Validation of a 2-Item Food Insecurity Screen Among Adult General Medicine Outpatients

Abstract

Food insecurity is defined as limited access to food and is associated with adverse physical, social, and emotional health outcomes. As social needs are addressed in heath care, efficient methods to identify patients living in food insecure households are necessary. A 2-item screen (HFSS-2) derived from the US Department of Agriculture Household Food Security Scale (HFSS-18) has been validated among parents of pediatric patients with a sensitivity of 97% and specificity of 83%. The objective was to validate the HFSS-2 in adult general medicine outpatients. HFSS-18 was administered to a sample of adult general medicine outpatients in Delaware from 2018 to 2019. The authors evaluated the sensitivity and specificity of the HFSS-2. Multivariable logistic regression was used to calculate convergent validity between the HFSS-18 and the HFSS-2. Three hundred ninety patients were approached with 295 (75%) enrolling in this study; 17.6% (52/295) were food insecure. A confirmatory response to either of the 2 items from the HFSS-2 had a sensitivity of 98% (95% CI: 94%, 100%) and specificity of 91% (95% CI: 87%, 94%). Food insecurity was associated with increased odds of coronary heart disease (adjusted odds ratio [AOR] 4.63; 95% CI: 1.55, 13.79; AOR 4.19; 95% CI: 1.51, 11.59) and diabetes (AOR 4.19; 95% CI: 1.94, 9.08; AOR 3.73; 95% CI: 1.83, 7.92) using both the HFSS-18 and the HFSS-2. HFSS-2 was found to be highly sensitive and specific. This is the first study to validate this tool in this population that the authors are aware of.

Introduction

Food insecurity is characterized as the limited and uncertain access to foods.¹ In 2015, 12.7% of the US population was food insecure.¹ The prevalence of food insecurity is higher among racial and ethnic minorities, low-income households, and those who have chronic illness.^1

–5 For example, rates of food insecurity among low-income patients with diabetes are 4 to 8 times greater than national averages with a prevalence of 40% to 100%.^6

–10 Identification of food insecurity can improve health and wellness.

Identifying food insecurity among patients with chronic disease is important because it can trigger interventions. Therefore, a reliable and convenient instrument is necessary. There are several validated instruments to assess food insecurity. The gold standard instrument is the 18-item US Department of Agriculture Household Food Security Scale (HFSS-18).¹¹ A 2-item version of HFSS-18 (HFSS-2) was developed for pediatric populations and their parents. This version is widely used across the United States to screen individuals for food insecurity. However, this instrument has not been validated among adult general medicine outpatients.

To the research team's knowledge, the literature on validation of HFSS-2 in the adult general medicine population is sparse. Therefore, the team examined the validity of the widely used HFSS-2 screening tool in an adult population in primary care outpatient settings. The objectives of this study were to estimate the prevalence of food insecurity in the general medicine population, describe differences in prevalence based on demographic and clinical characteristics, and validate the HFSS-2 within this patient population. The team hypothesized that the HFSS-2 would be a valid instrument to use among adult general medicine outpatients.

Methods

Participants

The research team conducted a survey study with a convenience sample of adult patients from 4 internal medicine and family medicine primary care offices at ChristianaCare Health System in New Castle County, Delaware from December 2018 to October 2019. Surveys were developed on REDCap, were read to participants, and data were collected on iPads by trained research assistants in clinic examination rooms to ensure privacy. If the interview was interrupted by a provider, the research assistants reapproached participants in the examination room at the conclusion of their appointment. To participate in the study, individuals had to be 18 years of age or older and English speaking. Additionally, subjects were excluded if the medical assistants felt they could not understand questions because of cognitive impairment. Informed consent was obtained from each participant at the time of the study. This study was reviewed and approved by the ChristianaCare Institutional Review Board.

Measures

This survey consisted of 40 closed-ended questions that included demographic, general health, and food insecurity measures and a question related to housing instability. Demographic questions consisted of the participant's name, date of birth, age, sex, race, ethnicity, and zip code. Housing stability was measured with the question: “What is your living situation today?” derived from the Protocol for Responding to and Assessing Patients' Assets, Risks, and Experiences (PRAPARE) assessment tool.¹² General health was measured using 10 questions related to self-reported chronic medical conditions, including cardiovascular disease, rheumatology disease, diabetes, and depression, derived from the 2018 Behavioral Risk Factor Surveillance System Questionnaire, Core Section 6 Chronic Health Conditions.¹³ Additional demographic information was collected from the electronic health record, such as marital status (married or other) and insurance status (Medicare, Medicaid, commercial, other). Food insecurity was assessed using HFSS-18 and was treated as a binary variable (food insecure vs food secure) with a raw score of ≥3 indicating food insecurity.¹⁴

HFSS-2 consists of the first 2 questions of HFSS-18

“We worried whether our food would run out before we got money to buy more. Was that often true, sometimes true, or never true for your household in the last 12 month?”

“The food we bought just didn't last and we didn't have money to get more. Was that often true, sometimes true, or never true for your household in the last 12 month?”¹⁴

Statistical analyses

Household food insecurity (binary assignment of yes/no) was assigned to each individual. The validity of HFSS-2 was then assessed in the patient sample against the gold standard HFSS-18. Demographic and clinical information was summarized using descriptive statistics. Patient characteristics were compared according to individuals' food insecurity status determined by HFFS-2 and HFSS-18. Means and proportions were compared between food-secure versus food-insecure individuals and were examined using t test, chi-square test, and Fisher exact test analysis, as appropriate.

A 2-by-2 table of food insecurity status was produced based on the results of HFSS-18 and HFSS-2. Sensitivity was determined as the ratio of food-insecure individuals correctly identified by HFSS-2 divided by the number of food-insecure individuals identified using the HFSS-18 tool. Specificity was determined as the ratio of individuals identified correctly by HFSS-2 as food secure divided by the number of food-secure individuals identified by the HFSS-18 tool. Positive and negative predictive values were calculated as well.¹⁵

Convergent validity of HFSS-2 and HFSS-18 was examined to determine whether the 2-item screen was associated with similar patient characteristics to the gold standard. This was tested by using 2 logistic regression models that examined the association between food insecurity as determined by each tool and various clinical conditions. Each model adjusted for insurance status, age, and race. Covariates were selected based on significant bivariate associations in the analysis, with a significance threshold <0.05, and previous literature on this topic.^11,16 Because of small cell counts, Medicaid and Medicare categories were combined and other insurance was dropped, resulting in a binary insurance category (Medicare/Medicaid vs private insurance) for multivariable analysis. All analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC).

Results

In total, 390 people were approached and 295 (75%) completed the survey. Table 1 provides descriptive statistics of the sample, and HFSS-2 and HFSS-18 scores. The average age of study participants was 53 years and more than 58% of the sample was female. Half of the sample identified as Black. The majority of the participants were housing secure. More than half of the sample were not married. Insurance coverage varied in this sample. Overall, a large proportion of study participants had commercial insurance. Comparisons of the prevalence of chronic diseases are shown in Table 1.

Table 1.

Comparison of Characteristics Across Food-Secure and Food-Insecure Items

Characteristics	Total (N = 295)	HFSS-2 food secure (n = 222)	HFSS-2 food insecure (n = 73)	P ^†	HFSS-18 food secure (n = 243)	HFSS-18 food insecure (n = 52)	P ^†
Age in years (Mean, SD, Min, Max)	53 (18.3, 18.0–93.0)	54 (18.4, 18.0–93.0)	47 (16.8, 18.0–80.0)	0.002^*	54 (18.5, 18.0–93.0)	48 (18.5, 19.0–80.0)	0.05^*
Sex, n (%)
Male	122 (41.6)	92 (41.6)	30 (41.7)	0.99	102 (42.2)	20 (39.2)	0.70
Female	171 (58.4)	129 (58.4)	42 (58.3)		140 (57.8)	31 (60.8)
Race, n (%)
White	123 (41.8)	126 (57.0)	23 (31.5)	<0.001^*	131 (54.1)	18 (34.6)	0.02^*
Black	149 (50.7)	82 (37.1)	41 (56.2)		96 (39.7)	27 (51.9)
Other	22 (7.5)	13 (5.9)	9 (12.3)		15 (6.2)	7 (13.5)
Living Situation, n (%)
Steady place to live	273 (93.8)	209 (95.9)	64 (87.6)	0.03^*	230 (96.2)	43 (82.7)	0.002^*
Steady place to live/worried about losing it	10 (3.4)	5 (2.3)	5 (6.9)		5 (2.1)	5 (9.6)
I do not have a steady place to live	8 (2.8)	4 (1.8)	4 (5.5)		4 (1.7)	4 (7.7)
Marital Status, n (%)
Married	106 (35.9)	82 (36.9)	24 (32.9)	0.53	89 (36.6)	17 (32.7)	0.60
Other	189 (64.1)	140 (63.1)	49 (67.1)		154 (63.4)	35 (67.3)
Insurance Coverage, n (%)
Commercial	115 (40.2)	95 (43.8)	20 (28.9)	<0.001^*	102 (42.8)	13 (27.1)	0.008^*
Medicaid	71 (24.8)	39 (18.0)	32 (46.4)		50 (21.0)	21 (43.7)
Medicare	90 (31.5)	76 (35.0)	14 (20.3)		78 (32.8)	12 (25.0)
Other Insurance	10 (3.5)	7 (3.2)	3 (4.4)		8 (3.4)	2 (4.2)


CHD, n (%)
Yes	29 (9.9)	19 (8.6)	10 (14.1)	0.18	21 (8.7)	8 (16.0)	0.12
No	263 (90.1)	202 (91.4)	61 (85.9)		221 (91.3)	42 (84.0)
Diabetes, n (%)
Yes	78 (26.9)	49 (22.4)	29 (40.8)	0.002^*	55 (22.9)	23 (46.0)	<0.001^*
No	212 (73.1)	170 (77.6)	42 (59.2)		185 (77.1)	27 (54.00)
Asthma, n (%)
Yes	65 (22.1)	43 (19.5)	22 (30.1)	0.06	49 (20.2)	16 (30.8)	0.10
No	229 (77.9)	178 (80.5)	51 (69.9)		193 (79.8)	36 (69.2)
Depression, n (%)
Yes	90 (31.0)	62 (28.3)	28 (39.4)	0.08	63 (26.4)	27 (52.9)	<0.001^*
No	200 (69.0)	157 (71.7)	43 (60.6)		176 (73.6)	24 (47.1)

†

t Test, chi-square, and Fishers exact test as appropriate for the variable's distribution.

Statistically significant at P < 0.05.

CHD, coronary heart disease; HFSS, Household Food Security Scale; SD, standard deviation.

Sample characteristics according to HFSS-18 and HFSS-2

According to the gold standard HFSS-18, 17.6% of the sample reported food insecurity. Individuals who reported food insecurity were younger than those who did not report food insecurity on both HFSS-18 and HFSS- 2 (Table 1). Additionally, study participants who were food insecure were more likely to be Black compared to those who were not food secure on both instruments. The percentage of women was similar for those who were food insecure and food secure according to both instruments as well. An analogous pattern to race was seen for individuals who report housing insecurity on both instruments. Those who were food insecure were less likely to have stable housing compared to food secure participants (Table 1).

Insurance status varied by food insecurity status according to both instruments. Those who were food insecure were more likely to have Medicaid coverage compared to those who were food secure (Table 1). Lastly, the prevalence of chronic disease also was higher in those who reported food insecurity. Significant differences in the prevalence of diabetes was seen on both instruments when comparing food-insecure and food-secure participants (P < 0.001 and P = 0.002, respectively). The difference in depression was significant when comparing food-insecure and food-secure participants according to the HFSS-18.

Test diagnostics of HFSS-2

Following the methods of Hager et al, the sensitivity, specificity, positive and negative predictive values, as well as the same combination of the first 2 questions of HFSS-18 were examined to assess the validity of HFFS-2 for the adult general medicine population.¹¹ A confirmatory response to question 1 of HFSS-2 yielded a sensitivity of 92% (95% confidence interval: 85–100) and a specificity of 95% (95% CI: 92–98). This combination also provided a positive predictive value of 80% (95% CI: 70–90) and a negative predictive value of 98% (95% CI: 97–100). A confirmatory response to question 2 of the HFSS-2 yielded a sensitivity and specificity of 88% (95% CI: 80–97) and 95% (95% CI: 92–97), respectively. The positive and negative predictive value for question 2 was 78% (95% CI: 67–89) and 97% (95% CI: 95–99), respectively. A confirmatory response to both questions 1 and 2 provided a sensitivity of 83% (95% CI: 72–93) and a specificity of 98% (95% CI: 97–100). The positive and negative predictive value for the question combination was 93% (95% CI: 86–100) and 96% (95% CI: 94–99), respectively. The sensitivity, specificity, and positive and negative predictive values for the combination of a confirmatory response to question 1 or question 2 are shown in Table 2.

Table 2.

HFSS-2 Test Diagnostics

	HFSS-2
Sensitivity, % (95% CI)	98 (94–100)
Specificity, % (95% CI)	91 (87–94)
Positive Predictive Value, % (95% CI)	70 (59–80)
Negative Predictive Value, % (95% CI)	99 (98–100)

CI, confidence interval; HFSS, Household Food Security Scale.

Convergent validity

Logistic regression models were used to assess the convergent validity of food insecurity status as determined by HFSS-18, HFSS-2, and self-reported chronic conditions previously associated with food insecurity (Table 3). Results show that associations among food insecurity and the chronic conditions identified were similar when using HFSS-2 and HFSS-18, thus showing convergent validity. Individuals who were food insecure, determined by the HFSS-2, had more than 4 times the odds of having coronary heart disease compared to those who were food secure and close to 4 times the odds of reporting diabetes.

Table 3.

Convergent Validity

Outcome	HFSS-2 aOR (95% CI)	HFSS-18 aOR (95% CI)
Coronary Heart Disease	4.19 (1.51–11.59)	4.63 (1.55–13.79)
Diabetes	3.73 (1.83–7.92)	4.19 (1.94–9.08)
Asthma	1.39 (0.71–2.73)	1.47 (0.70–3.10)
Depression	1.54 (0.82–2.92)	3.35 (1.65–6.79)

Adjusted for age, race, and insurance status.

aOR, adjusted odds ratio; CI, confidence interval; HFSS, Household Food Security Scale.

Results from HFSS-18 were similar and demonstrated a strong association between food insecurity, coronary heart disease, and diabetes. Depression was significantly associated with food insecurity; this result was not convergent with the HFSS-2 model (Table 3).

Discussion

In this study, 17.6% of participants surveyed at ChristianaCare primary care clinics reported food insecurity according to the gold standard HFSS-18. This is higher than the Delaware state prevalence, 11.9% as of 2015, but similar to the prevalence among a national sample of older adults.^1,5 These analyses demonstrated high sensitivity, specificity, and positive and negative predictive values for the HFSS-2, as well as convergent validity between HFSS-2 and the gold standard HFSS-18 for 3 of the 4 chronic disease associations. HFSS-2 did not demonstrate convergent validity with HFSS-18 regarding depression in this population, which is divergent from prior studies that have found that those who are food insecure have more than 3 times the odds of having major depression compared to those who were food secure.¹⁷ This analysis demonstrates higher sensitivity and specificity of HFSS-2 than previous publications and further supports the use of HFSS-2 for adults in the general medical population.^11,18

Screening for food insecurity and other social determinants of health, as well as connecting patients to assistance programs, have become a focal point for agencies such as the Centers for Medicare & Medicaid Services. Health care systems are increasingly beginning to be incentivized to screen for social determinants of health information in order to connect patients with the right resource.^19,20 This increase in population health focus necessitates the application of services outside the walls of the health care setting. Furthermore, increased food insecurity because of COVID-19 is a growing issue given the financial crisis brought about by the pandemic.²¹ This has brought an urgency to screen for social needs, such as food insecurity, in clinical settings.

Connecting patients to community-based social resources such as nutritional assistance has been shown to be effective in modifying poor health outcomes in adults.^2,22 For example, Seligman et al demonstrated that an intervention for food-insecure patients with diabetes that used a local food bank was associated with a reduction in hemoglobin A1c and an increase in fruit and vegetable intake, self-efficacy, and medication adherence.²² Moreover, another study found that referring patients with food insecurity to local social support resources resulted in reductions in blood pressure and low-density lipoprotein cholesterol.²³ Screening to ascertain the prevalence of food insecurity among patients will aide in highlighting strategies to modify poor health outcomes. The present study has demonstrated that HFSS-2 is a valid and acceptable instrument for use in adult general medicine outpatients during clinical care, many of whom may be at highest risk during the COVID-19 crisis.

Limitations

There were several limitations to this work. This study was cross-sectional and descriptive in nature. Exposure and outcome were determined at the same time; therefore, temporal associations could not be described.²⁴Additionally, the small sample size limited the ability to perform subgroup analyses with more granular categories of severity of food insecurity.¹⁴ These data also reflect one sample of adult general medicine outpatients in the state of Delaware. Although the primary care offices represented both urban and suburban communities, the validity of this screen should be tested in general medicine outpatient samples from multiple geographic areas. As with other surveys, questions are asked under the assumption of confidentiality. The research team does not know if participant responses were changed intentionally because of concerns of privacy or potential implications to medical care.

Conclusion

The widely used HFSS-2 is highly sensitive and specific in adult general medicine outpatients – a sicker population than the general adult population. This is the first study that the research team is aware of that validated this instrument in this important patient population. Although this investigation was conducted prior to the COVID-19 pandemic, it underscores that understanding the prevalence of food insecurity in the patient population allows for the development and bolstering of strategies to improve their health beyond the walls of health care institutions, especially during times such as the COVID-19 pandemic. Screening for this social factor adds to the growing scope of work addressing the social determinants of health.

Footnotes

Authors' Contributions

Cecelia Harrison, MPH: Project Administration, Resources, Data Curation, Investigation, Formal Analysis, and Writing – Original Draft, Visualization; Jennifer N. Goldstein, MD, MSc: Conceptualization, Methodology, Validation, Writing – Review & Editing; Adebayo Gbadebo, MBA: Data Curation, Software, Resources; Mia Papas, PhD, MS: Conceptualization, Methodology, Validation, Writing– Review & Editing, Supervision.

Acknowledgments

The authors express their gratitude to the research assistants on this project for their many hours screening across New Castle County, DE and the clinical staff whose patients were screened. We also are grateful to Lee Pachter, DO for reviewing this manuscript. This work also was supported by ChristianaCare's Office of Health Equity and Cultural Competence, led by Jacqueline Ortiz and Rose Kakoza, MD, MPH, Director of Community Health and Equity for Primary Care and Community Medicine.

Author Disclosure Statement

The authors declare that there are no conflicts of interest.

Funding Information

No funding was received for this article.

References

Coleman-Jensen

, Rabbitt

, Gregory

. Household food security in the United States in 2015. US Dep Agric Econ Res Serv, 2016; ERR-215:1–44.

Berkowitz

, Seligman

, Rigdon

, Meigs

, Basu

. Supplemental Nutrition Assistance Program (SNAP) participation and health care expenditures among low-income adults. JAMA Intern Med, 2017; 177:1642–1649.

Seligman

, Laraia

, Kushel

. Food insecurity is associated with chronic disease among low-income. J Nutr, 2009; 140:304–310.

Gundersen

, Ziliak

. Food insecurity and health outcomes. Health Aff, 2015; 34:1830–1839.

Bishop

, Wang

. Food insecurity, comorbidity, and mobility limitations among older U. S. adults: findings from the Health and Retirement Study and Health Care and Nutrition Study. Prev Med, 2018; 114:180–187.

Berkowitz

, Seligman

, Choudhry

. Treat or eat: food insecurity, cost-related medication underuse, and unmet needs. Am J Med, 2014; 127:303.e3–310.e3.

Patel

, Piette

, Resnicow

, Kowalski-Dobson

, Heisler

. Social determinants of health, cost-related nonadherence, and cost-reducing behaviors among adults with diabetes: findings from the National Health Interview Survey. Med Care, 2016; 54:796–803.

Moreno

, Morales

, Isiordia

, et al. Latinos with diabetes and food insecurity in an agricural community. Med Care, 2015; 53:423–429.

Knight

, Probst

, Liese

, Sercy

, Jones

. Household food insecurity and medication “scrimping” among US adults with diabetes. Prev Med, 2016; 83:41–45.

10.

Mayer

, McDonough

, Seligman

, Mitra

, Long

. Food insecurity, coping strategies and glucose control in low-income patients with diabetes. Public Health Nutr, 2016; 19:1103–1111.

11.

Hager

, Quigg

, Black

, et al. Development and validity of a 2-item screen to identify families at risk for food insecurity. Pediatrics, 2010; 126:e26–e32.

12.

NACHC (National Association of Community Health Centers). PRAPARE : Protocol for Responding to and Assessing Patient Assets, Risks, andExperiences. 2016.

13.

CDC. 2018 BRFSS Questionnaire. Published online 2018:1-109. https://www.cdc.gov/brfss/questionnaires/pdf-ques/2018_BRFSS_English_Questionnaire.pdf

14.

USDA. U.S. household food security survey module: three-stage design, with screeners. https://www.ers.usda.gov/media/8271/hh2012.pdf Accessed August 13, 2018 .

15.

Gordis

. Epidemiology. 5th ed. Elsevier, 2013. https://www-elsevier-com-s.web.bisu.edu.cn/books/epidemiology/gordis/978-1-4557-3733-8 Accessed September 5, 2018 .

16.

Berkowitz

, Basu

, Meigs

, Seligman

. Food insecurity and health care expenditures in the United States, 2011–2013. Health Serv Res, 2018; 53:1600–1620.

17.

Vozoris

, Tarasuk

. Household food insufficiency is associated with poorer health. J Nutr, 2003; 133:120–126.

18.

Gundersen

, Engelhard

, Crumbaugh

, Seligman

. Brief assessment of food insecurity accurately identifies high-risk US adults. Public Health Nutr, 2017; 20:1367–1371.

19.

Adler

, Glymour

, Fielding

. Addressing social determinants of health and health inequalities. JAMA, 2016; 316:1641–1642.

20.

Matulis

, Lloyd

The history, evolution, and future of Medicaid accountable care organizations. CHCS Brief, 2018:22.

21.

Baker

, Farrokhnia

, Meyer

, Pagel

, Yannelis

How does household spending respond to an epidemic? Consumption during the 2020 COVID-19 pandemic. SSRN Electron J 2020. DOI:10.2139/ssrn.3565521.

22.

Seligman

, Lyles

, Marshall

, et al. A pilot food bank intervention featuring diabetes-appropriate food improved glycemic control among clients in three states. Health Aff, 2015; 34:1956–1963.

23.

Berkowitz

, Hulberg

, Standish

, Reznor

, Atlas

. Improving chronic cardiometabolic disease management by addresing basic resource needs: a pragmatic evaulation. JAMA Intern Med, 2017; 177:139–148.

24.

Grimes

, Schulz

. Descriptive studies: what they can and cannot do. Lancet, 2002; 359:145–149.