Skip to main content
Frontiers in Nutrition logoLink to Frontiers in Nutrition
. 2026 Nov 3;13:1958938. doi: 10.3389/fnut.2026.1958938

Socioeconomic differences of psychological health and the potential benefit of eliminating food insecurity: longitudinal and counterfactual evidence from the health and retirement study

Shixing Feng 1,2,3,4,†, Tianyi Wang 3,†, Shuqi Zhu 5, Huaqiang Zhai 1,*, Jialin Liu 5,6,*
PMCID: PMC13775839  PMID: 43048474

Abstract

Background

Food insecurity (FI) is strongly socially patterned and may contribute to socioeconomic differences in psychological health. However, it remains unclear whether lower socioeconomic status (SES) groups would derive greater psychological health benefits from achieving food security and whether such differences reflect greater exposure to FI, greater vulnerability to FI, or both.

Methods

We analyzed 6,695 participants from the U.S. Health and Retirement Study. SES was constructed from household wealth, household income, and educational attainment. FI was assessed in 2013 using the U.S. Department of Agriculture Six-Item Short Form Food Security Survey Module. Psychological health was assessed using standardized, direction-reversed scores from the 8-item Center for Epidemiologic Studies Depression Scale in 2014, 2016, 2018, and 2020. Linear mixed-effects models examined longitudinal associations and temporal interactions. G-computation was used to estimate SES-specific population intervention benefits (PIB) and conditional average treatment effects (CATE) under hypothetical food-security scenarios.

Results

FI showed a pronounced socioeconomic gradient, affecting 40.8% of participants with low SES, 14.2% with middle SES, and 2.4% with high SES. Higher SES was associated with better psychological health, while FI was associated with poorer psychological health. The FI-related psychological health difference narrowed over time (FI × time: β = 0.065, 95% confidence interval [CI] 0.044–0.085; P < 0.001). The overall SES × FI × time interaction was significant (P = 0.014), indicating temporal heterogeneity across SES groups without a consistent socioeconomic gradient. Counterfactual analyses showed that the average PIB associated with eliminating economically driven barriers to adequate food access was greatest in the low-SES group (0.097 standard deviation [SD] units, 95% CI 0.068–0.126), compared with the middle-SES (0.023 SD, 95% CI 0.011–0.035) and high-SES groups (0.003 SD, 95% CI −0.002–0.007). CATEs showed less consistent SES differences over time.

Conclusions

FI was strongly concentrated among adults with lower SES and was consistently associated with poorer psychological health. The greater potential population-level benefit of achieving food security among lower-SES adults appeared to reflect primarily their greater burden of FI, with more limited evidence for consistently greater vulnerability to FI.

Keywords: counterfactual analysis, food insecurity, longitudinal analysis, psychological health, socioeconomic status

1. Introduction

Psychological health, a major global public health challenge (1), is socially patterned, with substantial differences across socioeconomic groups (2, 3). These disparities persist into midlife and older age, with lower socioeconomic status (SES) consistently associated with poorer psychological health and a higher risk of depression (3, 4).

Food insecurity (FI) may be particularly relevant to these socioeconomic differences in psychological health because it is closely tied to socioeconomic disadvantage while also being associated with psychological wellbeing. FI is defined as limited or uncertain availability of nutritionally adequate and safe foods, or limited or uncertain ability to acquire acceptable foods in socially acceptable ways (5). It is more prevalent among socioeconomically disadvantaged populations (6, 7) and has been consistently associated with poorer psychological health, including depressive symptoms and psychological distress (8).

The coexistence of these patterns raises an important question: whether the greater FI-related psychological health burden in lower-SES populations is explained only by their greater exposure to FI, or whether the psychological consequences of FI are also stronger among those with fewer socioeconomic resources. From a resource-buffering perspective, individuals with lower SES may have fewer economic, social, and material resources to buffer the stress associated with FI, potentially amplifying its adverse psychological consequences (9). Emerging empirical evidence also suggests that socioeconomic position may modify the association between FI and psychological health, although such evidence remains limited and context-specific (10). We therefore hypothesized that the adverse association between FI and psychological health would be stronger among individuals with lower SES.

However, previous studies have mainly examined whether FI is associated with poorer psychological health (11, 12). Longitudinal evidence in middle-aged and older adults remains comparatively limited (13), and few studies have distinguished whether a greater FI-related psychological health burden in lower-SES populations reflects greater exposure to FI, greater vulnerability to FI, or both. Consequently, it remains unclear whether reducing FI would yield greater population-level psychological health benefits among lower-SES groups and whether such benefits are driven primarily by the higher prevalence of FI or by stronger FI-related psychological health contrasts.

Therefore, using longitudinal data from the Health and Retirement Study (HRS), this study aimed to: (1) examine socioeconomic differences in psychological health trajectories and the prevalence of FI; (2) estimate whether a hypothetical elimination of FI could provide greater psychological health benefits for lower-SES groups; and (3) assess whether any greater benefit among lower-SES groups reflects their higher exposure to FI, a stronger psychological impact of FI, or both.

2. Materials and methods

2.1. Data source

This study used publicly released longitudinal data from the HRS, available to registered users subject to the HRS data-use conditions. The HRS is a nationally representative longitudinal cohort study conducted by the Institute for Social Research at the University of Michigan (14). It primarily surveys U.S. adults aged 50 years and older and their spouses or partners. Core interviews are conducted approximately every 2 years and collect repeated information on demographic characteristics, SES, household structure, health status, psychological wellbeing, and healthcare utilization.

2.2. Study population and sample selection

The study population was drawn from the 2013 Health Care and Nutrition Study (HCNS), a supplemental survey of HRS (15). The 2013 HCNS was administered to a random subsample of eligible HRS respondents; the sampling frame consisted of living HRS respondents who were not included in the 2013 Consumption and Activities Mail Survey. Questionnaires were mailed to 12,418 respondents, of whom 8,073 returned the HCNS questionnaire, corresponding to a simple response rate of 65%. The original HRS protocols were approved by the University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board. For the 2013 HCNS mail survey, participation was voluntary and consent was inferred from completion and return of the questionnaire, in accordance with HRS procedures.

Participants were then sequentially excluded if they had missing information required to construct the SES measure (n = 191), missing FI information (n = 494), missing 2012 Center for Epidemiologic Studies Depression Scale (CES-D) scores or no valid CES-D follow-up assessment between 2014 and 2020 (n = 507), or missing key baseline covariates (n = 186). The final analytic sample included 6,695 participants.

2.3. Socioeconomic status

SES was constructed using three dimensions: household wealth, household income, and educational attainment. Household wealth and household income were each categorized into low, middle, and high levels according to their distributions in the study sample. Educational attainment was categorized as less than high school, high school graduate, and some college or above.

Each dimension was assigned a score of 0, 1, or 2, with 0 indicating the most disadvantaged status and 2 indicating the most advantaged status. Scores across the three dimensions were summed to generate an overall SES score ranging from 0 to 6, with higher scores indicating more favorable socioeconomic conditions (16–18).

In the primary regression analyses, the SES score was modeled as a continuous variable to estimate the association of each 1-point increase in SES with psychological health levels and rates of change over time. A three-level SES variable was additionally constructed as low (0–2 points), middle (3–4 points), and high (5–6 points) and was used in descriptive analyses, trajectory analyses, and SES-stratified counterfactual analyses. These cut-off points were selected to preserve the ordinal integer structure of the composite SES score while yielding three ordered groups of broadly similar size. Because the continuous SES score was used in the primary regression analyses, the main association estimates did not depend on these categorical cut-off points. The continuous SES score and the three-level SES variable were not included simultaneously in the same model. Sensitivity analyses examined household wealth, household income, and educational attainment separately.

2.4. Food insecurity

FI was assessed in the 2013 HCNS using the U.S. Department of Agriculture (USDA) Six-Item Short Form Food Security Survey Module, which captures food-related hardship experienced during the previous 12 months (5, 19). The HCNS questionnaire presented five numbered questions covering food not lasting before there was money to buy more; inability to afford balanced meals; cutting or skipping meals because there was not enough money for food, including the frequency of this experience; eating less than needed because of lack of money; and being hungry but not eating because there was not enough money for food (Supplementary Table S1). Although presented as five numbered questions, the combined cut-or-skip-meals item contributes two scoring components, yielding a total food security score ranging from 0 to 6.

Following the USDA scoring algorithm, scores of 0–1 were classified as food secure or marginally food secure, scores of 2–4 as low food security, and scores of 5–6 as very low food security. In the primary analyses, FI was defined as a binary indicator, with scores of 2–6 classified as food insecure and scores of 0–1 as food secure. The continuous 0–6 FI score was additionally examined in sensitivity analyses to assess severity gradients.

2.5. Psychological health outcome

Psychological health was assessed using the HRS 8-item CES-D (Supplementary Table S2). The HRS 8-item CES-D uses binary response options to assess whether respondents experienced depressive symptoms during the previous week, including feeling depressed, lonely, sad, restless sleep, and difficulty getting going, as well as positive affect items such as feeling happy and enjoying life. After reverse coding the positively worded items, the eight items were summed to generate a raw CES-D score ranging from 0 to 8, with higher scores indicating a greater burden of depressive symptoms. HRS methodological documentation supports the use of the 8-item CES-D for longitudinal measurement across waves and indicates acceptable internal consistency (20, 21).

To place CES-D scores from different survey waves on a common longitudinal reference scale, scores from all waves were standardized using the mean and standard deviation (SD) of the 2012 baseline distribution as the fixed reference, as shown in Equation 1 (22).

CESDgoodzit=-CESDit-CESD¯2012SD(CESD2012) (1)

The standardized CES-D scores were then multiplied by −1 to reverse their direction. Thus, higher standardized scores indicate better psychological health, whereas lower scores indicate a greater burden of depressive symptoms.

2.6. Covariate

All baseline covariates were measured in the 2012 survey. The primary adjustment set included age, sex, race/ethnicity, current smoking status (yes or no), alcohol use (yes or no), body mass index (BMI), and partnership status. The 2012 CES-D score was additionally included in longitudinal models as a measure of prior psychological health.

Self-rated health, chronic disease burden, and health insurance status were included in the health-adjusted models. Self-rated health and chronic conditions may partly lie on the pathways linking socioeconomic disadvantage or FI to subsequent psychological health, while health insurance was treated as a healthcare access covariate. Self-rated health was assessed on a five-category scale ranging from excellent to poor. Chronic disease burden was calculated as the number of self-reported physician diagnoses of hypertension, diabetes, cancer, chronic lung disease, heart disease, stroke, and arthritis.

Because income, wealth, and educational attainment were used to construct the composite SES measure, they were not additionally entered as separate covariates in models in which the overall SES score was the exposure.

2.7. Statistical analysis

Baseline characteristics were described separately according to low, middle, and high SES, and according to the presence or absence of USDA-defined FI. Continuous variables were summarized as means and SDs. Categorical variables were summarized as counts and percentages. To assess potential selection bias, we compared baseline characteristics between participants included in the final analytic sample and participants who were excluded because of missing SES indicators, FI data, baseline CES-D data or follow-up psychological health outcomes, or covariates. Between-group differences in characteristics were evaluated using standardized mean differences (SMDs), rather than relying on P-values, which can be strongly influenced by sample size in large studies.

We also plotted mean CES-D trajectories from 2012 to 2020 according to SES category (low, middle, and high) and USDA-defined FI status (food secure vs. food insecure). The outcome was a standardized psychological health score assessed in 2012, 2014, 2016, 2018, and 2020, with higher values indicating better psychological health. These plots were intended to describe unadjusted overall trajectories and were not used to infer causal effects or independent associations.

Linear mixed-effects models (LMMs) were used to examine the longitudinal associations of SES and FI with psychological health trajectories (23). Models included participant-specific random intercepts to account for the correlation of repeated measurements within individuals and between-participant heterogeneity. Time was coded as 0, 1, 2, and 3, corresponding to the 2014, 2016, 2018, and 2020 follow-up waves, respectively, thereby assuming a linear change across follow-up waves. Models were fitted separately for SES and FI. Model 1 was adjusted for follow-up time and baseline CES-D score. Model 2 was further adjusted for age, sex, race/ethnicity, partnership status, current smoking status, alcohol use, and BMI. Model 3 additionally included an interaction term between the exposure (SES or FI) and follow-up time to assess whether the exposure-related difference in psychological health changed over follow-up. In health-adjusted sensitivity analyses, Model 4 extended Model 2 by additionally adjusting for self-rated health, the number of chronic conditions, and health insurance coverage, and Model 5 further included the corresponding exposure-by-time interaction term. In a separate analysis, a three-way interaction among SES, FI, and follow-up time was fitted to examine whether temporal changes in the FI–psychological health association differed across SES groups. The models assumed normally distributed random intercepts and residual errors and independence between participants.

Modified Poisson regression with a log link and robust sandwich variance estimation was used to estimate prevalence ratios (PRs) for the associations of SES with USDA-defined FI and individual food-insecurity indicators (24). Continuous and categorical SES measures were examined in separate models. All models were adjusted for prespecified covariates.

In the counterfactual analyses, Gaussian generalized estimating equations (GEE) were used to model standardized psychological health scores measured in 2014, 2016, 2018, and 2020 (25). The model included FI, baseline SES, follow-up year, all corresponding two- and three-way interactions, and the prespecified covariates, thereby accounting for within-participant correlation across repeated measurements while estimating population-averaged associations. G-computation was then used to estimate standardized mean psychological health scores under three FI scenarios (26). For each participant and follow-up year, psychological health was predicted under: (1) the observed scenario, in which FI was retained at its observed value; (2) a universal food-security scenario, in which FI was set to 0 for all participants; and (3) a universal food-insecurity scenario, in which FI was set to 1 for all participants. Baseline SES and all other covariates were retained at their observed values. Predicted outcomes were then averaged within each SES group and follow-up year.

Two principal counterfactual contrasts were calculated. The primary counterfactual estimand was the population intervention benefit (PIB), defined as the difference between predicted mean psychological health under universal food security and that under the observed FI distribution. PIB therefore quantifies the expected population-level improvement associated with eliminating observed FI within each SES group. The conditional average treatment effect (CATE) was additionally estimated as the difference between predicted mean psychological health under universal food security and universal food insecurity within each SES group, providing complementary information on whether the FI-related counterfactual contrast differed across SES groups. All estimates were expressed in SD units. SES heterogeneity was quantified as the difference in CATE between the low- and high-SES groups and assessed jointly across follow-up years using a Wald test. Robust 95% confidence intervals were obtained using the delta method. Interpretation of these standardized contrasts as causal effects requires conditional exchangeability, positivity, consistency, and correct specification of the outcome model (26).

As a sensitivity analysis, estimates were recalculated using wave-specific stabilized inverse-probability-of-observation weighting (IPOW) to reduce potential bias arising from differential missingness of follow-up psychological health outcomes (27). The probability of outcome observation at each follow-up wave was modeled using available observed information, and the resulting stabilized weights were truncated at the wave-specific 99th percentile to limit the influence of extreme weights.

All statistical tests were two-sided, and P < 0.05 was considered the threshold for statistical evidence. R software version 4.6.1 was used in this analysis.

3. Results

3.1. Participant characteristics

A total of 6,695 participants were included (Figure 1), of whom 2,129 (31.8%) were classified as low SES, 2,412 (36.0%) as middle SES, and 2,154 (32.2%) as high SES. Compared with participants in the high-SES group, those with low SES were less likely to have a spouse or partner and were more likely to smoke, abstain from alcohol, and have a higher BMI. They also had higher baseline CES-D scores and substantially greater FI burden. FI was present in 40.8% of participants with low SES, compared with 14.2% in the middle-SES group and 2.4% in the high-SES group (Table 1). Participants with greater FI severity were generally younger, more likely to be Hispanic or non-Hispanic Black, less likely to have a spouse or partner, more likely to smoke, and had a higher BMI. They also had markedly lower socioeconomic resources and poorer baseline psychological health (Supplementary Table S3).

Figure 1.

Flowchart outlining participant selection: the initial sample of 8,073 participants is reduced sequentially due to missing SES data (191 excluded), food insecurity questions (494 excluded), CES-D scores or follow-up (507 excluded), and key baseline covariates (186 excluded), yielding a final analytic sample of 6,695 participants.

Flowchart of participant selection. HCNS, Health Care and Nutrition Study; CES-D, Center for Epidemiologic Studies Depression Scale; SES, socioeconomic status.

Table 1.

Baseline characteristics of the analytic sample by socioeconomic status.

Characteristics Overall High SES Middle SES Low SES SMD
N 6,695 2,154 2,412 2,129
Age, mean (SD) 66.1 (10.7) 65.3 (9.9) 66.7 (10.8) 66.1 (11.2) 0.087
Sex, n (%)
Female 3,998(59.7) 1,161 (53.9) 1,423 (59.0) 1,414 (66.4) 0.172
Male 2,697(40.3) 993 (46.1) 989 (41.0) 715 (33.6)
Race/ethnicity, n (%)
Hispanic 759 (11.3) 95 (4.4) 215 (8.9) 449 (21.1) 0.603
Non-hispanic black 1,091 (16.3) 148 (6.9) 347 (14.4) 596 (28.0)
Non-hispanic white 4,624 (69.1) 1,844 (85.6) 1,771 (73.4) 1,009 (47.4)
Other 221 (3.3) 67 (3.1) 79 (3.3) 75 (3.5)
Spouse partner, n (%)
No 2,177 (32.5) 311 (14.4) 724 (30.0) 1,142 (53.6) 0.594
Yes 4,518 (67.5) 1,843 (85.6) 1,688 (70.0) 987 (46.4)
Current smoking, n (%)
No 5,859 (87.5) 2,040 (94.7) 2,144 (88.9) 1,675 (78.7) 0.326
Yes 836 (12.5) 114 (5.3) 268 (11.1) 454 (21.3)
Alcohol consumption, n (%)
No 2,900(43.3) 577 (26.8) 1,076 (44.6) 1,247 (58.6) 0.446
Yes 3,795(56.7) 1,577 (73.2) 1,336 (55.4) 882 (41.4)
BMI, kg/m2, mean (SD) 28.8 (6.0) 27.7 (5.2) 28.9 (6.0) 29.7 (6.7) 0.223
Self-rated health, n (%) 0.658
Poor 340 (5.1) 32 (1.4) 87 (3.6) 221 (10.4)
Fair 1,189 (17.8) 149 (6.9) 386 (16.0) 654 (30.7)
Good 2,175 (32.5) 604 (28.1) 844 (35.0) 727 (34.2)
Very good 2,283 (34.1) 999 (46.4) 863 (35.8) 421 (19.8)
Excellent 708 (10.6) 370 (17.2) 232 (9.6) 106 (5.0)
Chronic condition count 1.8 (1.3) 1.5 (1.2) 2.1 (1.4) 1.9 (1.3) 0.297
Health insurance, n (%) 0.238
None 2,579 (38.5) 988 (45.9) 928 (38.5) 663 (31.1)
Private 922 (13.8) 302 (14.0) 365 (15.1) 255 (12.0)
Public only 3,194 (47.8) 864 (40.1) 1,119 (46.4) 1,211 (56.9)
Wealth tertile, n (%)
Level 1 2,142 (32.0) 0 (0.0) 605 (25.1) 1,537 (72.2) 2.036
Level 2 2,232 (33.3) 423 (19.6) 1,243 (51.5) 566 (26.6)
Level 3 2,321 (34.7) 1,731 (80.4) 564 (23.4) 26 (1.2)
Income tertile, n (%) 2,097 (31.3) 0 (0.0) 433 (18.0) 1,664 (78.2) 2.283
Level 1
Level 2 2,219 (33.1) 398 (18.5) 1,374 (57.0) 447 (21.0)
Level 3 2,379 (35.5) 1,756 (81.5) 605 (25.1) 18 (0.8)
Educational attainment, n (%)
Low 945 (14.1) 0 (0.0) 107 (4.4) 838 (39.4) 1.415
Middle 2,338 (34.9) 264 (12.3) 1,096 (45.4) 978 (45.9)
High 3,412 (51.0) 1,890 (87.7) 1,209 (50.1) 313 (14.7)
High SES 2,154 (32.2) 2,154 (100.0) 0 (0.0) 0 (0.0)
Low SES 2,129 (31.8) 0 (0.0) 0 (0.0) 2,129 (100.0)
Middle SES 2,412 (36.0) 0 (0.0) 2,412 (100.0) 0 (0.0)
SES score (0–6), mean (SD) 3.4 (1.8) 5.5 (0.5) 3.5 (0.5) 1.3 (0.8) 4.628
CES-D score (0–8), mean (SD) 1.3 (1.9) 0.7 (1.3) 1.2 (1.8) 2.1 (2.2) 0.488
Food-insecurity score (0–6), mean (SD) 0.8 (1.7) 0.1 (0.7) 0.6 (1.5) 1.8 (2.2) 0.689
Food insecurity (binary), n (%)
No 5,433(81.2) 2,103 (97.6) 2,070 (85.8) 1,260 (59.2) 0.707
Yes 1,262(18.8) 51 (2.4) 342 (14.2) 869 (40.8)
Food insecurity (three levels), n (%)
Food secure/marginal 5,433(81.2) 2,103 (97.6) 2,070 (85.8) 1,260 (59.2) 0.707
Low food security 704 (10.5) 26 (1.2) 197 (8.2) 481 (22.6)
Very low food security 558 (8.3) 25 (1.2) 145 (6.0) 388 (18.2)

Values are presented as mean (SD) for continuous variables and n (%) for categorical variables. SMDs were used to quantify differences in baseline characteristics across SES groups. BMI, body mass index; CES-D, Center for Epidemiologic Studies Depression Scale; SD, standard deviation; SES, socioeconomic status; SMD, standardized mean difference; Z-score, standardized score.

Compared with participants included in the final analytic sample, excluded participants were older, had poorer self-rated health, a greater chronic disease burden, higher baseline CES-D scores, and were more socioeconomically disadvantaged. The prevalence of FI was nearly identical between excluded and included participants (18.9 vs. 18.8%; SMD <0.001), although the distribution of food-insecurity severity differed modestly (Supplementary Table S4).

3.2. Psychological health trajectories

Psychological health showed clear and persistent socioeconomic patterning across follow-up. Participants with low SES consistently had the lowest standardized psychological health scores (Figure 2A). Food-insecure participants consistently reported poorer psychological health than food-secure participants (Figure 2B).

Figure 2.

Two line graphs compare standardized CES-D depression scores from 2012 to 2020. Panel A shows low, middle, and high socioeconomic status (SES) groups with low SES consistently highest, high SES lowest, and all groups relatively stable. Panel B shows food secure and food insecure groups, with food insecure scores lower and a widening gap after 2014, highlighted by shaded confidence intervals.

Observed trajectories of psychological health by socioeconomic status and food insecurity. (A) Standardized psychological health scores from 2012 to 2020 among participants with low, middle, and high socioeconomic status (SES). (B) Standardized psychological health scores from 2012 to 2020 according to food security status assessed in 2013. Center for Epidemiologic Studies Depression Scale (CES-D) scores were standardized using the mean and standard deviation (SD) of the 2012 baseline sample and then multiplied by −1, such that higher scores indicate better psychological health. Shaded areas indicate 95% confidence intervals (CIs).

3.3. Longitudinal associations of SES and food insecurity with psychological health

Lower SES was consistently associated with a higher prevalence and greater severity of FI across alternative SES and FI measures (Supplementary Tables S5–S7). We subsequently examined the longitudinal associations of SES and FI with psychological health. In the fully adjusted models, a higher SES score was associated with better psychological health (β = 0.038; 95% confidence interval (CI), 0.025–0.050), whereas food insecurity was associated with poorer psychological health (β = −0.286; 95% CI, −0.336–−0.235). The SES advantage decreased over time (SES × time: β = −0.007; 95% CI, −0.011–−0.002), while the food-insecurity disadvantage narrowed over time (FI × time: β = 0.065; 95% CI, 0.044–0.085) (Table 2). The associations and time interactions remained materially unchanged after additional adjustment for baseline self-rated health, chronic disease burden, and health insurance coverage. These findings remained consistent when FI was modeled as a continuous severity score (Supplementary Table S8). Component-specific analyses showed poorer psychological health among participants with low wealth, income, or education; significant time interactions were observed for low wealth and low income, but not for low educational attainment (Supplementary Table S9).

Table 2.

Longitudinal associations of socioeconomic status and food insecurity with psychological health.

Variables Model 1 Model 2 Model 3 Model 4 Model 5
β (95% CI) P β (95% CI) P β (95% CI) P β (95% CI) P β (95% CI) P
Panel A. Socioeconomic status and psychological health
SES score, per 1-point increase 0.059 (0.050, 0.068) <0.001 0.049 (0.039, 0.060) <0.001 0.058 (0.046, 0.070) <0.001 0.029 (0.018, 0.039) <0.001 0.038 (0.025, 0.050) <0.001
Time, per 2-year wave −0.013 (−0.021), −0.005 0.002 −0.014 (−0.022, −0.006) <0.001 0.012 (−0.006, 0.030) 0.192 −0.015 (−0.023, −0.007) <0.001 0.010 (−0.008, 0.028) 0.270
SES score × time – – – – −0.007 (−0.012, −0.003) 0.002 – – −0.007 (−0.011, −0.002) 0.002
Panel B. food insecurity and psychological health
FI −0.265 (−0.306, −0.224) <0.001 −0.255 (−0.298, −0.211) <0.001 −0.337 (−0.388, −0.286) <0.001 −0.203 (−0.246, −0.160) <0.001 −0.286 (−0.336, −0.235) <0.001
Time, per 2-year wave −0.011 (−0.019, −0.003) 0.007 −0.013 (−0.021, −0.005) 0.001 −0.025 (−0.034, −0.016) <0.001 −0.014 (−0.023, −0.006) <0.001 −0.027 (−0.036, −0.018) <0.001
FI × time – – – – 0.065 (0.044, 0.085) <0.001 – – 0.065 (0.044, 0.085) <0.001

Values are β coefficients with 95% confidence intervals (CIs) estimated using linear mixed-effects models with participant-specific random intercepts. Model 1 included follow-up time, baseline CES-D score, and the exposure of interest. Model 2 additionally adjusted for age, sex, race/ethnicity, partnership status, current smoking, alcohol consumption, and body mass index. Model 3 additionally included the exposure × time interaction. Model 4 additionally adjusted for self-rated health, chronic disease burden, and health insurance coverage based on Model 2. Model 5 additionally included the exposure × time interaction based on Model 4. Higher standardized psychological health scores indicate better psychological health. Time was coded as 0, 1, 2, and 3 for the 2014, 2016, 2018, and 2020 waves, respectively; therefore, a 1-unit increase represents a 2-year interval. BMI, body mass index; CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; FI, food insecurity; SES, socioeconomic status; USDA, U.S. Department of Agriculture.

The adjusted trajectories showed that psychological health declined over time in the high- and middle-SES groups but remained relatively stable in the low-SES group (Figure 3A). Relative to the high-SES group, the 2-year change differed significantly in the low-SES group (β = 0.029, 95% CI 0.009–0.049; P = 0.005), but not in the middle-SES group (β = 0.014, 95% CI −0.005–0.033; P = 0.148). Participants with FI had poorer psychological health throughout follow-up, but the difference between food-insecure and food-secure participants narrowed over time (FI × time: β = 0.065 per 2-year follow-up wave, 95% CI 0.044–0.085; P < 0.001) (Figure 3B).

Figure 3.

Panel A displays a line graph showing adjusted psychological health scores from 2014 to 2020 for low, middle, and high socioeconomic status (SES) groups, with all scores declining and high SES remaining highest. Panel B shows adjusted psychological health scores over time comparing food secure to food insecure groups, with food insecurity associated with a steady increase and food security showing a gradual decline. Panel C presents three separate line graphs by SES—low, middle, and high—each depicting trends split by food security status, with food secure individuals consistently scoring higher, especially in high SES, and food insecurity associated with worsening or stagnant scores.

Adjusted trajectories of psychological health by socioeconomic status and food insecurity. Predicted mean trajectories of standardized psychological health scores from 2014 to 2020 were estimated using health-adjusted linear mixed-effects models. High SES, food-secure status, and the 2014 follow-up wave were used as the reference categories. Higher scores indicate better psychological health. (A) Trajectories by socioeconomic status (SES), illustrating the SES × time interaction. (B) Trajectories by food insecurity (FI) status, illustrating the FI × time interaction. (C) Trajectories by FI status within the low-, middle-, and high-SES groups, derived from the model including the SES × FI × time interaction. Models were adjusted for baseline CES-D score, age, sex, race/ethnicity, partnership status, smoking status, alcohol consumption, body mass index, self-rated health, chronic disease burden, and health insurance coverage. Shaded areas indicate 95% confidence intervals.

The overall SES × FI × time interaction was statistically significant (P = 0.014), indicating that the change over time in the FI-related psychological health difference varied across SES groups (Figure 3C). Specifically, this temporal change differed between the middle- and high-SES groups (β = 0.120, 95% CI 0.023–0.218; P = 0.016), but not between the low- and high-SES groups (β = 0.064, 95% CI −0.031–0.159; P = 0.185). Thus, the temporal pattern showed SES heterogeneity, but not a consistent gradient from low to high SES.

3.4. Counterfactual estimates of eliminating food insecurity

Counterfactual analyses showed that the PIB associated with eliminating economically driven barriers to adequate food access was consistently greatest among participants with low SES (Figure 4A). Averaged across follow-up, the PIB was 0.097 SD (95% CI 0.068–0.126) in the low-SES group, compared with 0.023 SD (95% CI 0.011–0.035) in the middle-SES group and 0.003 SD (95% CI −0.002–0.007) in the high-SES group (Figure 4C).

Figure 4.

Panel A shows a line graph of PIB in standard deviations for low, middle, and high SES groups from 2014 to 2020, with low SES consistently higher and decreasing over time. Panel B presents a line graph of CATE in standard deviations by SES and year, with variable values and wider confidence intervals. Panel C displays a forest plot of average PIB with confidence intervals for each SES group, showing higher values for low SES. Panel D provides a forest plot of differences in CATE by year with confidence intervals and p-values, indicating significant differences in some years.

Counterfactual estimates of psychological health under alternative food insecurity scenarios by socioeconomic status. (A) Year-specific population intervention benefit (PIB). (B) Year-specific conditional average treatment effect (CATE). (C) Average PIB across follow-up. (D) Difference in CATE between the low- and high-SES groups across follow-up years. Positive values indicate better psychological health. Points indicate point estimates, and error bars indicate 95% confidence intervals (CIs). Models were adjusted for baseline Center for Epidemiologic Studies Depression Scale (CES-D) score, age, sex, race/ethnicity, partnership status, smoking status, alcohol consumption, body mass index, self-rated health, chronic disease burden, and health insurance coverage.

SES-specific CATEs showed greater temporal variability (Figure 4B). The low- vs. high-SES difference in CATE was statistically significant in 2014 (0.288 SD, 95% CI 0.057–0.519; P = 0.015), but not at later waves. Nevertheless, the overall test across the four follow-up waves indicated heterogeneity in the low- vs. high-SES CATE difference over time (Wald P = 0.024) (Figure 4D). The IPOW sensitivity analysis showed that the findings were robust (Supplementary Figure S1). These findings suggest that the larger population-level benefit observed in the low-SES group was driven primarily by the greater burden of FI in this group, with less consistent evidence that the FI-related psychological health contrast itself was uniformly stronger across SES groups.

4. Discussion

In this longitudinal study of middle-aged and older adults, FI was strongly concentrated among participants with lower SES and was consistently associated with poorer psychological health. Higher SES was also associated with better psychological health. The overall SES × FI × time interaction was statistically significant, indicating heterogeneity in the temporal pattern of the FI–psychological health association across SES groups, although this heterogeneity did not follow a consistent socioeconomic gradient. Counterfactual analyses further showed that the PIB associated with eliminating economically driven barriers to adequate food access was consistently greatest among participants with low SES. SES-specific CATEs also showed evidence of heterogeneity, with a larger FI-related psychological health contrast in the low-SES group at some follow-up waves, although this pattern was not consistent over time.

The pronounced socioeconomic patterning of FI is consistent with the close dependence of food access on household economic resources (6–8, 28). Income determines the resources immediately available for routine food purchases, whereas wealth provides a financial buffer against income interruptions, medical expenses, and other unexpected costs that may compete with food expenditure (29, 30). Educational attainment may influence FI through longer-term socioeconomic pathways, including employment opportunities, earnings potential, health literacy, and access to social resources. Consistent with these distinct roles, the component-specific analyses showed that household wealth and income were strongly and consistently associated with FI, whereas the association with education was weaker after mutual adjustment. This heterogeneity across socioeconomic dimensions supports interpreting the composite SES measure as a summary of multidimensional socioeconomic circumstances rather than as a set of interchangeable components. More broadly, the concentration of FI may be associated with poorer psychological health through both nutritional and psychosocial mechanisms. Limited economic resources can constrain diet quality, disrupt regular eating patterns, and lead individuals to reduce portion sizes or skip meals, which may be particularly consequential for older adults managing chronic conditions that require stable food intake or specific dietary patterns (30, 31). Beyond nutritional consequences, uncertainty about whether food will last, repeated trade-offs between food and other essential expenses, and difficulty obtaining adequate meals may generate persistent stress, shame, uncertainty, and loss of control. Systematic reviews have consistently linked FI with depression, anxiety, psychological distress, and poorer mental wellbeing, including among older populations (32, 33), while qualitative evidence highlights the continuing financial and logistical burden of managing inadequate food resources (28).

The narrowing of SES- and FI-related differences over follow-up should not be interpreted as a straightforward improvement among disadvantaged groups. In the adjusted trajectories, psychological health declined in the high- and middle-SES groups but remained relatively stable in the low-SES group, while the difference between food-insecure and food-secure participants also narrowed over time. Similar convergence in socioeconomic differences at older ages has been reported previously and may partly reflect aging-related changes in health, retirement, income sources, social protection, or selective survival (34). For FI, however, interpretation of the temporal pattern requires additional caution because FI was assessed only once in 2013. Changes in FI status during follow-up may therefore have progressively weakened the correspondence between the 2013 exposure classification and participants' later circumstances. Previous longitudinal evidence suggests that changes in FI severity can accompany changes in psychological outcomes (35), so the observed narrowing may reflect both genuine temporal change and increasing discordance between baseline FI classification and subsequent FI status. The significant overall SES × FI × time interaction indicated that these temporal changes were not identical across SES groups; however, the difference was evident for middle vs. high SES but not for low vs. high SES. Thus, the data support SES heterogeneity in the temporal FI–psychological health association, but not a consistent socioeconomic gradient.

The counterfactual analyses help distinguish two related but conceptually different sources of the greater FI-related psychological health burden in lower-SES populations. The PIB reflects the expected population-level improvement associated with eliminating economically driven barriers to adequate food access from the observed distribution of FI; it therefore depends on both the prevalence of FI and the magnitude of the FI-related psychological health contrast. The consistently larger PIB in the low-SES group is thus consistent with the substantial concentration of FI in this population. By contrast, the CATE compares predicted psychological health under universal food security and universal food insecurity within each SES group and is less directly influenced by the observed prevalence of FI. CATEs showed evidence of SES heterogeneity, but the larger FI-related psychological health contrast in the low-SES group was not consistently observed across follow-up waves. Taken together, these findings suggest that the greater potential population-level benefit among participants with low SES was driven primarily by their greater burden of FI, with more limited and temporally variable evidence for differential vulnerability to FI across SES groups.

From a policy perspective, the counterfactual scenario is best understood as reducing economically driven barriers to adequate food access rather than eliminating all possible causes of FI. Such barriers may be addressed through measures that increase households' effective food purchasing power or directly improve access to food, including cash-like nutrition assistance, food subsidies or vouchers, and direct food provision. However, these approaches are not equivalent in their effects on FI, diet quality, or psychological health (36, 37). Accordingly, our counterfactual estimates should be interpreted as the potential psychological health benefit associated with achieving food security under the modeled scenario, rather than as the expected effect of any specific program.

This study has several strengths. First, the longitudinal design with repeated psychological health assessments allowed us to characterize socioeconomic differences in psychological health trajectories while also quantifying the marked socioeconomic patterning of FI. Second, the counterfactual analysis extended beyond conventional association estimates by quantifying the potential psychological health benefit associated with eliminating economically driven barriers to adequate food access within each SES group. Third, by jointly examining the PIB and CATE, we were able to distinguish whether the greater potential benefit among lower-SES groups was attributable primarily to their greater exposure to FI, to a larger FI-related psychological health contrast, or to both. Sensitivity analyses addressing baseline health adjustment and differential follow-up missingness further supported the robustness of these findings.

Several limitations should be considered. First, because this was an observational study, the counterfactual estimates rely on assumptions of conditional exchangeability, positivity, consistency, and correct specification of the outcome model. Residual or unmeasured confounding cannot be excluded, and potential bidirectionality between FI and psychological health may remain despite adjustment for prior psychological health. In addition, because FI was assessed only once, time-varying confounding could not be fully addressed; the counterfactual estimates should therefore be interpreted as standardized contrasts under the specified assumptions rather than as definitive effects of an intervention. Second, FI was measured only in 2013 and could not distinguish transient from persistent FI. Changes in FI status during follow-up may have led to increasing discordance between baseline FI classification and subsequent FI status, particularly at later waves. Third, the composite SES measure assigned equal weights to wealth, income, and educational attainment. Although component-specific analyses showed that the main socioeconomic patterning was not attributable to a single dimension, the composite score may still obscure heterogeneity across socioeconomic domains. Fourth, participants excluded because of missing data were generally older, more socioeconomically disadvantaged, and in poorer health than those included, which may have led to underrepresentation of more disadvantaged individuals. Moreover, because no HCNS-specific survey weight accounting jointly for HCNS subsampling and questionnaire nonresponse was available, the findings should be interpreted as pertaining primarily to the analytic HCNS sample rather than as nationally weighted estimates for the U.S. population of older adults. IPOW addressed missing psychological health outcomes during follow-up but did not correct for the original sampling design or selection into the HCNS respondent sample. Finally, psychological health was assessed using a self-reported CES-D symptom scale rather than a clinical diagnosis.

5. Conclusion

FI was strongly concentrated among middle-aged and older adults with lower SES and was consistently associated with poorer psychological health. Counterfactual analyses suggested that reducing economically driven barriers to adequate food access could yield the greatest population-level benefit among adults with low SES, driven mainly by their higher burden of FI rather than consistently stronger FI-related psychological health contrasts. These findings highlight food security as a potentially important target for addressing socioeconomic differences in psychological health.

Acknowledgments

We thank the Health and Retirement Study (HRS) for making the data used in this study available.

Funding Statement

The author(s) declared that financial support was received for this work and/or its publication. The HRS is sponsored by the National Institute on Aging (NIA; U01AG009740) and conducted by the University of Michigan. This funding supported the parent HRS study and did not constitute funding for the present secondary analysis. This work was supported by the Key R&D Program of the Shaanxi Provincial Health Commission (Grant No. 2026YF-46) and the Key Project of the Shaanxi Provincial Key R&D Program (Grant No. 2026SF-ZDGG-60). The funders had no role in the study design, data collection, analysis or interpretation, manuscript preparation, or the decision to submit the manuscript for publication.

Edited by: Jia Zhao, The University of Hong Kong, Hong Kong SAR, China

Reviewed by: Fawzia Zahidi, Charité University Medicine Berlin, Germany

Yajie Xiong, University of Notre Dame, United States

Abbreviations: BMI, body mass index; CATE, conditional average treatment effect; CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; FI, food insecurity; GEE, generalized estimating equations; HCNS, Health Care and Nutrition Study; HRS, Health and Retirement Study; IPOW, inverse-probability-of-observation weighting; LMM, linear mixed-effects model; PIB, population intervention benefit; PR, prevalence ratio; SD, standard deviation; SES, socioeconomic status; SMD, standardized mean difference; USDA, U.S. Department of Agriculture.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement

Ethical review and approval were not required for the present study in accordance with the applicable institutional requirements because it involved a secondary analysis of de-identified data. The original Health and Retirement Study data collection was approved by the University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board, and participants provided informed consent. No additional informed consent was required for the present secondary analysis. The study was conducted in accordance with the local legislation and institutional requirements.

Author contributions

SF: Writing – original draft, Conceptualization, Methodology. TW: Writing – original draft, Conceptualization, Formal analysis, Methodology. SZ: Writing – original draft, Formal analysis. HZ: Supervision, Writing – review & editing. JL: Writing – review & editing, Supervision.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2026.1958938/full#supplementary-material

Data_Sheet_1.docx (177KB, docx)

References

  • 1.Patel V, Saxena S, Lund C, Thornicroft G, Baingana F, Bolton P, et al. The Lancet Commission on global mental health and sustainable development (Erratum in: Lancet. 2018 Oct 27;392(10157):1518. doi: 10.1016/S0140-6736(18)32624-2). Lancet. (2018) 392:1553–98. doi: 10.1016/S0140-6736(18)31612-X [DOI] [PubMed] [Google Scholar]
  • 2.Lund C, Brooke-Sumner C, Baingana F, Baron EC, Breuer E, Chandra P, et al. Social determinants of mental disorders and the Sustainable Development Goals: a systematic review of reviews. Lancet Psychiatry. (2018) 5:357–69. doi: 10.1016/S2215-0366(18)30060-9 [DOI] [PubMed] [Google Scholar]
  • 3.Jespersen A, Madden RA, Whalley HC, Reynolds RM, Lawrie SM, McIntosh AM, et al. Socioeconomic status and depression-a systematic review. Epidemiol Rev. (2025) 47:mxaf011. doi: 10.1093/epirev/mxaf011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Kosciuszko M, Steptoe A, Ajnakina O. Genetic propensity, socioeconomic status, and trajectories of depression over a course of 14 years in older adults. Transl Psychiatry. (2023) 13:68. doi: 10.1038/s41398-023-02367-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.U.S. Department of Agriculture, Economic Research Service. Food Security in the U.S.—Measurement. Updated March 30, 2026. (Accessed September 30, 2026). [Google Scholar]
  • 6.Walker RJ, Garacci E, Dawson AZ, Williams JS, Ozieh M, Egede LE. Trends in food insecurity in the United States from 2011–2017: disparities by age, sex, race/ethnicity, and income. Popul Health Manag. (2021) 24:496–501. doi: 10.1089/pop.2020.0123 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Inoue M, Masa R, Prieto L, Baruah D, Kellermeyer K, Booker E, et al. Prevalence and Correlates of Food Insecurity Among Older Adults in the United States. J Gerontol Soc Work. (2024) 67:738–55. doi: 10.1080/01634372.2024.2339975 [DOI] [PubMed] [Google Scholar]
  • 8.Arenas DJ, Thomas A, Wang J, DeLisser HM. A systematic review and meta-analysis of depression, anxiety, and sleep disorders in US adults with food insecurity. J Gen Intern Med. (2019) 34:2874–82. doi: 10.1007/s11606-019-05202-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Bui CN, Peng C, Mutchler JE, Burr JA. Race and ethnic group disparities in emotional distress among older adults during the COVID-19 pandemic. Gerontologist. (2021) 61:262–72. doi: 10.1093/geront/gnaa217 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Militao EMA, Uthman OA, Salvador EM, Vinberg S, Macassa G. Association between socioeconomic position of the household head, food insecurity and psychological health: an application of propensity score matching. BMC Public Health. (2024) 24:2590. doi: 10.1186/s12889-024-20153-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Pourmotabbed A, Moradi S, Babaei A, Ghavami A, Mohammadi H, Jalili C, et al. Food insecurity and mental health: a systematic review and meta-analysis (Erratum in: Public Health Nutr. 2020 Jul;23(10):1854. doi: 10.1017/S1368980020001512). Public Health Nutr. (2020) 23:1778–90. doi: 10.1017/S136898001900435X [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Osei-Owusu C, Dhillon S, Luginaah I. The impact of food insecurity on mental health among older adults residing in low- and middle-income countries: a systematic review. PLoS ONE. (2024) 19:e0301046. doi: 10.1371/journal.pone.0301046 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Bergmans RS, Wegryn-Jones R. Examining associations of food insecurity with major depression among older adults in the wake of the Great Recession. Soc Sci Med. (2020) 258:113033. doi: 10.1016/j.socscimed.2020.113033 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JWR, Weir DR. Cohort profile: the health and retirement study (HRS). Int J Epidemiol. (2014) 43:576–85. doi: 10.1093/ije/dyu067 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Health and Retirement Study. 2013 Health Care and Nutrition Study (HCNS) (Final V5.0) [Data set]. Ann Arbor, MI: University of Michigan; 2018. [Google Scholar]
  • 16.Braveman PA, Cubbin C, Egerter S, Chideya S, Marchi KS, Metzler M, et al. Socioeconomic status in health research: one size does not fit all. JAMA. (2005) 294:2879–88. doi: 10.1001/jama.294.22.2879 [DOI] [PubMed] [Google Scholar]
  • 17.Lindberg MH, Chen G, Olsen JA, Abelsen B. Combining education and income into a socioeconomic position score for use in studies of health inequalities. BMC Public Health. (2022) 22:969. doi: 10.1186/s12889-022-13366-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Greco S, Ishizaka A, Tasiou M, Torrisi G. On the methodological framework of composite indices: a review of the issues of weighting, aggregation, and robustness. Soc Indic Res. (2019) 141:61–94. doi: 10.1007/s11205-017-1832-9 [DOI] [Google Scholar]
  • 19.Blumberg SJ, Bialostosky K, Hamilton WL, Briefel RR. The effectiveness of a short form of the household food security scale. Am J Public Health. (1999) 89:1231–4. doi: 10.2105/AJPH.89.8.1231 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Steffick DE. Documentation of Affective Functioning Measures in the Health and Retirement Study. HRS/AHEAD Documentation Report DR-005 Ann Arbor, MI: Institute for Social Research, University of Michigan. (2000). [Google Scholar]
  • 21.Turvey CL, Wallace RB, Herzog R. A revised CES-D measure of depressive symptoms and a DSM-based measure of major depressive episodes in the elderly. Int Psychogeriatr. (1999) 11:139–48. doi: 10.1017/S1041610299005694 [DOI] [PubMed] [Google Scholar]
  • 22.Ma Y, Liang L, Zheng F, Shi L, Zhong B, Xie W. Association between sleep duration and cognitive decline. JAMA Netw Open. (2020) 3:e2013573. doi: 10.1001/jamanetworkopen.2020.13573 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Laird NM, Ware JH. Random-effects models for longitudinal data. Biometrics. (1982) 38:963–74. doi: 10.2307/2529876 [DOI] [PubMed] [Google Scholar]
  • 24.Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol. (2004) 159:702–6. doi: 10.1093/aje/kwh090 [DOI] [PubMed] [Google Scholar]
  • 25.Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. (1986) 73:13–22. doi: 10.1093/biomet/73.1.13 [DOI] [Google Scholar]
  • 26.Hernán MA, Robins JM. Causal Inference: What If . Boca Raton, FL: Chapman & Hall/CRC. (2020). [Google Scholar]
  • 27.Seaman SR, White IR. Review of inverse probability weighting for dealing with missing data. Stat Methods Med Res. (2013) 22:278–95. doi: 10.1177/0962280210395740 [DOI] [PubMed] [Google Scholar]
  • 28.Gundersen C, Ziliak JP. Food insecurity and health outcomes. Health Aff. (2015) 34:1830–9. doi: 10.1377/hlthaff.2015.0645 [DOI] [PubMed] [Google Scholar]
  • 29.Leung CW, Wolfson JA. Food insecurity among older adults: 10-year national trends and associations with diet quality. J Am Geriatr Soc. (2021) 69:964–71. doi: 10.1111/jgs.16971 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Ridley M, Rao G, Schilbach F, Patel V. Poverty, depression, and anxiety: causal evidence and mechanisms. Science. (2020) 370:eaay0214. doi: 10.1126/science.aay0214 [DOI] [PubMed] [Google Scholar]
  • 31.Kirkbride JB, Anglin DM, Colman I, Dykxhoorn J, Jones PB, Patalay P, et al. The social determinants of mental health and disorder: evidence, prevention and recommendations. World Psychiatry. (2024) 23:58–90. doi: 10.1002/wps.21160 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Selvamani Y, Arokiasamy P. Association between food insecurity and perceived stress among older adults (50+) in six low- and middle-income countries. Aging Ment Health. (2022) 26:2339–47. doi: 10.1080/13607863.2021.1985965 [DOI] [PubMed] [Google Scholar]
  • 33.Mavegam Tango Assoumou BO, Coughenour C, Godbole A, McDonough I. Senior food insecurity in the USA: a systematic literature review. Public Health Nutr. (2023) 26:229–45. doi: 10.1017/S1368980022002415 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Bergmans RS, Zivin K, Mezuk B. Depression, food insecurity and diabetic morbidity: evidence from the health and retirement study. J Psychosom Res. (2019) 117:22–9. doi: 10.1016/j.jpsychores.2018.12.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Leung CW, Laraia BA, Feiner C, Solis K, Stewart AL, Adler NE, et al. The psychological distress of food insecurity: a qualitative study of the emotional experiences of parents and their coping strategies. J Acad Nutr Diet. (2022) 122:1903–1910.e2. doi: 10.1016/j.jand.2022.05.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Beckett M. Converging health inequalities in later life–an artifact of mortality selection. J Health Soc Behav. (2000) 41:106–19. doi: 10.2307/2676363 [DOI] [PubMed] [Google Scholar]
  • 37.Berkowitz SA, Palakshappa D, Seligman HK, Hanmer J. Changes in food insecurity and changes in patient-reported outcomes: a nationally representative cohort study. J Gen Intern Med. (2022) 37:3638–44. doi: 10.1007/s11606-021-07293-4 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data_Sheet_1.docx (177KB, docx)

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.


Articles from Frontiers in Nutrition are provided here courtesy of Frontiers Media SA

RESOURCES