Food security status of Native Hawaiians and Pacific Islanders in the US: analysis of a national survey

Abstract

Objective:

Document food insecurity prevalence among a nationally representative sample of Native Hawaiian and Pacific Islander (NHPI) adults and compare differences in food security status across races/ethnicities in the US.

Methods:

Using 2014 National Health Interview Survey (NHIS) and 2014 NHPI-NHIS data, food insecurity among NHPI is described, and a comparison of food security status across racial/ethnic groups is made using Rao-Scott chi-square and multinomial logistic regression.

Results:

Food insecurity prevalence was 20.5% among NHPI. NHPI had significantly higher odds of experiencing low and very low food security compared with Whites. Food insecurity among Hispanics, Blacks, and Other races/ethnicities was also significantly higher than that of Whites. Significant variation in food security status was observed by race/ethnicity (P<.001).

Conclusions and Implications:

This study provides the first documentation of food insecurity prevalence among NHPI and will inform chronic disease and nutrition research and programs conducted with NHPI communities in the US.

Food insecurity — a household-level condition of limited or uncertain access to adequate food due to lack of money or other resources — affects millions of households in the United States (US).¹ Food insecurity is a widely studied social determinant of health that has been associated with myriad health conditions,^2–8 inferior management of chronic diseases,^4,9–13 and numerous adverse health behaviors.^6,14–18 Food insecurity has also been shown to disproportionately affect racial/ethnic minority households.^1,19–23 However, no nationally representative study has documented food insecurity prevalence among Native Hawaiians and Pacific Islanders (NHPI) in the US. Given the high prevalence of diet-sensitive chronic diseases observed among NHPI adults, including obesity,^24–28 diabetes,^{24,25,29–31} hypertension,^{24,25,31–33} and obesity-related cancers,^25,34–36 it is important for researchers and policy-makers working with NHPI communities to understand food insecurity prevalence among this population.

Previous research documenting food insecurity among NHPI in the US has been limited to small non-representative samples of NHPI residing in Hawaii^37,38 and population-based surveys focused solely on NHPI residing in Hawaii.^39,40 The 3 most prominent sources of food insecurity data in the US do not report findings for NHPI. The US Department of Agriculture (USDA) conducts surveillance of food insecurity in the US annually,¹ using data gathered from the US Census Bureau’s Current Population Survey (CPS).⁴¹ However, USDA food insecurity reporting is limited to only 4 racial/ethnic categories: non-Hispanic Whites, non-Hispanic Blacks, Hispanics, and non-Hispanic Others.¹ Similarly, the Centers for Disease Control and Prevention’s (CDC) National Health Interview Survey (NHIS) and National Health and Nutrition Examination Survey (NHANES) capture population-based data on food insecurity; however, studies using these data typically only report on the same 4 racial/ethnic categories,^42–47 occasionally including Asians in analyses.^48,49 The lack of reporting on NHPI in USDA, NHIS, and NHANES studies has left an important gap in knowledge.

The purpose of this study is to fill this gap by documenting the food security status among a nationally representative sample of NHPI adults, and to compare differences in food security status across races/ethnicities.

METHODS

Data Sources

This cross-sectional study used data from the 2014 NHIS and the 2014 NHPI-NHIS. In 2014, the National Center for Health Statistics (NCHS) at the CDC launched the first and only NHIS focused exclusively on NHPI.⁵⁰ The questionnaire, based on the general NHIS, was administered to NHPI adults randomly selected from households in all 50 US states. In 2014, 2,590 NHPI adults aged 18 years and older and 36,697 adults aged 18 years and older were randomly sampled for the NHPI-NHIS and the NHIS, respectively. The surveys’ sampling procedures are designed to be representative of noninstitutionalized civilian adults living in the US. A more detailed description of the NHPI-NHIS and NHIS can be found elsewhere.^50–53

Study Population

The CDC defines NHPI as “A person having origins in any of the original peoples of Hawaii, Guam, Samoa, or other Pacific Islands“ (p. 7).⁵¹ From the NHPI-NHIS, 51% self-identified as NHPI alone, and 49% as NHPI in combination with 1 or more races. Adults in this latter group who did not consider NHPI as their primary race (n=418) were excluded from analyses. This resulted in a sample size of 2,172 NHPI respondents.

In the general NHIS, NHPI descent was not reported for confidentiality reasons as the ‘n’ was too small, so it was not possible to identify NHPI.⁵³ After accounting for missing data on the food security measure, non-Hispanic White (n=23,051), Hispanic (n=6,053), non-Hispanic Black/African-American (n=5,056), and non-Hispanic Asian (n=2,091) adults (hereafter referred to as Whites, Hispanics, Blacks, and Asians) from the general NHIS were included. American Indians/Alaska Natives and individuals of multiple races or with no primary race selected were categorized as “Other” (n=437). The final study sample comprised 38,860 respondents, representing 235,250,552 civilian noninstitutionalized adults aged 18 years and older living in the US.

Food Security

Household food security status was assessed in the NHIS and NHPI-NHIS via the 10-item U.S. Adult Food Security Survey Module (Table 1).⁵⁴ All respondents were asked the 10-item module, as opposed to the full battery of 18 items that are asked of households with children in the CPS. Furthermore, the NHIS and NHPI-NHIS food security module examines food security during the last 30 days, rather than during the past 12 months.

Table 1.

National Health Interview Survey Food Security Items, Response Options, and Scoring

Survey Items	Response Options	Score
1. We worried whether our food would run out before we got money to buy more	Often true	1
	Sometimes true	1
	Never true	0
2. The food we bought just didn’t last, and we didn’t have money to get more	Often true	1
	Sometimes true	1
	Never true	0
3. We couldn’t afford to eat balanced meals	Often true	1
	Sometimes true	1
	Never true	0
4. In the last 30 days, did you or other adults in your family ever cut the size of your meals or skip meals because there wasn’t enough money for food?	Yes	1
	No	0
5. (If “Yes” to Question 4) In the last 30 days, how many days did this happen? (open-ended, number of days, 1–30)	≥3 days	1
	≤2 days	0
6. In the last 30 days, did you ever eat less than you felt you should because there wasn’t enough money for food?	Yes	1
	No	0
7. In the last 30 days, were you ever hungry, bud didn’t eat because there wasn’t enough money for food?	Yes	1
	No	0
8. In the last 30 days, did you lose weight because there wasn’t enough money for food?	Yes	1
	No	0
9. In the last 30 days, did you or other adults in your household ever not eat for a whole day because there wasn’t enough money for food?	Yes	1
	No	0
10. (If “Yes” to Question 9) In the last 30 days, how many days did this happen? (open-ended, number of days, 1–30)	≥3 days	1
	≤2 days	0

Data source: NCHS, National Health Interview Survey, 2014. NCHS, Native Hawaiian and Pacific Islander National Health Interview Survey, 2014.

Note: All 10 items were dichotomized. A summative food security score (0–10) was created to represent the number of affirmative responses to the survey items. Answers of “Often true”, “Sometimes true”, and “Yes” were considered affirmative responses. Responses to items 5 and 10 regarding frequency of occurrence were considered affirmative if the respondent’s answer was greater than or equal to 3 days. Answers of “Don’t Know” and “Refused” were coded as ‘0’. The summative food security score was classified as: High food security (score=0); Marginal food security (score=1–2); Low food security (score=3–5); and Very low food security (score=6–10).

A food security score (0–10) was created to represent the number of affirmative responses to the food security items. Following procedures used by CDC and USDA, answers of “Often true”, “Sometimes true”, and “Yes” were considered affirmative responses.^53,55 All 10 items were dichotomized, and responses to items 5 and 10 regarding the frequency of occurrence in the past 30 days were considered affirmative if the respondent’s answer was greater than or equal to 3 days.⁵⁵ Following the approach used by CDC and USDA, the food security summative score was classified as: High food security (score=0); Marginal food security (score=1–2); Low food security (score=3–5); and Very low food security (score=6–10).^53,55

Sociodemographic factors

Since household food security is associated with sociodemographic factors other than race/ethnicity, analyses included additional sociodemographic variables available in the survey: age, sex, education level, marital status, uninsurance, English proficiency, nativity, and poverty level.

Survey-Based Cross-Race/Ethnicity Comparisons

The NHPI-NHIS and NHIS used the same questionnaire but different sampling methodology. For this reason, simply combining these 2 surveys into a single dataset may lead to biased estimates of standard errors. Kaminska and Lynn’s (2017) method was applied to compare and combine the 2 surveys in an unbiased way.⁵⁶ First, a stratum indicator that reflects the sampling strata from each survey was created. Stratum identifications were verified to remain unique after combining the 2 datasets. Second, because both surveys used a multi-stage design, a primary sampling unit (PSU) indicator was created to reflect multi-stage design with a unique value for each PSU. All PSUs were verified as still unique after combining both datasets. Third, sampling weights and variance estimation variables were taken into account, and confirmation that the point estimates had not changed after combining datasets was established. For all variables, the misspecification effect was assessed, which is the ratio of the true variance of a sample statistic under the complex sample design to the estimated variance when ignoring all or part of the sample design.⁵⁶ The inverse of the misspecification effect was computed to evaluate the factor by which the variance of the estimates was underestimated or overestimated. All specifications indicated by the NCHS regarding the use of sampling weights and the steps delineated by Kaminska and Lynn (2017) for combining datasets with distinct sampling designs were followed.^51,53,56 This allowed statistical analyses of food security status to be conducted across race/ethnicity.

Statistical Analyses

Stata/SE 15.1⁵⁷ svy procedures were used to estimate parameters and adjust for NHIS complex stratified-multistage-area-probability sampling design. The Rao-Scott chi-square (χ²) test of independence, which accounts for sampling design, was used to determine statistically significant differences in household food security status across race/ethnicity.⁵⁸

Associations between household food security and race/ethnicity were examined by conducting multinomial logistic regression. Two separate models were estimated: Model 1 with race/ethnicity as the sole main independent predictor; and Model 2 with race/ethnicity along with additional sociodemographic variables. Model 2 allowed for examination of the extent to which the influence of race/ethnicity on household food security status was attenuated when the influence of other potential factors are considered.

High food security was selected as the base outcome, and Whites were chosen as the reference group. Whites were chosen as the reference group to facilitate comparisons with other food security studies, and to allow for the reporting of specific odds ratios for NHPI as compared to Whites. Although multinomial logistic regression is typically used to estimate nominal response variables, it was used here as an alternative to ordinal regression since the proportional odds assumption was violated.⁵⁹ Statistical analyses included only complete cases (i.e., no imputation of missing data was used). Statistical significance was determined at α=.05.

All study procedures described herein were reviewed and exempted by the University of Arkansas for Medical Sciences’s Institutional Review Board.

RESULTS

Approximately 1 in 5 (20.5%) NHPI reported low (12.5%) or very low (8.0%) food security (Table 2). Only 68.2% of NHPI experienced high food security, a proportion similar to Blacks and Other races/ethnicities. A lower proportion of Whites experienced low food security or very low food security than all other races/ethnicities, except for Asians. There was a significant association between levels of food security and race/ethnicity — Omnibus χ² was significant (P<.001).

Table 2.

Food Security Status across Race/Ethnicity in the US (Weighted Proportions and 95% Confidence Intervals)^a

Food Security Status	NHPI n=2,172	White n=25,051	Hispanic n=6,053	Black n=5,056	Asian n=2,091	Other n=437	Total n=38,860
High food security	68.2	87.2	71.7	68.2	89.3	66.1	82.4
	(64.7–71.6)	(86.5–87.8)	(70.1–73.2)	(66.5–69.7)	(87.3–91.0)	(58.1–73.4)	(81.8–83.0)
Marginal food security	11.3	5.2	11.4	11.8	5.6	10.8	7.0
	(9.1–13.9)	(4.7–5.6)	(10.5–12.4)	(10.7–13.0)	(4.4–7.2)	(6.6–17.2)	(6.6–7.3)
Low food security	12.5	4.0	11.0	11.4	3.8	10.9	6.0
	(10.2–15.2)	(3.7–4.4)	(10.0–12.0)	(10.2–12.6)	(3.0–4.9)	(6.8–17.0)	(5.7–6.4)
Very low food security	8.0	3.7	5.9	8.7	1.3	12.2	4.6
	(6.1–10.4)	(3.3–4.0)	(5.2–6.7)	(7.9–9.6)	(0.7–2.2)	(8.1–18.0)	(4.3–4.9)

Data source: NCHS, National Health Interview Survey, 2014. NCHS, Native Hawaiian and Pacific Islander National Health Interview Survey, 2014.

Note: 95% confidence intervals are in parentheses. n=unweighted sample size. NHPI=Native Hawaiian and Pacific Islander.

^aRao-Scott chi-square test: F(15, 313=56.47; P<.001).

Associations were further expressed in terms of odds-ratios (OR) (Table 3). Looking at the association between household food security status and race/ethnicity alone (Model 1), NHPI were 4 times as likely to experience low food security as Whites (OR: 4.0; 95% confidence intervals [CI]: 3.1–5.1). The odds of experiencing low food security versus high food security for Blacks, Hispanics, and people of Other races/ethnicities were, respectively, 3.6, 3.3, and 3.5 times as large as the odds for Whites.

Table 3.

Associations between Race/Ethnicity and Food Security Status in the US: Results from Multinomial Logistic Regression Models (Odds Ratios and 95% Confidence Intervals)

MODEL 1†
Race/Ethnicity	High Food Securitya	Marginal Food Security	Low Food Security	Very Low Food Security
Whiteb	–	–	–	–
NHPI	–	2.8 (2.2–3.6)***	4.0 (3.1–5.1)***	2.8 (2.0–3.8)***
Hispanic	–	2.7 (2.3–3.1)***	3.3 (2.9–3.8)***	2.0 (1.7–2.3)***
Black	–	2.9 (2.5–3.4)***	3.6 (3.1–4.2)***	3.0 (2.6–3.6)***
Asian	–	1.1 (0.8–1.4)	0.9 (0.7–1.2)	0.3 (0.2–0.6)***
Other	–	2.8 (1.6–4.8)***	3.5 (2.0–6.1)***	4.4 (2.7–7.1)***
MODEL 2‡
Race/Ethnicity	High Food Securitya	Marginal Food Security	Low Food Security	Very Low Food Security
Whiteb	–	–	–	–
NHPI	–	2.3 (1.8–3.1)***	3.5 (2.7–4.5)***	2.6 (1.7–3.8)***
Hispanic	–	1.7 (1.4–2.1)***	1.7 (1.4–2.1)***	1.5 (1.1–1.9)**
Black	–	2.0 (1.6–2.3)***	2.3 (1.9–2.7)***	1.9 (1.6–2.4)***
Asian	–	1.1 (0.8–1.6)	0.8 (0.6–2.1)	0.4 (0.2–0.9)*
Other	–	1.8 (0.9–3.4)	2.4 (1.2–4.6)*	2.6 (1.4–4.8)**

Data source: NCHS, National Health Interview Survey, 2014. NCHS, Native Hawaiian and Pacific Islander National Health Interview Survey, 2014.

Note: 95% confidence intervals are in parentheses. NHPI=Native Hawaiian and Pacific Islander.

^aHigh food security is the base outcome.

^bWhites are the reference group.

^†Model 1 presents the associations between race/ethnicity and household food security status.

^‡Model 2 presents the associations between race/ethnicity and household food security status holding constant sociodemographic variables (age, sex, education, marital status, uninsurance, English proficiency, born in the US, Federal Poverty Level). Odds-ratios for the sociodemographic factors are reported in Supplemental material.

^*P<.05;

^**P<.01;

^***P<.001

NHPIs were 2.8 times as likely to experience very low food security as Whites (OR: 2.8; CI: 2.0–3.8). In contrast, Asians were less likely to experience very low food security than Whites (OR: 0.3; CI: 0.2–0.6). As compared with Whites, Blacks, Hispanics, and Other races/ethnicities face triple (OR: 3.0; CI: 2.6–3.6), double (OR: 2.0; CI: 1.7–2.3), and quadruple (OR: 4.4; CI: 2.7–7.1) the respective odds of experiencing very low food security versus high food security.

When additional sociodemographic covariates were held constant in multivariable regression (Model 2), the magnitude of the estimate for race/ethnicity was attenuated across levels of food security status (Supplemental Table 1). Yet, the statistical significance of the association between race/ethnicity and household food security status found in Model 1 remained. For very low food security, NHPIs were still twice as likely to experience very low food security as Whites (OR: 2.6; CI: 1.7–3.8). Asians were still less likely to experience very low food security than Whites (OR: 0.4; CI: 0.2–0.9). Blacks, Hispanics, and Other races/ethnicities were still more likely than Whites to experience very low food security versus high food security (OR: 1.9; CI: 1.6–2.4), (OR: 1.5; CI: 1.1–1.9), and (OR: 2.6; CI: 1.4–4.8), respectively.

DISCUSSION

In the US, NHPI are the second fastest growing population⁶⁰ and have been underrepresented in health research.⁶¹ This study provides the first documentation of household food insecurity prevalence among NHPI in the US using a nationally representative sample. Prevalence of low and very low food security among NHPI were significantly higher than those of Whites, Asians, and Hispanics, and similar to those of Blacks and Other races/ethnicities. Whites and Asians had the highest levels of food security, while Blacks and Hispanics had much lower levels of food security. These findings are consistent with other studies that find similar patterns of food insecurity among national samples.^1,19–23 Furthermore, the findings remained relatively consistent across racial/ethnic categories when sociodemographic covariates were added to the regression model.

However, the findings of the present study differ slightly from the pattern of results in USDA’s Household Food Security in the United States in 2014 report.⁶² Compared to USDA reporting, the present study’s results appear to underestimate food insecurity among the 3 largest racial/ethnic groups. Specifically, USDA reported food insecurity prevalence of 10.5% for Whites, 26.1% for Blacks, and 22.4% for Hispanics,⁶² compared with 7.7%, 20.1%, and 16.9%, respectively, observed in this study. The majority of these observed differences are likely attributable to the different reference periods used by the measures (30 days for NHIS and 12 months for CPS). Given the seasonal and episodic nature of food insecurity,^63,64 it follows that 2 measures using different reference periods would produce differing estimates. In fact, Nord (2002) examined the differences between the estimates produced by measures of each time period.⁶⁵ Using a “ratio of prevalences” across 3 years of the CPS data, Nord (2002) found a 30-day measure produced estimates of very low food security (formerly named food insecurity with hunger) approximately 70% (66.1–74.4%) of those observed when using the 12-month measure.⁶⁵ Nord’s (2002) approximately 70% ratio aligns fairly closely with the ratios between the present study’s estimates for overall food insecurity and the estimates of overall food insecurity reported by USDA for 2014: 75% ratio for all food insecure households; 73% for food insecure Whites; 77% for food insecure Blacks; and 75% for food insecure Hispanics. These results suggest food insecurity prevalence for NHPI measured using a 12-month period could be as high as 25–30%.

Some differences may also be related to differences between the 10-item adult module used by the NHIS and the 18-item household module used by the CPS. While it is well established that having children in a household increases household food insecurity prevalence,¹ it is unclear whether asking only the adult module, as opposed to the household module, significantly impacts estimates for households with children. Asking the complete battery of items tends to only identify the most extreme cases of food insecurity, as the cutoff for being classified as being food insecure is the same for both measures (≥3 affirmative responses).¹ In fact, Bickel et al. (1995) refer to the final 8 items for households with children as “disproportionately the more severe questions” (pg. 22).⁶⁶

Food insecurity estimates may also be impacted by how the respondent views the concepts of “family201D and “households” as presented in the food security module. Examining food insecurity among American Indians, Gundersen (2008) notes the probability of answering affirmatively to food security items may be increased if the respondent’s definition of household is different, based on cultural differences.⁶⁷ In fact, past research indicates NHPI have reported broader views of family and households that extend to distant relatives and individuals that may not always reside in the household,⁶⁸ which may impact food insecurity estimates among this group. If this finding holds true for respondents to the NHPI-NHIS, then the estimates reported here may be slightly inflated due to cultural differences in how a household is defined.

The primary limitation to this study, as well as the USDA food security estimates, is that the surveys rely on cross-sectional data to describe food security. This limitation is common to many surveys, including the widely cited annual Household Food Security in the United States reports,¹ but it is important to acknowledge that following a sample of households over time may yield a different pattern of results for NHPI and the other races/ethnicities than the cross-sectional analysis. In addition, at the time of analysis, the NHPI-NHIS data were already 5 years old. However, the public use data set was not released until 2017, and there is no other nationally representative survey that samples sufficient numbers of NHPI in order to characterize household food security status.

Another limitation of the present study is the decision to designate non-Hispanic Whites as the reference group for analyses, rather than NHPI. This approach focuses attention on differences between NHPI and Whites and limits comparisons between NHPI and other races and ethnicities. This approach was followed for consistency with the approach used in the vast majority of food security research (i.e., non-Hispanic Whites as reference group), which allows for comparisons of this study to past studies. It also allows for the direct reporting of specific odds ratios for NHPI, which will be of particular interest to readers of this study.

Future studies using national surveys, including the CPS and NHANES, would benefit from including analysis of NHPI when possible. Although CDC has announced no plans for a follow-up NHPI-NHIS, the findings of this study support continued oversampling of NHPI in order to monitor changes in food insecurity and other key health indicators. In addition, other groups—including American Indians/Alaska Natives and individuals of multiple races—who are often included in “Other” races/ethnicities in food security reporting would benefit from targeted oversampling in future population-based surveys.

IMPLICATIONS FOR RESEARCH AND PRACTICE

This study fills an important gap in the literature by documenting food insecurity prevalence among a nationally representative sample of NHPI adults. NHPI in the US experience high rates of diet-sensitive chronic diseases,^24–36 and the results of the present study will inform researchers working with NHPI communities to prevent and manage these conditions. By documenting food insecurity prevalence among NHPI in a national sample, this study will serve as an important foundation for research and programs focused on addressing the high prevalence of food insecurity observed in this study. The disparities in food insecurity between NHPI and other races and ethnicities found in this study illustrate the need for researchers to identify and test strategies to improve food security among NHPI groups, taking into account the cultural preferences of specific communities.^68,69 The present study’s findings must be considered in the context of past research identifying a range of health and socioeconomic disparities for NHPI, including increased chronic disease risk^24–36 and barriers to health care access.^70–72 In that context, the prevalences and disparities documented in this study will be immediately useful to policy-makers, health professionals, nutrition educators, and food systems researchers to inform policy and practice to improve NHPI communities’ access to sufficient healthy foods.