Investigating the Association between English Proficiency and General Health among Marshallese Adults in the United States

Abstract

Prior studies have not examined the association between English proficiency and health among Marshallese. Cross-sectional data from 378 Marshallese adults enrolled in a Diabetes Prevention Program study in Arkansas and Oklahoma were used to document English proficiency, assess the relationship between sociodemographic health-related characteristics and English proficiency, and examine the association between English proficiency and general health. Approximately one-fifth of participants reported LEP. One-fifth of participants reported being in fair or poor health. General health, age group, education, household food insufficiency, inability to afford necessities, and comorbidities were significantly associated with LEP. Participants with LEP were seventy-eight percent less likely to report better general health compared with those who reported speaking English very well. LEP was high among this sample of Marshallese adults. Findings highlight the need for additional Marshallese clinical interpretation and translation services.

Introduction

Between 2000 and 2010, Native Hawaiians and Pacific Islanders (NHPI) were the second largest increasing population in the United States (US).¹ Southern and Midwestern states (e.g., Arkansas, Kansas, Missouri, and Oklahoma) experienced marked growth in their Pacific Islander migrant communities.¹ For example, the Marshallese – a Micronesian Pacific Islander population – grew by two-hundred and fifty-two percent in Arkansas.¹ Arkansas now has the largest Marshallese community in the continental US.¹

Marshallese adults have high rates of type 2 diabetes^{2, 3}, cardiovascular disease⁴, cancer⁴, and poor social determinants of health.³ One community-based study conducted with 401 Marshallese adults in Arkansas found that seventy-one percent had HbA1c levels indicative of pre-diabetes or type 2 diabetes, the majority of participants (89.7%) were overweight or obese, and most of the sample (80.3%) had blood pressure readings indicative of pre-hypertension or hypertension.³ Despite these well-documented disparities, little is known with regard to how Marshallese adults perceive and rate their overall health status. The one available study found among Marshallese adults with type 2 diabetes (n=221), approximately one-fifth of the sample (18.9%) rated their health as fair or poor and only a small number of participants (3.7%) rated their health as excellent.⁵

Limited English proficiency (LEP) is a prominent issue among immigrant and migrant populations that is frequently associated with difficulties navigating the health care system^{6, 7} and poorer self-reported health.^{8, 9} One study of the Micronesian population in Hawaii found that about half of participants (47.1%) reported having “numerous problems” related to language (i.e., limited English language skills impeded them from obtaining needed services, and relatives often acted as translators).¹⁰ Qualitative studies have shown that language is a common barrier for Marshallese and healthcare providers in Arkansas.^{11, 12} Even when Marshallese translators are available, many medical terms do not have a direct translation.¹² Although language barriers have been noted in qualitative studies, no study to date has quantitatively documented the prevalence of LEP among Marshallese in Arkansas or investigated its relationship to health status.

Purpose of the Study

The aims of the present study are to: 1) document English proficiency; 2) determine what sociodemographic and health-related characteristics are associated with English proficiency; and 3) determine whether English proficiency is associated with general health.

Methods

Participants

This study used cross-sectional data from Marshallese participants recruited for a Diabetes Prevention Program study. To be eligible, participants had to be Marshallese, aged 18 years or older, and have a BMI ≥25. All participants were recruited by bilingual Marshallese staff. Additional details regarding study design and recruitment are published elsewhere.^{13, 14} Study protocols were approved by the University of Arkansas for Medical Sciences Institutional Review Board (IRB#207034).

Data Collection

All data were collected in-person by bilingual (Marshallese and English) research study staff. Survey data were collected in Marshallese or English, dependent on participants’ preference.

Measures

Sociodemographic and health-related characteristics were assessed using valid and reliable survey questions, including several items from the Behavioral Risk Factor Surveillance System (BRFSS).

English proficiency was assessed by a single item: “How well do you speak English? Very Well; Well; Not Well; or Not at All.” Responses of “Not at All” and “Not Well” were combined for statistical analysis purposes and considered LEP.

General health status was assessed by asking the participants: “Would you say that in general your health is: Excellent; Good; Fair; or Poor?” Due to the extremely low number of responses of “Poor”, responses were combined to Fair/Poor.

Sociodemographic variables included in this study are self-reported age, sex, education level, marital status, employment status, health insurance, household food sufficiency status, and ability to pay for necessities (financial strain). Health-related characteristics included BMI and comorbidities. BMI ((weight in pounds/[height in inches]2)*703) was assessed by measuring participants’ weight (without shoes) to the nearest .2 pound using a calibrated digital scale and measuring participants’ height (without shoes) to the nearest .25 inch using a stadiometer. Comorbidities indicate the total number of diagnosed health conditions (heart disease, hypertension, cancer, and type 2 diabetes) reported by participants.

Analysis

All analyses were carried out using SAS/STATv14.2. Descriptive statistics were calculated for all measures. Chi-square tests were used to assess the relationship between English proficiency and sociodemographic variables. Spearman rank correlations were used to assess the relationship between English proficiency and BMI and comorbidities. The association between English proficiency and general health was assessed using ordinal logistic regression, adjusting for sociodemographic and health-related characteristics (age, sex, education level, marital status, employment status, health insurance, household food sufficiency status, ability to afford necessities, BMI, and number of comorbidities). All variables except for sex met the proportional odds assumption. Sex was treated as an unstructured covariate, allowing odds ratios to vary by level of the outcome variable. Results were considered significant at the alpha level of .05.

Results

A total of 378 Marshallese adults were recruited and enrolled. Table 1 provides participant sociodemographic and health outcomes, and characteristics by English proficiency level. One-fifth (20.9%) of participants reported they spoke English not at all/not well. One-fifth (20.2%) of participants reported being in fair/poor health. Significant associations were observed between English proficiency and general health status (χ²(4)=18.8, p=<.001). A larger proportion of those in excellent health reported speaking English very well (30.6%) compared with those in good (13.6%) or fair/poor (11.8%) health. A larger proportion of those in fair/poor health reported speaking English not at all/not well (30.3%) compared with those in good (21.5%) or excellent (4.1%) health. Significant associations were also observed between LEP and age group, education level, household food sufficiency status, ability to pay for necessities, and comorbidities. Specifically, larger proportions of older adults, those with less education, those from food insufficient households, those with difficulties paying for necessities, and those with more comorbidities reported speaking English not at all/not well (Table 1).

Table 1.

Participant Socio-Demographic and Health-Related Characteristics, overall and by English Proficiency levels, N=378

Baseline Characteristics	Total n (%)	English Proficiency
		Not at all/Not Well n (%)	Well n (%)	Very Well n (%)	χ2
English Proficiency
Not at all/Not Well	79 (20.9)	-	-	-
Well	241 (63.8)	-	-	-
Very Well	58 (15.3)	-	-	-
General Health Status					18.8***
Fair/Poor	76 (20.2)	23 (30.3)	44 (57.9)	9 (11.8)
Good	251 (66.8)	54 (21.5)	163 (64.9)	34 (13.6)
Excellent	49 (13.0)	2 (4.1)	32 (65.3)	15 (30.6)
Age					22.3**
18–34	98 (25.9)	15 (15.5)	56 (57.7)	26 (26.8)
35–44	140 (36.9)	24 (17.1)	101 (72.1)	15 (10.7)
45–54	84 (22.2)	25 (29.8)	46 (54.8)	13 (15.5)
55+	57 (15.0)	15 (26.3)	38 (66.7)	4 (7.0)
Sex					5.13
Female	214 (56.6)	53 (24.8)	127 (59.4)	34 (15.9)
Male	164 (43.4)	26 (15.9)	114 (69.5)	24 (14.6)
Education Level					81.9***
Less than HS	184 (48.7)	65 (35.3)	112 (60.9)	7 (3.8)
HS Diploma/GED	132 (34.9)	14 (10.6)	92 (69.7)	26 (19.7)
Beyond HS	62 (16.4)	0 (0.0)	37 (59.7)	25 (40.3)
Marital Status a					3.9
Single	77 (20.4)	17 (22.1)	43 (55.8)	17 (22.1)
Married	301 (79.6)	62 (20.6)	198 (65.8)	41 (13.6)
Employment Status b					3.0
Employed	206 (54.5)	37 (18.0)	139 (67.5)	30 (14.6)
Unemployed	172 (45.5)	42 (24.4)	102 (59.3)	28 (16.3)
Health Insurance					0.3
Yes	214 (57.5)	35 (22.2)	100 (63.3)	23 (14.6)
No	158 (42.5)	43 (20.1)	138 (64.5)	33 (15.4)
Household Food Sufficiency Status					15.1***
Food Insufficient	97 (25.7)	32 (33.0)	58 (59.8)	7 (7.2)
Food Sufficient	281 (74.3)	47 (16.7)	183 (65.1)	51 (18.2)
Ability to Pay for Necessities					22.3***
Not Hard at All	66 (17.8)	14 (21.2)	33 (50.0)	19 (28.8)
Somewhat Hard	281 (75.7)	55 (19.6)	196 (69.8)	30 (10.7)
Very Hard	24 (6.5)	10 (41.7)	11 (45.8)	3 (12.5)
	Mean±SD				r †
Body Mass Index	33.7±5.4				0.02
Comorbidities c	1.54±0.74				−0.21***

Notes: SD, standard deviation; HS, high school. Percentages may not total 100 due to rounding.

^aSingle includes response options: Single, Divorced/Separated, Widowed, and A member of an unmarried couple.

^bEmployed includes response options: Full Time, Part Time, and Self-Employed; Unemployed includes response options: Unable to find work for 1 year or more, Unable to find work for less than 1 year, Retired, Student, Disabled, and Take care of home/family.

^cComorbidities is the sum of self-reported previous diagnoses of heart disease, hypertension, cancer, and/or type 2 diabetes.

^*p<.05;

^**p<.01;

^***p<.001

^†Spearman rank correlations conducted for continuous variables.

Table 2 provides the results from the ordinal logistic regression. After controlling for sociodemographic and health variables, there was a significant association between English proficiency and general health status. Specifically, participants who reported speaking English not at all/not well were seventy-eight percent less likely to report better general health (excellent health or good health vs. fair/poor health) compared with those who reported speaking English very well (OR=.22, CI:.09, .54). Participants who reported speaking English well were sixty-six percent (66%) less likely to report better general health (excellent health or good health vs. fair/poor health) compared with those who reported speaking English very well (OR=.34, CI=.16, .70). Significant associations were also found between general health and age group, sex, education level, household food sufficiency status, and comorbidities (Table 2).

Table 2.

Association between General Health Status and English Language Proficiency, controlling for Sociodemographics and Health-Related Characteristics – Results of Ordinal Logistic Regression

Measures	OR	95% CI	p-value
English Proficiency
Not Well/Not at All	0.22	0.09, 0.54	0.001
Well	0.34	0.16, 0.70	0.003
Very Well†	-	-	-
Age (years)
18–34	1.80	0.78, 4.17	0.172
35–44	2.33	1.07, 5.07	0.033
45–54	1.17	0.54, 2.52	0.699
55+†	-	-	-
Sex ‡
Female (Excellent v. Fair/Poor)	1.33	0.68, 2.60	0.411
Female (Good v. Fair/Poor)	0.43	0.23, 0.81	0.009
Male†	-	-	-
Education Level
Less than High School	2.15	1.03, 4.50	0.042
High School Diploma/GED	1.49	0.72, 3.07	0.283
Beyond High School†	-	-	-
Marital Status a
Single	0.74	0.42, 1.32	0.303
Married†	-	-	-
Employment Status b
Employed	1.48	0.86, 2.57	0.161
Unemployed†	-	-	-
Health Insurance
Yes	1.39	0.81, 2.37	0.229
No†	-	-	-
Household Food Sufficiency Status
Food Insufficient	0.43	0.25, 0.76	0.003
Food Sufficient†	-	-	-
Ability to Afford Necessities
Not Hard at All	1.17	0.41, 3.31	0.766
Somewhat Hard	1.83	0.74, 4.54	0.190
Very Hard†	-	-	-
Body Mass Index	1.01	0.97, 1.05	0.720
Comorbidities c	0.64	0.45, 0.92	0.020

Notes: Statistically significant p-values are bolded. OR=Odds Ratio; CI=Confidence Intervals.

^†=reference category.

^‡The sex variable did not meet the proportional odds assumption; therefore, it was treated as an unstructured covariate, allowing odds ratios to vary by level of the outcome variable.

^aSingle includes response options: Single, Divorced/Separated, Widowed, and A member of an unmarried couple.

^bEmployed includes response options: Full Time, Part Time, and Self-Employed; Unemployed includes response options: Unable to find work for 1 year or more, Unable to find work for less than 1 year, Retired, Student, Disabled, and Take care of home/family.

^cComorbidities is the sum of self-reported previous diagnoses of heart disease, hypertension, cancer, and/or type 2 diabetes.

Discussion

The purpose of this study was to document English proficiency among Marshallese adults, determine which characteristics were associated with English proficiency, and examine the association between English proficiency and general health. In total, the majority (84.7%) of the sample reporting speaking English less than very well, and one-fifth (20.9%) reported they spoke English not well/not at all. These proportions are much higher than the proportions among US residents who speak a second language, among whom only a small percentage (8.3%) report speaking English less than very well.¹⁵ The study findings are consistent with prior research that has found LEP to be a common issue among Marshallese migrants in other parts of the US.^{10, 16} Our results also add valuable quantitative data that reinforce the findings of prior qualitative studies with Marshallese community members and healthcare professionals in Arkansas. These studies have shown that language barriers impact delivery of primary and prenatal care, contribute to medication non-adherence, and often cause patients to skip appointments when interpreters are not available.^{11, 12, 17–21}

A significant association was found between English proficiency and general health. Among those who reported being in fair/poor health, almost one-third (30.3%) spoke English not well/not at all. This is similar to findings in a prior US population-based study that found significantly more LEP individuals were in fair/poor health, compared with those who were English-proficient.²² A significant association was also found between age group and English proficiency. Similar to previous studies, younger adults are often found to be more English proficient than older adults, especially among migrant/immigrant communities.²³ Comorbidities were found to be negatively associated with English proficiency, with those who had LEP reporting more comorbidities. This finding is consistent with previous research, which find comorbidities to be more common among those with LEP.^{6, 24}

Regression results indicated participants who reported speaking English very well were significantly more likely to report being in better health, compared with those who reported speaking English well and those who speak English not well/not at all. To our knowledge, this is the first time this association has been documented among Marshallese adults. Prior research has documented similar associations between LEP and general health in other populations, including Asian, Latinx, and African immigrants.^25–30 Prior research has also found that patients with LEP were more likely to have low health literacy and receive lower quality care compared to those without LEP.^{27, 31–34} This study did not capture measures of health literacy or satisfaction with care received. Future studies should explore the relationship between LEP and health literacy, as well as the relationship between LEP and satisfaction with healthcare among Marshallese patients in the US.

Limitations

The findings of this study should be interpreted by considering two important limitations. First, the study sample consists solely of Marshallese adults with BMI ≥25, which limits the generalizability of the results to other populations, as well as to normal/healthy weight Marshallese adults. Second, the study is limited to analyzing cross-sectional data, so causal inferences cannot be made about the observed associations (e.g., poor English skills cause poor health). Despite these limitations, this is the first study to examine the relationship between LEP and general health status among any Marshallese sample.

Conclusions

This study documented LEP among Marshallese adults, and provided previously unavailable data on the Marshallese community. The findings of this study will be useful for healthcare providers and researchers working with Marshallese communities in the US. Given the health disparities faced by the community, this study highlights the critical need for clinical interpretation and translation services.^{35, 36} Past research has cited language concordance between patients and providers as a means of improving patient care and outcomes^{37, 38}; however, this is not presently a viable solution, given the small number of trained providers in the US who speak Marshallese. Previous efforts undertaken with the community such as training community health workers³⁹ and providing cultural competency trainings for healthcare providers^{40, 41} should be expanded, as they may help to improve patient care and, ultimately, improve the stark health disparities faced by the Marshallese community.