Abstract
Background:
The genesis of schizophrenia is multifactorial, including biological and environmental risk factors. We tested for an interactive effect between early-onset schizophrenia (EOS) and social class of origins (socioeconomic status (SES)). Data were further analyzed for a possible connection to type of schizophrenic symptoms.
Sampling/Methods:
Data for the study are taken from the medical records of 642 patients from a large state hospital in the northeastern United States. Clinical assessments were divided into positive and negative symptomatology through application of the Scale for the Assessment of Negative Symptoms (SANS), the Scale for the Assessment of Positive Symptoms (SAPS) and the Positive and Negative Syndrome Scale (PANSS). Detailed information about age of onset and SES of origin was obtained through Social Service Assessment interviews.
Results:
We uncovered a significant impact of EOS among the poor that elevates risk for negative symptomatology.
Conclusion:
Poor SES alone does not increase the likelihood of EOS, but it magnifies the deleterious effect of EOS on negative symptoms. Future research on these variables may inform the relative contribution of each.
Introduction
The origins of schizophrenia include both psychosocial and biological risk factors. Indeed, our own research has uncovered the etiological role of genetic loading (Jones, Gallagher, McFalls, & Pisa, 2008), prenatal stressors (Gallagher, McFalls, Jones, & Pisa, 1999), child abuse (Gallagher & Jones, 2016a), childhood neglect (Gallagher & Jones, 2016b), stressful life events (Gallagher, Jones, & Pardes, 2016) and obstetrical complications (Jones, Gallagher, Moss, & McFalls, 2011). Most of these findings have uncovered an additional risk for negative symptoms if the schizophrenic person was born into poverty. Is there a similar connection between social class of origin (socioeconomic status (SES)), schizophrenic symptoms and early age of onset (early-onset schizophrenia (EOS))? That is the central question of this study.
The typical time frame for the initial episode of schizophrenia is during late adolescence or early adulthood (Jablensky, 2000). From that point, it is characterized as following ‘a relapsing course for life’ (Davies, 1994; Walker, Kestler, Bollini, & Hochman, 2004). Although schizophrenia is rare in childhood (Fleischhaker et al., 2005), there are reports of EOS as young as 7 years of age (Hoff et al., 1996). Two meta-analyses of studies including EOS as a variable reported that the typical definition of EOS is before age 18 (Estherberg, Trotman, Holtzman, Compton, & Walker, 2010; Rajji, Ismail, & Mulsant, 2009). That is the way EOS is defined in this study.
Although studies of EOS are not common (Puig et al., 2012), almost every one of those studies clearly connect EOS with various forms of severe impairment (Rabinowitz, Levine, & Häfner, 2006). These include less favorable outcome (Asarnow, Tompson, & McGrath, 2004; Fleischhaker et al., 2005) and more severe cognitive and developmental deficits ( Rajji et al., 2009). Additionally, EOS has been associated with a host of brain abnormalities (Epstein et al., 2014; Gogtay, Vyas, Testa, Wood, & Pantelis, 2011; Hoff et al., 1996). Neuroimaging studies reveal unique insights into EOS. One such study reports that the area of brain dysfunction is especially prominent in the prefrontal cortex (Koike et al., 2011). Most germane to this study are reports of the association of EOS with severe negative symptoms (Puig et al., 2012; Rabinowitz et al., 2006; Schneider, Corrigall, Hayes, Kyriakopoulos, & Frangou, 2014).
This study assesses severity of EOS according to types of symptoms: negative form and positive form. We employ standardized measures of both groups of symptoms. Negative symptomatology is a condition with greater impairment (McGlashan & Fenton, 1992), poor prognosis and diminished responsiveness to medications (Edwards, McGarry, Waddell, & Harrigan, 1999; Malla, Norman, & Manchanda, 2004). Some contend that schizophrenia with negative symptoms has such poor outcome because it is linked with an abnormality in the physical structure of the brain (Andreasen, Arndt, Alliger, Miller, & Flaum, 1995; Nasrallah & Weinberger, 1990). This may explain why onset of illness is typically insidious for patients with negative symptoms (van Os, Jones, Sham, Bebbington, & Murray, 1998).
EOS is not solely connected to biological abnormalities. It is also associated with environmental risk factors (Rabinowitz et al., 2006). Indeed, some report low SES of origin as a contributing factor (Yeo et al., 2014). Researchers in this area theorize that the interconnection between EOS and low SES of origin results from poor individuals shouldering a disproportionate amount of stress (Grzywacz, Almeida, Neupert, & Ettner, 2004; Read, 2010). These include extreme exposures to violence and parental substance abuse (Parke & Clark-Stewart, 2003), among others. Parental neglect and lack of adequate supervision are also reported more frequently in low SES families as well as greater exposure to a range of environmental toxins (Schell, 1997). Other ‘mechanisms of harm’ connected to living in poverty are chaotic surroundings, inside and outside the home, few material and psychosocial investments to stimulate development, overcrowding, noise, substandard housing conditions, persistent ‘daily hassles’ (Grzywacz et al., 2004) and diminished prenatal/overall healthcare (Gallagher, Jones, McFalls, & Pisa, 2007). Although this is probably only a partial list, it clearly demonstrates an elevated risk for exposure to stressors (Evans & English, 2002) among impoverished individuals compared to their nonpoor counterparts. The sheer quantity and severity of these stressful life experiences support an extensive empirical and theoretical literature on the ‘psychological costs of growing up poor’ (Dearing, 2008). For these reasons, SES of origin is a key variable in this study.
Specifically, we hypothesize that there is an elevated risk of negative symptoms among EOS patients born into the lower class. To date, no published report has tested for these particular associations. It is the first such study of its kind.
Methods
Participants
Data for this study are taken from the cumulative anonymous medical records of 642 schizophrenic patients discharged from Norristown State Hospital (NSH) in Pennsylvania, USA, between 1984 and 1990. Diagnostic procedures employed multidisciplinary evaluations with periodic review. Specific criteria for index diagnosis were based on the Diagnostic and Statistical Manual of Mental Disorders (DSM, 3rd ed.; American Psychiatric Association, 1987).
Upon admission, patients were evaluated by staff psychiatrists and other members of a multidisciplinary team within 48 hours for diagnosis and treatment plan purposes. Later, diagnostic reviews were conducted for each patient every 3 months or as needed during hospitalization. Since some patients were discharged and readmitted over time, we employed a combination of three NSH operational measures to enhance longitudinal analysis of symptom stability. The measures include clinical assessments by NSH staff at intake and during last hospital stay as well as DSM diagnosis at last discharge.
Clinical assessments
In addition to diagnosis by DSM standards, NSH staff professionals further categorized patients into negative and positive subtypes as described above (Carpenter, 1994). Chart materials with detailed patient symptomatology enhanced subtyping in this study. A number of positive and negative scales have been retrospectively applied from chart materials in addition to subtyping drawn from patient files. They include the Scale for the Assessment of Negative Symptoms (SANS) (Andreasen & Olsen, 1982), the Scale for the Assessment of Positive Symptoms (SAPS) (Andreasen & Olsen, 1982) and the Positive and Negative Syndrome Scale (PANSS) (Kay, Fiszbein, & Opler, 1987).
Although some may question the validity and reliability of chart-based assessments of negative and positive symptoms, we do not think these are serious problems in this study. It was standard procedure at NSH to require that interview observation of the patient be completely and directly recorded onto the charts. NSH staff professionals solely conducted the clinical assessments. We retrospectively applied the identical assessments to our sample. Both the original assessments at NSH and our replication of those assessments were conducted independently of patient history of EOS. It has been documented that the retrospective applications of the SANS, SAPS and PANSS can be completed based on chart materials if the latter are sufficiently detailed (Fenton, McGlashan, Victor, & Blyer, 1997; Kay, Fiszbein, & Opler, 1987). Such was the case in this study.
One of the issues we faced was how to deal with diagnoses that changed over time. This proved to be a minor problem since this type of discrepancy rarely occurred and, when it did occur, we simply eliminated the case from the sample. Thus, as stated above, diagnosis was operationalized from three temporal sources: clinical assessment at first intake, during last hospital stay and DSM diagnosis at last discharge. The temporal points of these measurements not only permit the observation of symptom stability over time but also reflect reports that schizophrenic patients who show persistent positive or negative symptoms are important subgroups within the overall schizophrenic population (Malla et al., 2004).
Three independent raters who are experts in the field conducted negative/positive assessments. Consensus was reached on the classification of all included cases. Thus, interrater reliability is 100% because, in the rare instances where there was disagreement, the cases were dropped. The end result is that the sample only includes patients who clearly presented as negative or positive and were known to have/have not showed symptoms of schizophrenia before age 18. No evidence of extrapyramidal complications is present.
SES classification
Epidemiological analyses of SES and risk of schizophrenia are often confounded by the ‘social stress’ versus ‘social drift’ controversy. This study is not compromised by that debate, since SES is only assessed at the time of patients’ births. This provides a direct measure of a potential risk factor connected to SES of family of origin (Bresnahan & Susser, 2003).
Information about the SES of the patients as well as age of onset is contained in the ‘social history’ section of their hospital records. The social history typically includes detailed accounts of the family into which the patient was born. The histories were compiled at intake by psychiatric social workers from personal interviews with first-degree relatives. These standardized interviews regularly provide specific information about SES, such as the occupation, income status and level of education of the family head(s).
Researchers using occupational scales to rate SES frequently dichotomize social class into higher and lower class groups (Brown, Susser, Jandorf, & Bromet, 2000; Hollingshead, 1975). Consistently, our sample is bifurcated into low SES and high SES categories at the time of the patient’s birth. SES classification was facilitated by the application of the Occupational Distributions of the U.S. Bureau of the Census (Census of Population, 1971). This classification scheme is comprehensive and also temporally matches the time span of our data set. Whenever nonoccupational information was available in the social histories (e.g. ‘the family was well-off financially’), it was used to facilitate the dichotomous classification. Low SES includes the indigent, the unemployed and unskilled laborers; high SES generally consists of skilled laborers and above. Thus, the final sample comprised 642 cases that have been carefully delineated by schizophrenic subtype, age of onset and SES of origin.
Data analysis
This study is a convergence of two streams of literature about EOS. The first concerns symptom severity, specifically positive symptoms versus (more severe) negative symptoms. This dichotomous dependent variable is tested directly with a dichotomous independent variable divided by evidence versus no evidence of EOS. The first step in the analysis, therefore, is a simple 2 × 2 cross-tabulation of these two variables.
Step two involves the incorporation of the second literature stream about SES. As explained above, this variable has also been bifurcated into low versus high. The full analysis examines the simultaneous effect of the two dichotomous independent variables EOS and SES on the dependent variable schizophrenia subtype (coded 0 = positive symptoms, 1 = negative symptoms). The statistic of choice is analysis of variance (ANOVA), which can specify these categorical variables and also afford a direct test of statistical interaction.
Results
Table 1 commences the analysis by cross-tabulating EOS and schizophrenia subtype. In the top row for clients with no evidence of EOS, 73.0% are classified with positive and 27% with negative symptoms. In the evidence of EOS row, the split is 57.1% versus 42.9%, respectively. The apparent greater risk of negative subtype for EOS clients is verified by a statistically significant chi-square of 4.129 (p = .042). This confirms the expectation of the previous literature.
Evidence of early onset by schizophrenic symptoms.
N = 642.
To extend that confirmation, Figure 1 displays the results of an ANOVA incorporating SES. Recognizing that the height of the dots represents the distribution of negative versus positive symptoms, there is an apparent discrepancy in the course of this risk by SES. For high SES clients, early onset actually seems to drop the risk of negative subtype (the right-hand dot is below the left-hand dot). Low SES clients, however, show a very similar risk of negative symptoms when there is no evidence of EOS (the two dots at left), but a substantial rise in such risk where there is such evidence (the highest dot at right).

ANOVA of negative symptoms risk by EOS and SES.
This radical divergence of risk by SES is a textbook interaction effect, which tests out at an F-value of 6.747 (p = .10). For high SES clients, EOS is associated with a minor drop in negative subtype risk; for low SES clients, by statistically significant contrast, EOS is associated with a major uptick in such risk.
Discussion
Our study strongly suggests that the impact of EOS is elevated among the poor with negative symptomatology. No other study has tested for an interaction between EOS, type of schizophrenia and SES of origin. However, other researchers report a connection between some of these variables. EOS is widely accepted as carrying strong clinical and prognostic features (Öngür, Lin, & Cohen, 2009). Not only is it a major risk factor that worsens symptoms but also it is specifically known to be associated with negative symptoms (Hoff et al., 1996) and linearly corresponds with the severity of the course of the illness (Rabinowitz et al., 2006). Consistently, EOS is also associated with insidious onset, reduced social skills, long-term psychiatric treatment (Schneider et al., 2014) and unfavorable prognosis (Compton, Berez, & Walker, 2014). In fact, we only found one study of EOS that did not link it with worse symptoms and dismal prognosis (Amminger et al., 2011).
What is the role of SES as an independent variable in this study? Separate analyses (not shown) indicate that SES alone does not increase the likelihood of EOS but it somehow magnifies the deleterious effect of EOS on negative symptoms. This is only true for those of low SES of origin. For the nonpoor, the effect is the opposite. Earlier we listed a number of independent variables connected with being born poor and negative symptomatology (Compton et al., 2014; Gallagher, Jones, McFalls, & Pisa, 2006). EOS now occupies a place on that list. Future research on these variables may inform us of the relative contribution of each. We hypothesize that those studies will justify more analyses about the antecedents of schizophrenia (such as low SES of origin) rather than the consequences of age of onset.
This study has some clear limitations. First, it is limited by its provisional nature. Second, the analysis utilizes chart-based assessments. As noted earlier, we do not believe that validity and reliability are compromised by this fact. Third, we only examine the combined effect of SES of origin and age of onset on type of symptomatology whereas a number of other factors, such as genetic loading and prenatal stressors, may be at work as well. Fourth, we are not able to measure the specific effect of SES and EOS separately but only as a combined variable. Fifth, we are not able to control for variables such as gender and race because of sample size. Sixth, our data set, while extremely detailed, only includes patients from a single state hospital in the northeastern United States. Consequently, the cross-cultural representativeness of the sample (N = 642) cannot be exhaustively tested.
The study’s strengths are threefold. First, it measures schizophrenia by subtype, whereas many analyses of risk for schizophrenia fail to specify diagnosis. Some simply refer to schizophrenia as a general condition while others clump the entire group of schizophrenia spectrum disorders into one category. Second, it is the only study to test for an interaction between EOS, schizophrenic symptoms and SES of origin. Third, it uncovers more of the complex interplay of environmental stressors on the genesis of schizophrenia and calls for more studies of this type.
Footnotes
Acknowledgements
The authors gratefully acknowledge the work of Marge DiLullo, Director of Medical Records at the state hospital, for her painstaking efforts in collecting detailed clinical and social histories of the patients for this study. This study is dedicated to her memory. The authors are also grateful to Susan A. Bur for her technical support.
Funding
The author(s) received no financial support for the research, authorship and/or publication of this article.
