Health Insurance Status and Symptoms of Psychological Distress among Low-income Urban Women

Abstract

Although numerous studies have considered the effects of having health insurance on access to health care, physical health, and mortality risk, the association between insurance coverage and mental health has been surprisingly understudied. Building on previous work, we use data collected from a two-year follow-up of low-income women living in Boston, Chicago, and San Antonio to estimate a series of latent fixed-effects regression models assessing the association between insurance status and symptoms of psychological distress. We find that having any insurance and private insurance is unrelated to depression, anxiety, and somatization. Having public insurance is unrelated to depression and somatization, but there is some evidence that having public insurance is associated with greater anxiety. Although not a direct test of the Affordable Care Act, our results suggest that the expansion of coverage may have a limited impact on symptoms of psychological distress among low-income urban women with children.

Keywords

health insurance mental health psychological distress women low-income urban

Although having health insurance should, in theory, promote health, interpretations of the empirical evidence are often divided and contentious. While some believe that “there is a substantial body of research supporting the hypothesis that having health insurance improves health” (Hadley 2003:60S) and that “the absence of health insurance is hazardous to your health” (Institute of Medicine 2009:ix), others believe that “no current consensus exists on the causal effect of gaining or losing health insurance on health care utilization and health outcomes” (Freeman et al. 2008:1023) and that “there is little evidence to suggest that extending insurance coverage to all adults would have a large effect on the number of deaths in the United States” (Kronick 2009:1211). Such glaring interpretive discrepancies may be surprising or even troubling due to the simple fact that “most of us forgot to ask a fundamental question, because we were certain we already knew the answer” (McArdle 2010:35).

Although numerous studies have considered the effects of having health insurance on access to health care, physical health, and mortality risk (Freeman et al. 2008; Hadley 2003; Levy and Meltzer 2008; McWilliams 2009), the association between insurance coverage and mental health has been surprisingly understudied (Finkelstein et al., 2012; McWilliams 2009). For example, in one comprehensive review of the literature, McWilliams (2009:480) notes that “no recent observational or quasi-experimental studies of the effects of uninsurance have focused primarily on mental health outcomes.” In this paper, we ask the fundamental question: Does having health insurance protect against symptoms of psychological distress?

In the pages that follow, we review the most rigorous empirical studies of the association between insurance status and mental health. We also present a competing framework to explain (1) why insurance should protect against psychological distress and (2) why insurance should be unrelated to distress. We then employ longitudinal data collected from a sample of low-income urban women to formally assess the main effect association between insurance status and psychological distress.

Theoretical Background

Previous Empirical Studies

Freeman and colleagues (2008) identify three general research designs that are capable of estimating the causal effect of insurance status on health-related outcomes. These designs involve the use of longitudinal cohorts, instrumental variables, and experiments. Limited to these designs, our review revealed eight studies of insurance coverage and mental health. Consistent with the general tendencies of the broader literature, the empirical evidence concerning the mental health consequences of insurance coverage can be characterized as limited and mixed.

While some studies using longitudinal cohorts and instrumental variables show that having some form of health insurance is associated with lower levels of depression (McWilliams et al. 2007a; Tian et al. 2012) and nonspecific psychological distress (Lave et al. 1998), at least one longitudinal study suggests that health insurance is unrelated to psychological distress (Probst et al. 2008). Although the seminal RAND Health Insurance Experiment shows that level of cost sharing for health care is unrelated to psychological well-being and psychological distress (Newhouse 1996), the more recent Oregon Health Insurance Experiment suggests that random assignment to Medicaid can contribute to greater happiness and fewer symptoms of depression (Baicker et al. 2013; Finkelstein et al. 2012). There is also some indirect (but compelling) evidence to suggest that state laws mandating mental health benefits (mental health parity legislation) are unrelated to adult suicide rates (Klick and Markowitz 2006).

Why Insurance Should Protect against Psychological Distress

Although insurance coverage is thought to support physical health and longevity by increasing access to health care (McWilliams 2009), the exact mechanisms through which insurance might contribute to better mental health are currently undertheorized and untested. While the effects of insurance coverage on physical health and mortality risk are almost entirely indirect through access to health care, insurance could be directly and/or indirectly associated with lower levels of psychological distress.

The direct link between insurance status and mental health refers to the immediate psychological consequences of living with or without coverage. On the one hand, living without the security of any insurance could contribute to symptoms of psychological distress (e.g., feelings of fear and worry). On the other hand, living with insurance and knowing that one’s family is covered in the event of illness could contribute to psychological well-being (e.g., feelings of happiness and tranquility). According to the Robert Wood Johnson Foundation’s (2009) Health Care Consumer Confidence Index, nearly one in four (23.7 percent) Americans are worried that they will lose their health insurance coverage in the next 12 months, and nearly half of all Americans (46 percent) are worried that they would not be able to afford treatment if they or someone in their family became seriously ill. In one study of families of newly insured children, Lave and colleagues (1998) found that nearly three in four (73.5 percent) parents were “worried, scared, and stressed out” during the period in which their children lacked coverage. Over half (61 percent) of these same parents reported that the acquisition of coverage for their children “led to peace of mind, reduced worry, and lowered stress.”

The indirect link between insurance status and mental health refers to downstream mechanisms, like access to mental health care and secondary stressors that stem from living without coverage (e.g., financial debt). The direct link between insurance and psychological distress is more or less imagined (e.g., people can fear the prospect of being uninsured). The indirect link between insurance and mental health is perhaps more important because it is material. As with outcomes related to physical health and mortality risk, lacking insurance coverage could undermine mental health by limiting access to mental health care (National Alliance on Mental Illness 2008; Strine et al. 2011). Wells and colleagues (2002) show that the uninsured are more likely to forgo needed mental health care than their publicly and privately insured counterparts. They also report that the uninsured exhibit lower rates of active treatment (given need) and lower levels of satisfaction with overall care, care for emotional or mental health problems, and care for substance abuse problems.

Along with limiting access to mental health care, lacking coverage could contribute to psychological distress by exposing individuals to new stressful conditions. The concept of stress proliferation is especially useful for explaining the mental health consequences of living without insurance. Stress proliferation occurs when stressful conditions contribute to the development of new or secondary conditions that are experienced as stressful (Pearlin et al. 2005). For example, Ross and Mirowsky (2000:313) explain that “if medical insurance proves to yield measurable health benefits, they may come mostly from the reduction of economic hardship.” Indeed, several studies show that the uninsured are disproportionately exposed to a range of financial difficulties, including greater medical expenditures and medical debt and more financial strain within the family (Baicker et al. 2013; Finkelstein and McKnight 2008; Finkelstein et al. 2012; Institute of Medicine 2003; Kaiser Family Foundation 2014; Lave et al. 1998; McWilliams et al. 2007b; Ross and Mirowsky 2000; Schoen and DesRoches 2000). Although state mental health parity legislation is seemingly unrelated to adult suicide rates, there is at least some evidence to suggest that such laws are effective in reducing financial strain (e.g., out-of-pocket spending) within families of children with mental health care needs (Barry and Busch 2007). Because the 2008 Mental Health Parity and Addiction Equity Act (MHPAEA) and the 2010 Patient Protection and Affordable Care Act (ACA) extend benefits and coverage, they are also expected to support mental health by reducing financial stress and increasing access to care (Golberstein and Busch 2013; Mechanic 2011; Rowan, McAlpine, and Blewett 2013).

When reviewing the ways in which insurance might protect against psychological distress, it is important to consider the unique effects of public and private insurance plans. Freeman and colleagues (2008:1024) note that “variations in insurance plans—depth of coverage, cost sharing arrangements, and economic incentives—have the potential to affect utilization and/or outcomes.” Because public and private plans tend to have different levels of coverage, previous studies often distinguish between these two broad classes of insurance (Kasper, Giovannini, and Hoffman 2000; Kaiser Family Foundation 2014; Ross and Mirowsky 2000; Sorlie et al. 1994; Wells et al. 2002; Young et al. 2001). To the extent that private plans provide greater coverage than public plans, private insurance may be more likely than public insurance to promote mental health. Some research suggests that having private insurance is associated with more favorable health outcomes compared to having public insurance (Kasper, Giovannini, and Hoffman 2000; Wells et al. 2002). Other studies actually show that having public insurance is associated with worse health outcomes compared with having private insurance or no insurance at all (Ross and Mirowsky 2000; Sorlie et al. 1994). At this time, there is no established theoretical rationale to explain why public insurance might undermine health. Instead, this so-called “Medicaid conundrum” is often described as an artifact of health selection and/or model misspecification (Hadley 2003). This seems reasonable given that the publicly insured are often older, poorer, and sicker than the privately insured.

Why Insurance Should Be Unrelated to Psychological Distress

The reasons for why insurance could be directly and/or indirectly associated with better mental health are intuitive and compelling. However, given the mixture of previous empirical studies, it is important to at least consider the possibility that insurance coverage could be unrelated to mental health. In fact, there are several reasons for why health insurance may be ineffective in protecting against psychological distress.

When mental health issues arise, having insurance could conceivably aid in recovery. However, this secondary role of insurance is largely contingent upon the effectiveness of the most common treatments. Although some cognitive behavioral approaches have been shown to improve mental health (Roy-Byrne et al. 2010), psychoactive drugs are often (not always) ineffective and sometimes counterproductive (Glenmullen 2000; Liebert and Gavey 2009; Mechanic 2011; Mirowsky and Ross 2003; Weilburg et al. 2003). As stated, the key theoretical link between insurance and health is health care (i.e., health insurance buys care, and care improves health). We do not challenge the idea that health insurance buys care, but we do challenge the idea that health care always improves health. According to this model, health insurance is effective in promoting health to the extent that the treatments insurance buys promote health. If psychoactive drugs are often (not always) ineffective in treating mental health, insurance may not protect against psychological distress through this increasingly widespread form of care. Theoretically, the mental health benefits of insurance could be undermined by the limits of mental health care.

Even if medical treatments for common mental health conditions were consistently effective, coverage for mental health conditions is often limited and costly. Mental health parity legislation has sought to address glaring benefit disparities, but there are significant gaps in state and federal mandates (Barry and Busch 2007; Golberstein and Busch 2013; Sarata 2011; Strine et al. 2011). Although the 2008 MHPAEA and the 2010 ACA address these benefit disparities to some extent, the high cost of mental health services continues to limit access to needed care (Mechanic 2011; Rowan, McAlpine, and Blewett 2013). The key point is that having insurance is unlikely to protect against psychological distress if mental health coverage is weak or nonexistent or cost prohibitive.

Any problems associated with coverage may be compounded by the fact that most people with diagnosable mental health conditions never seek treatment (Kessler et al. 1996; Regier et al. 1993; Sherbourne, Dwight-Johnson, and Klap 2001). In this case, the stigma associated with mental health issues serves as a cultural barrier to help-seeking (Mechanic 2011). Partly due to these cultural constraints, mental health parity legislation (Bao and Sturm 2004; Barry and Busch 2007; Pacula and Sturm 2000), cost sharing for health care (Newhouse 1996), and having health insurance (Sherbourne, Dwight-Johnson, and Klap 2001; Wang et al. 2005; Young et al. 2001) are often unrelated to the use of needed mental health care (but see Harris, Carpenter, and Bao [2006] for an exception concerning mental health parity legislation).

Finally, having insurance is unlikely to benefit mental health when access to care is limited by broader structural barriers. Bird and Rieker’s (2008) constrained-choices theory places health-related behavior in the context of evolving labor markets, extended work hours, unpredictable work schedules, and conflicting institutional demands of work and family roles. For example, a recent study by the Kaiser Family Foundation (2014) shows that many women—especially low-income women with children—are unable to attain needed health care because they cannot find time to go to the doctor, cannot take time off of work, have problems getting child care, and have transportation problems. Regular appointments and therapy sessions with mental health professionals take time and planning. Broader structural barriers may limit the pursuit of mental health care by constraining resources, like free time and flexible schedules. This may be especially true for theoretically relevant subgroups (e.g., low-income women with children).

The Current Study

The current study builds on previous work by using data collected from a two-year follow-up of low-income urban women to estimate a series of fixed-effects regression models assessing the main effect association between insurance status and symptoms of psychological distress. We focus on low-income families for three reasons. The first reason is theoretical. Given that low-income families are especially likely to be among the uninsured (Anderson and Eamon 2005; Institute of Medicine 2001; S. Schroeder 2007), they are most likely to be affected by the expansion of Medicaid through the ACA (Baicker et al. 2013; Finkelstein et al. 2012; Golberstein and Busch 2013). The second reason is methodological. Because there is considerable volatility in the health insurance status of low-income families, we have the unique opportunity to assess the mental health consequences of intra-individual changes in coverage. The third reason is related to policy. Our analyses may serve as an early indication of whether insurance reform, or more specifically, extending coverage, is likely to protect against psychological distress among low-income urban women with children.

In addition to focusing on an important subpopulation, our work contributes to previous research by examining the effects of health insurance coverage on several forms of psychological distress, including depression, anxiety, and somatization. Although there is some evidence to suggest that having some form of insurance is associated with lower levels of depression (Baicker et al. 2013; Finkelstein et al. 2012; McWilliams et al. 2007a; Tian et al. 2012), it is unclear whether anxiety or somatization might also vary according to insurance status.

Method

Data

To formally test the link between health insurance status and symptoms of psychological distress, we use data from the Welfare, Children, and Families (WCF) project (see http://www.jhu.edu/~welfare/). The WCF project is a household-based, stratified random sample of 2,402 low-income women living in low-income neighborhoods in Boston, Chicago, and San Antonio. The WCF first sampled census blocks (or neighborhoods) with at least 20 percent of residents below the federal poverty line based on the 1990 census. Within these neighborhoods, households under 200 percent of the poverty line were sampled, with an oversample of households below 100 percent of the poverty line. Because one of the goals of the WCF project is to assess the impact of welfare policy and work on children, households were screened for the presence of children. Households with at least one infant or child (ages 0–4) or young adolescent (ages 10–14) were sampled. The children’s caregivers, all women, were interviewed face-to-face. We refer to the caregivers as “women with children” instead of “mothers” because some caretakers did not identify themselves as the child’s parent. The data were collected in 1999 with a follow-up in 2001. The baseline response rate was 75 percent, and 89 percent of the original sample were reinterviewed. The overall respondent-level response rate is 75 percent, with city-specific response rates of 74 percent (Boston), 71 percent (Chicago), and 79 percent (San Antonio). Subsequent analyses are weighted to account for variations in sample sizes across cities.

Measures

Psychological Distress

We use the Brief Symptom Inventory (BSI-18) to assess symptoms of psychological distress (Derogatis 2000). The BSI-18 is widely used as a screening tool for psychological distress in community-based samples, to inform clinical decisions, and to assess treatment progress in clinical samples (Derogatis 2000; Deoragatis and Fitzpatrick 2004; Zabora et al. 2001). In subsequent analyses, we employ a continuous specification of distress symptoms because it is more sensitive to the detection of emotional suffering than categorical diagnostic categories (Mirowsky and Ross 2003). Table 1 presents survey items and component loadings for the Depression, Anxiety, and Somatization subscales. Depression is measured as the mean response to items 1 through 6 (α = 0.84). Anxiety is measured as the mean response to items 7 through 12 (α = 0.84). Somatization, the physiological manifestation of distress, is measured as the mean response to items 13 through 18 (α = 0.80). Respondents were asked to indicate how much in the past seven days they were distressed or bothered by each symptom. Response categories for each item range from 0 = not at all to 4 = extremely. The component loadings presented in Table 1 were estimated using principal components analysis, specifying a minimum eigenvalue of 1.00, with varimax rotation. Eigenvalues for each component exceed 3.00, and all component loadings exceed 0.60. These measures of distress have demonstrated construct validity in previous research through associations with criminal behavior, heavy drinking, and self-rated physical health (Hill and Angel 2005; Hill, Ross, and Angel 2005; R. Schroeder et al. 2010).

Table 1.

Baseline Items and Component Loadings for the Brief Symptom Inventory–18 Subscales.

Item	Depression	Anxiety	Somatization
01. Feeling no interest in things	0.72
02. Feeling lonely	0.78
03. Feeling blue	0.79
04. Feelings of worthlessness	0.81
05. Feeling hopeless about the future	0.80
06. Thoughts of ending your life	0.64
07. Nervousness or shakiness		0.75
08. Feeling tense or keyed up		0.70
09. Suddenly feeling scared for no reason		0.76
10. Spells of terror or panic		0.78
11. Feeling so restless you couldn’t sit still		0.75
12. Feeling fearful		0.79
13. Faintness or dizziness			0.67
14. Pains in your heart or chest			0.69
15. Nausea or upset stomach			0.67
16. Trouble getting your breath			0.71
17. Numbness or tingling in parts of your body			0.72
18. Feeling weak in parts of your body			0.81
Eigenvalues	3.44	3.44	3.04
Alpha reliability	0.84	0.84	0.80

Source. Welfare, Children, and Families (1999; http://www.jhu.edu/~welfare/).

Note. N = 2,357. All respondents are women. Component loadings were estimated using principal components analysis, specifying a minimum eigenvalue of 1.00, with varimax rotation.

Health Insurance Status

We assess health insurance status in two ways. We first asked respondents to indicate whether they were covered by any type of health insurance plan or program that pays for at least some of their medical expenses. Responses to this item were dummy coded no (0) or yes (1). Following previous research (Ross and Mirowsky 2000), we also distinguish between public and private insurance. All respondents who reported having some form of insurance were asked to indicate their provider. Original response categories included (1) a plan through an employer or union; (2) a plan paid for directly; (3) Medicaid; (4) Champus, VA, or other military plan; and (5) another plan or program. Respondents were classified as having public insurance if they selected coverage through Medicaid or some form of military plan. Respondents were classified as having private insurance if they selected coverage through an employer, union, or direct payment. The small number of respondents who selected another plan or program (n = 60) were categorized as public or private according to their open-ended responses. For example, responses like “Medicare,” “Army,” and “Government Medical Plan” were coded as public. Responses like “Blue Cross Blue Shield,” “husband’s job,” or “through husband” were coded as private. All unfamiliar plans were researched to determine if they were public or private. The appendix provides a complete description of our classifications.

To account for heterogeneity in coverage, subsequent regression analyses assess changes in (1) any insurance, (2) public insurance, and (3) private insurance. Because we measure health insurance status at only two points in time (1999 and 2001), we are unable to assess continuity of coverage in a direct way. As a precaution, we control for whether the respondent had any lapses in health insurance coverage over the study period. At each wave, respondents were asked to indicate whether there was any time in the past 12 months when they were not covered by any health insurance program. In subsequent analyses, we assess lapses in (1) any insurance, (2) public insurance, and (3) private insurance.

Time-varying Characteristics

Previous research has identified a range of time-varying social and health-behavioral characteristics that are correlated with health insurance status and health-related outcomes (Anderson and Eamon 2005; Angel, Frias, and Hill 2005; Freeman et al. 2008; Jerant, Fiscella, and Franks 2012; Tian et al. 2012). Following this work, subsequent regression analyses control for age (in years), education (0 = none, 1 = high school, 2 = associate degree, 3 = college degree or higher), employment status (1 = currently employed, 0 = otherwise), household income (square root of continuous monthly income), welfare status (1 = currently receiving public assistance, 0 = otherwise), marital status (1 = currently married/spouse in house, 0 = otherwise), number of children (1 to 6 or more, top-coded continuous variable), and self-rated physical health (0 = poor, 1 = fair, 2 = good, 3 = very good, 4 = excellent). Jerant and colleagues (2012) report that smoking and body mass are related to changes in coverage, but we are unable to assess these characteristics over time. Instead, we control for changes in past year intoxication frequency (0 = never, 1 = once or twice, 2 = several times/often) because smoking, body mass, and mental health are generally correlated with drinking behavior (Breslow and Smothers 2005; Hill and Angel 2005; McKee et al. 2006).

Time-invariant Factors

In subsequent fixed effects regression analyses, all time-invariant characteristics are controlled by design. Nevertheless, subsequent descriptive analyses include race and ethnicity (four dummy variables capturing non-Hispanic Whites, non-Hispanic Blacks, Mexican Americans, other Hispanics) and city of residence (three dummy variables capturing residence in Boston, Chicago, and San Antonio).

Analytic Strategy

Our analyses begin with the presentation of descriptive statistics for selected study variables, including ranges, means, and standard deviations (Table 2). In accordance with the research designs specified by Freeman and colleagues (2008), we use a latent-variable fixed-effects regression model to estimate the association between insurance status and psychological distress. Table 3 presents the results for any insurance coverage. Table 4 presents the results for public and private coverage (any insurance disaggregated).

Table 2.

Descriptive Statistics.

Variable	Range	Mean	Standard Deviation
Insurance status
Any (1999)	0–1	0.67
Public (1999)	0–1	0.44
Private (1999)	0–1	0.23
Acquired any (2001)	0–1	0.15
Acquired public (2001)	0–1	0.10
Acquired private (2001)	0–1	0.05
Lost any (2001)	0–1	0.12
Lost public (2001)	0–1	0.08
Lost private (2001)	0–1	0.04
Any insurance lapse (1999)	0–1	0.08
Public lapse (1999)	0–1	0.05
Private lapse (1999)	0–1	0.03
Psychological distress
Depression (1999)	0–4	0.41	0.60
Δ Depression (2001–1999)	−3–3	−0.02	0.58
Anxiety (1999)	0–4	0.31	0.51
Δ Anxiety (2001–1999)	−3–3	−0.02	0.48
Somatization (1999)	0–4	0.34	0.51
Δ Somatization (2001–1999)	−3–3	−0.02	0.51
Time-varying factors (1999)
Age	18–74	33.00	9.61
Education	0–3	0.80	0.74
Employed	0–1	0.43
Household income	0–7,800	1,094	898
Receiving welfare	0–1	0.29
Married	0–1	0.30
Number of children	0–6	2.72	1.40
Intoxication frequency	0–2	0.33	0.57
Self-rated health	0–4	2.34	1.12
Time-invariant factors (1999)
Non-Hispanic White	0–1	0.04
Non-Hispanic Black	0–1	0.42
Mexican American	0–1	0.34
Other Hispanic	0–1	0.20
Boston residence	0–1	0.33
Chicago residence	0–1	0.33
San Antonio residence	0–1	0.33

Source. Welfare, Children, and Families (1999, 2001; http://www.jhu.edu/~welfare/).

Note. N = 2,357 (1999). N = 2,081 (2001). All respondents are women.

Table 3.

Fixed Effects Regression of Psychological Distress on Any Insurance and Time-varying Factors.

Variable	Depression	Anxiety	Somatization
Any insurance	0.014 (0.022)	0.027 (0.020)	0.014 (0.019)
Insurance lapse	−0.018 (0.033)	−0.030 (0.028)	0.039 (0.027)
Age	−0.005 (0.001)***	0.000 (0.001)	−0.002 (0.001)*
Education	−0.000 (0.015)	−0.009 (0.014)	−0.019 (0.012)
Employed	−0.060 (0.021)**	−0.040 (0.019)*	−0.026 (0.018)
Household income	0.000 (0.001)	−0.001 (0.001)	−0.001 (0.001)
Receiving welfare	0.016 (0.023)	0.013 (0.020)	0.045 (0.019)*
Married	−0.104 (0.030)**	−0.056 (0.026)*	−0.058 (0.025)*
Number of children	0.026 (0.009)**	0.028 (0.008)***	0.020 (0.007)**
Intoxication frequency	0.175 (0.017)***	0.112 (0.015)***	0.047 (0.014)**
Self-rated health	−0.113 (0.009)***	−0.098 (0.008)***	−0.141 (0.008)***
Model χ²	114.493***	134.004***	133.274***
RMSEA	0.034	0.038	0.038
SRMR	0.014	0.016	0.016
CFI	0.931	0.918	0.918

Source. Welfare, Children, and Families (1999, 2001; http://www.jhu.edu/~welfare/).

Note. N = 2,044. All respondents are women. Shown are unstandardized coefficients with standard errors in parentheses. RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual; CFI = comparative fit index.

p < .05. **p < .01. ***p < .001 (two-tailed tests).

Table 4.

Fixed Effects Regression of Distress on Public and Private Insurance and Time-varying Factors.

Variable	Depression	Anxiety	Somatization
Public insurance	0.044 (0.024)	0.048 (0.021)*	0.038 (0.020)
Private insurance	−0.058 (0.031)	−0.022 (0.027)	−0.045 (0.026)
Public lapse	−0.030 (0.041)	−0.047 (0.035)	0.025 (0.034)
Private lapse	0.010 (0.053)	0.004 (0.046)	0.079 (0.045)
Age	−0.005 (0.001)***	−0.000 (0.001)	−0.002 (0.001)
Education	0.003 (0.015)	−0.007 (0.014)	−0.017 (0.012)
Employed	−0.046 (0.022)*	−0.031 (0.019)	−0.015 (0.018)
Household income	0.000 (0.001)	−0.001 (0.001)	0.000 (0.001)
Receiving welfare	−0.002 (0.024)	0.003 (0.021)	0.033 (0.020)
Married	−0.093 (0.030)**	−0.049 (0.027)	−0.049 (0.025)*
Number of children	0.025 (0.009)**	0.028 (0.008)***	0.020 (0.007)**
Intoxication frequency	0.176 (0.017)***	0.112 (0.015)***	0.047 (0.014)**
Self-rated health	−0.112 (0.009)***	−0.097 (0.008)***	−0.140 (0.008)***
Model χ²	114.493***	142.040***	137.695***
RMSEA	0.034	0.035	0.035
SRMR	0.014	0.014	0.014
CFI	0.931	0.917	0.920

Source. Welfare, Children, and Families (1999, 2001; http://www.jhu.edu/~welfare/).

p < .05. **p < .01. ***p < .001 (two-tailed tests).

We use Mplus 6.0 to identify our regression models through the analysis of respondents who experience changes in insurance status and psychological distress between waves (Allison 2009). Fixed-effects models are unique because they control for all time-invariant characteristics that are correlated with both insurance status and distress symptoms (e.g., race and city of residence). One limitation of fixed-effects models is that they cannot control for bias due to omitted time-varying characteristics. In subsequent analyses, we control for several relevant time-varying characteristics, including age, education, employment status, household income, welfare status, marital status, number of children, frequency of intoxication, and self-rated health. Given the relatively short two-year study period, we are confident that unobserved time-varying characteristics are unlikely to create any substantial bias.

Random-effects Models

In unreported analyses, we calculated a series of Hausman (1978) tests in Stata 12(SE) to assess the viability of the random effects model for our data. Each of these tests rejected the null that the random-effects coefficients are equivalent to the fixed-effects coefficients. These results suggest that a key assumption of the random-effects model—that unobservable characteristics are uncorrelated with the included variables—is not satisfied. Although subsequent analyses are limited to fixed-effects regression models, estimates from our random-effects models are available upon request.

Sample Attrition

When considering changes in outcomes over time, it is customary to examine the issue of bias due to sample attrition. The primary concern is whether or not there are any systematic changes in the sample across waves. To formally assess this issue, we estimated a binary logistic regression model predicting the log odds of sample attrition (results not shown). The dependent variable in this case is dummy coded such that respondents who completed questionnaires for both waves were given a value of 0 and those who completed the wave 1 questionnaire were given a value of only 1. The independent variables include insurance status, psychological distress, and all time-varying and time-invariant characteristics. Approximately 11 percent of the sample (267 respondents) was lost to follow-up. The logistic regression results show very little evidence of systematic sample attrition. Only the coefficient for self-rated health (odds ratio = 2.92, p < .05) was statistically significant at conventional levels. Briefly, this result suggests that the odds of attrition are higher for healthier respondents. Important for our research design, none of the insurance or distress variables are statistically significant predictors of attrition.

Results

Descriptive Analyses

Table 2 provides descriptive statistics for all study variables. At baseline, most respondents (67 percent) report having some form of health insurance. Approximately 44 percent of the sample reported having public insurance, while 23 percent of respondents reported having private insurance. We also observed some notable volatility in insurance status over time. Nearly 8 percent of respondents reported a lapse in any insurance. Lapses in public insurance (5 percent) were slightly more prevalent than lapses in private insurance (3 percent). Nearly 15 percent of the sample acquired some form of health insurance. The acquisition of public insurance (10 percent) was more common than the acquisition of private insurance (5 percent). At baseline, the average respondent reported low levels of depression, anxiety, and somatization. The negative means for the distress change scores indicate slight improvements in symptoms over the study period.

Fixed-effects Analyses

Table 3 presents the results of our latent fixed-effects regressions of psychological distress on any insurance. This table contains several model fit indices. All root mean square error of approximation and standardized mean square residual estimates fall below the standard of 0.05, and the comparative fit index estimates exceed the standard of 0.90 (Byrne 2012; West, Taylor, and Wu 2012). Taken together, these results indicate adequate or good model fit. The coefficients in Table 3 reveal a consistent null pattern: Having any insurance is unrelated to depression, anxiety, and somatization. The coefficients for any insurance lapses are also statistically indistinguishable from zero.

Table 4 presents the results of our latent fixed-effects regressions of mental health on public and private insurance (any insurance disaggregated). Once again, the model fit indices can be characterized as adequate or good. The coefficients in Table 4 confirm that having private insurance is unrelated to depression, anxiety, and somatization. Public insurance is also unrelated to depression and somatization. However, we observe that having public insurance is associated with greater anxiety. Consistent with the results presented in Table 3, the coefficients for lapses in public and private insurance are statistically indistinguishable from zero.

Although we are not primarily concerned with the results for our time-varying covariates, we would like to describe several patterns that are consistent across distress outcomes in Tables 3 and 4. On the one hand, being in a marital relationship and having better self-rated physical health are associated with fewer symptoms of depression, anxiety, and somatization. On the other hand, having more children and being drunk more often are associated with greater depression, anxiety, and somatization.

Discussion

Although numerous studies have considered the effects of having health insurance on access to health care, physical health, and mortality risk, the association between insurance coverage and mental health has been surprisingly understudied. While some studies show that having some form of health insurance is associated with greater happiness, lower rates of depression, and lower levels of nonspecific psychological distress, other studies report null associations with these same outcomes. Building on previous work, we used data collected from a two-year follow-up of low-income women living in Boston, Chicago, and San Antonio to estimate a series of latent fixed-effects regression models assessing the association between insurance status and symptoms of psychological distress.

For the most part, we observe that having health insurance is unrelated to symptoms of psychological distress. Our longitudinal analyses show that having any insurance and private insurance is unrelated to depression, anxiety, and somatization. Although having public insurance is unrelated to depression and somatization, there is some evidence to suggest that having public insurance is associated with greater anxiety.

Our extensive null findings support previous studies of suicide (Klick and Markowitz 2006), nonspecific psychological distress (Newhouse 1996; Probst et al. 2008), happiness (Baicker et al. 2013), and overall psychological well-being (Newhouse 1996). Our finding suggesting that public insurance might increase anxiety is generally consistent with the often-cited “Medicaid conundrum” (Hadley 2003; Ross and Mirowsky 2000; Sorlie et al. 1994). However, this pattern is likely the result of some unobserved heterogeneity (e.g., health selection). To the best of our knowledge, we are the first to formally test whether health insurance status is associated with symptoms of anxiety and somatization.

The present study is characterized by several limitations. The primary limitation of the present study is our observational design. Experimental and quasiexperimental designs (e.g., the RAND Health Insurance Experiment and the Oregon Health Insurance Experiment) are the gold standard for establishing cause-and-effect relationships. The results of our fixed-effects regression models are suggestive because they control for all time-invariant characteristics and a range of relevant time-varying characteristics, but we cannot exclude the possibilities of reverse causation and bias due to omitted time-varying characteristics. We considered the possibility of employing instrumental variable techniques to identify the causal effects of insurance status. However, valid instrumental variables are generally hard to find because they require that changes in the instrument be associated with changes in X but not changes in Y. Unfortunately, we were unable to locate any instruments that are correlated with insurance status but not psychological distress.

We must acknowledge that our ability to detect any statistically significant associations for our insurance variables may be compromised by sample size and the estimation constraints of the fixed-effects regression model. Only 15 percent of the sample acquired insurance over the study period, and only 12 percent lost coverage. Statistical power is a concern when these relatively small samples sizes are combined with the within-subjects design of the fixed-effects regression model. For these reasons, our key results are likely to be conservative and should be interpreted accordingly. We would like to note that the results of our fixed-effects regressions are generally consistent with our preliminary cross-sectional regressions (available upon request) that employed a larger sample. It is also interesting that our only statistically significant effect for insurance status is observed after the any insurance category is disaggregated into public and private (the sample sizes for public and private insurance are separately smaller than the sample size for the combined any insurance category).

One important measurement issue is our rather coarse distinction between public and private insurance. The public-private distinction may be useful to some extent, but it ignores potentially important within-plan heterogeneity. Freeman and colleagues (2008) acknowledge that insurance plans can vary according to depth of coverage, level of cost sharing, and degree of economic incentives. It would be ideal for studies with larger sample sizes to incorporate information obtained from more detailed assessments of insurance coverage. More precise measurements of insurance status (e.g., level of mental health coverage and affordability) could contribute to more consistent patterns in the empirical literature.

Another potential limitation is our measurement of distress symptoms and the low levels of reported symptoms in our sample. We noted three primary reasons why mental health and symptoms of psychological distress might vary according to health insurance status. The first reason is that lacking coverage is stressful in itself. If being uninsured is directly stressful, people without coverage should report more symptoms of distress. The second reason is that lacking health insurance leads to stress proliferation (e.g., additional stressors like financial hardship). If lacking insurance contributes to new stressful conditions, then people without coverage should report more symptoms of distress. Although we believe our measurements and sample characteristics are reasonably suited to these first two processes, we are less certain about our ability to capture the third process—the fact that lacking coverage limits access to mental health care. When we assume that lacking health insurance is stressful in itself or indirectly stressful through stress proliferation processes, distress symptoms are appropriate because they are sensitive to a range of human suffering. However, if we assume that mental health care is the primary reason for why health insurance is associated with mental health, our analyses would be limited by our use of symptom inventories and the generally low levels of reported symptoms. The idea is that people with chronic and severe symptoms may be more likely to seek care and that diagnoses are often required to receive formal services. This suggests that health insurance would mainly benefit people with the poorest mental health—those people with the greatest access to the mechanism of mental health care. If mental health care is the primary link between health insurance status and mental health, our analyses would conceivably underestimate the effectiveness of coverage. The degree of underestimation is nevertheless conditional upon the true salience of the mental health care mechanism. In our view, the stress mechanisms are most viable because the link between stress and mental health is well established. The care mechanism is less likely because mental health treatments (e.g., psychoactive drugs) are often ineffective in treating people with serious mental health issues. If stress mechanisms are most salient, our general conclusions are appropriate. If the care mechanism is most salient, our general conclusions would be conservative and should be interpreted with this in mind.

We must also acknowledge that the sample is restricted to low-income urban women with children living in Boston, Chicago, and San Antonio. Although we have noted the theoretical and empirical strengths of our sample, we cannot generalize our results beyond these specific population parameters. Our data are also over 13 years old. It is unclear whether any changes in insurance programs and health care delivery over this period would alter the substantive conclusions of our analyses.

Clearly, the internal and external validity of our results are contingent upon replication and further testing of more recent data collected from higher-income women, women without children, men, and people living in other cities and rural areas. Having said this, we are encouraged that our results are generally consistent with the findings of several previous studies that use different methodologies (e.g., instrumental variables and experiments) and different data sources, including older studies, more recent studies, and studies of diverse populations.

Conclusion

In this paper, we tested whether having health insurance is protective against symptoms of psychological distress. Despite the limitations of our study, we are confident that insurance coverage is generally unrelated to symptoms of depression, anxiety, and somatization among low-income women with children living in Boston, Chicago, and San Antonio. Additional research is needed to examine associations with a broader range of mental health outcomes, including diagnostic assessments, constituent emotions (e.g., anger and fear), and behavioral disorders (e.g., chemical dependence and antisocial personality). It is also important for future work to move beyond the main effects of insurance status to explore theoretically relevant subgroup variations. For example, does having insurance moderate the association between poor physical health and mental health? To what extent do factors like stigma, access to care, and utilization patterns condition the effectiveness of health insurance? Finally, we must seriously consider the possibility that health insurance is unrelated to mental health. Of course, this realization would force us to emphasize and confront the fundamental problems of social inequality that contribute to the development of mental health issues in the first place. In terms of policy, our results indirectly suggest that the ACA and Medicaid expansion may have a limited effect on the mental health of low-income urban women with children. These policies may facilitate access to needed treatments and reduce financial stress associated with out-of-pocket expenses, but any indirect impact on mental health is uncertain.

Footnotes

Appendix

Coding of Open-ended Responses to Insurance Status

Plans Classified as Public	Plans Classified as Private
Army	Blue Cross/Blue Shield
Boston Health Net (Free Care)	Brigham and Women’s Clinic
Boston Medical Center	Carried insurance into retirement
Care-Link (San Antonio)	Harba Per Green
Center Care	Harvard Pilgrim
Codman Square Health	Husband’s job or retirement
Department of Child and Family Services	Dominican Republic Insurance
Free plan from hospital	Intercontinental De Segur
Government medical plan	Medical center coverage
Healthy Start	Own expense
Medicare	Proof Care Insurance
Neighborhood health plan	University hospital/medical center plan
Secure Horizon
Through a health center
Upham’s Corner Health
Wesley Community Services

Acknowledgements

The authors thank Jill Quadagno, Catherine Ross, Johanna Catherine Maclean, William Avison (editor), and the anonymous reviewers for critical and constructive comments on earlier versions of this article.

References

Allison

Paul

. 2009. Fixed Effects Regression Models. Thousand Oaks, CA: Sage.

Anderson

Steven

Eamon

Mary

. 2005. “Stability of Health Care Coverage among Low-income Working Women.” Health & Social Work 30:7-17.

Angel

Ronald

Frias

Sonia

Hill

Terrence

. 2005. “Determinants of Household Insurance Coverage among Low-income Families from Boston, Chicago, and San Antonio: Evidence from the Three-city Study.” Social Science Quarterly 86:1338-53.

Baicker

Katherine

Taubman

Sarah

Allen

Heidi

Bernstein

Mira

Gruber

Jonathan

Newhouse

Joseph

Schneider

Eric C.

Wright

Bill J.

Zaslavsky

Alan M.

Finkelstein

Amy N.

2013. “The Oregon Experiment: Effects of Medicaid on Clinical Outcomes.” New England Journal of Medicine 368:1713-22.

Bao

Yuhua

Sturm

Roland

. 2004. “The Effects of State Mental Health Parity Legislation on Perceived Quality of Insurance Coverage, Perceived Access to Care, and Use of Mental Health Specialty Care.” Health Services Research 39:1361-77.

Barry

Coleen

Busch

Susan

. 2007. “Do State Parity Laws Reduce the Financial Burden on Families of Children with Mental Health Care Needs?” Health Services Research 42:1061-84.

Bird

Chloe

Rieker

Patricia

. 2008. Gender and Health: The Effects of Constrained Choices and Social Policies. New York: Cambridge University Press.

Breslow

Rosalind

Smothers

Barbara

. 2005. “Drinking Patterns and Body Mass Index in Never Smokers National Health Interview Survey, 1997-2001.” American Journal of Epidemiology 161:368-76.

Byrne

Barbara

. 2012. Structural Equation Modeling with Mplus: Basic Concepts, Applications, and Programming. New York: Taylor & Francis.

10.

Derogatis

Leonard

. 2000. Brief Symptom Inventory–18 (BSI-18): Administration, Scoring, and Procedures Manual. Minneapolis, MN: National Computer Systems.

11.

Derogatis

Leonard

Fitzpatrick

Maureen

. 2004. “The SCL-90-R, the Brief Symptom Inventory (BSI), and the BSI-18.” Pp. 1-41 in The Use of Psychological Testing for Treatment Planning and Outcomes Assessment: Instruments for Adults, Vol. 3, edited by Maruish

Mahwah, NJ: Lawrence Erlbaum.

12.

Finkelstein

Amy

McKnight

Robin

. 2008. “What Did Medicare Do? The Initial Impact of Medicare on Mortality and Out of Pocket Medical Spending.” Journal of Public Economics 92:1644-68.

13.

Finkelstein

Amy

Taubman

Sarah

Wright

Bill

Bernstein

Mira

Gruber

Jonathan

Newhouse

Joseph

Allen

Heidi

Baicker

Katherine

. 2012. “The Oregon Health Insurance Experiment: Evidence from the First Year.” Quarterly Journal of Economics 127:1057-1106.

14.

Freeman

Joseph

Kadiyala

Srikanth

Bell

Janice

Martin

Diane

. 2008. “The Causal Effect of Health Insurance on Utilization and Outcomes in Adults: A Systematic Review of US Studies.” Medical Care 46:1023-32.

15.

Glenmullen

Joseph

. 2000. Prozac Backlash: Overcoming the Dangers of Prozac, Zoloft, Paxil, and Other Antidepressants with Safe, Effective Alternatives. New York: Simon & Schuster.

16.

Golberstein

Ezra

Busch

Susan

. 2013. “Two Steps Forward, One Step Back? Implications of the Supreme Court’s Health Reform Ruling for Individuals with Mental Illness.” JAMA Psychiatry 70:567-68.

17.

Hadley

Jack

. 2003. “Sicker and Poorer: The Consequences of Being Uninsured: A Review of the Research on the Relationship between Health Insurance, Medical Care Use, Health, Work, and Income.” Medical Care Research and Review 60:3S-75S.

18.

Harris

Katherine

Carpenter

Christopher

Bao

Yuhua

. 2006. “The Effects of State Parity Laws on the Use of Mental Health Care.” Medical Care 44:499-505.

19.

Hausman

Jerry

. 1978. “Specification Tests in Econometrics.” Econometrica 46:1251-71.

20.

Hill

Terrence

Angel

Ronald

. 2005. “Neighborhood Disorder, Psychological Distress, and Heavy Drinking.” Social Science & Medicine 61:965-75.

21.

Hill

Terrence

Ross

Catherine

Angel

Ronald

. 2005. “Neighborhood Disorder, Psychophysiological Distress, and Health.” Journal of Health and Social Behavior 46:170-86.

22.

Institute of Medicine. 2001. Coverage Matters: Insurance and Health Care. Retrieved September 21, 2013 (http://www.iom.edu/Reports/2001/Coverage-Matters-Insurance-and-Health-Care.aspx).

23.

Institute of Medicine. 2003. Hidden Costs, Value Lost: Uninsurance in America. Retrieved August 13, 2013 (http://www.iom.edu/Reports/2003/Hidden-Costs-Value-Lost-Uninsurance-in-America.aspx).

24.

Institute of Medicine. 2009. America’s Uninsured Crisis: Consequences for Health and Health Care. Retrieved September 21, 2013 (http://www.iom.edu/Reports/2009/Americas-Uninsured-Crisis-Consequences-for-Health-and-Health-Care.aspx).

25.

Jerant

Anthony

Fiscella

Kevin

Franks

Peter

. 2012. “Health Characteristics Associated with Gaining and Losing Private and Public Health Insurance: A National Study.” Medical Care 50:145-51.

26.

Kaiser Family Foundation. 2014. Women and Health Care in the Early Years of the Affordable Care Act Key: Findings from the 2013 Kaiser Women’s Health Survey. Retrieved June 5, 2014 (http://kaiserfamilyfoundation.files.wordpress.com/2014/05/8590-women-and-health-care-inthe-early-years-of-the-affordable-care-act.pdf).

27.

Kasper

Judith

Giovannini

Terence

Hoffman

Catherine

. 2000. “Gaining and Losing Health Insurance: Strengthening the Evidence for Effects on Access to Care and Health Outcomes.” Medical Care Research and Review 57:298-318.

28.

Kessler

Ronald

Nelson

Christopher

McGonagle

Katherine

Edlund

Mark

Frank

Richard

Leaf

Philip

. 1996. “The Epidemiology of Co-occurring Addictive and Mental Disorders: Implications for Prevention and Service Utilization.” American Journal of Orthopsychiatry 66:17-31.

29.

Klick

Jonathan

Markowitz

Sara

. 2006. “Are Mental Health Insurance Mandates Effective? Evidence from Suicides.” Health Economics 15:83-97.

30.

Kronick

Richard

. 2009. “Health Insurance Coverage and Mortality Revisited.” Health Research and Educational Trust 44:1211-31.

31.

Lave

Judith

Keane

Christopher

Lin

Chyongchiou

Ricci

Edmund

Amersbach

Gabriele

LaVallee

Charles

. 1998. “Impact of a Children’s Health Insurance Program on Newly Enrolled Children.” JAMA: The Journal of the American Medical Association 279:1820-25.

32.

Levy

Helen

Meltzer

David

. 2008. “The Impact of Health Insurance on Health.” Annual Review of Public Health 29:399-409.

33.

Liebert

Rachel

Gavey

Nicola

. 2009. “‘There Are Always Two Sides to These Things’: Managing the Dilemma of Serious Adverse Effects from SSRIs.” Social Science & Medicine 68:1882-91.

34.

Mechanic

David

. 2011. “Behavioral Health and Health Care Reform.” Journal of Health Politics, Policy and Law 36:527-31.

35.

McKee

Sherry

Krishnan-Sarin

Suchitra

Shi

Julia

Mase

Tricia

O’Malley

Stephanie

. 2006. “Modeling the Effect of Alcohol on Smoking Lapse Behavior.” Psychopharmacology 189:201-210.

36.

McArdle

Megan

. 2010. “Myth Diagnosis.” March 1, 2010. Available at The Atlantic. Retrieved August 19, 2013 (http://www.theatlantic.com/magazine/archive/2010/03/myth-diagnosis/307905/?single_page=true).

37.

McWilliams

J. Michael

. 2009. “Health Consequences of Uninsurance among Adults in the United States: Recent Evidence and Implications.” Milbank Quarterly 87:443-94.

38.

McWilliams

J. Michael

Zaslavsky

Alan

Meara

Ellen

Ayanian

John

. 2007a. “Health of Previously Uninsured Adults after Acquiring Medicare Coverage.” Journal of the American Medical Association 298:2886-94.

39.

McWilliams

J. Michael

Zaslavsky

Alan

Meara

Ellen

Ayanian

John

. 2007b. “Use of Health Services by Previously Uninsured Medicare Beneficiaries.” New England Journal of Medicine 357:143-53.

40.

Mirowsky

John

Ross

Catherine

. 2003. Social Causes of Psychological Distress. Hawthorne, NY: Aldine de Gruyter.

41.

National Alliance on Mental Illness. 2008. Coverage for All: Inclusion of Mental Illness and Substance Use Disorders in State Healthcare Reform Initiatives. Retrieved September 23, 2013 (http://www.nami.org/Template.cfm?Section=Issue_Spotlights&template=/ContentManagement/ContentDisplay.cfm&ContentID=119024).

42.

Newhouse

Joseph

. 1996. Free for All? Lessons from the RAND Health Insurance Experiment. Cambridge, MA: Harvard University Press.

43.

Pacula

Rosalie

Sturm

Roland

. 2000. “Mental Health Parity Legislation: Much Ado about Nothing?” Health Services Research 35:263-75.

44.

Pearlin

Leonard

Schieman

Scott

Fazio

Elena

Meersman

Stephen

. 2005. “Stress, Health, and the Life Course: Some Conceptual Perspectives.” Journal of Health and Social Behavior 46:205-19.

45.

Probst

Janice

Wang

Jong-Yi

Moore

Charity

Powell

M. Paige

Martin

Amy

. 2008. “Continuity of Health Insurance Coverage and Perceived Health at Age 40.” Medical Care 65:450-77.

46.

Regier

Darrel

Narrow

William

Rae

Donald

Manderscheid

Ronald

Locke

Ben

Goodwin

Frederick

. 1993. “The De Facto US Mental and Addictive Disorder Service System: Epidemiologic Catchment Area Prospective One-year Prevalence Rates of Disorders and Services.” Archives of General Psychiatry 50:85-94.

47.

Robert Wood Johnson Foundation. 2009. Health Care Consumer Confidence Index: A Monthly Survey of Americans’ Attitudes about Health Care. Retrieved July 9, 2013 (http://rwjf.org/content/dam/farm/reports/reports/2009/rwjf42110).

48.

Ross

Catherine

Mirowsky

John

. 2000. “Does Medical Insurance Contribute to Socioeconomic Differentials in Health?” Milbank Quarterly 78:291-321.

49.

Rowan

Kathleen

McAlpine

Donna

Blewett

Lynn

. 2013. “Access and Cost Barriers to Mental Health Care, by Insurance Status, 1999-2010.” Health Affairs 32:1723-30.

50.

Roy-Byrne

Peter

Craske

Michelle

Sullivan

Greer

Rose

Raphael

Edlund

Mark

Lang

Ariel

Bystritsky

Alexander

Welch

Stacy Shaw

Chavira

Denise A.

Golinelli

Daniela

Campbell-Sills

Laura

Sherbourne

Cathy D.

Stein

Murray B.

2010. “Delivery of Evidence-based Treatment for Multiple Anxiety Disorders in Primary Care.” Journal of the American Medical Association 303:1921-28.

51.

Sarata

Amanda

. 2011. Mental Health Parity and the Patient Protection and Affordable Care Act of 2010. Retrieved August 3, 2013 (http://artowerhcc.com/artower/resources/BHA%20~%208.pdf).

52.

Schoen

Cathy

DesRoches

Catherine

. 2000. “Uninsured and Unstably Insured: The Importance of Continuous Insurance Coverage.” Health Services Research 35:187-206.

53.

Schroeder

Ryan

Hill

Terrence

Haynes

Stacy

Bradley

Christopher

. 2011. “Physical Health and Crime: An Application of General Strain Theory.” Journal of Criminal Justice 39:21-29.

54.

Schroeder

Steven A.

2007. “We Can Do Better: Improving the Health of the American People.” New England Journal of Medicine 357(12):1221-28.

55.

Sherbourne

Cathy Donald

Dwight-Johnson

Megan

Klap

Ruth

. 2001. “Psychological Distress, Unmet Need, and Barriers to Mental Health Care for Women.” Women’s Health Issues 11:231-43.

56.

Sorlie

Paul

Johnson

Norman

Backlund

Eric

Bradham

Douglas

. 1994. “Mortality in the Uninsured Compared with That in Persons with Public and Private Health Insurance.” Archives of Internal Medicine 154:2409-16.

57.

Strine

Tara

Zack

Matthew

Dhingra

Satvinder

Druss

Benjamin

Simoes

Eduardo

. 2011. “Uninsurance among Nonelderly Adults with and without Frequent Mental and Physical Distress in the United States.” Psychiatric Services 62:1131-37.

58.

Tian

Donghua

Zhiyong

Wang

Xiaohua

Guo

Jing

Fan

Zhang

Xiulan

Chan

Cecilia Lai-Wan

. 2012. “The Role of Basic Health Insurance on Depression: An Epidemiological Cohort Study of a Randomized Community Sample in Northwest China.” Biomedical Central Psychiatry 12:151-62.

59.

Wang

Phillip

Lane

Michael

Olfson

Mark

Pincus

Harold

Wells

Kenneth

Kessler

Ronald

. 2005. “Twelve-month Use of Mental Health Services in the United States.” Archives of General Psychiatry 62:629-40.

60.

Weilburg

Jeffrey

O’Leary

Kathleen

Meigs

James

Hennen

John

Stafford

Randall

. 2003. “Evaluation of the Adequacy of Outpatient Antidepressant Treatment.” Psychiatric Services 54(9):1233-39.

61.

Wells

Kenneth

Sherbourne

Cathy Donald

Sturm

Roland

Young

Alexander S.

Burnam

M. Audrey

. 2002. “Alcohol, Drug Abuse, and Mental Health Care for Uninsured and Insured Adults.” Health Services Research 37:1055-66.

62.

West

Stephen

Taylor

Aaron

Wei

. 2012. “Model Fit and Model Selection in Structural Equation Modeling.” Pp. 209-31 in Handbook of Structural Equation Modeling, edited by Hoyle

New York: Guilford Press.

63.

Young

Alexander S.

Klap

Ruth

Sherbourne

Cathy D.

Wells

Kenneth B.

2001. “The Quality of Care for Depressive and Anxiety Disorders in the United States.” Archives of General Psychiatry 58(1):55-61.

64.

Zabora

James

Brintzenhofeszoc

Karlynn

Jacobsen

Paul

Curbow

Barbara

Piantadosi

Steven

Hooker

Craig

Owens

Albert

Derogatis

Leonard

. 2001. “A New Psychosocial Screening Instrument for Use with Cancer Patients.” Psychosomatics 42:241-46.