Abstract
Characterizing food pantry (FP) clients’ FP usage patterns may provide opportunities to tailor health-related interventions. Respondents (n=245) at seven FPs reported their frequency and reliance on FPs and their sociodemographics, health status, and health-related trade-offs. Clients were categorized via latent class analysis. Higher FP usage was associated with being older, having a household member with heart disease, and putting off buying medicine to buy food. Lower FP usage was associated with higher levels of education and having a household member with cancer. Findings highlight the potential importance of measuring FP clients’ degree of FP use.
Keywords: Food Insecurity, Food Pantry, Heart Disease, Chronic Disease, Latent Class Analysis
INTRODUCTION
Approximately 13.7 million US households (10.5%) were food insecure in 2019.1 Food insecurity is defined as having difficulty providing sufficient food for all household members due to lack of resources.1 Food insecurity has been linked to exacerbation and prevalence of several chronic diseases, including hypertension, diabetes, cancer, depression, anxiety, asthma, chronic obstructive pulmonary disease, hyperlipidemia, and coronary heart disease.1–3 Food insecure households face many risk factors for disease, including poor diet quality.4,5 Understanding the strategies through which food insecure households are accessing food is essential to address their health needs.
One strategy used by food insecure households to acquire food is accessing food pantries (FPs).6,7 FPs are local hunger relief organizations that assist food insecure households by distributing unprepared food for offsite consumption.7 In 2019, 27.7% of food insecure US households accessed FPs.7 Most FP clients are food insecure, and they experience a number of socioeconomic needs (e.g. difficulties with housing, utilities, or medical bills).6,8 FP use has also been associated with a range of health risks, including obesity9 and poor diet quality.10 As a result, many FP clients engage in health-related trade-offs, foregoing healthcare in order to conserve money to buy food.6,11 For example, FP use has been associated with medication underuse or foregoing needed medication.6,12,13
In research, FP use is often categorized according to a simple yes/no binary variable where a household is a FP user or not. Specific FP use patterns have not been well characterized. Many studies do not summarize client characteristics according to duration, frequency, or other indicator of intensity of FP use. The few studies that present those data have focused primarily on duration, characterizing long-term (e.g. ≥24 months) vs. short-term users14–16 or distinguishing between long-term higher frequency and long-term lower frequency users.17 Long-term users tend to be older than short-term users;17 however, apart from mobility challenges,16 neither the health status nor health-related trade-offs of long-term vs. short-term users have been documented.
Duration of use is insufficient to characterize the extent of clients’ current reliance on FPs. The degree of households’ food insecurity fluctuates with changes in income and employment status.18 These fluctuations affect the strategies households use to acquire food. There is a knowledge gap with respect to documenting the differences between FP clients who continuously rely on FPs to provide much of their households’ food most months vs. FP clients who occasionally rely on pantries for smaller proportions of their households’ food over time. The present study seeks to document how households characterized by high FP use differ from those with lower use patterns.
Research Objectives
The objective of this study was to characterize FP use patterns of clients from a sample of FPs in Arkansas. First, we identified categories of FP use based on: (a) self-reported number of months that respondent households obtained food from FPs in the past 12 months and (b) self-reported proportion of respondent households’ food obtained from FPs in the past month. Second, we examined associations between the FP use categories and selected predictors (sociodemographic characteristics, health status, and health-related trade-offs).
METHODS
Respondents and Recruitment
The research team partnered with seven FPs to recruit their clients as participants. FPs were located in Washington or Benton counties in Northwest Arkansas, where there are approximately 119 food bank affiliated FPs. The seven FPs were chosen by the research team to provide representation of diverse types of FPs in the region: these FPs varied according to the number of clients served per month, their clients’ racial/ethnic demographics, their hours and days of operation, location across the major cities in the region, and the extent to which clients received food pre-selected by the FPs or had choice over food received.
Respondents were recruited in the seven food pantries by bilingual (English/Spanish or English/Marshallese) or English-speaking research staff with extensive research training. Research staff approached FP clients who were in line for food pick up and asked them to participate.
Respondents were 18 years or older and spoke English, Spanish, or Marshallese. To avoid duplicate responses, clients were asked if they or anyone in their household had participated in the survey previously and, if so, were not surveyed again.
Survey data collection was completed in July and August 2018. Surveys were administered in English, Spanish, or Marshallese, according to each respondent’s preference. Surveys were administered face-to-face with tablet computers using the Research Electronic Data Capture (REDCap)19,20 mobile application. Respondents were compensated with a $10 gift card for completing the survey.
Survey
A 27-item survey was used to assess food pantry clients’ food insecurity, FP usage and duration of use, household food received from FPs, Supplemental Nutrition Assistance Program (SNAP) benefits, and individual and household health characteristics, medical expenses and trade-offs, and demographics. The survey included yes/no items, ordinal response scales, multiple choice items, and open-ended items in order to capture clients’ responses.
Survey items were selected based on the research team’s previous work in FPs in Arkansas,21–24 Maryland,25 and Minnesota.26,27 Items were selected (and in some cases adapted) from existing survey instruments (see descriptions of items below), including the Hunger Vital Sign,28,29 Feeding America’s Hunger in America Client Survey 2014,6 B’More Healthy: Communities for Kids Adult Impact Questionnaire,30 Food Independence, Security, and Health client survey,31 and the 2018 Behavioral Risk Factor Surveillance Survey.32 To select and adapt items, the research team selected an initial list of items to adapt and then completed three rounds of refinement and adaptation before a final survey draft was approved via consensus. Item adaptation focused on providing face-valid categorical and ordinal response options with which FP clients would be familiar and could quickly complete. The survey instrument is included as a supplementary file. Items central to analysis are described below.
FP Usage
Respondents were asked “during how many of the past 12 months (including this one) did your household get food from any food pantry?,”6 and their responses were recorded numerically from 0–12. For this question, respondents were asked to provide their best estimate if they were uncertain of the exact number.
Household Food Received from FPs
Respondents were asked how much of their household’s food came from FPs in the past 30 days.6 Responses were on an ordinal scale where 1=only a few days’ worth of food in a month, 2=1–2 weeks’ worth of food in a month, 3=more than half of our food in a month, or 4=almost all of our food in a month.
Food Insecurity
Respondents completed the two-item Hunger Vital Sign food security screener,28,29 a valid, sensitive, and specific instrument adapted from longer United States Department of Agriculture (USDA) household food security measures. The two items are: “within the past 12 months, we worried whether our food would run out before we got the money to buy more” and “within the past 12 months, the food we bought just didn’t last and we didn’t have money to get more.” Response options were “often true,” “sometimes true,” or “never true.” A household is considered food insecure when a respondent responds affirmatively to question 1 and/or question 2.
SNAP Benefits
Respondents were asked a yes/no question: “In the past 12 months…did you or anyone in your household receive SNAP benefits?”
Health Conditions
Respondents were asked if a doctor or nurse ever told them that they have: high blood pressure, overweight or obesity, type 2 diabetes, heart disease, or cancer.30 For each health condition, response categories were no, yes, and I don’t know. Clients were then asked about the same conditions but were told to consider if anyone in their household (not including themselves) had ever been told that they had any of the previous conditions.
Medical Expenses and Trade-offs
Respondents were asked a yes/no question: “In the past 12 months, did you or anyone in your household have problems paying or were unable to pay any medical bills? Include bills for doctors, dentists, hospitals, therapists, medication, equipment, nursing home, or home care.”31 Respondents were asked two questions about trade-offs related to food and medicine: “In the past 12 months, how often did you or anyone in your household put off buying medicines or take less medicine than you were supposed to so that you would have money to buy food?” and “In the past 12 months, how often did you or anyone in your household put off buying food so that you would have money to buy your medicines?”31 Responses options for these two questions were 1=never, 2=only 1 or 2 months, 3=some months but not every month, 4=almost every month, and 5=every month. In analyses, the response options 4=almost every month and 5=every month were grouped into a single category.
Health Care Coverage and Demographic Variables
Respondents were asked to indicate whether or not they had any kind of health care coverage and to provide their age, race/ethnicity, sex, educational attainment, household composition, and employment status. Those items were adapted from the 2018 Behavioral Risk Factor Surveillance Survey.32 Respondents were also asked to describe their current housing status (stable housing vs. no/temporary housing).6
Analytic Strategy
Descriptive Statistics
Frequency distributions and descriptive statistics (e.g. medians and interquartile ranges) were computed.
Latent Class Analysis
Latent class analysis (LCA) was used to identify behavior patterns of FP clients. LCA is a modeling technique that identifies subgroups—or latent classes—within a population. The construct is latent in the sense that it is unobservable but inferred from individuals’ membership to a class measured with indicators. Latent classes are mutually exclusive and exhaustive.33
Two indicators were used to create the classes: (1) Proportion of food from FPs measured on an ordinal scale: “Thinking of all the food pantries you visited in the past 30 days (including this one), how much of your household’s food would you say was from food pantries?” (range: 1–4) and (2) Number of recent months respondents obtained food from FPs measured on a continuous scale: “During how many in the past 12 months (including this one) did your household get food from a food pantry?” (range: 1–12).
A specified latent class model was fit to the data. Existence of distinct homogenous categories of FP clients and ordinality of the classes was assumed (e.g. that we would find classes representing more reliance on FPs vs. less reliance on FPs). Information was generated to select the number of classes for the model by running a 1-class model followed by 2-class, 3-class, and 4-class models. Class interpretability and goodness-of-fit (Akaike’s information criterion [AIC], Schwarz’s Bayesian information criterion [BIC]) were used to determine the optimal model due to the lack of a strong theory to make a solid assumption about the number of classes to extract. BIC performed better than AIC to differentiate the models and is more widely accepted with LCA.34
Based on model goodness-of-fit, a 3-class model was selected (i.e., three latent subgroups in the population of adults visiting FP in northwest Arkansas). These categories exhibited a natural ordering on two dimensions: the number of recent months visiting FPs and the proportion of food from FPs. This new ordinal variable was labeled as FP Use, with Low, Medium, and High categories. We obtained the probability that respondents of each class had in answering questions about the proportion of food from FP and average number of recent months they obtained food from FP. Standard errors for the probabilities were derived from the delta method.
Ordinal Logistic Regression
We used ordinal logistic regression (OLR) to examine associations between independent variables (sociodemographic characteristics of consumers, health status, and health-related trade-offs) and the dependent variable the resultant latent class FP Use variable. The models did not incorporate the pantry site from which the respondents were recruited. We were unwilling to assign each respondent to a single site because we encountered some of the respondents at multiple sites (although each respondent only participated once).
OLR is based on the proportional odds (PO) model, which assumes that each predictor has the same effects across categories of the ordinal dependent variable. The non-significant test Brant test for the OLR indicated that the assumption of PO was not violated, χ2(24)=26.69;p<0.319.35 We thus analyzed the model as ordinal. For each polytomous variable, we obtained an overall p-value for the variable as a predictor. For any variables with an overall p-value <0.05, we examined each category of the variable as a predictor compared to a reference category. To assess overall adequacy of the ordinal model, we calculated two goodness-of-fit statistics (ordinal Hosmer-Lemeshow test and the Lipsitz Likelihood-ratio-tests).36,37
All analyses were conducted using STATA 16/SE.38 Statistical significance was determined at alpha α=0.05.
RESULTS
Descriptive Analysis
The survey recruitment rate was 83.6%; 287 clients were approached, 247 clients agreed to participate, and 40 refused. Two respondents discontinued participation after only a few questions because they were called to pick up their food, so data from 245 respondents are included in all analyses. Table 1 presents the self-reported sociodemographic characteristics, health status, and health-related trade-offs of respondents. Respondents’ mean age was 46.5 years, and 71.0% of respondents were female. The majority (70.2%) of respondents had completed high school/GED or had at least some college education. The majority of respondents were not working, either because they were not employed (39.6%) or were unable to work (25.3%). Most respondents (94.7%) were classified as food insecure by responding affirmatively to at least one of the two Hunger Vital Sign items. A minority of respondents (39.1%) reported receiving SNAP benefits. Many respondents reported that they or someone in their household has high blood pressure (57.1%), overweight/obesity (41.2%), type 2 diabetes (29.9%), or heart disease (23.6%). The majority of respondents (57.8%) indicated they have been receiving food from FPs for at least two years. Several respondents indicated that they put off buying medicines to save money for food (36.8%) or put off buying food to save money for medicines (24.0%) more than two months in the past year.
Table 1.
Respondents’ Demographic Characteristics, Health Status, and Trade-Offs Presented by Food Pantry Use Categories
| SOCIODEMOGRAPHIC AND HEALTH CHARACTERISTICS | Food Pantry Use Classes | |||
|---|---|---|---|---|
|
| ||||
| Low | Medium | High | Total | |
| N=86 | N=69 | N=90 | N=245 | |
| Age (Mean ± SD) | 41.1 ± 13.9 | 46.9 ± 13.3 | 51.5 ± 14.7 | 46.5 ± 14.7 |
| Race/Ethnicity | ||||
| Non-White | 41 (47.7%) | 27 (39.1%) | 36 (40.0%) | 104 (42.4%) |
| White | 45 (52.3%) | 42 (60.9%) | 54 (60.0%) | 141 (57.6%) |
| Sex | ||||
| Male | 27 (31.4%) | 18 (26.1%) | 26 (28.9%) | 71 (29.0%) |
| Female | 59 (68.6%) | 51 (73.9%) | 64 (71.1%) | 174 (71.0%) |
| Health Care Coverage * | ||||
| No | 23 (27.1%) | 19 (27.9%) | 28 (31.1%) | 70 (28.8%) |
| Yes | 62 (72.9%) | 49 (72.1%) | 62 (68.9%) | 173 (71.2%) |
| Education | ||||
| Never/Completed grades 1–8 | 6 (7.0%) | 6 (8.7%) | 19 (21.1%) | 31 (12.7%) |
| Completed grades 9–11 | 8 (9.3%) | 21 (30.4%) | 13 (14.4%) | 42 (17.1%) |
| Completed grade 12/GED | 43 (50.0%) | 27 (39.1%) | 36 (40.0%) | 106 (43.3%) |
| Completed one or more years of college | 29 (33.7%) | 15 (21.7%) | 22 (24.4%) | 66 (26.9%) |
| Employment Status | ||||
| Employed | 30 (34.9%) | 15 (21.7%) | 21 (23.3%) | 66 (26.9%) |
| Not employed | 34 (39.5%) | 31 (44.9%) | 32 (35.6%) | 97 (39.6%) |
| Retired | 3 (3.5%) | 7 (10.1%) | 10 (11.1%) | 20 (8.2%) |
| Unable to work | 19 (22.1%) | 16 (23.2%) | 27 (30.0%) | 62 (25.3%) |
| Housing Status | ||||
| Stable housing | 78 (90.7%) | 63 (91.3%) | 79 (88.0%) | 220 (89.9%) |
| Temporary or no housing | 8 (9.3%) | 6 (8.7%) | 11 (12.0%) | 25 (10.1%) |
| Food Security Status | ||||
| Food secure | 4 (4.7%) | 3 (4.3%) | 6 (6.5%) | 13 (5.3%) |
| Food insecure | 82 (95.3) | 66 (95.7%) | 84 (93.5%) | 232 (94.7%) |
| SNAP Benefits * | ||||
| Did not receive SNAP benefits in past year | 52 (61.9%) | 35 (50.7%) | 61 (67.8%) | 148 (60.9%) |
| Received SNAP benefits in past year | 32 (38.1%) | 34 (49.3%) | 29 (32.2%) | 95 (39.1%) |
| Number of Children in Household (Mean ± SD) | 1.7 ± 2.1 | 1.6 ± 1.8 | 1.5 ± 1.8 | 1.6 ± 1.9 |
|
HEALTH CONDITIONS Anyone in household (including respondent) has: |
||||
| High blood pressure | ||||
| No | 37 (43.0%) | 29 (42.0%) | 39 (43.3%) | 105 (42.9%) |
| Yes | 49 (57.0%) | 40 (58.0%) | 51 (56.7%) | 140 (57.1%) |
| Overweight/obesity | ||||
| No | 51 (59.3%) | 41 (59.4%) | 52 (57.8%) | 144 (58.8%) |
| Yes | 35 (40.7%) | 28 (40.6%) | 38 (42.2%) | 101 (41.2%) |
| Type 2 diabetes * | ||||
| No | 61 (70.9%) | 49 (71.0%) | 61 (68.5%) | 171 (70.1%) |
| Yes | 25 (29.1%) | 20 (29.0%) | 28 (31.5%) | 73 (29.9%) |
| Heart disease * | ||||
| No | 69 (81.2%) | 55 (79.7%) | 61 (69.3%) | 185 (76.4%) |
| Yes | 16 (18.8%) | 14 (20.3%) | 27 (30.7%) | 57 (23.6%) |
| Cancer * | ||||
| No | 70 (81.4%) | 58 (84.1%) | 76 (86.4%) | 204 (84.0%) |
| Yes | 16 (18.6%) | 11 (15.9%) | 12 (13.6%) | 39 (16.0%) |
| MEDICAL BILLS | ||||
| In past 12 months, did household have problems paying or were unable to pay medical bills? | ||||
| No | 47 (54.7%) | 37 (53.6%) | 34 (38.2%) | 118 (48.4%) |
| Yes | 39 (45.3%) | 32 (46.4%) | 55 (61.8%) | 126 (51.6%) |
|
TRADE-OFFS In the past 12 months, how often did respondents put off… |
||||
| Buying/taking medications to afford food? | ||||
| Never | 50 (58.1%) | 34 (49.3%) | 38 (42.2%) | 122 (49.8%) |
| Only 1 or 2 months | 16 (18.6%) | 7 (10.1%) | 10 (11.1%) | 33 (13.5%) |
| Some months but not every month | 10 (11.6%) | 14 (20.3%) | 23 (25.6%) | 47 (19.2%) |
| Almost every month/every month | 10 (11.6%) | 14 (20.3%) | 19 (21.1%) | 43 (17.6%) |
| Buying food to afford medications? | ||||
| Never | 62 (72.1%) | 45 (65.2%) | 49 (54.4%) | 156 (63.7%) |
| Only 1 or 2 months | 7 (8.1%) | 9 (13.0%) | 14 (15.6%) | 30 (12.2%) |
| Some months but not every month | 14 (16.3%) | 11 (15.9%) | 16 (17.8%) | 41 (16.7%) |
| Almost every month/every month | 3 (3.5%) | 4 (5.8%) | 11 (12.2%) | 18 (7.3%) |
SNAP, Supplemental Nutrition Assistance Program.
Respondents answering “I don’t know” were omitted from percentage calculations.
This contingency table shows cell frequencies and percentages for categorical variables. Missing data were not taken into account in percentage computation. Means and standard deviations (SD) are presented for continuous variables (age and number of children in household).
Latent Class Analysis to Identify Patterns of FP Use
Information Criterion statistics (AIC, BIC) were obtained after fitting each LCA model to gauge model fit for the 1-class (AIC = 1970.12; BIC = 1987.67), 2-class (AIC = 1828.68; BIC = 1863.77), 3-class (AIC = 1724.83; BIC = 1777.48), and 4-class (AIC = 1708.74; BIC = 1778.93) models. As described in the Analytic Strategy, the 3-class model was selected (i.e., FP Use, with ordinal Low, Medium, and High categories) based on the lowest BIC and class interpretability.
Table 2 presents the probabilities for FP clients to belong to a specific class as well as the class-specific response probabilities and means of each item.
Table 2.
Probabilities for Food Pantry Clients to Belong to a Specific Pantry Use Category and Item-Level Category-Specific Response Probabilities and Means: Results from Latent Class Analysis
| Probability | Mean | SE | p-value | 95% CI | ||
|---|---|---|---|---|---|---|
| Low Food Pantry Users | 0.35 † | 0.25 | 0.42 | |||
| Proportion of food from FP | ||||||
| Only a few days’ worth of food in a month | 0.54 | 0.06 | 0.43 | 0.65 | ||
| 1–2 weeks’ worth of food in a month | 0.32 | 0.05 | 0.23 | 0.43 | ||
| More than half of our food in a month | 0.09 | 0.03 | 0.04 | 0.17 | ||
| Almost all of our food in a month | 0.05 | 0.03 | 0.02 | 0.13 | ||
| Number of months respondents obtained food from FP (past 12 months) | 1.98 | 0.12 | <0.001 | 1.75 | 2.22 | |
|
| ||||||
| Medium Food Pantry Users | 0.37 † | 0.31 | 0.41 | |||
| Proportion of food from FP | ||||||
| Only a few days’ worth of food in a month | 0.43 | 0.06 | 0.31 | 0.56 | ||
| 1–2 weeks’ worth of food in a month | 0.39 | 0.06 | 0.27 | 0.51 | ||
| More than half of our food in a month | 0.15 | 0.04 | 0.08 | 0.26 | ||
| Almost all of our food in a month | 0.03 | 0.02 | 0.01 | 0.13 | ||
| Number of months respondents obtained food from FP (past 12 months) | 5.79 | 0.15 | <0.001 | 5.50 | 6.09 | |
|
| ||||||
| High Food Pantry Users | 0.28 † | 0.22 | 0.34 | |||
| Proportion of food from FP | ||||||
| Only a few days’ worth of food in a month | 0.25 | 0.05 | 0.17 | 0.35 | ||
| 1–2 weeks’ worth of food in a month | 0.34 | 0.05 | 0.25 | 0.45 | ||
| More than half of our food in a month | 0.23 | 0.04 | 0.15 | 0.33 | ||
| Almost all of our food in a month | 0.18 | 0.04 | 0.11 | 0.27 | ||
| Number of months respondents obtained food from FP (past 12 months) | 11.85 | 0.1 | <0.001 | 11.65 | 12.05 | |
FP, food pantry; SE, standard error; CI, confidence intervals.
For the continuous variable number of months respondents obtained food from FP, the latent class analysis is based on linear regression. For the ordinal variable proportion of food from FP, the latent class analysis is based on ordinal logistic regression. The probability statistics for Low, Medium, and High Food Pantry Users indicates the probability of a respondent from the total sample to be classified in that grouping. The probability statistics for proportion of food from FP indicates the relative probability of that classification for respondents in that pantry usage grouping.
Designates the proportion of respondents assigned to a classification. This assignment was based on the respondent’s most likely latent class membership.
LCA revealed a 35% chance that FP clients fall into the Low FP users category, a 37% chance that they fall into the Medium FP users category, and a 28% chance that they fall into the High FP users category.
Low FP users obtained food from a FP in a mean of 1.98 months (95% confidence intervals [CI]: 1.75–2.22) during the past 12 months. Medium FP users obtained food from a FP in 5.79 months (CI: 5.50–6.09) during the past 12 months. High FP users obtained food from FP in 11.85 months (CI: 11.65–12.05) during the past 12 months.
Low FP users had a 14% chance of obtaining between half and almost all of their households’ food from a FP in the past 30 days. Medium FP users had an 18% chance of obtaining between half and almost all of their households’ food from a FP in the past 30 days. High FP users had a 41% chance of obtaining between half and almost all of their households’ food from a FP in the past 30 days. High FP users also had an 18% chance (CI: 0.11–0.27) of almost all of their food coming from a FP in the past 30 days.
Variables Associated with FP Use
Table 3 presents ordinal logistic regression analysis from a set of predictor variables to estimate the ordinal dependent variable class of FP clients. We interpreted estimates of the fitted model in terms of proportional odds ratios (OR) with CIs. Neither the ordinal Hosmer-Lemeshow (statistic=17.156, df(17); p=0.4438) nor the Lipsitz Likelihood-ratio-test (statistic=9.613, df(9); p=0.3828) indicated lack of fit.
Table 3.
Determinants of Food Pantry Use Patterns of Clients in Arkansas, 2018: Results from Ordinal Logistic Regression
| Measures | OR | 95% CI | p-value |
|---|---|---|---|
| Age | 1.05 | 1.03, 1.08 | <0.001* |
| Race/Ethnicity | 1.68 | 0.90, 3.15 | 0.105 |
| Sex | 1.18 | 0.63, 2.20 | 0.592 |
| Health Care Coverage | 0.81 | 0.43, 1.54 | 0.524 |
| Education | |||
| Never/Completed grades 1–8† | — | — | — |
| Completed grades 9–11 | 0.86 | 0.31, 2.39 | 0.770 |
| Completed grade 12/GED | 0.37 | 0.15, 0.93 | 0.036* |
| Completed one or more years of college | 0.30 | 0.11, 0.81 | 0.017* |
| Number of Children | 1.06 | 0.90, 1.25 | 0.506 |
| Employment Status | 1.01 | 0.77, 1.31 | 0.960 |
| Housing Status | 2.01 | 0.76, 5.31 | 0.158 |
| SNAP Benefits | 0.75 | 0.43, 1.31 | 0.312 |
|
HEALTH CONDITIONS Anyone in household (including respondent) has: |
|||
| Overweight/obesity | 1.11 | 0.61, 2.02 | 0.714 |
| Heart disease | 2.07 | 1.01, 4.62 | 0.048* |
| High blood pressure | 0.70 | 0.37, 1.27 | 0.235 |
| Type 2 diabetes | 0.70 | 0.40, 1.32 | 0.266 |
| Cancer | 0.35 | 0.15, 0.78 | 0.010* |
|
TRADE-OFFS
In the past 12 months, how often did respondents put off… |
|||
| Buying/taking medications to afford food? | |||
| Never† | — | — | — |
| Only 1 or 2 months | 1.23 | 0.52, 2.90 | 0.632 |
| Some months but not every month | 2.89 | 1.27, 6.56 | 0.011* |
| Almost every month/every month | 2.44 | 1.05, 5.64 | 0.038* |
| Buying food to afford medications? | 1.17 | 0.84, 1.61 | 0.354 |
| Problems Paying Medical Bills, Past 12 Months | 1.71 | 0.95, 3.09 | 0.074 |
OR, Odds Ratio; CI, Confidence Intervals
Statistical significance at p<0.05
Reference category
Food insecurity was not entered in the regression model because of the very high proportion of food insecure clients (i.e., there was not enough variability in this variable). Regression was based on complete cases analysis.
Older respondents used FPs more. The odds of shifting into higher use categories (Low to Medium or Medium to High) increased by 5% (proportional OR=1.05; CI: 1.03–1.08; p<0.001) for each year increase in age, when all the other predictors remained constant.
More educated respondents used FPs less. The odds of shifting into a higher FP use category (Low to Medium or Medium to High) were 63% lower among respondents with a high school diploma/GED compared to those who had completed no higher than eighth grade, when all the other predictors remained constant (proportional OR=0.37; CI: 0.15–0.93; p=0.036). The odds of shifting into a higher FP use category (Low to Medium or Medium to High) were 70% smaller among respondents with education beyond high school compared to those who had completed no higher than eighth grade, when all the other predictors remained constant (proportional OR=0.30; CI: 0.11–0.81; p=0.017).
Respondents in households in which someone (either the respondent or other household member) has heart disease used FPs more. The odds of shifting into a higher FP use category (Low to Medium or Medium to High) for households with heart disease were 2.07 times as high as for households without heart disease, when holding all the other predictors constant (CI: 1.01–4.62; p=0.048).
Respondents in households in which someone (either the respondent or other household member) has cancer used FPs less. The odds of shifting into a higher FP use category (Low to Medium or Medium to High) were 65% lower among households who reported cancer compared to those who did not, when holding all the other predictors constant (proportional OR=0.35; CI: 0.15–0.78; p=0.010).
Respondents who reported more frequently putting off buying medicine to buy food used FPs more. The odds of shifting into a higher use category (Low to Medium or Medium to High ) for respondents who put off buying medicine to buy food some months but not every month were 2.89 times as high as for those who never engaged in this trade-off, when holding all the other predictors constant (CI: 1.27–6.56; p=0.011). The odds of shifting into a higher use category (Low to Medium or Medium to High) for respondents who put off buying medicine to buy food almost every month or every month were 2.44 times as high as for those who never engaged in this trade-off, when holding all the other predictors constant (CI: 1.05–5.64; p=0.038).
There were no differences among the class categories with regard to race/ethnicity, sex, health care coverage, employment status, housing status, number of children in household, being a SNAP participant, having difficulty paying medical bills, putting off buying food to buy medicine, or having or living with someone who has overweight/obesity, high blood pressure, or type 2 diabetes (all p-values >0.05).
DISCUSSION
The high, medium, and low FP user categories identified via LCA differed meaningfully by the number of months in the past year respondents obtained food from FPs and the proportion of their households’ food they reported receiving from FPs. High users visited FPs more regularly and reported receiving a greater proportion of their food from FPs than did medium and low users. Similar to previous findings about frequency of FP use,16,17 older respondents had increasingly greater odds of engaging in higher use of FPs.
Similar to previous studies,6,11 Arkansas FP clients reported regularly putting off buying medicines in order to save money for food. This study adds important new information as the first study to document that putting off buying medicine to buy food was associated with the extent to which people rely on FPs. Putting off buying medicine to buy food was associated with higher use of FPs. Like previous studies,6,8 the present study found that a majority of food pantry users report difficulty paying medical bills; however, difficulty paying medical bills was not related to increased usage of FPs. Together these findings suggest that the categories of FP usage identified in the present study may predict a broad set of characteristics (e.g. age, education, putting off buying medicine to buy food) related to food insecurity.
Similar to previous studies of FP clients, the sample in the present study was 71.0% female, had a mean age of 46.5 years,6 and many (29.9%) had a household member with type 2 diabetes.6 However, race/ethnicity, sex, and type 2 diabetes status did not predict households’ degree of recent FP use.
The positive association between the degree of recent FP use and reporting heart disease in the household aligns with previous studies showing associations between food insecurity and heart disease.2 The negative association between degree of recent FP use and having a household member with cancer was unexpected. Previous studies have shown associations between food insecurity and cancer.2 Cancer had not previously been explicitly connected to FP use patterns. Recent reviews24,39 of disease prevention and management interventions in FPs identified only one cancer-focused intervention40 and one heart disease-focused intervention.41 Further research is needed to replicate the associations among heart disease, cancer, and FP use and, if they replicate, to understand what factors affect FP use among households in which a member has heart disease or cancer.
Limitations
Limitations of the present study include those typical of survey studies that rely upon a convenience sample of respondents. Despite a high response rate, respondents may not have been representative of the entire population of FP clients in this region. For example, the high percentage of female respondents in the sample may have affected findings related to sex. In addition, in the present study, 94.7% of FP clients were food insecure, in contrast to ~65% of FP clients in USDA’s Household Food Security in the United States in 2019 report.7 This difference may be in part due to the different measurement instruments in the present study (i.e., Hunger Vital Sign 2-item screener) compared with the USDA report, and it may be related to the present survey recruiting a local convenience sample of FP clients rather than a national probability sample of households.1,42
A second limitation is that data were collected at a single point in time from each respondent. This cross-sectional approach does not provide insight into whether specific behaviors or circumstances (e.g. difficulty paying medical bills or presence of heart disease) preceded, followed, or began simultaneously with changes in degree of recent FP use. A third limitation is that this survey relied exclusively on self-report data of food pantry use and health-related variables. Future research can evaluate the robustness of the present study’s findings using pantry records and health records. A final limitation is that the distances between the High, Medium, and Low user categories are unknown. However, the Brant test revealed that the proportional odds assumption was not violated, and the goodness of fit tests of the fitted model did not give evidence of lack of fit. Therefore, our assumption of ordinality of the categories seems appropriate.
Conclusions
While studies have documented how food insecure households differ from food secure households by sociodemographic characteristics, health status, and health-related trade-offs, 1,6–8,12 this is the first study to categorize degree of recent FP use and its associations with FP users’ characteristics, making a significant contribution to the literature. The three categories of FP users (i.e., high, medium, and low users) were associated with important sociodemographic, behavioral, and health-related characteristics, including age, education, putting off buying medicine to buy food, and presence of heart disease or cancer in the household. This study addresses a gap in knowledge by identifying a straightforward, two-item approach to differentiate users’ reliance on FPs. Existing studies have sorted participants into simple binary categories (e.g. pantry users vs. non-users) or sorted based primarily on duration of use, rather than combining frequency of visits and proportion of household food obtained. With further validation and refinement, the classes identified in the present study may provide a tool to understand how recent reliance on FPs is related to clients’ current diet quality and health status. This approach could eventually inform development of interventions tailored to improve the health of high, medium, and low FP users, respectively.
Supplementary Material
Acknowledgements:
The authors would like to thank our partners at Northwest Arkansas Food Bank and their partner agencies who facilitated recruitment of participants.
Financial Support:
Research reported in this publication was supported by the National Institute of General Medical Sciences of the NIH (#5P20GM109096). Additional support was provided by a Translational Research Institute grant from the National Center for Advancing Translational Sciences of the NIH (#UL1TR003107). The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the funders.
Footnotes
Ethical Standards Disclosure: This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research study participants were reviewed by the Institutional Review Board at the University of Arkansas for Medical Sciences, which approved this study as exempt (IRB#217560).
REFERENCES
- 1.Coleman-Jensen A, Rabbitt MP, Gregory CA, Singh A. Household Food Security in the United States in 2019 2020. ERR-275.
- 2.Gregory CA, Coleman-Jensen A. Food Insecurity, Chronic Disease, and Health Among Working-Age Adults 2017. Accessed November 29, 2018. https://www.ers.usda.gov/webdocs/publications/84467/err-235.pdf
- 3.Russell JC, Flood VM, Yeatman H, Wang JJ, Mitchell P. Food insecurity and poor diet quality are associated with reduced quality of life in older adults. Article. Nutrition & Dietetics Feb 2016;73(1):50–58. doi: 10.1111/1747-0080.12263 [DOI] [Google Scholar]
- 4.Nguyen BT, Shuval K, Bertmann F, Yaroch AL. The Supplemental Nutrition Assistance Program, Food Insecurity, Dietary Quality, and Obesity Among U.S. Adults. Am J Public Health Jul 2015;105(7):1453–9. doi: 10.2105/ajph.2015.302580 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hanson KL, Connor LM. Food insecurity and dietary quality in US adults and children: a systematic review. Am J Clin Nutr Aug 2014;100(2):684–92. doi: 10.3945/ajcn.114.084525 [DOI] [PubMed] [Google Scholar]
- 6.Weinfield N, Mills G, Borger C, et al. Hunger in America 2014: National Report Prepared for Feeding America 2014. http://help.feedingamerica.org/HungerInAmerica/hunger-in-america-2014-full-report.pdf
- 7.Coleman-Jensen A, Rabbitt MP, Gregory CA, Singh A. Statistical supplement to household food security in the United States in 2019 2020. AP-084.
- 8.Gundersen C, Engelhard E, Hake M. The Determinants of Food Insecurity among Food Bank Clients in the United States. Journal of Consumer Affairs 2017;51(3):501–518. doi: 10.1111/joca.12157 [DOI] [Google Scholar]
- 9.Robaina KA, Martin KS. Food Insecurity, Poor Diet Quality, and Obesity among Food Pantry Participants in Hartford, CT. Article. Journal of Nutrition Education and Behavior Mar-Apr 2013;45(2):159–164. doi: 10.1016/j.jneb.2012.07.001 [DOI] [PubMed] [Google Scholar]
- 10.Simmet A, Depa J, Tinnemann P, Stroebele-Benschop N. The Dietary Quality of Food Pantry Users: A Systematic Review of Existing Literature. J Acad Nutr Diet Oct 2017;117(4):563–576. doi: 10.1016/j.jand.2016.08.014 [DOI] [PubMed] [Google Scholar]
- 11.Spees CK, Alwood A, Wolf KN, Rusnak S, Taylor CA. Poor adherence to preventive health care and cancer screening guidelines among food pantry clients. Journal of Hunger and Environmental Nutrition 2016;12(1):123–135. doi: 10.1080/19320248.2015.1095143 [DOI] [Google Scholar]
- 12.Berkowitz SA, Seligman HK, Choudhry NK. Treat or eat: food insecurity, cost-related medication underuse, and unmet needs. Am J Med Apr 2014;127(4):303–310.e3. doi: 10.1016/j.amjmed.2014.01.002 [DOI] [PubMed] [Google Scholar]
- 13.Knight CK, Probst JC, Liese AD, Sercy E, Jones SJ. Household food insecurity and medication “scrimping” among US adults with diabetes. Preventive Medicine. 2016/02/01/ 2016;83:41–45. doi: 10.1016/j.ypmed.2015.11.031 [DOI] [PubMed] [Google Scholar]
- 14.Kicinski LR. Characteristics of short and long-term food pantry users. Sociological Review 2012;26:58–74. [Google Scholar]
- 15.Pollard CM, Booth S, Jancey J, et al. Long-Term Food Insecurity, Hunger and Risky Food Acquisition Practices: A Cross-Sectional Study of Food Charity Recipients in an Australian Capital City. Int J Environ Res Public Health 08 2019;16(15)doi: 10.3390/ijerph16152749 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Black JL, Seto D. Examining Patterns of Food Bank Use Over Twenty-Five Years in Vancouver, Canada. Voluntas 2020;31(5):853–869. doi: 10.1007/s11266-018-0039-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kaiser ML, Cafer AM. Exploring Long-term Food Pantry Use: Differences Between Persistent and Prolonged Typologies of Use. Journal of Hunger & Environmental Nutrition 2017/01/02 2017;12(1):46–63. doi: 10.1080/19320248.2016.1157554 [DOI] [Google Scholar]
- 18.Loopstra R, Tarasuk V. Severity of household food insecurity is sensitive to change in household income and employment status among low-income families. J Nutr Aug 2013;143(8):1316–23. doi: 10.3945/jn.113.175414 [DOI] [PubMed] [Google Scholar]
- 19.Harris P, Taylor R, Thielke R, Payne J, Gonzalez N, Conde J. Research electronic data capture (REDCap)--a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Infrom 2009;42(2):377–81. doi: 10.1016/j.jbi.2008.08.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Harris PA, Taylor R, Minor BL, et al. The REDCap consortium: Building an international community of software platform partners. J Biomed Inform 07 2019;95:103208. doi: 10.1016/j.jbi.2019.103208 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Long CR, Rowland B, McElfish PA. Intervention to improve access to fresh fruits and vegetables among Arkansas food pantry clients. Prev Chronic Dis 2019;16:E09. doi: 10.5888/pcd16.180155 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rowland B, Mayes K, Faitak B, Stephens RM, Long CR, McElfish PA. Improving Health while Alleviating Hunger: Best Practices of a Successful Hunger Relief Organization. Curr Dev Nutr Sep 2018;2(9):nzy057. doi: 10.1093/cdn/nzy057 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Long CR, Narcisse MR, Rowland B, et al. Written Nutrition Guidelines, Client Choice Distribution, and Adequate Refrigerator Storage Are Positively Associated with Increased Offerings of Feeding America’s Detailed Foods to Encourage (F2E) in a Large Sample of Arkansas Food Pantries. J Acad Nutr Diet Oct 2019;doi: 10.1016/j.jand.2019.08.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Long CR, Rowland B, Steelman SC, McElfish PA. Outcomes of disease prevention and management interventions in food pantries and food banks: a scoping review. BMJ Open Aug 2019;9(8):e029236. doi: 10.1136/bmjopen-2019-029236 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Yan S, Caspi C, Trude ACB, Gunen B, Gittelsohn J. How urban food pantries are stocked and food is distributed: food pantry manager perspectives from Baltimore. Journal of Hunger & Environmental Nutrition 2020;doi: 10.1080/19320248.2020.1729285 [DOI] [Google Scholar]
- 26.Caspi CE, Davey C, Friebur R, Nanney MS. Results of a Pilot Intervention in Food Shelves to Improve Healthy Eating and Cooking Skills Among Adults Experiencing Food Insecurity. J Hunger Environ Nutr 2017;12(1):77–88. doi: 10.1080/19320248.2015.1095146 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Caspi CE, Grannon KY, Wang Q, Nanney MS, King RP. Refining and implementing the Food Assortment Scoring Tool (FAST) in food pantries. Public Health Nutr Oct 2018;21(14):2548–2557. doi: 10.1017/s1368980018001362 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Hager ER, Quigg AM, Black MM, et al. Development and validity of a 2-item screen to identify families at risk for food insecurity. Pediatrics Jul 2010;126(1):e26–32. doi: 10.1542/peds.2009-3146 [DOI] [PubMed] [Google Scholar]
- 29.Makelarski JA, Abramsohn E, Benjamin JH, Du S, Lindau ST. Diagnostic Accuracy of Two Food Insecurity Screeners Recommended for Use in Health Care Settings. Am J Public Health 11 2017;107(11):1812–1817. doi: 10.2105/AJPH.2017.304033 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Vedovato GM, Surkan PJ, Jones-Smith J, et al. Food insecurity, overweight and obesity among low-income African-American families in Baltimore City: associations with food-related perceptions. Public Health Nutr 06 2016;19(8):1405–16. doi: 10.1017/S1368980015002888 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Wetherill MS, Williams MB, White KC, Li J, Vidrine JI, Vidrine DJ. Food pantries as partners in population health: assessing organizational and personnel readiness for delivering nutrition-focused charitable food assistance. J Hunger Environ Nutr 2019;14(1–2):50–69. doi: 10.1080/19320248.2018.1512931 [DOI] [Google Scholar]
- 32.Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System (BRFSS) Accessed February 19, 2018, 2019. http://www.cdc.gov/brfss/
- 33.Hagenaars JA, McCutcheon AL, eds. Applied Latent Class Analysis Cambridge University Press; 2002. [Google Scholar]
- 34.Nylund KL, Asparouhov T, Muthén BO. Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study. Structural Equation Modeling: A Multidisciplinary Journal 2007;14(4):535–569. doi: 10.1080/10705510701575396 [DOI] [Google Scholar]
- 35.Brant R Assessing proportionality in the proportional odds model for ordinal logistic regression. Biometrics Dec 1990;46(4):1171–8. [PubMed] [Google Scholar]
- 36.Fagerland MW, Hosmer DW. Tests for goodness of fit in ordinal logistic regression models. Journal of Statistical Computation and Simulation 2016;86(17):3398–3418. doi: 10.1080/00949655.2016.1156682 [DOI] [Google Scholar]
- 37.Lipsitz SR, Fitzmaurice GM, Molenberghs G. Goodness-of-fit tests for ordinal response regression models. Applied Statistics 1996;45(2):175–190. doi: 10.2307/2986153 [DOI] [Google Scholar]
- 38.STATA. Version 16 StataCorp LLC; 2019. Accessed July 24, 2020. https://www.stata.com/
- 39.An R, Wang J, Liu J, Shen J, Loehmer E, McCaffrey J. A systematic review of food pantry-based interventions in the USA. Public Health Nutr Jun 2019;22(9):1704–1716. doi: 10.1017/s1368980019000144 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Bencivenga M, DeRubis S, Leach P, Lotito L, Shoemaker C, Lengerich EJ. Community partnerships, food pantries, and an evidence-based intervention to increase mammography among rural women. J Rural Health Winter 2008;24(1):91–5. doi: 10.1111/j.1748-0361.2008.00142.x [DOI] [PubMed] [Google Scholar]
- 41.Greder K, Garasky S, Klein S. Research to action: a campus-community partnership to address health issues of the food insecure. Journal of Extension 2007;45(6):6FEA4. [Google Scholar]
- 42.Current Population Survey (CPS) Methodology Accessed August 22, 2020, https://www.census.gov/programs-surveys/cps/technical-documentation/methodology.html
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.