	0.5 (0.42)	1.4 (0.21)	0.20
Total sugar (g)	ref	−0.9 (0.65)	12.2 (0.0008)	0.04
	Odds ratios (95% confidence intervals)
Sweet beveragesd	1.0	1.6 (0.8–3.5)	2.7 (0.9–8.1)	0.07
Varietye	1.0	1.1 (0.8–1.6)	1.3 (0.7–2.6)	0.39
Adequacyf	1.0	0.9 (0.5–1.4)	0.6 (0.2–1.4)	0.27
Moderatione	1.0	0.6 (0.4–0.8)	0.8 (0.3–1.4)	0.02
Balancee	1.0	1.1 (0.8–1.7)	1.5 (0.7–2.9)	0.28
DQI-I scoref	1.0	0.7 (0.5–1.1)	0.7 (0.3–1.5)	0.15

^aAdjusted for age, marital status, birthplace, and score on General Ethnicity Questionnaire – American version. Models for energy density, total fat, and total sugar also adjusted for energy intake.

^bJoint education/occupation category 1 = 0–8 years of school and machine operator, farm, craft, or service worker, or not employed, 3 = at least some college and manager, administrator, or professional, 2 = all others.

^cModeled in linear regression.

^dModeled in logistic regression for binary outcome (consumer vs. non-consumer).

^eModeled in logistic regression for polychotomous outcome.

^fModeled in logistic regression for binary outcome (above vs. below median).

DISCUSSION

The most notable finding in this sample of Chinese immigrant women was that SEP, as indicated by either education or occupation, was associated with lower dietary moderation. It was also associated with having higher energy density and percent of energy from fat, higher intake of energy and sugar, and greater likelihood of consumption of sugar-sweetened beverages.

The finding that higher education was associated with less dietary moderation confirms an association we observed previously in a separate sample of Chinese immigrant women ⁹. The dietary moderation component of the DQI-I score is of particular interest because it evaluates dietary factors that are related to chronic diseases and that may require restriction ³¹. Better educated women may have greater access, knowledge, opportunity, and/or resources to obtain more foods, leading to less dietary moderation, although this was not reflected in higher dietary variety or adequacy scores in the current sample as it was in the previous sample ⁹. In fact, in additional analyses in the current sample, we observed an inverse association between education and dietary adequacy when energy was included as a covariate in the logistic regression model (results not shown), contradicting our previous finding that education was associated with higher adequacy ⁹. This suggests that the more educated women in the current sample did not have less adequate diets per se because they had higher energy intake overall. However, their diets were less adequate (i.e., less nutrient-dense) for a given level of energy intake, as indicated by the inverse association between education and dietary adequacy when energy intake was controlled for.

Higher SEP has been linked to better diet and health, and lower SEP to poorer diet and health, in higher-income nations ^1–8. In low-income, developing countries, however, the opposite is true. Using income elasticity to quantify effects of income on food consumption in China, for example, Du et al. ³³ found that increased income was linked to a greater likelihood of consuming poultry and beef, and to a high-fat diet overall. In another analysis, Kim et al. ²³ found that markers of a healthy lifestyle (based on diet quality, physical activity, smoking, alcohol use) increased with income and education in the US but decreased with these SEP indicators in China. Recent reviews of the literature on SEP and obesity in countries at different stages of economic development demonstrate a similar phenomenon: Risk of obesity tends to decrease with SEP in higher-income countries but to increase with SEP in lower-income countries ^{8, 34}; associations are mixed in middle-income countries, indicating a shift in the burden of obesity from high to low SEP individuals.

Our findings suggest that among immigrants to higher-income countries, the association of SEP with diet follows the pattern of their country of origin. Consistent with this, a study conducted in the Netherlands observed a different association of SEP with diet between immigrant and non-immigrant residents; education was positively associated with diet quality among ethnic Dutch but not consistently among Surinamese immigrants ³⁵. Another study, while not conducted among immigrants, also demonstrates that the association of SEP with diet and health is not homogeneous across all population subgroups within a country. Among adults in the lowest quintile of income in seven states in Mexico, SEP, quantified using six different indicators, was positively associated with obesity, although the association between SEP and obesity is inverse in Mexico overall; the association was mediated in part by increased consumption of alcohol and soda ³⁶.

In higher-income countries, better diet quality with higher SEP may be due to better health knowledge and to greater financial capacity to purchase more nutritious, and costlier, foods ^{8, 23}. McLaren ⁸, citing the sociologist Pierre Bourdieu, also suggests that SEP, or ‘class,’ can be viewed as a ‘constellation of attributes’ including thinness and a healthy lifestyle, and that valuing such attributes may be internalized to distinguish among classes. In lower-income countries, however, the economic/material rather than social dimension of SEP or class is more important ⁸, and SEP reflects greater purchasing power, access to a larger variety of foods, and pursuit of new, not necessarily healthful dietary norms, desirable because of current or prior popularity of these products among individuals of higher SEP ^{8, 23, 36}. The implication for US immigrants is that an improved SEP, reflecting greater economic resources, may not have beneficial effects on diet and health as it does in the general US population. A reversal of this association – that is, for SEP to be associated with improved diet and health – may require acculturation in the form of internalization of the ‘constellation of attributes,’ including social and health-related attributes, that define higher SEP in the US. Additional analyses stratified on level of acculturation in the current sample of women did not reveal clear evidence that higher SEP was differently associated with better diet among women who were more acculturated (not shown). However, level of acculturation in the sample was low overall, and a greater range in level of acculturation may be required to detect such a difference in effect. Such acculturation may be difficult to achieve through health education. Providing immigrants with the opportunity and resources to achieve full social and economic integration would allow them to internalize class-related norms and to enjoy the health advantages that accompany higher SEP in the US.

Some limitations of our study are worth mentioning. First, this was a convenience sample of women of generally low acculturation and low SEP, raising the possibility of limited variability and generalizability. Nevertheless, the fact that we were able to detect significant differences in dietary intake even within this relatively homogeneous group points to the strength of effect of SEP on dietary behaviors and merits confirmation in other samples. Second, the sample is limited in size, although this allowed for a notable strength of the study – the ability to collect detailed, quantitative dietary data from participants in the form of multiple dietary recall interviews. We used 48-hour recalls rather than 24-hour recalls in order to balance the need to capture greater intra-individual variability in dietary intake with the logistical difficulty of trying to reach participants for interviews on multiple days. Additional analyses comparing the distributions of several nutrient variables based on days 2 and 4 of the four dietary recall days (the equivalent of two regular 24-hour recalls) with distributions based on all four days from the two 48-hour recalls showed comparable means, and the expected increase in variance in distributions based on only two rather than four days of recalls. Multiple comparisons are a potential limitation in this analysis, although the variables examined were selected a priori based on our research question and the prior literature.

In addition, we measured only education level and occupation as indicators of SEP, and the extent to which these adequately represented SEP in this sample is unclear. Associations with occupation may have been attenuated as a result of the relative homogeneity of our sample with respect to occupation. The appropriateness of using husband’s occupation as the basis for categorization is another concern ³⁷, although our preliminary analyses showed stronger results when based on husband’s occupation than when based on the woman’s own occupation (results not shown). We speculate that, in some cases, the husband’s occupation might more closely reflect true SEP or social status than the woman’s occupation, which may account for the stronger findings when using husband’s occupation. Further, because we did not collect information on income, we were unable to evaluate whether associations of SEP with diet might be due to differences in financial resources and purchasing power. In her review of SEP and obesity, McLaren ⁸ found that income and material wealth were the SEP indicators most strongly associated with obesity in countries undergoing a nutrition transition. Recent work also suggests that traditional indicators of SEP may not capture all dimensions of SEP in immigrant populations; education, for example, may not translate into the same gains in status and resources as it would in a non-immigrant population ³⁸. Examining a variety of different SEP indicators capturing other dimensions of SEP (e.g., wealth, subjective social status) may provide further insight into mechanisms by which SEP influences dietary behaviors.

Finally, the associations reported here are cross-sectional only. We are aware of no longitudinal data describing social mobility or changes in SEP or in relation to dietary behaviors or health in a Chinese immigrant sample. Whether change in SEP over time influences dietary changes deserves consideration in future work ³⁹.

In summary, we found significant differences in dietary intake and dietary moderation across categories of SEP in this sample of Chinese immigrant women. These findings suggest that dietary differences in immigrants may not, as often supposed, be completely attributable to differences in level of acculturation but may in fact be due partly to differences in SEP. Our findings also demonstrate that higher SEP in immigrants is not associated with better diet; rather, an association of SEP with less dietary moderation follows the pattern of their country of origin, in this case a lower-income country undergoing the nutrition transition. While these findings require confirmation in longitudinal studies, they suggest the importance among US immigrants of establishing and internalizing a change in norms regarding desirability of a healthy diet and lifestyle, regardless of whether the healthy diet is achieved through traditional or American eating habits.

Appendix

Appendix.

Components and scoring criteria for the Diet Quality Index – International (DQI-I)^a, adapted from Kim et al.³¹

Component	Maximum score	Scoring criteria
Variety
Overall food group variety (meat/poultry/fish/eggs; dairy/beans; grain; fruit; vegetable)	15	15: ≥1 serving/day from each food group
		12: Any 1 food group missing
		9: Any 2 food groups missing
		6: Any 3 food groups missing
		3: ≥4 food groups missing
		0: None from any food groups
Within-group variety for protein source (meat, poultry, fish, dairy, beans, eggs)	5	5: Meaningful consumption (≥0.5 serving/day) from ≥3 different sources
		3: 2 different sources
		1: 1 source
Maximum score	20	0: None
Adequacyb
Vegetable group	5	5: ≥3–5 servings/day
		0: 0 servings/day
Fruit group	5	5: ≥2–4 servings/day
		0: 0 servings/day
Grain group	5	5: ≥6–11 servings/day
		0: 0 servings/day
Fiber	5	5: ≥20–30 g/day
		0: 0 g/day
Protein	5	5: ≥10% of energy
		0: 0% of energy
Iron	5	5: ≥100% RDA
		0: 0% RDA
Calcium	5	5: ≥100% AI
		0: 0% AI
Vitamin C	5	5: ≥100% RDA
		0: 0% RDA
Maximum score	40
Moderation
Total fat	6	6: ≤20% of total energy
		3: >20–30% of total energy
		0: >30% of total energy
Saturated fat	6	6: ≤7% of total energy
		3: >7–10% of total energy
		0: >10% of total energy
Cholesterol	6	6: ≤300 mg/day
		3: >300–400 mg/day
		0: >400 mg/day
Sodium	6	6: ≤2400 mg/day
		3: >2400–3400 mg/day
		0: >3400 mg/day
Empty calorie foodsc	6	6: ≤3% of total energy
		3: >3–10% of total energy
		0: >10% of total energy
Maximum score	30
Overall balance
Macronutrient ratio (carbohydrate : protein : fat)	6	6: 55–65 : 10–15 : 15–25
		4: 52–<55 or >65–68 : 9–<10 or >15–16 : 13–<15 or >25–27
		2: 50–<52 or >68–70 : 8–<9 or >16–17 : 12–<13 or >27–30
		0: Otherwise
Fatty acid ratio	4	4: P:S 1–1.5 and M:S 1–1.5
		2: P:S 0.8–<1 or >1.5–1.7 and M:S 0.8–<1 or >1.5–1.7
		0: Otherwise
Maximum score	10

^aDQI-I, Diet Quality Index-International; RDA, Recommended Dietary Allowance; AI, Adequate Intake; P:S, ratio of polyunsaturated to saturated fatty acid intake; M:S, ratio of monounsaturated to saturated fatty acid intake.

^bAll sub-scores coded as continuous. Recommended intake of food groups depending on three levels of energy intake (≤1900 kcal, >1900–2500 kcal, and >2500 kcal). Nutrients percentage attainment of dietary recommended intakes.

^cDefined as foods for which sum of nutrient densities across 15 nutrients (protein, vitamin A, thiamin, riboflavin, vitamin B-6, vitamin B-12, niacin, folate, vitamin C, vitamin E, calcium, phosphorus, iron, magnesium, and zinc) is <1. Nutrient density calculated as (nutrient content / recommended nutrient intake) / (energy content / recommended energy intake). Recommended nutrient intake levels varied by age. Recommended energy intake was based on level of physical activity reported by participant.