Suffering related to persistent anxiety has been an important focus of the mental health community since the nineteenth century (Kessler et al. 2004). Although recent trends equate anxiety with disorder, anxiety in the human condition is ubiquitous. Anxiety has been the focus of philosophers such as Kierkegaard, philosophical movements such as existentialism, artists such as Edvard Munch, and literary writers such as Proust and Poe. The poet W. H. Auden described post World War II as the “age of anxiety” and described the anxiety of this era as a natural response to the devastation of modern war, the development of nuclear arsenals, concentration camps and related atrocities. It is the distress of anxiety that makes it a social work concern. In fact, the root meaning of the word anxiety is to “vex or trouble.” Given the ubiquitous nature of anxiety, it is important to establish whether states of anxiety should come under the rubric of mental disorder or be addressed in other ways.

Several scholars have recently focused on the conflation of pain and suffering with mental disorder (Horwitz and Wakefield 2007). According to these scholars, psychiatry has developed overly inclusive definitions of disorder. Currently disorders are based on the Diagnostic and Statistical Manual of Mental Disorders (DSM). The DSM uses symptom based criteria to determine disorder and recent versions do not give attention to context or etiology. A major confound is that the same symptoms that determine disorder often occur in response to life events in the absence of disorder. Generalized Anxiety Disorder (GAD) is one such example. Social workers, who use a person-in-environment lens, are concerned because the role of environmental social conditions in the etiology of psychological distress are often overlooked or underestimated.

A review of iterations of the DSM shows that there has never been conceptual clarity in the definition of GAD, although it has undergone extensive changes since the advent of the initial classification system over 30 years ago. Conceptual shifts in different versions of the DSM are emblematic of the confusion surrounding this diagnostic category (Brown et al. 1993). According to the psychiatric literature, the viability of GAD to stand as a distinct disorder has been questionable. At times it has been conceptualized as a composite of symptoms, which consists of a residual from other disorders. DSM-I (1952) depicted anxiety as the flag ship of neurotic disorders. This conceptualization subsequently evolved into a subcategory of “anxiety neurosis” in DSM-II (1968). GAD became a diagnostic category in its own right in the Research Diagnostic Criteria (RDC), which defined it as a disorder distinct from panic disorder. This was a prelude for the inclusion of GAD in DSM-III (1980) as a separate diagnostic category albeit a residual category incorporating situations where the criteria for other categories could not be met. In DSM-III GAD was defined as an occurrence for at least 1 month duration of generalized, persistent anxiety with the presence of symptoms from at least three of four categories including: motor tension, autonomic hyperactivity, apprehensive expectation, vigilance and scanning. Ongoing empirical research identified the need to refine the diagnostic criteria to take into account both the etiology and contextual basis in which the symptoms occurred; these changes were included in DSM-III-R (1987). DSM-III-R criteria were: excessive and/or unrealistic worry in two or more areas unrelated to other Axis I diagnostic categories, associated symptom criteria meeting 6 symptoms out of 18, an increase in duration from 1 to 6 months and finally, and perhaps most significantly, that it no longer be considered a residual category (Hettema et al. 2004).

DSM-IV (1994) changes intended to make GAD more accessible. In DSM-IV, GAD was defined by the presence of these symptoms: excessive, difficult to control worry about a number of life events/activities, accompanied by at least three of six associated symptoms of negative affect/tension (i.e., restlessness/feeling keyed up or on edge, fatigability, concentration difficulties, irritability, muscle tension, and sleep disturbance). Duration criteria required that symptoms be present for at least 6 months and result in impairment in social, occupational, or other important areas of functioning. The differential diagnosis guidelines specified that GAD should not be assigned if its features occurred during the course of a mood disorder, posttraumatic stress disorder, psychotic disorder, pervasive developmental disorder or if they were related to substance abuse. Notably absent from the diagnostic criteria after DSM-III-R was an evaluation of the social contextual environment in which symptoms occurred. A historical review of the nosology of mental disorder from the time of ancient Greece up until the twentieth century indicates that context consistently was integral to the determination of disorder. In other words affective reactions that arose in the absence of situations that would produce them were essential factors in determining the presence of disorder (Horwitz and Wakefield 2007). Thus, the deletion of context in versions following DSM-III-R was a radical conceptual shift. Moreover, it was neither anchored in evidence nor logic.

One of the most notable facts about GAD is that it is commonly diagnosed in the socially disadvantaged, especially those who live in poverty (Blazer et al. 1991; Brawman-Mintzer and Lydiard 1996; Kessler and Wittchen 2002; Wittchen et al. 1994). Prior research indicates that GAD has a tendency to aggregate in families, but genetics have not been found to play a substantial role (McLaughlin et al. 2008). For example, children of parents with an anxiety disorder are seven times more likely to develop an anxious condition than those children with parents lacking an anxiety disorder (Turner et al. 1987). Recent prospective data have shown that a number of environmental risk factors measured in childhood were associated with the future onset of GAD. These included quality of home environment, especially low socioeconomic status, maltreatment and maternal internalizing symptoms (Moffitt et al. 2007). It is notable that these factors were found to be specifically associated with the development of GAD and did not represent global risk factors for any other type of psychopathology (Moffitt et al. 2007). Despite evidence reported from some twin studies of low heritability of anxiety (Kendler et al. 1992; Torgensen 1983), a meta-analysis conducted by Hettema et al. (2001) contended that while GAD aggregates in families, hypotheses about genetic factors were not confirmed.

Thus this paper argues that DSM definitions have led to an expansion in the domains of what is considered disorder to such an extent that in some cases the normal vicissitudes of existence have been pathologized. Moreover, social factors, which are importantly involved in many samples used to study GAD, have been overlooked and remain unrecognized. The purpose of this study was to explore the relationship between maternal anxiety and poverty as well as to investigate the aggregation of GAD in poor families by examining anxiety, parental stress and types of maternal involvement as a possible pathway for the development of anxiety in offspring. We posed the fundamental question: Is this GAD or is this the stress of poverty?

Methodology

This was a secondary analysis of data from the Fragile Families and Child Wellbeing Study (N = 4,898). The Fragile Families and Child Well-being Study, is a nationally representative birth-cohort study designed to provide longitudinal information about new parents and their biological children in urban areas. The Fragile Families study follows 4,898 children born between 1998 and 2000 (including 3,712 children born to unmarried parents and 1,186 children born to married parents). The sampled cities were among 77 U.S. cities with a population over 200,000. Sixteen cities were selected using a stratified random sampling frame and later five cities were added. Within the cities, hospitals with over 1,000 non-marital births a year were randomly sampled. Births were sampled until preset quotas were reached. The unwed subsample was over-sampled to obtain better data on unwed parents (Reichman et al. 2001). At baseline, the mean age of mothers was 25 (range = 14–50), and fathers were on average 2 years older. The typical age profile of mothers was skewed toward 18 years and up because of IRB restrictions at certain hospitals that did not allow minors to be questioned. In addition, mothers who planned to give their baby up for adoption, those who were too ill, and those who could not speak English or Spanish well enough to complete the interview were also excluded (Reichman et al. 2001).

Baseline interviews were conducted shortly after the child’s birth with follow-up interviews when the child was approximately 1, 3 and 5 years of age. Response rates were 87 % for unmarried mothers and 82 % for married mothers. The attrition rate from the baseline surveys to the 3-year follow-up home observations was 22.89 %.

The present study used data from the 3-year follow-up surveys and 3-year home observations since one of our primary research questions was to investigate the role of poverty and maternal GAD in children’s mental health at a point when it was possible to adequately investigate child outcomes. Prior studies have investigated children’s’ mental health status at an early age (Duggal et al. 2001; Fanti and Henrich 2010; Sterba et al. 2007).

Sample

Inclusion criteria of this study were: random sample, and dyads living together most time. Thus, we included cases from 16 cities and omitted cases from 5 cities to avoid a bias in generalizing the results back to the population. Applying the inclusion criteria, the sample included 2,338 dyads of mothers (Mean age in years = 27.97, SD = 6.06, range = 16–50) and their children (Mean age in months = 35.34, SD = 2.34, range = 30–46) out of the eligible cases at the 3-year surveys and home observations.

Weights were used to make the sample nationally representative of both unwed and wed births. Jackknifing, a replication sampling variance estimation technique was applied using a set of replicate weights to estimate variance. Replicate weights are a series of variables that contain the information necessary for correctly computing the standard errors of point estimates when analyzing complex survey data. The replicate weights in the Fragile Families and Child Wellbeing Study account for the strata and the differential sampling rates characteristic of the primary sampling unit (Carlson and Mathematica Policy Research, April 2008). After both weighting and jackknifing, the sample consisted of 863,162 mothers (Mean age in years = 29.78, Jackknife SE = .16) and their children (Mean age in months = 34.45, Jackknife SE = .11). The weighted children were comprised of 55 % boys and 45 % girls. The weighted percentage of mothers, who were diagnosed with GAD at the third year, was 4.4 %. Other characteristics of the population were: non-Hispanic White (37 %), non-Hispanic Black (22 %), Hispanic/Latino (13 %), and Other (29 %). For mother’s education at the child’s birth, 41 % had some college or more, 30 % equal to high school, and 29 % less than high school. With regard to relationship status, 60 % were married, 18 % cohabiting, 6 % romantically involved, and 16 % not married, not romantically involved.

Measures

Generalized Anxiety Disorder (GAD)

The Composite International Diagnostic Interview—Short Form (CIDI-SF: Kessler et al. 1998) was used to assess GAD. The CIDI-SF is based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) and is a standardized instrument for use in epidemiological, cross-cultural, and other research studies. GAD was measured using categorical variables coded 0 = no, 1 = yes. A probability of GAD equal to 1 was assigned when the following four diagnostic stems were met. The diagnostic stems were: an anxious period lasting at least 6 months during the past 12 months; further qualifiers, for example were: “During that/this period, (was/is) your worry stronger than in other people?” and “(Did/Do) you worry most days?” all of which were necessary qualifiers for DSM-IV GAD criterion A; lack of control over this anxious period, which was necessary to meet DSM-IV GAD criterion B, and at least three types of psychological symptoms, when they were worried or anxious, which is necessary to meet DSM-IV GAD criterion C. Psychological symptoms were assessed using the following questions: Restless? (Were/Are) you keyed up or on edge? (Were/Are) you easily tired? (Did/Do) you have difficulty keeping your mind on what you were doing? (Were/Are) you more irritable than usual? (Did/Do) you have tense, sore or aching muscles? (Did/Do) you have trouble falling asleep or staying asleep?

Child’s Internalizing Behavior Problems

Two types of internalizing behavior problems, withdrawn and anxious/depressed symptoms were assessed using the Child Behavior Check List (CBCL: Achenbach and Rescorla 2000). Mothers rated as 0 (not true in the last 6 months), 1 (somewhat or sometimes true), or 2 (very true or often true) for their child.

The 8 item withdrawn symptoms included: Acts too young for his/her age, avoids looking at others in the eye, does not answer when people talk, refuses to play games, is unresponsive to affection, shows little affection toward people, demonstrates little interest in things around the child, and is uninvolved with others. The 7 item anxious/depressed symptoms included: Looks unhappy without good reason, feels nervous, feels overtired, feels self-conscious or easily embarrassed, is too shy or timid, feels too fearful or anxious, and feels unhappy, sad, or depressed. The scores of the withdrawn symptom items were summed to derive a total score (ranged from 0 to 16), and higher scores indicated more withdrawn symptoms (internal consistency reliability = .67). The scores of the anxious/depressed symptom items were summed to derive a total score (ranged from 0 to 14), and higher scores indicated more anxious/depressed symptoms (internal consistency reliability = .61).

Poverty

The five items were used to assess poverty: Was your telephone service ever disconnected? Did you receive free food/meals? Did you not pay full rent/mortgage? Did you not pay full gas/oil/electric city bills? Did you move in with other people because of financial problems? Responses were in the form of yes (1) or no (0). Responses to the five items were summed to derive a total score (ranged from 0 to 5), and higher scores indicated more poverty. Lyons et al. (2005) conducted a confirmatory factor analysis for the items asking about food, rent, utilities, and clothes to obtain a latent factor of financial strain. Results were: food, R 2 = .78, for rent, R 2 = .61, for utilities, R 2 = .57, and for clothes, R 2 = .77; internal consistency reliability was .54.

Maternal Parenting Stress

Őstberg and Hagakull (2000) conceptualized parenting stress as a condition where the different aspects of parenthood result in a perceived discrepancy between situational demands and personal resources. The items used in the Fragile Family Studies data were borrowed and/or adapted from the Early Head Start Study’s items regarding parenting stress. In the present study, parenting stress was represented by the aggregation of 5 items asking about parental stress related to the child. Mothers responded to a 5 point likert-type scale as follows: (1 = strongly agree and 5 = strongly disagree). The items that tapped stress were: Did you find yourself giving up more of you to meet child’s need than expected, do you feel trapped by parental responsibility, are you unable to do new and different things since having the child, are you almost never able to do things you like to do since having the child, and do you feel that having the child caused more problems than you expected in your relationships with men. Each score was reverse coded and summed to derive a total score (from 0 to 20), and higher scores indicated more parenting stress. The internal consistency was .75.

Maternal Parenting

Maternal parenting behaviors toward the child were conceptualized as the mother’s engagement in parent–child interactions that are generally practiced every day and that are assumed to represent maternal sensitivity to the child’s signal. Two types of parenting behaviors were measured: nurturing parenting was conceptualized as positively stimulating, caring, and affection-showing behaviors; and punitive parenting was conceptualized as interfering, negative-emotion discharging, and punishing behaviors. Mothers parenting behaviors were measured during home observation using the Home Observation Measurement of the Environment (HOME: Caldwell and Bradley 2001). The 6 items which tapped nurturing parenting included: Spontaneously vocalized to the child at least twice, responded verbally to the child’s vocalizations, spontaneously praised the child at least twice, told the child the name of an object or person, mother’s voice conveyed positive feelings toward the child, and mother caressed or kissed the child at least once during the visit. The 5 items for punitive parenting were: Shouted at the child, expressed annoyance/hostility toward child, slapped/spanked, scolded/criticized, and interfered with the child more than three times during the visit. Each item was scored as 0 (no) or 1 (yes). The scores of nurturing parenting items were summed to derive a total score (from 0 to 6), and higher scores indicated more nurturing parenting. The internal consistency reliability was .74. The scores of punitive parenting items were summed to derive a total score (from 0 to 5), and higher scores indicated more punitive parenting. The internal consistency reliability was .76.

Analytic Strategies

The analysis for the relationship between GAD and the poverty consisted of logistic regression using Stata 11.1. Structural Equation Modeling (SEM) was conducted using Mplus 5 to test the conceptual model (shown in Fig. 1). Parameter estimates were calculated via WLSMV (Weighted Least Square Parameter Estimates) that use a full weight matrix (Muthén and Muthén 2010). Given the categorical nature of some of the data and following Muthen’s recommendation, RMSEA (Root Mean Square Residual) was used to test model fit. As for RMSEA, MacCallum et al. (1996) have used .01, .05, and .08 to indicate excellent, good, and mediocre fit respectively.

Fig. 1
figure 1

The conceptual model. RMSEA = .000; **p < .01; ***p < .001

Full size image

Results

Results of the logistic regression showed that mothers who received free food had a 2.5 odds of being in the GAD group (p < .001). Mothers who had problems paying their utilities had a 2.44 (p < .001) odds of being in the GAD group; mothers who had to move in with others had a 1.90 (p < .04) odds of being in the GAD group. Having the phone disconnected and being unable to pay rent were not significant and were a minority in the sample.

Conceptual Model Testing

A test of the conceptual model results are presented in Fig. 1. There were no problems in convergence, inadmissible solutions, or under identification of the model. Fit indices for the model were good (RMSEA = .000). As shown the coefficient for the relationship between poverty and generalized anxiety disorder was .027 (p < .01). Maternal parenting stress was positively associated with punitive parenting, r = .217 (p < .01). In turn, punitive parenting was negatively associated with nurturing parenting, r = −.334 (p < .001). Children’s withdrawn and anxious/depressed symptoms were positively associated, r = .025 (p < .01). Finally, parenting stress was positively associated with both children’s being withdrawn (r = .045, p < .01) and anxious/depressed symptoms (r = .034, p < .01).

Discussion

This study examined the relationship between poverty and generalized anxiety disorder. The findings confirmed the hypotheses that the poorest mothers had greater odds of being classified as having generalized anxiety disorder. While we agree that high levels of stress, over long periods of time can lead to psychological problems, we conclude that our findings fall under and support Wakefield’s “harmful dysfunction” concept of mental disorder. The harmful dysfunction concept posits that dysfunction is an inability of some internal mechanism to perform its natural function (Wakefield 1992). In the case of poor mothers, there is no evidence for a malfunction of some internal mechanism, there is physical need in the real world that is unmet and produces anxiety. This is an important distinction to make because interventions are predicated on ways in which problems are conceptualized. Our findings suggest that anxiety in poor mothers is usually not psychiatric, but a reaction to severe environmental deficits. Thus assessment should include careful attention to the contextual factors and environmental deficits as important precipitants and or concomitants to symptom presentation.

Our conceptual model examined the relationships between poverty, maternal parenting, stress, and child behaviors in order to test the idea that poverty leads to maternal stress, which affects or impedes positive parenting in such a way that the child may be adversely affected. While this is not a new idea in the parenting literature, it does investigate the question of aggregation of anxiety in poor families. Our findings showed that the path from anxiety to parenting stress was not supported. This suggests that mothers can be poor and anxious, but still provide positive parenting for their children. Nonetheless the picture is complicated as the path from stress to negative parenting was supported. This finding also has implications for practice. It suggests that in addition to concrete support, poor mothers may benefit from interventions designed to increase their capacity for coping with stress in addition to training in positive parenting skills.

The broader implications of our findings underscore the conceptual confusion in the GAD nosology, which occur in part due to the lack of attention to context factors. Wakefield (1994) has noted that the conceptual blurring of mental disorder with the normal responses to the vicissitudes of life masks the occurrence of true mental disorder and acts as an impediment to scholarship about true disorder.

In sum, our finding that poverty was positively associated with GAD supports our hypotheses that poverty is positively associated with anxiety. Moreover, these results support Wakefield’s conceptual definition of disorder as harmful dysfunction, e.g., symptoms in the absence of situations that would produce them; symptoms caused by some dysfunction in normal functioning. Additionally, our findings, while small, have implications for epidemiological studies of mental disorder, which are often used to justify mental health service need. Most epidemiological studies conducted in the U.S. justify the use of symptom-based criteria to calculate prevalence data in population studies (see Ron Kessler’s work). Such practices reinforce the general assumption that symptoms represent disorder and often blur true information about prevalence. Under these confounding conditions, it is hard for science to move forward.

Furthermore, labeling an individual with a diagnosis especially if it is inaccurate has a serious social stigma. Many people, who do not have a mental disorder, seek mental health services perhaps as a point of entry to receive additional sources of support for non-psychiatric issues. Thus, we endorse views that propose that the need for services should not be predicated on the basis of mental disorder. The current symptom—based conceptions of mental disorder expand the range of conditions that can be the legitimate objects of psychiatric management.

Finally, a limitation of this study contains the small number of mother with GAD because our data is drawn from a community sample. Thus, although the relation between GAD and poverty was significant, it was not substantive. Future research should be conducted with clinical samples, which would provide a better understanding of the role of mother’s GAD and poverty in children’s internalizing symptoms as well as investigate whether long-term poverty results in true brain changes which result in generalized anxiety disorder.