Household Food Insufficiency, Financial Strain, Work–Family Spillover, and Depressive Symptoms in the Working Class: The Work, Family, and Health Network Study

Abstract

Objectives. We evaluated the association of household-level stressors with depressive symptoms among low-wage nursing home employees.

Methods. Data were collected in 2006 and 2007 from 452 multiethnic primary and nonprimary wage earners in 4 facilities in Massachusetts. We used logistic regression to estimate the association of depressive symptoms with household financial strain, food insufficiency, and work–family spillover (preoccupation with work-related concerns while at home and vice versa).

Results. Depressive symptoms were significantly associated with household financial strain (odds ratio [OR] = 1.82; 95% confidence interval [CI] = 1.03, 3.21) and food insufficiency (OR = 2.10; 95% CI = 1.10, 4.18). Among primary earners, stratified analyses showed that food insufficiency was associated with depressive symptoms (OR = 3.60; 95% CI = 1.42, 9.11) but financial strain was not. Among nonprimary wage earners, depressive symptoms correlated with financial strain (OR = 3.65; 95% CI = 1.48, 9.01) and work–family spillover (OR = 3.22; 95% CI = 1.11, 9.35).

Conclusions. Household financial strain, food insufficiency, and work–family spillover are pervasive problems for working populations, but associations vary by primary wage earner status. The prevalence of food insufficiency among full-time employees was striking and might have a detrimental influence on depressive symptoms and the health of working-class families.

Depression is among the most commonly experienced disorders and is a leading cause of disability worldwide.¹ Increasing evidence suggests that depression is a leading cause of sickness-related absence in the labor force and is a concern for employers and employees alike.² Women, people in lower socioeconomic positions, and racial/ethnic minorities have higher rates of depression than the general population, whether they are in the workforce or not.³ Some evidence suggests, however, that the higher prevalence of depression observed in disadvantaged groups may stem from stressors associated with their common experiences rather than from race/ethnicity, income, or gender per se.^3–5

Several studies have established that the presence of persistent negative and stressful experiences may lead to depression.^6–9 Evidence shows that work-related strain, specifically job strain (high demand/low control) and emotional strain, is associated with depression, especially among caregiving workers.^7–9 The work–family literature suggests that work–family spillover (preoccupation with work impinging on home life or preoccupation with personal responsibilities impinging on work) may be correlated with depression.¹⁰ Fewer studies, however, have explored the contribution of both household- and work-related stressors to depression.

The inability to provide for one's family despite working full time could be a significant source of stress.¹¹ Indeed, household financial strain and food insufficiency (sometimes or often not having enough food to eat) are considered particularly stressful,^12–15 especially for low-income populations.^16,17 Studies have found separate associations between financial strain and food insufficiency and adverse mental health outcomes.^12,17–22 However, these studies concentrated on populations with obvious disadvantages, such as the unemployed and elderly people with disabilities. Thus, the extent and deleterious effects of both household financial strain and food insufficiency—along with work–family spillover—on working-class households have not been fully examined.

Nursing home workers are a growing part of the workforce who may face higher rates of household food insufficiency, financial strain, and work–family spillover.^23,24 Among nursing assistants—the biggest work group in nursing homes, and among the lowest paid—the proportion of women is estimated to be 80% to 90%; most are single mothers and are thus the primary wage earners for their families.^23,25 Nursing home workers are more likely to be recent immigrants who may not be aware of or eligible for government benefits.²⁴ A majority of these low-wage earners are also members of racial/ethnic minority groups.²⁵

Research has not fully explored the relationship between household- and work-related stressors and mental health outcomes, particularly among working-class households. We examined work and home conditions—household financial strain, food insufficiency, and work–family spillover—associated with depressive symptoms in a low-wage, multiethnic group of women and men employed in the long-term care industry. Because previous studies did not examine these factors jointly, our first objective was to assess the multivariate association of each exposure with depressive symptoms. We then determined whether the association between these exposures remained in a model incorporating all 3 main variables. We hypothesized that depressive symptoms associated with these variables would be more significant and stronger for people who were the primary income providers for their households than for nonprimary wage earners.

METHODS

Our study sample was from the Boston site of the Work, Family, and Health Network study, which was designed to investigate work and family demands and experiences of lower-wage and racially/ethnically diverse workforces in 4 nursing homes in Massachusetts. Employees at each nursing home (n = 590) were invited to participate in the survey, and 76.6% (n = 452) responded. We excluded 36 people who were missing main outcomes or predictors for a sample size of 416. Trained research assistants administered a survey in English, Spanish, or Haitian Creole during employees’ work shifts from September 2006 to July 2007. Interviews took approximately 40 minutes, and participants received prepaid debit cards as incentives.

Measures

Outcome.

We assessed depressive symptoms with a shortened 11-item version of the Center for Epidemiologic Studies Depression Scale (CES-D), which assesses past-week depressive symptomatology.²⁶ Items retained in the shortened form represent the same major symptom components captured by the 20-item CES-D.²⁷ The short-form items have demonstrated reliability and correlated favorably (r = 0.88–0.95) with scores from the original CES-D among diverse samples.^27,28 Because a cutoff of 10 was established for the CES-D10,²⁹ we computed the interitem correlation for the 11 items and eliminated the item with the lowest interitem correlation. Internal consistency reliability of the scale within our population was α = 0.781. We then evaluated depressive symptomatology as a dichotomous response in our analyses, and we considered a score of 10 or higher indicative of depressive symptomatology.

Primary predictors.

We assessed household financial strain with a question (validated to measure subjective adequacy of household resources) adapted from the Established Populations for the Epidemiologic Study of the Elderly studies^30,31: “How would you describe the money situation in your household right now?” Answer options were “comfortable with extra,” “enough but no extra,” “have to cut back,” and “cannot make ends meet.” We created a dichotomous measure that found financial strain if a respondent chose either of the last 2 options.

We assessed household food insufficiency by asking, “In the past 12 months, how often has the following statement been true in your household: The food we bought ran out and we didn't have money to get more.” Answer options were “never true,” “sometimes true,” and “often true.” In accordance with studies that have validated this question,¹³ we categorized respondents as experiencing food insufficiency unless they chose the first response.

We assessed work–family spillover in the past month with 2 questions about the frequency of being preoccupied with work at home or with personal responsibilities at work. We considered respondents as experiencing work–family spillover if they responded to either question with “often.”

Potential confounders.

We measured workplace social support with a scale created by summing values of 4 separate items that asked how often respondents received support from their colleagues and supervisors (separately) and how often individuals within these categories were willing to listen to the respondents’ work-related problems (α = 0.75). Response categories for each item ranged from 0 to 3, allowing an index range of 0 to 12. Job strain was derived from measures evaluating job control and job demands.³² We evaluated the distribution of job control and job demands across our sample, and we categorized respondents as having high-strain jobs if their job demand versus control score was less than the median. Two questions assessed pain in the past 4 weeks. We created dichotomous categories, with 1 indicating that pain occurred “very often” or “almost every day.”

We included age as a continuous variable in the models, centered at the mean value. Marital status was dichotomous: categories were currently married or living with a partner and single or divorced. Nativity indicated the country where respondents were born and was modeled as a dichotomous variable of US-born versus foreign-born. Gender indicated whether the respondent was male or female. We stratified education into 4 categories: less than high school, high school or general equivalency diploma, more than high school (including some college or an associate's degree), and college graduate. Race indicated respondents' self-categorization as American Indian, Asian, Black, Hawaiian/Pacific Islander, White, or other. Respondents could choose multiple categories. For analysis, we further aggregated race into non-Hispanic White, non-Hispanic Black, and Hispanic or non-Hispanic other. We used indicator variables for these categories in our model.

Income gauged yearly household income from all sources, in categories from 1 (< $10 000) to 9 (≥ $80 000). We also asked respondents how many people were currently supported by this income, even if they did not live in the same household. Distribution of sociodemographic and outcome variables did not differ by missing income data (n = 50). We imputed household income for those whose income information was missing. We computed each of the missing values as the median income of its site and occupational group. To calculate an adjusted income value, we divided the median values of each range by the square root of the number of people the respondent reported as supported by that income.³³ We then calculated tertiles and compared low and middle income to high income with indicator variables for analysis.

Respondents were asked, “Are you the primary wage earner in your family?” We coded those who answered yes as primary wage earners and those who answered no or equal as nonprimary wage earners. We asked respondents about the number and ages of children in their household and coded this information dichotomously for analysis indicating whether any children younger than 19 years lived with the respondent.

Statistical Analysis

We conducted analyses with SAS version 9.2 (SAS Institute Inc, Cary, NC). We first examined the univariate descriptive distribution of covariates of interest; then we compared them by depressive symptom status with χ² statistics for categorical variables and t tests for normal continuous variables. In building models, we considered demographic variables that have been empirically or theoretically associated with depressive symptoms. In addition, we controlled for pain and job strain, which might be associated with predictors and have been shown to be associated with depressive symptoms.^8,34

Model building began with analysis of the crude association of predictor and control variables with depressive symptoms. Next, we examined associations of each of the main predictors, with controls for demographic variables and potential confounders. Final models for the overall population incorporated all predictors and controlled for potential confounders. To detect differences in the association of the predictors with depressive symptoms by primary wage earner status, we stratified the final models by wage earner status. Because of potential within-cluster (worksite) correlation caused by recruitment strategy, all regression analyses controlled for the random effect of worksite and were conducted with SAS GLIMMIX, since that is the preferred modeling technique when the number of clusters is small.³⁵

RESULTS

Associations of sociodemographic characteristics with depressive symptoms are shown in Table 1. We found a 26% prevalence of depressive symptoms among the study participants. The highest prevalence (52.2%) occurred in the 16% of households classified as food insufficient, followed by the 31.5% of households that experienced financial strain (41.2% prevalence). Almost 60% of participants who reported that they could not make ends meet and approximately 32% of those who had to cut back had depressive symptoms. Nearly half (49.1%) of respondents who reported sometimes running out of food and two thirds (66.7%) who reported often running out of food had depressive symptoms.

TABLE 1—

Sociodemographic Characteristics of Participants: Work, Family, and Health Network Study, Massachusetts, 2006–2007

Characteristic	No. (%) or Mean ±SD	Depressive Symptoms, %	P
Overall	416 (100.0)	26.0
Gender
Male	71 (17.1)	18.3	.106
Female	345 (82.9)	27.5
Age, y
< 25	56 (13.5)	30.4	.096
26–35	96 (23.1)	25.0
36–45	117 (28.1)	33.3
46–55	87 (20.9)	20.7
≥ 56	60 (14.4)	16.7
Marital status
Single/divorced	182 (43.8)	28.6	.284
Married/living with partner	234 (56.3)	23.9
Race/ethnicity
White	185 (44.5)	18.9	.008
Black	143 (34.4)	35.0
Hispanic	31 (7.5)	32.3
Other	57 (13.7)	22.8
Education
< high school	52 (12.5)	23.1	.082
High school or GED	126 (30.3)	33.3
Some college or 2-y degree	168 (40.4)	25.0
≥ college graduate	70 (16.8)	17.1
Adjusted household income, $
Low (< 21 000)	140 (33.7)	37.1	<.001
Middle (21 000–37 500)	147 (35.3)	22.5
High (> 37 500)	129 (31.0)	17.8
Immigration status
US-born	195 (46.9)	21.0	.031
Foreign-born	221 (53.1)	30.3
Financial strain
Comfortable with extra	139 (33.4)	15.8	< .001
Enough but no extra	146 (35.1)	21.9
Have to cut back	85 (20.4)	31.8
Cannot make ends meet	46 (11.1)	58.7
Food insufficiency
Never	349 (83.9)	20.9	< .001
Sometimes	55 (13.2)	49.1
Often	12 (2.9)	66.7
Work–family spillover
Yes	85 (20.4)	38.8	.002
No	331 (79.6)	22.7
Frequent pain
Yes	61 (14.7)	37.7	.024
No	355 (85.3)	23.9
High job strain
Yes	101 (24.3)	37.6	.002
No	315 (75.7)	22.2
Child aged < 19 y in household
Yes	199 (47.8)	21.6	.052
No	217 (52.2)	30.0
Primary wage earner
No	187 (45.0)	24.1	.425
Yes	229 (55.1)	27.5
Social support at job
Overall	9.87 ±2.41
Depressed	10.22 ±2.15		< .001
Not depressed	8.87 ±2.83

Note. GED = general equivalency diploma.

We found much higher prevalence of financial strain (40% vs 20%) and food insufficiency (23% vs 7%) among primary earners than among nonprimary wage earners (Table 2). It is noteworthy that almost half of the primary wage earners who reported financial strain also reported food insufficiency. In fact, all 3 variables with a financial component (food insufficiency, financial strain, and household income) were significantly correlated with one another. These variables were more highly correlated among primary wage earners; the correlation coefficient was 0.54 between financial strain and food insecurity, 0.42 between financial strain and household income, and 0.36 between food insecurity and household income; all with P values of less than .001. Among nonprimary wage earners, the coefficient was 0.29 between financial strain and food insecurity, 0.35 between financial strain and household income, and 0.25 between food insecurity and household income (all P < .001). Primary wage earners were also more likely than nonprimary earners to be single, to belong to a minority race/ethnicity, to have less than a high school education, and to have a low household income.

TABLE 2—

Characteristics of Primary and Nonprimary Wage Earners: Work, Family, and Health Network Study, Massachusetts, 2006–2007

Characteristic	Primary Wage Earner (n = 229), No. (%)	Nonprimary Wage Earner (n = 187), No. (%)
Food insufficiency and financial strain
Both food insufficiency and financial strain	44 (19.21)	8 (4.28)
Food insufficiency only	9 (3.93)	6 (3.21)
Financial strain only	48 (20.96)	31 (16.58)
Neither food insufficiency nor financial strain	128 (55.9)	142 (75.94)
Gender
Male	36 (15.72)	35 (18.72)
Female	193 (84.28)	152 (81.28)
Marital status
Single/divorced	110 (48.03)	72 (38.5)
Married/living with partner	119 (51.97)	115 (61.5)
Race/ethnicity
White	91 (39.74)	94 (50.27)
Black	86 (37.55)	57 (30.48)
Hispanic	21 (9.17)	10 (5.35)
Other	31 (13.54)	26 (13.9)
Education
< high school	37 (16.16)	15 (8.02)
High school or GED	67 (29.26)	59 (31.55)
Some college or 2-y degree	90 (39.3)	78 (41.71)
≥ college graduate	35 (15.28)	35 (18.72)
Adjusted household income, $
Low (< 21 000)	98 (42.97)	42 (22.46)
Middle (21 000–37 500)	81 (35.37)	66 (35.29)
High (> 37 500)	50 (21.83)	79 (42.25)
Immigration status
US-born	99 (43.23)	96 (51.34)
Foreign-born	130 (56.77)	91 (48.66)
Child aged < 19 y in household
Yes	117 (51.09)	100 (53.48)
No	112 (48.91)	87 (46.52)

Note. GED = general equivalency diploma.

In both the crude and adjusted models (Table 3), food insufficiency was associated with increased odds of depressive symptoms. In the multivariable model, which adjusted for sociodemographic variables, pain, and job strain but not for financial strain and work–family spillover, participants who reported food insufficiency had almost 3 times the odds of having depressive symptoms compared with those who did not report food insufficiency (odds ratio [OR] = 2.78; 95% confidence interval [CI] = 1.47, 5.27). Financial strain was also associated with depressive symptoms in crude and adjusted models. Participants who reported financial strain had more than twice the adjusted odds of depressive symptoms compared with those who did not report financial strain (OR = 2.34; 95% CI = 1.38, 3.96; model 3). Likewise, work–family spillover was associated with depressive symptoms in both crude and adjusted models; respondents who reported work–family spillover had nearly twice the adjusted odds of depressive symptoms compared with those who did not (OR = 1.78; 95% CI = 1.01, 3.12; model 4).

TABLE 3—

Relationship of Food Insufficiency, Financial Insufficiency, and Work–Family Spillover to Depressive Symptoms in the Overall Population: Work, Family, and Health Network Study, Massachusetts, 2006–2007

Predictor	Model 1, OR (95% CI)	Model 2, OR (95% CI)	Model 3, OR (95% CI)	Model 4, OR (95% CI)	Model 5, OR (95% CI)
Food insufficiency					2.10 (1.10, 4.18)
Yes	3.88 (2.25, 6.68)	2.78 (1.47, 5.27)
No (Ref)	1.00	1.00			1.00
Financial strain
Yes	2.94 (1.87, 4.67)		2.34 (1.38, 3.96)		1.82 (1.03, 3.21)
No (Ref)	1.00		1.00		1.00
Work–family spillover
Yes	2.08 (1.26, 3.44)			1.78 (1.01, 3.12)	1.60 (0.90, 2.84)
No (Ref)	1.00			1.00	1.00
Job strain
Yes	2.18 (1.35, 3.53)	1.16 (0.64, 2.09)	1.31 (0.73, 2.36)	1.18 (0.65, 2.13)	1.12 (0.61, 2.06)
No (Ref)	1.00	1.00	1.00	1.00	1.00
Frequent pain
Yes	1.98 (1.13, 3.47)	2.58 (1.36, 4.89)	2.39 (1.25, 4.58)	2.38 (1.25, 4.53)	2.33 (1.21, 4.48)
No (Ref)	1.00	1.00	1.00	1.00	1.00
Social support (continuous)	0.81 (0.74, 0.89)	0.83 (0.75, 0.92)	0.83 (0.75, 0.92)	0.82 (0.74, 0.91)	0.83 (0.75, 0.92)
Adjusted household income, $				.
Low (< 21 000)	2.55 (1.45, 4.47)	1.09 (0.52, 2.26)	1.06 (0.51, 2.23)	1.40 (0.68, 2.88)	0.93 (0.44, 1.99)
Middle (21 000–37 500)	1.35 (0.75, 2.41)	0.86 (0.44, 1.69)	0.87 (0.44, 1.72)	0.95 (0.48, 1.87)	0.81 (0.41, 1.61)
High (> 37 500) (Ref)	1.00	1.00	1.00	1.00	1.00
Gender
Female	0.61 (0.86, 3.01)	1.95 (0.97, 3.93)	1.92 (0.94, 3.91)	1.99 (0.98, 4.06)	1.91 (0.93, 3.89)
Male (Ref)	1.00	1.00	1.00	1.00	1.00
Education
≤ high school or GED	1.44 (0.93, 2.22)	0.99 (0.57, 1.70)	1.08 (0.63, 1.85)	1.07 (0.62, 1.84)	1.01 (0.58, 1.76)
> high school (Ref)	1.00	1.00	1.00	1.00	1.00
Race/ethnicity
Non-Hispanic Black	2.24 (0.36, 3.69)	2.52 (0.98, 6.49)	2.31 (0.88, 6.07)	2.56 (0.99, 6.66)	2.35 (0.89, 6.20)
Hispanic/other race	1.52 (0.84, 2.74)	1.71 (0.69, 4.21)	1.66 (0.66, 4.16)	1.63 (0.65, 4.07)	1.68 (0.67, 4.22)
Non-Hispanic White (Ref)	1.00	1.00	1.00	1.00	1.00
Immigrant status
Foreign born	1.58 (1.01, 2.46)	0.64 (0.28, 1.47)	0.66 (0.28, 1.55)	0.68 (0.29, 1.57)	0.67 (0.29, 1.57)
US born (Ref)	1.00	1.00	1.00	1.00	1.00
Model statistics
−2LL		1966.79	1979.46	1974.23	1984.64

Note. CI = confidence interval; GED = general equivalency diploma; LL = log of likelihood; OR = odds ratio. Model 1 was the crude model; models 2–5 also controlled for age, marital status, and presence of a child younger than 19 years in the household (none of which were significant at the .05 level). The main predictors in models 2–4 were food insufficiency, financial strain, and work–family spillover, respectively. Model 5 included all predictors.

The final model incorporated all 3 primary predictors (food insufficiency, financial strain, work–family spillover) and all control variables. Food insufficiency remained significantly associated with depressive symptoms in this model, but the odds ratio was attenuated. Participants who reported food insufficiency had twice the odds of depressive symptoms compared with those who did not report food insufficiency (OR = 2.10; 95% CI = 1.10, 4.18). Likewise, financial strain continued to be positively associated with depressive symptoms in this model (OR = 1.82; 95% CI = 1.03, 3.21). The association of work–family spillover with depressive symptoms was still positive but became statistically nonsignificant with inclusion of all primary predictors (OR = 1.60; 95% CI = 0.90, 2.84).

In analyses stratified by whether the respondent was the primary wage earner in the household, we observed differences in the association of food insufficiency, financial strain, and work–family spillover with depressive symptoms (Table 4). Among primary wage earners, depressive symptoms were 3.6 times as likely (95% CI = 1.42, 9.11) in those who reported food insufficiency as in those who did not. Among nonprimary earners, food insufficiency was not significantly associated with depressive symptoms, but financial strain (OR = 3.65; 95% CI = 1.48, 9.01) and work–family spillover (OR = 3.22; 95% CI = 1.11, 9.35) were.

TABLE 4—

Relationship of Depressive Symptoms to Food Insufficiency and Financial Strain, by Primary Wage Earner Status: Work, Family, and Health Network Study, Massachusetts, 2006–2007

Predictors	Primary Wage Earner,Model 1, OR (95% CI)	Nonprimary Wage Earner,Model 2, OR (95% CI)
Food insufficiency
Yes	3.60 (1.42, 9.11)	1.77 (0.44, 7.08)
No (Ref)	1.00	1.00
Financial strain
Yes	1.34 (0.59, 3.05)	3.65 (1.48, 9.01)
No (Ref)	1.00	1.00
Work–family spillover
Yes	1.30 (0.59, 2.88)	3.22 (1.11, 9.35)
No (Ref)	1.00	1.00
Job strain
Yes	1.55 (0.64, 3.72)	0.70 (0.24, 2.00)
No (Ref)	1.00	1.00
Frequent pain
Yes	2.93 (1.21, 7.06)	1.35 (0.40, 4.50)
No (Ref)	1.00	1.00
Social support (continuous)	0.83 (0.71, 0.97)	0.81 (0.70, 0.94)

Note. CI = confidence interval; OR = odds ratio. Models 1 and 2 controlled for age, adjusted household income, marital status, gender, educational attainment, race/ethnicity, immigrant status, and presence of a child younger than 19 years in the household (none of which were significant at the .05 level).

DISCUSSION

Our sample of low-wage nursing home workers had a high prevalence of food insufficiency and financial strain and reported many symptoms of depression. Their jobs, providing direct care to frail and ill elderly persons, are often described as stressful. We examined the role of household food insufficiency, financial strain, and work–family spillover on depressive symptoms in these employees. As with other studies of workers who are mostly female, belong to minority racial/ethnic groups, and have low socioeconomic status,³⁶ our data revealed a high prevalence of depressive symptoms. Although these sociodemographic variables significantly predicted depressive symptoms in crude models, the effect estimates attenuated and became nonsignificant after control for household- and work-related stressors. These results suggest that the work, financial, and familial circumstances faced by these employees may partly account for their depressive symptoms.

Despite the general prosperity of the United States, financial strain and food insufficiency were pervasive in this working population. Food insufficiency has been noted in the disabled elderly²¹ and welfare recipients,^12,22 but studies in the United States have not examined the presence of food insufficiency in working populations nor the potential contribution of this stressor to adverse mental health outcomes. Consistent with studies that have examined the associations separately, our data showed that food insufficiency, financial strain, and work–family spillover were associated with depressive symptoms.^4,17–22

Although separate studies have investigated the association of food insufficiency, financial strain, and work–family spillover with health outcomes, ours was the first to investigate all 3 household- and work-related variables in the same analysis. Stratified analyses additionally demonstrated that the associations differed by primary wage earner status. A key strength of our study was our ability to collect, and thus control for, data on both work- and household-related factors that are theoretically or empirically associated with depressive symptoms. Also, we controlled for the effects of pain, which may be related to key predictors and has been shown to be associated with depressive symptoms in workers who perform heavy manual labor.³⁴ Our sample was composed primarily of low-wage workers from a variety of racial/ethnic groups, providing empirical data on the experiences of these understudied employees.³⁷

Both household food insufficiency and financial strain are functions of available resources versus household needs; therefore, both conditions have an apparent relation with income. However, financial strain and food insufficiency capture fundamentally different dimensions of economic hardship from income because they measure the ability of households to borrow and save money.³⁸ Although the measures were closely associated with household income, their association with symptoms of depression remained after adjustment for income and other indicators of socioeconomic position, none of which had significant effects. This finding is consistent with other studies that have found a significant association between household food insufficiency and financial strain and mental health, despite revealing no income effect.^5,39 Not all studies have reported similar associations. A prospective study in the United Kingdom did not find an association between financial strain and depression, but it did not consider food insufficiency or stratify its sample by wage-earner status,¹⁵ making direct comparisons with our study difficult.

The lack of an association between financial strain and depressive symptoms among primary wage earners in our study is likely attributable to the close statistical association between food insufficiency and financial strain. Most of the respondents who were food insufficient were also financially strained, and half of those who were financially strained were food insufficient. This colinearity prevented us from statistically observing the independent effect of financial strain among primary wage earners.

Work–family spillover was associated with depressive symptoms only in nonprimary wage earners. This may be because primary wage earners concentrate on instrumental and material resources for their families and nonprimary wage earners are more concerned with emotional and caregiving support. Thus, nonprimary earners may be more vulnerable to the effects of work–family spillover.

Limitations

It is possible that persons who are depressed are less motivated and are thus less likely to occupy the types of jobs that provide household food and financial security and control over work–family issues. Furthermore, they may manage their finances differently. Because our study was cross-sectional, it was difficult to disentangle the causal pathways among these conditions. Our group's current work is longitudinal and will be able to lend further insights into the dynamics of these relationships.

The association between food insufficiency and depressive symptoms may be cyclical because depressive symptoms may limit the ability to become food sufficient, and food insufficiency may then contribute to depressive symptoms. We were unable to further tease out the differences between the different categories of financial strain and food insufficiency beyond a dichotomous outcome (e.g., between “have to cut back” and “cannot make ends meet”) because our study was not powered to show these differences.

Almost all of those who reported food insufficiency were primary wage earners. They had a high co-occurrence of financial strain and food insufficiency, and that made it mathematically difficult to separate the effects of these 2 factors in this group. The nonprimary wage earners were unlikely to report household food insufficiency; thus, we did not have enough power to see the effects of this exposure for these respondents.

To decrease response burden on study participants, all of whom were interviewed during work hours, we developed abbreviated measures of many of the variables. However, these measures have all been validated and have been found to be associated with health status in other studies.^3,13,15,40

Conclusions

Our results have several public health implications. Depression is associated with absenteeism and turnover among workers,⁴¹ which in turn translate to poorer care for residents in direct care settings.^42–44 Because improvement in care quality at nursing homes is an important public health priority, reducing workers’ depressive symptoms and their associated effects may have positive results for both workers and nursing home residents. Further life course implications are possible because food insufficiency is a household-level stressor that has been associated with overweight, behavioral problems, and poor mental health in children and adolescents.^40,45–47

The presence of food insufficiency in this low-wage sample could have further implications for obesity in this group because several studies have reported a seemingly paradoxical relation between food insufficiency and obesity whereby those who report food insufficiency also report obesity.^48–51 The presence of food insufficiency and depressive symptoms in this already vulnerable group is noteworthy because of the potential interplay between these conditions. At least 1 study has suggested that food insufficiency works through depression to influence overweight in household members.⁵²

Financial strain, food insufficiency, and symptoms of depression were common in our multiethnic sample of nursing home workers. Discussions on economic hardships have often focused on the health of the unemployed. As our study shows, financial strain can also be a pervasive issue for working men and women. For primary wage earners, household food insufficiency was associated with tripled odds of depressive symptoms—the leading cause of disability worldwide. The toll of these circumstances on already socially and economically disadvantaged women and men is of major public health significance.