Schizophrenia Research 156 (2014) 115–121

Contents lists available at ScienceDirect

Schizophrenia Research

j ourna l homepage: www.e lsev ie r .com/ locate /schres

Identifying aspects of neighbourhood deprivation associated withincreased incidence of schizophrenia

Vishal Bhavsar a,⁎, Jane Boydell a, Robin Murray a, Paddy Power b

a Institute of Psychiatry, De Crespigny Park Road, London SE5 8AF, United Kingdomb St. Patrick's Hospital, James Street, Dublin 8, Ireland

⁎ Corresponding author.E-mail addresses: Vishal.2.bhavsar@kcl.ac.uk (V. Bhavs

(J. Boydell), Robin.murray@kcl.ac.uk (R. Murray).

http://dx.doi.org/10.1016/j.schres.2014.03.0140920-9964/Crown Copyright © 2014 Published by Elsevie

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 11 April 2013Received in revised form 17 February 2014Accepted 16 March 2014Available online 14 April 2014

Keywords:EpidemiologySchizophreniaIncidenceDeprivation

Background: Several studies have found an association between area deprivation and incidence of schizophrenia.However, not all studies have concurred and definitions of deprivation have varied between studies. Relativedeprivation and inequality seem to be particularly important, but which aspects of deprivation or how this effectmight operate is not known.Methods: The Lambeth Early Onset case register is a database of all cases of first episode psychosis aged 16 to35 years from the London Borough of Lambeth, a highly urban area. We identified 405 people with firstonset schizophrenia who presented between 2000 and 2007. We calculated the overall incidence of first onsetschizophrenia and tested for an association with area-level deprivation, using a multi-domain index of depriva-tion (IMD 2004). Specific analyses into associations with individual sub-domains of deprivation were thenundertaken.

Results: Incidence rates, directly standardized for age and gender, were calculated for Lambeth at two geograph-ical levels (small and large neighbourhood level). The Poisson regression model predicting incidence rate ratiosfor schizophrenia using overall deprivation scorewas statistically significant at both levels after adjusting for eth-nicity, ethnic density, population density and population turnover. The incidence rate ratio for electoral warddeprivation was 1.03 (95% CI = 1.004–1.04) and for the super output area deprivation was 1.04 (95% CI =1.02–1.06). The individual domains of crime, employment deprivation and educational deprivation were statis-tically significant predictors of incidence but, after adjusting for the other domains aswell as age, gender, ethnic-ity and population density, only crime and educational deprivation, remained statistically significant. Lowincome, poor housing and deprived living environment did not predict incidence.Conclusions: In a highly urban area, an association was found between area-level deprivation and incidence ofschizophrenia, after controlling for age, gender, ethnicity and population density; high crime and low levels ofeducation accounted for this. As both of these are potentiallymodifiable, this suggests a possiblemeans to reducethe incidence of schizophrenia.

Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved.

1. Introduction

There is considerable interest in the relationship between the socialenvironment and psychosis(Allardyce and Boydell, 2006). This focus issupported by the now repeated observation that schizophrenia inci-dence varies considerably between neighbourhoods in a single area(Boydell et al., 2001; March et al., 2008). Kirkbride et al. (2007) founda wide variation in the incidence of broadly defined psychosis betweenelectoral wards in South East London; applying Bayesian methods ap-propriate for modelling the spatial patterning of rare events, rate ratios

ar), Jane.boydell@kcl.ac.uk

r B.V. All rights reserved.

were found to range between aminimum of 0.48 to a maximum of 2.33relative to the average. There is now good evidence for the role of con-textual exposures on schizophrenia rates, for example, in relation toethnic density (Boydell et al., 2001; Kirkbride et al., 2008; Veling et al.,2008), social fragmentation (Allardyce et al., 2005) and social isolation(van Os et al., 2000). (See Fig. 1.)

Higher rates of schizophrenia have been found in deprived areas, butfindings have been inconsistent (Harrison et al., 1995; Boardman et al.,1997; Koppel and McGuffin, 1999).Thornicroft et al. (1993) providedevidence that the effect might not hold in rural areas. Giggs (1973)examined the geographical distribution of all patients admitted forschizophrenia in Nottingham. An increased concentration of cases inthe inner city was observed, particularly in areas with concentratedlow social status, high unemployment and single-person households.The association seems to be stronger between relative deprivation and

0 1 20.5 Miles 0 1 20.5 Miles

0 1 20.5 Miles0 1 20.5 Miles

SOA deprivation in quintiles

Least deprived SOAs

Most deprived SOAs

Ward deprivation in quintiles

SOA incidence by quintiles

Ward deprivation by quintiles Ward incidence by quintile

Lowest incidence SOAs

Highest incidence SOAs

Least deprived wards

Most deprived wards

Lowest incidence wards

Highest incidence wards

Fig. 1.Maps of Lambeth displaying ward and SOA incidence of schizophrenia and ward and SOA deprivation score.

116 V. Bhavsar et al. / Schizophrenia Research 156 (2014) 115–121

schizophrenia incidence rather than absolute deprivation. Inequalitywithin neighbourhoods has been shown to predict schizophrenia inci-dence (Boydell et al., 2004). Kirkbride et al. (2012) found that area-level relative deprivation and inequality were strongly associated with

incidence of non-affective psychosis after adjusting for individual-levelsocioeconomic status. This suggests a true ecological phenomenon.

Socioeconomic deprivation is a biologically plausible causal factorfor psychosis (Morgan et al., 2010). However, epidemiological evidence

117V. Bhavsar et al. / Schizophrenia Research 156 (2014) 115–121

for a causal effect of socioeconomic status on risk for schizophrenia isequivocal. Early studies indicated that schizophrenia cases were morelikely to come from lower social groups (Hollingshead and Redlich,1958; Dohrenwend, 1990). Hare (1956)found amarked preponderanceof schizophrenia in people from the lowest two social classes. There hasbeen long recognition that the association between socioeconomic sta-tus and schizophrenia is due in part to the social decline of individualswith the established disorder (Goldberg and Morrison, 1963). More re-cently, a case–control study in South London found that cases weremore likely to have been born in deprived areas and to have fatherswith manual occupations (Castle et al., 1993). It also appears that bothbeing born in a deprived area and being of a low social class interactto increase risk for the disorder (Harrison et al., 2003). Werner et al.(2007) observed an association with schizophrenia risk for both pater-nal socioeconomic status and community level socioeconomic statusat birth. However, in Byrne et al.'s (2004) study of the Danish NationalRegister, low socioeconomic status was strongly associated with onset,but the picture at birth wasmore complicated—larger families, parentalsocioeconomic status and low parental income were associated with aslightly increased risk, but overall wealth was not. Mulvany et al.(2001) found that schizophrenia was slightly more common in highersocial classes but found an effect of age, with cases from lower socialclasses presenting later in life. Large birth cohort studies have notfoundan effect of socioeconomic status at birth on risk for schizophrenia(Jones et al., 1993, 1994). At an individual level then, the relationshipbetween social deprivation and risk for psychosis is complex and mayvary over the life course.

In this study,we examine schizophrenia in people aged 16–35 years,employing a recently defined geographical construct, super outputareas (Bond and Insalaco, 2007), that allows identification and analysisof “pockets” of deprivation nested within larger areas as well as themore traditional electoral wards. We employed a deprivation indexthat was derived from UK census indices (ODPM, 2004) and whichaims tomeasure the overall deprivation of an area, aswell as seven con-stituent “sub-domains” of crime, education, living environment, healthand disability, housing, income, employment and does not includeage, gender or ethnicity.

We aimed to determine whether deprivation predicted incidence ofschizophrenia after adjusting for other environmental predictors and, ifso, which aspects of deprivation were predictive.

2. Methods

2.1. Case finding

We generated a list of all patients (aged 16–35 years) presenting forthe first time with possible schizophrenia to mental health services inthe London borough of Lambeth (population 267,000) using the referralregisters of the Lambeth Early Onset (LEO) service (Power et al., 2007).Clinical records were examined to ensure that patients were true inci-dent cases and were residents in Lambeth at the time of contact. Casespresenting outside the study period were excluded, as were patientswho were not resident in Lambeth at the time of first contact. Thecase list was cross-checked with the case-log of another first-episodestudy in progress, the GAP study (Di Forti et al., 2009), whose study pe-riod (2004–2008) overlapped with ours. A leakage study carried out aspart of a concomitant randomised trial did not identify any leakage ofcases (Power et al., 2007).

Thirteen hundred and forty-two referrals of suspected psychosiswere made to LEO during the period 1/1/2000–31/12/2007. Seven hun-dred and forty-nine were assessed as suffering from first-episode psy-chosis (defined as a week or more of unremitting positive psychoticsymptoms). Fifty-three cases were excluded because the first assess-ment and diagnosis of psychosis by mental health services was madeoutside the study period (27 before and 26 after). Forty-one caseswere excluded because they were either homeless or residing outside

Lambeth at first contact. Seventeen cases were excluded as they wereoutside the study's target age range (16–35 year olds) at first contact.This left 638 Lambeth residents aged 16–35 years who presented forthe first time to mental health services with their first episode of psy-chosis during the 8-year study period. Routine demographics weregathered for all incident cases. The 16-point ethnic categories from the2001 UK census were grouped into White, Black Caribbean, BlackAfrican, Asian and others. Individual socioeconomic status in 11 group-ingswas assigned according to theONS socioeconomic classification ap-plied in the UK Census.

2.2. Diagnostic procedure

OPCRIT items were used to generate standardized diagnoses fromthe first 6 months of case records, applying the RDC criteria for“narrowly defined” schizophrenia (McGuffin et al., 1984) to all thefirst-episode psychosis patients (n = 638) identified by the case-finding exercise above. This work was carried out by authors VB andJB. Forty-six cases were excluded because not enough evidence wasavailable in the clinical files to generate an RDCdiagnosis of Schizophre-nia. One hundred and eighty-seven cases were excluded because theydid not fulfil RDC criteria for schizophrenia (minimumof 2 weeks of un-remitting non-affective positive psychotic symptoms). This left 405cases presenting with first-episode schizophrenia.

2.3. Super output areas

Super output areas (SOAs) are geographic units designed to aid thereporting of area-statistics in the UK. Each SOA contains amean popula-tion of 1500 (minimum 1000). SOA boundaries were generated in 2004by a computerized aggregation of census output areas, accounting forpopulation size, proximity and social homogeneity. SOAs are nestedwithin electoral wards, allowing the analysis of incidence rates at bothSOA and ward level. Lambeth is made up of 177 SOAs in 21 electoralwards. Lambeth SOAs ranged in population from 1100 to 1770, withan interquartile range of 1500–1600 (Vickers and Rees, 2007).

2.4. Indices of deprivation

We used the Index of Multiple Deprivation (IMD) 2004 (ODPM,2004) as the measure of area-level deprivation. The IMD for England isa weighted, exponentially transformed aggregate score of seven indicesof deprivation. A bigger score indicates a more deprived area. A singlescore is assigned to each SOA on the basis of its seven domains of depri-vation (income deprivation; employment deprivation; health depriva-tion and disability; education, skills and training deprivation; barriersto housing and services; living environment deprivation; crime depriva-tion) (see Table 1).

Population-weighted total deprivation scores (excluding the healthdeprivation domain, as this would have been influenced by the out-come) were generated for each electoral ward by totalling the scoresfor the contained SOAs and weighting the aggregate score accordingto the contribution of each SOA to the ward population. The IMD 2004has been used to analyse health outcomes in a variety of settings, in-cluding primary care (Ashworth et al., 2007), cancer care (Stephenset al., 2005) and cardiovascular health (Ferguson et al., 2010). TheIMD2004 should represent an advance over previousmeasures becauseit does not incorporate demographic indices, such as age, gender or eth-nicity, thereby allowing for the testing of these factors separately fromdeprivation (ODPM, 2004).

2.5. Population at risk

Population estimates for each SOA in Lambeth were obtained fromthe Office of National Statistics (ONS), stratified by gender and age, formid-2001 and mid-2006. The population for the intervening years

Table 1To show indices making up the English Indices of Deprivation 2004(derived mainly from 2001 census data) courtesy of Office of Deputy Prime Minister 2004.

Domain name Indices making up the domain Weighting and rangein area

1. Income Deprivation Domain • Adults and children in Income Support households• Adults and children in Income Based Job Seekers Allowance households• Adults and children in Working Families Tax Credit households whose equivalized income(excluding housing benefits) is below 60% of median before housing costs• Adults and children in Disabled Person's Tax Credit households whose equivalized income(excluding housing benefits) is below 60% of median before housing costs• National Asylum Support Service supported asylum seekers in England in receipt of subsistenceonly and accommodation support

22.5%Range: 0.03–24

2. Employment Deprivation Domain • Unemployment claimant count (JUVOS) of women aged 18–59 years and men aged 18–64 yearsaveraged over 4 quarters• Incapacity Benefit claimants women aged 18–59 years and men aged 18–64 years• Severe Disablement Allowance claimants women aged 18–59 years and men aged 18–64 years• Participants in New Deal for the 18–24 s who are not included in the claimant count• Participants in New Deal for 25+ who are not included in the claimant count• Participants in New Deal for Lone Parents aged 18 years and over

22.5%Range: 0–0.46

3. Health Deprivation and Disability Domain • Years of potential life lost• Comparative illness and disability ratio• Measures of emergency admissions to hospital• Adults under 60 suffering from mood or anxiety disorders

13.5%Not used in study

4. Education, Skills and Training Deprivation Domain Subdomain: children/young people• Average points score of children at Key Stage 2• Average points score of children at Key Stage 3• Average points score of children at Key Stage 4• Proportion of young people not staying on in school or school level education above 16• Proportion of those aged under 21 not entering Higher Education• Secondary school absence rateSubdomain: skills• Proportions of working age adults (aged 25–54 years) in the area with no or low qualifications

13.5%Range: 2–38

5. Barriers to Housing and Services Domain Subdomain wider barriers• Household overcrowding• LA level percentage of households for whom a decision on their application for assistance under• The homeless provisions of housing legislation has been made, assigned to SOAs• Difficulty of access to owner occupationSubdomain: geographical barriers• Road distance to GP premises• Road distance to a supermarket or convenience store• Road distance to a primary school• Road distance to a Post Office

9.3%Range: 28–45

6. Crime Domain • Burglary (4 recorded crime offence types)• Theft (5 recorded crime offence types)• Criminal damage (10 recorded crime offence types)• Violence (14 recorded crime offence types)

9.3%Range: −0.39 to 2.5

7. Living Environment Deprivation Domain Subdomain: The “indoor” living environment• Social and private housing in poor condition• Houses without central heatingSubdomain: The “outdoor” living environment• Air quality• Road traffic accidents involving injury to pedestrians and cyclists

9.3%Range: 8–79

118 V. Bhavsar et al. / Schizophrenia Research 156 (2014) 115–121

was based on linear interpolation and linear extrapolation was used forthe years 2000 and 2007. The ONS also supplied data estimating thetotal proportion of ethnic minorities and population density (personsper square kilometre) in each SOA. At the ward level, population esti-mates stratified by age, gender, ethnic group and socioeconomic statuswere obtained from the 2001 census. Own ethnic group density andtotal population density were calculated at this level. Further, popula-tion turnover datawere gathered, but thiswas available only at themid-dle super output area (MSOA) level.

2.6. Data analysis

We carried out direct standardization using the DSTDIZE procedurein STATA 11 (2009). The weighted average of the study-area popula-tions was used as the standard. This produced age- and sex-adjustedrates, thus enabling comparison of incidence across SOAs and wards.The distribution of the incidence data was checked and the mean num-ber of cases (stratified by the predictors to be tested) was noted to be

very similar to the variance, at both levels. Poisson regressionwas there-fore chosen as themost appropriate modelling technique. Zero-inflatedPoisson regression was initially used, as there was an excess of zerocases. Deprivation and social classwere tested as the variables to predictthe excess of zero cases in the zero-inflated models. Zero-inflated andnon zero-inflated Poisson regression modelling were then comparedusing the Vuong test. If there was no improvement with the zero-inflated model, it was not used.

With cases in super output areas as the outcome, age (dichotomizedinto groups with ages 16–25 and 26–35), gender, deprivation, propor-tion of Black and minority ethnic groups, population density and popu-lation turnover (the rate of in- or out-migratory moves per 1,000resident population)were entered as predictors using the ZIP commandin STATA. Each variable was added in a stepwise fashion and the likeli-hood ratio test used to test whether the addition improved the model.The same analysis was repeated at ward level with individual ethnicityandwith proportion of own ethnic group replacing proportion of ethnicminorities, as this was available at ward level. The health domain was

Table 3Effect of deprivation on incidence rate ratios (IRR) of OPCRIT derived RDC schizophrenia insuper output areas (SOA) in Lambeth.

Predictor IRR p 95% CI

Sex 0.49 b0.001 0.39–0.6Age 0.25 b0.001 0.2–0.3Population density 0.99 0.59 0.99–1.0Proportion of ethnic minorities 1.01 0.25 0.99–1.02SOA deprivation 1.04 b0.001 1.02–1.06

119V. Bhavsar et al. / Schizophrenia Research 156 (2014) 115–121

omitted from all analyses and did not contribute to ward or SOA totaldeprivation scores as this could have been influenced by the outcome.

3. Results

3.1. Deprivation analysis

Four hundred and five cases presented with first-onset schizophre-nia between January 2000 and December 2007. Of these, 277 (68%)were male, and 128 (32%) were female.

The overall incidence was 54.6 cases per 100,000 person years. Thecrude incidence within wards varied between 18.2 and 76.7 per100,000 person years. Incidence adjusted for age and gender withinwards ranged from 16.8 to 73.8 per 100,000 person years, using directstandardisation by means of the DSTDIZE command in STATA. At SOAlevel, crude incidence ranged from 0 to 207 per 100,000 person yearsand incidence adjusted by the same method ranged from 0 to 187 per100,000 person years.

Fig. 1 displays incidence and deprivation at SOA and ward levels. Atward level, association was found between the electoral ward depriva-tion score and incidence of schizophrenia (see Table 2). The zero-inflated Poisson regression model predicting incidence rate ratios forschizophrenia cases using social class as the predictor of zerocases was statistically significant and an improvement (Vuong z = 4.55,p = b0.0001). The incidence rate ratio for ward deprivation was 1.03(95% CI = 1.004–1.04, p = 0.02). This means that the incidence rateratio increased by 1.03 (3%) for each unit increase of deprivation (on ascale of theoretically 1–100, range in study area 23.49–46.29). At wardlevel, the effect of deprivation was adjusted for ethnicity, the proportionof own ethnic group and the interaction between them. Population den-sity and population turnover did not improve the model.

A further zero-inflated Poisson analysis was carried out atward levelto include individual level social classwhilst also using social class as thepredictor of zero cases. The zero-inflated model was statistically signif-icant and an improvement (Vuong z= 2.69, p= b0.0003). Social classwas treated as a continuous variable in this model. This demonstratedan effect of individual social class after adjusting for area level depriva-tion and vice versa, after adjusting for age, gender, ethnicity and propor-tion of own ethnic group. Again population density and populationturnover did not improve themodel. It was not possible to test for inter-actions between individual social class and area level deprivation orbetween individual ethnicity and proportion of own ethnic group inthis model. IRR was 1.15 for individual social class (95% CI = 1.11–1.2,p b 0.0001) and 1.04 for ward deprivation (95% CI = 1.02–1.06,p b 0.0001). It was not possible to repeat this analysis at SOA level be-cause the population datawere not stratified by social class at this level.

At SOA level, there was no advantage to the use of the zero-inflatedmodel (Vuong test z = 0.05, p = 0.48), so random-effects Poisson re-gression modelling was carried out using the XTPOISSON command inSTATA. The IRR was very similar for the SOA level compared to theward level (see Tables 2 (Ward) and 3 (SOA)). The addition of

Table 2Effect of deprivation on incidence rate ratios (IRR) of OPCRIT-derived RDC schizophrenia at ele

Predictor

Age*Sex**Black Caribbean ethnicity***Black African ethnicity***Interaction between Black Caribbean ethnicity and proportion of Black Caribbean peopleInteraction between Black African ethnicity and proportion of Black African peopleWard Deprivation

Main ethnic groups only presented social class predicted zero cases—1.26 (95% CI = 2.3 to− .*The younger age group 16–25 years was used as the reference group.** Male was used as the reference group.*** With white British as the reference group.

population density and proportion of ethnic minorities did not improvethe model.

3.2. Subdomain analyses

We then analysed the individual subdomains (available at SOAlevel) to look for associations between six of the seven individualcomponents of the overall deprivation score (excluding the health do-main) and incidence of schizophrenia again using random-effectsPoisson regression modelling. All correlations between the domainshad correlation coefficients b0.4. Of six domains tested individually,three domains reached statistical significance after adjusting for ageand gender but not the other domains. These were crime (IRR = 1.28,95% CI = 1.06–1.56, p = 0.011), education deprivation (IRR = 1.02,95% CI = 1.003–1.04, p = 0.02) and unemployment (IRR = 7.61, 95%CI = 1.53–37.62, p = 0.013). In a further model, the association of thesix domains with schizophrenia incidence was tested, adjusting forthe other domains as well as age gender, population density and pro-portion of ethnicminorities. This showed that crime and education dep-rivation predicted incidence. These results are displayed in Table 4.

4. Discussion

4.1. Findings

Using a first-contact approach within a tightly defined catchmentarea, and a deprivation index that did not include demographic factorssuch as ethnicity, we found a significant association between depriva-tion at the SOA level and schizophrenia incidence after adjusting forage, gender, ethnicity, ethnic density, population density and popula-tion turnover. The choice of geographical unit, the SOA as opposed tothe electoral ward, had almost no influence on the strength of the ob-served relationship between a multi-domain measure of deprivationand incidence of schizophrenia.

4.2. Limitations

The direction of causality between themain exposure, area depriva-tion, and outcome, psychosis incidence rate, cannot be determined. Wehave not accounted for reverse causality, i.e., that cases with psychosisare more likely to relocate into more deprived areas, but we did adjust

ctoral ward level in Lambeth.

IRR p 95% CI

0.53 b0.001 0.43–0.660.44 b0.001 0.35–0.55

12.01 0.014 1.66–86.775.9 0.069 0.87–40.70.003 0.038 b0.001–0.720.04 0.28 b0.001–12.471.026 0.02 1.004–1.04

21), p = 0.02.

Table 4RDC Schizophrenia incidence rate ratios for deprivation domains adjusted for age andgender and the other domains.

IMD domain IRR p 95% CI

Crime 1.26 0.02 1.03 1.53Employment deprivation 2.44 0.26 0.51 11.7Income 0.93 0.54 0.73 1.18Living environment 0.99 0.18 0.98 1.01Education, skills and training deprivation 1.02 0.046 1.001 1.03Barriers to housing and services 1.02 0.42 0.98 1.06

Note: Adjusting for population density and proportion of ethnic minorities did notimprove the model.

120 V. Bhavsar et al. / Schizophrenia Research 156 (2014) 115–121

for population turnover. First contact with mental health services forpsychosis was taken as onset rather than actual time of illness onsetas this is difficult to date retrospectively. No information was availableon how long cases had been resident in their respective areas. Longitu-dinal studies, ideally using serial measurements of deprivation and pro-spective case-finding, would be needed to overcome this.

Case ascertainment was reliant on the LEO case register, a databasecompiled of referrals with psychosis to mental health services, ratherthan a general-population-based case-finding survey. It is unlikely thatdifferential case ascertainment would have influenced the results inthe direction of our findings. There is a theoretical possibility thatsome cases of first-onset schizophrenia remained in the private sectorfor the entire duration of their illness and did not come to the attentionof the state sector, although given the paucity of such services inLambeth, the private sector is unlikely to have contributed a sufficientnumber of cases to influence the findings.

The study is ecological in nature so we do not know, for example, ifthe people who were victims of crime were the people who becamepsychotic.

4.3. Comparison with other work and possible mechanisms

Interpreting ecological studies of social organization and schizo-phrenia is dependent to some degree on how we define the areasused to calculate population-based incidence rates (Diez Roux, 2001).It has been pointed out that the use of electoral-ward-based data raiseproblems (Macintyre et al., 2002) asward boundaries are subject to reg-ular change, population sizes vary considerably and they might containa great deal of internal variation in the social environment, especiallylevels of deprivation (Vickers and Rees, 2007). Smaller areas are likelyto contain less heterogeneity in characteristics, allowingmoremeaning-ful inferences to be drawn about the spatial ranges overwhich social ex-posures may act, although we found little difference between the SOAand ward level analyses.

Studies of neighbourhood organization and schizophrenia rateshave employed a number of indicators for the social conditions ofareas, ranging from the use of individual census indices selected fornon-correlation with other census variables (Hare, 1956), through tocomponent analysis of whole census data (Giggs, 1986). A further prob-lem is that some measures of deprivation have included demographicindices such as ethnicity and single person households as markers ofdeprivation. The IMD scoring system applies a particularway of viewingsocial deprivation, as a continuous, multi-dimensional construct andcompeting models of deprivation measurement abound (Morrisand Carstairs, 1991). Previously, Koppel and McGuffin (1999), in anelectoral-ward-based study, found that the strength of the associationbetween social deprivation and admissions for psychosis was notaffected by the choice of particular indices of deprivation used, includ-ing single census indicators. However, the IMD 2004 offers a moresophisticated index that incorporates measures of social and environ-mental deprivation such as access to services and crime aswell asmate-rial factors. Conceptualising deprivation as a multi-domain construct,

changing over time, may represent a methodological advance in theway that social deprivation is quantified at the area level. For psychiatricepidemiology, it is also an advance in that it does not include ethnicity inthe measure.

Our central finding, that the deprivation of an area predicts the rateof first episode psychosis, fits with the previous literature on psychosisrates and social structure (Boydell et al., 2004; Zammit et al., 2010).Area deprivation may exert effects via contextual mechanisms, involv-ing such social fragmentation, threat, or lack of trust (Allardyce et al.,2005). An ongoing challenge in this literature is the development ofvalid methods used to measure the social environment, and to investi-gatewhich aspects are of particular importance, in order to generate hy-potheses for how the effects of area deprivation might operate. Theapproach used in this study was to explore whether particular domainsof deprivation were particularly important. This approach had the ad-vantage of avoiding bias introduced by differential weighting of the dif-ferent domains within the overall score. Of seven regression modelsbuilt to test the individual domains after adjusting for age and gender,three reached statistical significance (crime, education and employ-ment). When adjusted for the other domains, employment deprivationdid not reach statistical significance although the IRR was still raised.This suggests that some of the area level effect of high unemploymentwas confounded by crime and poor education.

The crime domain is defined as “incidence of recorded crime,representing the occurrence of personal and material victimisation ata small area level, regardless of the presence or absence of other typesof deprivation (such as income deprivation) in the area” (ODPM,2004). The statistical effect of crime on incidence may be explained bya number of hypotheses. It is unlikely that the level of criminalitywould increase in an areawhere there aremore individualswith schizo-phrenia, as the absolute numbers are low. Alternatively, the effect ofcrime on incidence may genuinely be causal, that is, that exposure tocriminal activity at the area level increases the likelihood of becomingpsychotic, perhaps via mechanisms linked to social defeat (Selten andCantor-Graae, 2007) or experience of threat (Bebbington et al., 2004).Living in a threatening area may generate social stress, and increasethe likelihood of paranoia, for example. The link between unemploy-ment and psychosis is well established (Boardman et al., 2003;Marwaha and Johnson, 2004). Agerbo et al. (2004) found that social de-cline, in the form of unemployment, living alone and receipt of benefits,preceded first admission for schizophrenia by up to 19 years. However,this does not exclude unemployment as an important contextual expo-sure or aetiological factor for schizophrenia. So far, there has been littleresearch into unemployment as an aetiological factor for the develop-ment of schizophrenia (Mallett et al., 2002). This may derive from thechallenges of accounting for social drift and the difficulties ofdisentangling unemployment from the broader material deprivationof populations (Dutta et al., 2008). The finding that population educa-tional deprivation might explain some of the effect of deprivation onschizophrenia rates has not previously been demonstrated as far aswe know.

It is interesting to note that the material deprivation domains suchas income and housing did not predict incidence of schizophrenia andthis might explain some of the previous contradictory findings. Itmight be that poverty itself is not predictive but may become sowhen associated with social disorganisation and threat. This mayalso explain why rural deprivation has not been associated with in-creased rates of schizophrenia (Thornicroft et al., 1993) and whyadjusting for deprivation statistically reduces the urbanicity effect(Allardyce et al., 2005). The importance of these findings is thatthey point towards modifiable population level risk factors for a se-vere and costly mental illness.

Role of the funding sourceV. B. was supported by the award of an Academic Clinical Fellowship by the National

Institute of Health Research.

121V. Bhavsar et al. / Schizophrenia Research 156 (2014) 115–121

ContributorsV. B. and P. P. were responsible for the design of the study. V. B. established andmain-

tained the incidence database. Data analysiswas carried out by V. B. and J. B. V. B.wrote thepaper with the assistance of J. B. and R. M.

Conflict of interestThe authors declare no conflicts of interest pertaining to this study.

AcknowledgementsThe authors acknowledge the help of the UK Office of National Statistics in the supply

of population estimates and the Office of the Deputy Prime Minister for deprivation dataused in this study.

References

Agerbo, E., Byrne, M., Eaton, W.W., Mortensen, P.B., 2004. Marital and labor market statusin the long run in schizophrenia. Arch. Gen. Psychiatry 61 (1), 28–33.

Allardyce, J., Boydell, J., 2006. Review: the wider social environment and schizophrenia.Schizophr. Bull. 32 (4), 592–598.

Allardyce, J., Gilmour, H., Atkinson, J., Rapson, T., Bishop, J., McCreadie, R.G., 2005. Socialfragmentation, deprivation and urbanicity: relation to first-admission rates for psy-choses. Br. J. Psychiatry J. Ment. Sci. 187, 401–406.

Ashworth, M., Seed, P., Armstrong, D., Durbaba, S., Jones, R., 2007. The relationship be-tween social deprivation and the quality of primary care: a national survey using in-dicators from the UK Quality and Outcomes Framework. Br. J. Gen. Pract. 57 (539),441–448.

Bebbington, P.E., Bhugra, D., Brugha, T., Singleton, N., Farrell, M., Jenkins, R., Lewis, G.,Meltzer, H., 2004. Psychosis, victimisation and childhood disadvantage: evidencefrom the second British National Survey of Psychiatric Morbidity. Br. J. Psychiatry J.Ment. Sci. 185, 220–226.

Boardman, A.P., Hodgson, R.E., Lewis, M., Allen, K., 1997. Social indicators and the predic-tion of psychiatric admission in different diagnostic groups. Br. J. Psychiatry 171 (5),457–462.

Boardman, J., Grove, B., Perkins, R., Shepherd, G., 2003. Work and employment for peoplewith psychiatric disabilities. Br. J. Psychiatry 182 (6), 467–468.

Bond, S., Insalaco, F., 2007. Area classification of super output areas and datazones. ProjectFinal ReportONS, London.

Boydell, J., van Os, J., McKenzie, K., Allardyce, J., Goel, R., McCreadie, R.G., Murray, R.M.,2001. Incidence of schizophrenia in ethnic minorities in London: ecological studyinto interactions with environment. BMJ 323 (7325), 1336–1338.

Boydell, J., van Os, J., McKenzie, K., Murray, R.M., 2004. The association of inequality withthe incidence of schizophrenia–an ecological study. Soc. Psychiatry Psychiatr.Epidemiol. 39 (8), 597–599.

Byrne, M., Agerbo, E., Eaton, W.W., Mortensen, P.B., 2004. Parental socio-economic statusand risk of first admission with schizophrenia—a Danish national register basedstudy. Soc. Psychiatry Psychiatr. Epidemiol. 39 (2), 87–96.

Castle, D., Scott, K., Wessely, S., Murray, R.M., 1993. Does social deprivation during gesta-tion and early life predispose to later schizophrenia? Soc. Psychiatry Psychiatr.Epidemiol. 28 (1), 1–4.

Di Forti, M., Morgan, C., Dazzan, P., Pariante, C., Mondelli, V., Marques, T.R., Handley, R.,Luzi, S., Russo, M., Paparelli, A., Butt, A., Stilo, S.A., Wiffen, B., Powell, J., Murray, R.M.,2009. High-potency cannabis and the risk of psychosis. Br. J. Psychiatry J. Ment. Sci.195 (6), 488–491.

Diez Roux, A.V., 2001. Investigating neighborhood and area effects on health. Am. J. PublicHealth 91 (11), 1783–1789.

Dohrenwend, B.P., 1990. Socioeconomic status (SES) and psychiatric disorders. Soc.Psychiatry Psychiatr. Epidemiol. 25 (1), 41–47.

Dutta, R., Murray, R., Boydell, J., 2008. Unemployment and risk of psychosis in black andethnic minority groups. Schizophr. Res. 102 (1, Supplement 2), 164–165.

Ferguson, H.J.M., Nightingale, P., Pathak, R., Jayatunga, A.P., 2010. The influence of socio-economic deprivation on rates of major lower limb amputation secondary to periph-eral arterial disease. Eur. J. Vasc. Endovasc. Surg. 40 (1), 76–80.

Giggs, J., 1973. The distribution of schizophrenics in Nottingham. Trans. Inst. Br. Geogr.55–76.

Giggs, J.A., 1986. Mental disorders and ecological structure in Nottingham. Soc. Sci. Med.23 (10), 945–961 (1982).

Goldberg, E.M., Morrison, S.L., 1963. Schizophrenia and social class. Br. J. Psychiatry J.Ment. Sci. 109, 785–802.

Hare, E.H., 1956. Mental illness and social conditions in Bristol. J. Ment. Sci. 102 (427),349–357.

Harrison, J., Barrow, S., Creed, F., 1995. Social deprivation and psychiatric admission ratesamong different diagnostic groups. Br. J. Psychiatry J. Ment. Sci. 167 (4), 456–462.

Harrison, G., Fouskakis, D., Rasmussen, F., Tynelius, P., Sipos, A., Gunnell, D., 2003.Association between psychotic disorder and urban place of birth is not mediated byobstetric complications or childhood socio-economic position: a cohort study.Psychol. Med. 33 (4), 723–731.

Hollingshead, A.B., Redlich, F.C., 1958. Social Class and Mental Illness: Community Study.Jones, P.B., Bebbington, P., Foerster, A., Lewis, S.W., Murray, R.M., Russell, A., Sham, P.C.,

Toone, B.K., Wilkins, S., 1993. Premorbid social underachievement in schizophrenia.Results from the Camberwell Collaborative Psychosis Study. Br. J. Psychiatry J. Ment.Sci. 162, 65–71.

Jones, P., Rodgers, B., Murray, R., Marmot, M., 1994. Child development risk factors foradult schizophrenia in the British 1946 birth cohort. Lancet 344 (8934), 1398–1402.

Kirkbride, J.B., Morgan, C., Fearon, P., Dazzan, P., Murray, R.M., Jones, P.B., 2007.Neighbourhood-level effects on psychoses: re-examining the role of context. Psychol.Med. 37 (10), 1413–1425.

Kirkbride, J.B., Barker, D., Cowden, F., Stamps, R., Yang, M., Jones, P.B., Coid, J.W., 2008.Psychoses, ethnicity and socio-economic status. Br. J. Psychiatry J. Ment. Sci. 193(1), 18–24.

Kirkbride, J.B., Errazuriz, A., Croudace, T.J., Morgan, C., Jackson, D., Boydell, J., Murray, R.M.,Jones, P.B., 2012. Incidence of schizophrenia and other psychoses in England, 1950–2009: a systematic review and meta-analyses. PLoS One 7 (3), e31660.

Koppel, S., McGuffin, P., 1999. Socio-economic factors that predict psychiatric admissionsat a local level. Psychol. Med. 29 (5), 1235–1241.

Macintyre, S., Ellaway, A., Cummins, S., 2002. Place effects on health: how can we concep-tualise, operationalise and measure them? Soc. Sci. Med. 55 (1), 125–139.

Mallett, R., Leff, J., Bhugra, D., Pang, D., Zhao, J.H., 2002. Social environment, ethnicity andschizophrenia: a case–control study. Soc. Psychiatry Psychiatr. Epidemiol. 37 (7),329–335.

March, D., Hatch, S.L., Morgan, C., Kirkbride, J.B., Bresnahan, M., Fearon, P., Susser, E., 2008.Psychosis and place. Epidemiol. Rev. 30, 84–100.

Marwaha, S., Johnson, S., 2004. Schizophrenia and employment. Soc. Psychiatry Psychiatr.Epidemiol. 39 (5), 337–349.

McGuffin, P., Farmer, A.E., Gottesman, I.I., Murray, R.M., Reveley, A.M., 1984. Twin concor-dance for operationally defined schizophrenia. Confirmation of familiality and herita-bility. Arch. Gen. Psychiatry 41 (6), 541–545.

Morgan, C., Charalambides, M., Hutchinson, G., Murray, R.M., 2010. Migration, ethnicity,and psychosis: toward a sociodevelopmentalmodel. Schizophr. Bull. 36 (4), 655–664.

Morris, R., Carstairs, V., 1991. Which deprivation? A comparison of selected deprivationindexes. J. Public Health 13 (4), 318–326.

Mulvany, F., O'Callaghan, E., Takei, N., Byrne, M., Fearon, P., Larkin, C., 2001. Effect of socialclass at birth on risk and presentation of schizophrenia: case–control study. Br. Med. J.323 (7326), 1398–1401.

ODPM, 2004. Indices of Deprivation 2004 (London).Power, P.,McGuire, P., Iacoponi, E., Garety, P., Morris, E., Valmaggia, L., Grafton, D., Craig, T.,

2007. Lambeth Early Onset (LEO) and Outreach & Support in South London (OASIS)service. Early Interv. Psychiatry 1 (1), 97–103.

Selten, J.P., Cantor-Graae, E., 2007. Hypothesis: social defeat is a risk factor for schizophre-nia? Br. J. Psychiatry Suppl. 51, s9–s12.

STATA 11, 2009. College Station, Texas, USA.Stephens, M.R., Blackshaw, G.R., Lewis, W.G., Edwards, P., Barry, J.D., Hopper, N.A., Allison,

M.C., 2005. Influence of socio-economic deprivation on outcomes for patients diag-nosed with gastric cancer. Scand. J. Gastroenterol. 40 (11), 1351–1357.

Thornicroft, G., Bisoffi, G., De Salvia, D., Tansella, M., 1993. Urban-rural differences in theassociations between social deprivation and psychiatric service utilization in schizo-phrenia and all diagnoses: a case-register study in Northern Italy. Psychol. Med. 23(2), 487–496.

van Os, J., Driessen, G., Gunther, N., Delespaul, P., 2000. Neighbourhood variation in inci-dence of schizophrenia. Evidence for person-environment interaction. Br. J. Psychia-try J. Ment. Sci. 176, 243–248.

Veling, W., Susser, E., van Os, J., Mackenbach, J.P., Selten, J.P., Hoek, H.W., 2008. Ethnicdensity of neighborhoods and incidence of psychotic disorders among immigrants.Am. J. Psychiatry 165 (1), 66–73.

Vickers, D., Rees, P., 2007. Creating the UK National Statistics 2001 output area classifica-tion. J. R. Stat. Soc.: Ser. A (Stat. Soc.) 170 (2), 379–403.

Werner, S., Malaspina, D., Rabinowitz, J., 2007. Socioeconomic status at birth is associatedwith risk of schizophrenia: population-basedmultilevel study. Schizophr. Bull. 33 (6),1373–1378.

Zammit, S., Lewis, G., Rasbash, J., Dalman, C., Gustafsson, J.E., Allebeck, P., 2010. Individ-uals, schools, and neighborhood: a multilevel longitudinal study of variation in inci-dence of psychotic disorders. Arch. Gen. Psychiatry 67 (9), 914–922.