Persistent Food Insecurity Among Older Adult Cancer Survivors: A National Cohort Study

Abstract

OBJECTIVES

To describe the trajectories of food insecurity and examine their associations with sociodemographic and health-related factors in older adult cancer survivors.

SAMPLE & SETTING

Data from 2015 to 2021 from the National Health and Aging Trends Study, a nationally representative cohort study of community-dwelling older adults aged 65 years or older, were extracted and analyzed.

METHODS & VARIABLES

Food insecurity was annually measured based on five self-reported items. Group-based trajectory modeling was used to identify food insecurity trajectory groups. Data analysis accounted for the complex survey design and analytic weights.

RESULTS

The sample consisted of 1,935 older adult cancer survivors. The weighted prevalence of food insecurity ranged from 2.46% to 4.73% from 2015 to 2021. The following two food insecurity trajectory groups were identified: low–stable (n = 1,796, 93%) and medium–stable (n = 139, 7%). Individuals who were younger, female, not non-Hispanic and White, and physically frail, and those with lower household income and higher levels of anxiety and depression, were more likely to experience persistent food insecurity.

IMPLICATIONS FOR NURSING

Food security in older adult survivors is persistent and distributed inequitably among those who are female, not non-Hispanic and White, and younger, and those with lower household income.

Food insecurity, defined as a lack of continuous access through socially acceptable means to nutritious and safe foods, is a common and important issue (Coleman-Jensen et al., 2022; Pooler et al., 2019). It is particularly relevant to cancer survivors because of their fundamental need for enough food and appropriate nutrition. Food insecurity is negatively associated with treatment adherence; cancer survivors with food insecurity are more likely to forgo, delay, or make changes to treatment (McDougall et al., 2020). However, cancer and anticancer treatments frequently lead to metabolic disorders, diet-related symptoms (e.g., lack of appetite, nausea, vomiting), and malnutrition (de van der Schueren et al., 2018). In addition, the financial toxicity of cancer and its treatment may exacerbate food insecurity because of limited access to healthy food and nutrition (Charkhchi et al., 2018), highlighting the importance of food insecurity management throughout the cancer care continuum.

Data regarding food security among older adult cancer survivors are scarce (Rock et al., 2022). Previous studies have reported a broad range of food insecurity prevalence (8%–55%) among patients with cancer (Charkhchi et al., 2018; Patel et al., 2019; Trego et al., 2019), which is higher than in community-dwelling older adults as a whole (4.3%) (Tucher et al., 2021). However, the longitudinal change in and duration of food insecurity remains unclear. Cancer survivors may experience chronic food insecurity, may be unable to obtain nutritious safe foods for an extended period, or may have transient or episodic food insecurity before and after cancer diagnosis (Tucher et al., 2021). Possible inequities exist because food insecurity is influenced by factors like race, ethnicity, income, and disability (Rock et al., 2022), but these factors have not been established in cancer survivors. Understanding the dynamic nature of food insecurity status and its influencing factors could facilitate targeted food insecurity screening and assistance programs for older adult cancer survivors at higher risk for persistent food insecurity. This study aimed to describe the trajectories of food insecurity among older adult cancer survivors and examine the associated sociodemographic and health-related factors among this population using a nationally representative sample from 2015 to 2021.

Methods

Data Source and Study Population

The National Health and Aging Trends Study (NHATS) is an annual, nationally representative longitudinal study of older adults aged 65 years or older who are enrolled in Medicare (Freedman et al., 2022). The initial sample was first interviewed in 2011 (round 1), with an oversampling of individuals aged 90 years or older and non-Hispanic Black adults to reduce the small sample bias; the sample was replenished in 2015 (round 5). The overall weighted response rates ranged from 71.3% to 96% from 2011 (round 1) through 2021 (round 11) (Freedman et al., 2022). Analytic weighting was performed each round to account for the complex sample design of the NHATS and to adjust for potential nonresponse bias. All participants provided informed consent, and the NHATS was approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

This study used data from rounds 5–11 (2015–2021). Participants were included if they met the following criteria: (a) having a self-reported diagnosis of cancer in 2015 or in prior rounds of the NHATS, (b) being community-dwelling in round 5, and (c) having complete data about food insecurity in round 5. Of 8,334 participants, 1,996 were community-dwelling cancer survivors. Individuals with missing food insecurity data in 2015 (n = 61) were excluded, resulting in 1,935 participants.

Measurements

Food insecurity was measured annually from rounds 5–11 based on the following five items on the NHATS: (a) going without groceries because of limited ability or social support, (b) going without hot meals because of functional limitation or a lack of help, (c) going without eating because of an inability to feed oneself or a lack of support, (d) skipping meals because of insufficient food or money, and (e) experiencing more than a few days of skipped meals (Tucher et al., 2021). Participants scored one point per criterion and otherwise scored zero; the total score ranged from 0 to 5. Participants were classified as either food insecure (one or more points) or food secure (zero points).

Potential sociodemographic and health-related factors associated with food insecurity trajectory affiliations were determined according to previous studies about food insecurity (Nagata et al., 2019; Rock et al., 2022). Sociodemographic characteristics collected at round 5 were age (categorized in five-year increments); sex (self-reported male or female); and race and ethnicity (self-reported Non-Hispanic White, African American, Hispanic, or other). Considering the existing relationships between food security and socioeconomic factors (Rock et al., 2022), this study also used data about marital status (married or not married); education level (less than high school, high school diploma, some college or equivalent, or college degree or higher); and annual income (less than $15,000, $15,000–$29,999, $30,000–$44,999, $45,000–$60,000, or more than $60,000). Health-related characteristics were as follows:

• ■ Obesity (body mass index of 30 kg/m² or greater or body mass index of less than 30 kg/m²)

• ■ Number of comorbidities (ranging from 0 to 8 based on self-reported history of heart attack, heart disease, hypertension, arthritis, osteoporosis, diabetes, lung disease, and/or stroke)

• ■ Anxiety and depression (assessed by the Patient Health Questionnaire–4, a four-item, four-point Likert-type scale measuring anxiety and depression, with responses ranging from 0 [not at all] to 3 [nearly every day] and higher total scores indicating worse anxiety and depression [Löwe et al., 2010])

• ■ Physical frailty (frail or not frail, as assessed by the Fried frailty phenotype, a method for evaluating frailty with the following five criteria: unintentional weight loss, poor handgrip strength, self-reported exhaustion, slow walking speed, and low physical activity; frailty is indicated if three of five criteria are met [Bandeen-Roche et al., 2015])

• ■ Cognition (probable dementia, possible dementia, or no dementia, as assessed by self-reported dementia diagnosis, the Ascertain Dementia 8 dementia screening interview, or performance on cognitive tests of memory, orientation, and executive function [Kasper et al., 2013])

Statistical Analysis

Group-based trajectory modeling was used to identify food insecurity trajectories across seven years. Group-based trajectory modeling is a method of clustering participants who follow similar developmental patterns into several groups and profiling their longitudinal trajectories (Nagin & Odgers, 2010). The analyses were performed in two steps (Jones & Nagin, 2013). First, participants were hypothesized to be clustered into two to five latent trajectory groups, models were fitted with cubic polynomial terms, and the optimal number of trajectories was selected based on a lower Bayesian information criterion. As shown in Supplementary Table 1 online, the Bayesian information criterion value was lowest when the number of trajectory groups was two. Second, the number of trajectory groups was fixed at two, and different polynomial forms (intercept only, linear, quadratic, or cubic terms) for each trajectory group were tested to determine the shape of the trajectories. The optimal functional form for each trajectory was determined using the chi-square test with p < 0.05. The results indicated that the polynomial terms of the optimal functional form were zero (see Supplementary Table 2 online). The model was considered adequate if the average group posterior probabilities (probability of membership for individuals who were assigned to each group) were greater than 70% and if the odds of correct classification (the ratio of the odds of classification based on the maximum posterior probability classification) were five or greater (Nagin & Odgers, 2010; Niyonkuru et al., 2013).

Complex survey design and sampling weights were incorporated in the subsequent analyses of the data. Taylor series linearization methods were used for variance estimation. Proportions for categorical variables and means and standard errors (SEs) for continuous variables were used to describe the sociodemographic and health-related characteristics at baseline. Bivariate analyses were used to compare the baseline characteristics of distinct trajectory groups. Rao–Scott chi-square tests were used for categorical variables, and design-based F tests were used for continuous variables. The predictors of trajectory affiliations were examined using design-based logistic regression analysis, and adjusted odds ratios (AORs) and 95% confidence intervals (CIs) were calculated. To examine the robustness of the results, a sensitivity analysis was conducted by excluding participants who did not have food security data through all seven rounds of the NHATS (n = 913) and repeating the analyses. All analyses were performed using Stata/SE, version 17.0, with a two-sided p value of less than 0.05 considered to be statistically significant.

Results

Sample Characteristics

Overall, 1,935 participants were assessed in this study, representing a population of 9.84 million older adult cancer survivors in the United States. The majority of the participants were aged 75 years or older (n = 1,352, 52%), were non-Hispanic White (n = 1,525, 87%), were married (n = 1,037, 60%), were not obese (n = 1,396, 70%), were not frail (n = 1,250, 82%), had at least some college education (n = 1,037, 60%), and had no dementia (n = 1,562, 86%) (see Table 1). The average number of chronic conditions was 2.38 (SE = 0.44), and the mean score for anxiety and depression was 5.78 (SE = 0.08).

TABLE 1.

Baseline Sample Characteristics by Trajectory Group

Characteristic	Total (N = 1,935)		Low–Stable Risk (N = 1,838)		Medium–Stable Risk (N = 97)		pa
	X̄	SE	X̄	SE	X̄	SE
Number of chronic conditions	2.38	0.44	2.33	0.04	3.22	0.16	< 0.001
Anxiety and depression score	5.78	0.08	5.66	0.07	7.96	0.39	< 0.001
Characteristic	n	Weighted % b	n	Weighted % b	n	Weighted % b	p a
Age (years)							0.213
65–69	177	22	165	22	12	28
70–74	406	26	377	26	29	34
75–79	426	21	409	21	17	16
80–84	439	16	423	17	16	10
85–89	305	10	293	10	12	6
90 or older	182	5	171	5	11	5
Sex							< 0.001
Female	989	50	273	49	18	76
Male	946	50	57	51	5	24
Race and ethnicity							0.006
African American	291	5	273	5	18	6
Hispanic	62	4	57	4	5	8
Non-Hispanic White	1,525	87	1,461	88	64	70
Other	48	3	43	3	5	9
Missing	9	1	4	–	5	6
Marital status							< 0.001
Married	1,037	60	1,007	62	30	32
Not married	898	40	831	38	67	68
Education level							0.005
Less than high school	367	15	346	15	21	18
High school diploma	524	26	498	26	26	24
Some college or equivalent	493	26	459	26	34	39
College degree or higher	544	32	532	33	12	13
Missing	7	1	3	–	4	1
Annual income ($)							< 0.001
Less than 15,000	296	12	262	10	34	34
15,000–29,999	470	21	444	21	26	25
30,000–44,999	351	18	333	18	18	22
45,000–60,000	220	12	214	12	6	5
More than 60,000	598	38	585	39	13	15
Obesity							0.126
No	1,396	70	1,335	70	61	60
Yes	514	29	479	29	35	38
Missing	25	1	24	1	1	2
Physical frailty							< 0.001
No	1,250	82	1,212	83	38	48
Yes	377	18	338	17	39	52
Missing	308	15	288	14	20	20
Cognition							0.938
No dementia	1,562	86	1,482	86	80	87
Possible dementia	168	7	159	7	9	7
Probable dementia	203	7	195	7	8	6
Missing	2	–	2	–	–	–

^aP values compared participants across groups using Rao–Scott chi-square tests or design-based F tests as appropriate.

^bPercentages were adjusted for National Health and Aging Trends Study sampling weights.

BMI—body mass index; PHQ-4—Patient Health Questionnaire–4; SE—standard error

Note. Number of comorbidities ranged from 0 to 8 based on self-reported history of heart attack, heart disease, hypertension, arthritis, osteoporosis, diabetes, lung disease, and/or stroke. Anxiety and depression were assessed using the PHQ-4, a 4-item, 4-point Likert-type scale with responses ranging from 0 (not at all) to 3 (nearly every day); higher total scores indicate worse anxiety and depression. Obesity was defined as BMI of 30 kg/m² or greater. Physical frailty was measured using the Fried frailty phenotype with the following 5 criteria: unintentional weight loss, poor handgrip strength, exhaustion, slow walking speed, and low physical activity; frailty is indicated if 3 of these criteria are met. Cognition was assessed by self-reported dementia diagnosis, the Ascertain Dementia 8 dementia screening interview, or performance on cognitive tests of memory, orientation, and executive function.

Note. Data from 2015 to 2021 were sourced from the National Health and Aging Trends Study, an annual, nationally representative longitudinal study of older adults aged 65 years or older enrolled in Medicare (Freedman et al., 2022). National estimates accounted for complex survey design.

Note. Group-based trajectory modeling was used to determine food insecurity risk groups. The low–stable risk group had a low, persistent risk and the medium–stable risk group had a medium, persistent risk.

Note. Baseline sample characteristics used actual group size rather than estimated group size, which was used for trajectory modeling.

Note. Because of rounding, percentages may not total 100.

Food Insecurity Trajectories

Supplementary Table 3 online provides the proportions with SEs of food insecurity in each survey year; the median weighted prevalence of food insecurity from rounds 5–11 was 3.62% (range = 2.46%–4.73%). As shown in Figure 1, the optimally fitting trajectory model had the following two trajectory groups: the low–stable risk group, with an estimated group size of 93% of older adult cancer survivors (n = 1,796) who maintained a low risk of food insecurity near zero during the data collection period, and the medium–stable risk group with 7% (n = 139) who maintained a medium risk of food insecurity during the same period. Of note, this model suggested using estimated group sizes rather than actual group sizes for trajectory modeling. Both trajectories were flat, and no transition was observed. The goodness-of-fit model indices showed relatively adequate model fit. The average group posterior probability and odds of correct classification of the low–stable risk group were 97.02% and 2.52, respectively; of the medium–stable risk group, they were 87.14% and 87.26, respectively.

FIGURE 1.

Food Insecurity Risk Groups

Note. Group-based trajectory modeling was used to determine food insecurity risk groups in data from the National Health and Aging Trends Study, an annual, nationally representative longitudinal study of older adults aged 65 years or older enrolled in Medicare. Dashed lines represent 95% confidence intervals for each group.

Compared with the low–stable risk group, participants in the medium–stable risk group were more likely to be female, not of non-Hispanic White race and ethnicity, not married, and frail, and have a lower education level, lower annual income, greater number of chronic conditions, and a higher level of anxiety and depression (p < 0.05). No significant differences were found between trajectory groups in the distribution of age, obesity, or cognition. Sensitivity analysis showed similar trajectories and proportions as the total sample; the low–stable risk group was 94% of the sensitivity analysis sample, and the medium–stable risk group was 6% of that sample (see Supplementary Figure 1 online).

Predictors of Trajectory Affiliations

Associations between sociodemographic and health-related predictors and distinct trajectory affiliations are shown in Figure 2. Patients who were younger, female (AOR = 2.37, 95% CI [1.08, 5.2]), not non-Hispanic and White (AOR = 3.94, 95% CI [1.26, 12.28]), frail (AOR = 3.24, 95% CI [1.49, 7.02]), or had a higher level of anxiety and depression (AOR = 1.16, 95% CI [1.04, 1.29]) or lower annual income were more likely to have a medium–stable risk of food insecurity.

FIGURE 2.

Associations Between Sample Characteristic Predictors and Distinct Trajectory Affiliations

AOR—adjusted odds ratio; ref—reference

Note. Risk estimates were adjusted for age, sex, race and ethnicity, marital status, education level, annual income, obesity, number of chronic conditions, anxiety and depression score, frailty, and cognition. The logistic regression model accounted for complex survey design and weighting, and the low–stable risk group is the ref group.

Note. AORs to the right of 1 (dashed line) represent greater likelihood of association between that predictor and higher risk of food insecurity. Solid lines represent 95% confidence intervals for each AOR.

Discussion

To the investigators’ knowledge, this is the first study to explore the trajectories of food insecurity in the population of older adult cancer survivors. In the context of cancer care, food insecurity is an important yet under-recognized health-related social risk factor. This study reveals the persistent nature of food insecurity and suggests nonmodifiable and modifiable factors that may inform early detection and personalized interventions to mitigate this issue.

This study found a 3.62% median prevalence (2.46%–4.73% from 2015 to 2021) of food insecurity in older adult cancer survivors, which is similar to that of community-dwelling older adults (Tucher et al., 2021). Prior studies of cancer survivors have reported a higher prevalence of food insecurity (8%–55%) (Charkhchi et al., 2018; Patel et al., 2019; Simmons et al., 2006; Trego et al., 2019), and the divergence may be explained by the older age and coverage by Medicare insurance of participants in the current study. In terms of the longitudinal change in food insecurity, the current study’s findings suggest that it may persist for years without intervention. Prior studies of different populations reported inconsistent findings. For example, a study of children reported the following two food insecurity trajectories: high-risk trajectory (3.6%) and low-risk trajectory (96.4%) (Paquin et al., 2021). A study of two disadvantaged communities reported the following three food insecurity trajectories: persistent food security (27%), transient food insecurity (36%), and persistent food insecurity (37%) (Liese et al., 2021). Food insecurity was found to be persistent in the current study, which indicates that programs are essential to address food insecurity and help patients eat well throughout the cancer care continuum. Possible programs to support food-insecure patients include clinical screening for food insecurity (e.g., household food security surveys like The Hunger Vital Sign™) (Makelarski et al., 2017), local food banks to mitigate food insecurity directly by providing food, and prescription payment assistance programs to relieve food budgets by allowing for the shifting of household expenses (Patel et al., 2019; Raber et al., 2022). Universal screening can empower providers to begin addressing food insecurity in their practices by encouraging enrollment in federal benefits programs, providing information on local food depositories, and offering referrals to social workers. Policymakers may consider advocating for expanded food assistance programs or insurance coverage for nutritional counseling to mitigate food insecurity in older adult cancer survivors.

The burden of food insecurity is distributed inequitably across the population. This study supports previous findings that disparities in food security, including being female, not non-Hispanic and White, and younger, and having lower household income, may indicate a higher risk of food insecurity (Raber et al., 2022). Efforts to screen for and address food insecurity need to be prioritized to eliminate these disparities and inequities. Patients with frailty and mental disorders may also experience greater risk of food insecurity (Schwartz et al., 2019; Tucher et al., 2021). Personalized programs supporting food security, combined with interventions targeting physical and mental impairment, show promise as more effective methods to address food insecurity.

Limitations

This study has several limitations. First, participants in this study were community-dwelling Medicare beneficiaries with completed data about food insecurity in round 5 of the NHATS, which limits the generalizability of these findings regarding older adult cancer survivors who reside in other settings (e.g., long-term care facilities, assisted living facilities) or who do not have health insurance. Second, food insecurity in this study was self-reported, which may indicate recall bias. Third, some cancer-specific information (e.g., cancer stage, treatment type) was not available in this dataset, so cancer-specific factors associated with food insecurity were not examined. Future longitudinal studies of a more diverse older adult cancer population are needed to track food insecurity objectively and incorporate more detailed cancer-specific data.

Conclusion

In the context of cancer care, food insecurity is an important yet understudied health issue. This study explored the prevalence and trajectories of food insecurity in the population of older adult cancer survivors and revealed the low prevalence (3.62%) but persistent state of food insecurity in this population. The burden of food insecurity is distributed inequitably among patients who are female, not non-Hispanic and White, and younger, and who have lower household income, highlighting the importance of strategies to eliminate these disparities and inequities. Modifiable factors (e.g., frailty, anxiety, depression) may help inform personalized interventions to improve physical and mental health while supporting food security.

KNOWLEDGE TRANSLATION.

• ■ Participants were clustered into the following two food insecurity trajectory groups: low–stable risk and medium–stable risk.

• ■ The prevalence of food insecurity found in this study was lower than that reported in previous studies and other populations with cancer.

• ■ The burden of food insecurity is distributed inequitably among older adult cancer survivors, which highlights the importance of strategies to eliminate these disparities and inequities.

QUESTION GUIDE FOR A JOURNAL CLUB.

Journal clubs can help to increase and translate findings to clinical practice, education, administration, and research. Use the following questions to start discussion at your next journal club meeting. Then, take time to recap the discussion and make plans to proceed with suggested strategies.

1. What are some factors specific to cancer diagnosis and treatment that may influence food insecurity?

2. What assessment questions can you ask to screen for food insecurities among older cancer survivors?

3. How can healthcare facilities address food insecurities when people are in need?

Visit https://bit.ly/3C7ar0z for details on creating and participating in a journal club. Contact pubONF@ons.org for assistance or feedback. Photocopying of the article for discussion purposes is permitted.