Evidence of a Dose-Response Relationship Between Urbanicity During Upbringing and Schizophrenia Risk

Background  Many studies have demonstrated that an urban birth or upbringing increases schizophrenia risk, but no studies have been able to distinguish between these effects. The objectives of this study were to discriminate the effect of urbanicity at birth from an effect of urbanicity during upbringing, and to identify particularly vulnerable age periods and a possible dose-response relationship between urbanicity during upbringing and schizophrenia risk.

Methods  Using data from the Danish Civil Registration System, we established a population-based cohort of 1.89 million people, which included information on place of birth, place of residence during upbringing, and the accumulated number of years lived in each category of the 5-level degree of urbanization during upbringing. Schizophrenia in cohort members and mental illness in a parent or sibling were identified by linkage with the Danish Psychiatric Central Register.

Results  Individuals living in a higher degree of urbanization than 5 years earlier had a 1.40-fold (95% confidence interval, 1.28-1.51) increased risk, while individuals living in a lower degree of urbanization than 5 years earlier had a 0.82-fold (95% confidence interval, 0.77-0.88) decreased risk of schizophrenia. For fixed urbanicity at the 15th birthday, risk increased with increasing degree of urbanization at birth. Furthermore, the more years lived in the higher the degree of urbanization, the greater the risk. Individuals who lived their first 15 years in the highest category of the 5-level urbanicity had a 2.75-fold (95% confidence interval, 2.31-3.28) increased risk of schizophrenia.

Conclusion  Continuous, or repeated, exposures during upbringing that occur more frequently in urbanized areas may be responsible for the association between urbanization and schizophrenia risk.

ALTHOUGH A family history of schizophrenia is the best established risk factor for the disease,¹ it may account for only a small proportion of the population occurrence of schizophrenia.^2,3 Other factors, such as urbanicity at birth and upbringing, are associated with an increased risk also,^2-4 and causal factors underlying this association may account for a much higher proportion of the population occurrence of the disease. Although the causes of these urban-rural differences are unknown, they have been hypothesized to include, eg, obstetric complications, infections, diet, toxic exposures, household crowding, breastfeeding, social class, and an artifact due to migration.^5,6

To reduce the number of possible candidates responsible for this association, the objective of this study was to investigate at what age during upbringing (including birth) children were most vulnerable to urbanicity. However, since urbanicity at birth and urbanicity during upbringing are strongly associated, a large study population, including information on place of birth and upbringing, is needed to discriminate the effect of urbanicity at birth from an effect of urbanicity during upbringing, and vice versa.⁷ To our knowledge, studies based on such data have not been published previously. This study uses a large population-based sample of the Danish population, including complete information on all permanent addresses at which cohort members had lived since 1971, to investigate the relative impact of urbanicity at birth and during upbringing. Furthermore, the hypothesis that there is a dose-response relationship between urbanicity during upbringing and schizophrenia risk is investigated.

We used data from the Danish Civil Registration System⁸ to obtain a large and representative set of data on Danish persons, which for all persons included current and historical information on address in Denmark and Greenland and emigrations and immigrations to and from other countries, together with exact dates for changes of residence. All citizens in Denmark are obliged to inform the authorities about any change of permanent address within 5 days. Failure to supply this information will result in the inability to receive a supplementary benefit (eg, unemployment, sickness or disablement benefits, and educational aid from public funds), to go to a day nursery, to go to nursery school, to attend primary and lower secondary school, to obtain free national health care, and to obtain a tax deduction card (required to have paid work). Therefore, it is unlikely that this mandatory information is not reported. This information is complete from January 1, 1971. Our study cohort consists of all persons with known maternal identity who were born in Denmark between January 1, 1956, and December 31, 1983, and who were alive at the 15th birthday (1.89 million persons).

The study population and their mothers, fathers, and siblings were linked with the Danish Psychiatric Central Register,⁹ which has been computerized since April 1, 1969. The Danish Psychiatric Central Register contains data on all admissions to Danish psychiatric inpatient facilities and includes data on approximately 450 000 persons and 1.6 million admissions. From 1995 onward, information on outpatient visits to psychiatric departments was included in the register. There are no private psychiatric departments in Denmark. From April 1, 1969, to December 31, 1993, the diagnostic system used was the International Classification of Diseases, 8th Revision (ICD-8),¹⁰ and from January 1, 1994, the diagnostic system used was the International Classification of Diseases, 10th Revision (ICD-10).¹¹ Cohort members were classified as having schizophrenia if they had been admitted to a psychiatric hospital or had been in outpatient care with a diagnosis of the disorder (ICD-8 code 295 or ICD-10 code F20). The date of onset was defined as the first day of the first contact (inpatient or outpatient) with a diagnosis of schizophrenia. Parents and siblings were categorized hierarchically with a history of schizophrenia (ICD-8 code 295 or ICD-10 code F20), schizophrenialike psychoses (ICD-8 code 297, 298.39, or 301.83 or ICD-10 codes F21-F29), or other mental disorders (any ICD-8 or ICD-10 diagnosis) if they had been admitted to a psychiatric hospital or had been in outpatient care with one of these diagnoses.

The 276 municipalities in Denmark were classified according to the degree of urbanization¹²: capital, capital suburb, provincial city with more than 100 000 inhabitants, provincial town with more than 10 000 inhabitants, or rural areas. Denmark is a small homogeneous country with a population of 5.3 million people and a total area of 43 000 km². The capital, capital suburb, provincial city, provincial town, and rural area hold 5220, 845, 470, 180, and 55 people per square kilometer, respectively.¹³ Distances are small in Denmark; most people live within 25 km of a city with more than 30 000 inhabitants, and are even closer to a psychiatric hospital.

Using data from the Danish Civil Registration System, for each person in the cohort we obtained information on (1) the degree of urbanization in 1-year age points from birth to the 15th birthday; (2) the number of changes of the address, the municipality, and the degree of urbanization in 1-year age bands from birth to the 15th birthday; and (3) the accumulated number of years each person born in 1971 or later had been living in each degree of urbanization from birth to the 5th birthday, from the 5th to the 10th birthday, and from the 10th to the 15th birthday. The reason we considered change of residence was that initial analyses suggested that to investigate the association between schizophrenia and urbanization we had to control for an increased risk associated with change of residence. Furthermore, to avoid the potential impact of selective migration to cities in the prodromal phase of schizophrenia, only residence during upbringing was considered. By upbringing, we are referring to the period from birth to the 15th birthday. Information on urbanicity at birth, urbanicity during upbringing, and change of residence is independent of the disease status. A total of 1.89 million persons were followed up from their 15th birthday or from April 1, 1970, whichever came later, until the date of onset of schizophrenia, the date of death, the date of emigration, or December 31, 1998, whichever came first.

The relative risk of schizophrenia was estimated by log-linear Poisson regression¹⁴ with the GENMOD procedure, using SAS statistical software, version 6.12.¹⁵ All relative risks were adjusted for age and its interaction with sex, calendar year, and history of mental illness in a parent or sibling. Age, calendar year, and history of mental illness in siblings were treated as time-dependent variables,¹⁶ whereas history of mental illness in a parent was treated as a variable that was independent of time. Age was categorized using the following cut points: 15, 16, 17, 18, 19, 20, 22, 24, 26, 28, 30, 35, and 40 years. Calendar year was categorized using the following cut points: 1971, 1976, 1979, 1982, 1985, 1988, 1991, 1993, 1994, 1995, 1996, 1997, and 1998. P values were based on findings from the likelihood ratio tests, and 95% confidence limits were calculated by the Wald test.¹⁶ The adjusted-score test¹⁷ suggested that the regression models were not subject to overdispersion. Apart from the reduction in the size of the study population, omission of adjustment for seasonality, maternal and paternal age, and the inclusion of information on permanent address for all cohort members since 1971, the material described is identical to that used in our previous study,³ where we found that the effects of urbanicity at birth and mental illness in a family member were invariant to the diagnostic system and the inclusion of outpatient information.

To evaluate the hypothesis that there is a dose-response relationship between urbanicity during upbringing and schizophrenia risk, we used a statistical model for the accumulated number of years each person had been living in the capital, the capital suburb, the provincial city, and the provincial town. It implies that those who always lived in the rural area during upbringing were chosen as the reference category, and that the effect of these 4 variables measures the effect of exchanging upbringing in the rural area with upbringing in the corresponding degree of urbanization.

By urbanicity at some age point, we refer only to the degree of urbanization at the given age point, whereas by urbanicity during upbringing, we refer to the full sequence of degrees of urbanization of places of residence from birth to the 15th birthday. Furthermore, by a model for urbanicity during upbringing, we refer to a model for the accumulated number of years each person has lived in the capital, the capital suburb, the provincial city, and the provincial town during upbringing.

A total of 8235 persons developed schizophrenia during the 27.1 million person-years of follow-up (Table 1). Urbanicity at birth is known for all persons, whereas information on urbanicity at a given age depends on the year of birth, as information on place of residence was accessible only from 1971 onward.

The relative risk of schizophrenia increases with increasing age at change of the address or the municipality (Figure 1). However, change of address within the same municipality had no influence on schizophrenia risk. Furthermore, the effect of changes of the municipality within the same degree of urbanization increased with increasing age (data not shown). Therefore, the confounding effect of change of residence can be described by changes of the municipality.

We classified change of municipality by 4 variables, counting the number of changes of the municipality from birth to the 5th birthday, from the 5th to the 10th birthday, from the 10th to the 13th birthday, and from the 13th to the 15th birthday (Table 2). Compared with persons with no changes of the municipality from the 5th to the 10th birthday, those with one change of the municipality had a relative risk of 1.18. The effect of the number of changes of municipality increased with increasing age and increasing number of changes. Overall, these 4 variables had a significant effect (P<.001), and the model with these 4 variables had a significantly better fit than the model in which the change of municipality was described in 1-year bands (Figure 1).

Individuals who at birth lived in the capital, the capital suburb, the provincial city, or the provincial town had a relative risk of schizophrenia of 2.24 (95% confidence interval [CI], 1.92-2.61), 1.71 (95% CI, 1.46-2.00), 1.62 (95% CI, 1.36-1.92), and 1.27 (95% CI, 1.10-1.47), respectively, compared with individuals who lived in the rural area (Figure 2). This difference remained almost constant for ages from birth to the 15th birthday.

Among those who were born in the capital suburb, those who at the 5th birthday lived in a higher, the same, or a lower degree of urbanization than at birth had a relative risk of schizophrenia of 2.01, 1.82, or 1.55, respectively (Table 3, first adjustment), compared with those who at birth and at the 5th birthday lived in the rural area. In general, living in a higher degree of urbanization at the 5th birthday than at birth increased risk, while living in a lower degree of urbanization at the 5th birthday than at birth decreased risk. This tendency was not modified by adjustment for change of the municipality (Table 3, second adjustment).

Living in a higher or lower degree of urbanization at the 10th birthday than at the 5th birthday increased risk, but risk was increased more if living in a higher than in a lower degree of urbanization (Table 3, first adjustment). When these estimates were adjusted for the number of changes of the municipality, living in a higher degree of urbanization at the 10th birthday than at the 5th birthday increased risk, while living in a lower degree of urbanization at the 10th birthday than at the 5th birthday decreased risk (Table 3, second adjustment). The same tendency was found when comparing place of residence at the 10th birthday with place of residence at the 15th birthday. Overall, living in a higher degree of urbanization than 5 years earlier increased risk 1.40-fold (95% CI, 1.28-1.51), while living in a lower degree of urbanization than 5 years earlier decreased risk 0.82-fold (95% CI, 0.77-0.88). Furthermore, for fixed urbanicity at birth, risk increases with increasing degree of urbanization at place of residence at the 15th birthday, and for fixed urbanicity at the 15th birthday, risk increases with increasing degree of urbanization at place of birth (data not shown).

A total of 807 000 people in the cohort were born in 1971 or later and, therefore, these people had accessible information on urbanicity during upbringing. Among them, 1553 persons developed schizophrenia during the 5.6 million person-years at risk (Table 4). Compared with those who had always been living in the rural area, those who had always been living in the capital had a relative risk of 2.62. In general, the more years lived in the higher the degree of urbanization, the greater the risk of schizophrenia.

Estimates in the second column of Table 5 measure the risk per year associated with exchanging residence in the rural area with residence in each degree of urbanization, and estimates in the third column measure the risk associated with exchanging all residence (15 years) in the rural area with residence in each degree of urbanization. The relative risk per 15 years is calculated by raising the relative risk per year to the 15th power, eg, 1.0697¹⁵ = 2.75. Compared with those who had always been living in the rural area during upbringing, those who had always been living in the capital during upbringing had a relative risk of 2.75. According to this model, a person who had been living 7 years in the capital and 8 years in the capital suburb during upbringing had a relative risk of 2.12 (1.0697⁷ × 1.0355⁸) compared with a person who had always been living in the rural area (not accounting for the increased risk associated with change of residence). In general, risk increases with increasing degree of urbanization, and inherent in the log-linear model is that risk increases with increasing number of years lived in each degree of urbanization. The categorical model (Table 4) could be simplified to the log-linear model (Table 5) for urbanicity during upbringing (P = .63). In the rest of this article, we use the log-linear model to describe the association between schizophrenia and urbanization. Furthermore, this model had a significantly better fit than any of the models for the modifying effects of urbanicity between the successive age points of birth, 5th, 10th, and 15th birthday (Table 3).

When urbanicity at birth and urbanicity during upbringing were adjusted mutually, the effect of urbanicity at birth vanished (P = .80), while the effect of urbanicity during upbringing was not modified and remained strongly significant (P<.001). This means that urbanicity during upbringing explains the effect of urbanicity at birth, ie, urbanicity at birth is a proxy for urbanicity during upbringing. The effect of urbanicity during upbringing did not differ significantly by age at residence (P = .08) or sex (P = .79) (data not shown), meaning that there was no evidence of age at residence or sex differences in the vulnerability to upbringing in urbanized areas. Furthermore, there were no urban-rural differences in age of onset (P = .21).

In any of the models presented herein, the effect of change of municipality remained constant.

Places of residence at nearby ages are strongly associated, meaning that for most people, place of residence at the fifth birthday was identical to place of residence at the fourth and sixth birthday. This correlation impedes the interpretation of the finding of the constant effect of urbanicity according to age (Figure 2). However, suppose children were only vulnerable to urbanicity at the fifth birthday and not to urbanicity at birth or urbanicity at any other age, then the reason many studies (eg, Mortensen and others^2,3,6) have found an association between urbanicity at birth and schizophrenia risk is that urbanicity at birth is a proxy for urbanicity at the fifth birthday. Furthermore, because of attenuation caused by misclassification,¹⁸ the effect of this proxy variable would be lower than the effect of urbanicity at the fifth birthday. Therefore, if children were most fragile to urbanicity at some single age point during upbringing, then urbanicity would have a higher effect for this age point than for the nearby age points. However, since the effect of urbanicity (Figure 2) does not depend on age at residence, these data show no indication of any ages particularly vulnerable to residence in urbanized areas during upbringing.

The risk associated with 15 years of residence in any urbanicity was greater than the risk associated with living in the same urbanicity as 5 years earlier (Table 3, second adjustment, and Table 5), which in turn was greater than the effect of urbanicity at birth or at some age point (Figure 2). This indicates that the greater the refinement of the modeling of urbanicity, the greater the risk, and that not only urbanicity at birth or at some age point but also the effect of living in the same urbanicity as 5 years earlier are proxies for urbanicity during upbringing. In conclusion, the log-linear model for the accumulated number of years lived in each degree of urbanization during upbringing is the best model describing the association between urbanization and schizophrenia risk. Furthermore, it provides evidence of a dose-response relationship between urbanicity during upbringing and schizophrenia risk.

The possible etiological factors that might explain our findings remain unknown. However, the finding of a dose-response relationship between schizophrenia risk and urbanicity during upbringing lends support to a causal association. Our findings may suggest that constant, cumulative, or repeated exposures during upbringing occurring more frequently in urbanized areas are responsible for the association between urbanization and schizophrenia risk. The potential explanations for the urban-rural differences shift from factors influencing children at or around birth to factors influencing children continuously, or repeatedly, throughout upbringing. Factors such as obstetric complications, prenatal infections, and breastfeeding are, therefore, less likely explanations for these urban-rural differences, although they may well affect risk independently of urbanicity. Also, studies have suggested that household crowding,¹⁹ obstetric complications,²⁰ parental social class,²¹ and prenatal exposure to influenza²² do not explain the urban-rural differences in schizophrenia risk.

When Danish children move from one municipality to another, they usually change school. Therefore, it is interesting that change of municipality had a greater effect on schizophrenia risk than change of address, while change of address within the same municipality had no effect (Figure 1). We can only speculate regarding the underlying causes of these findings; however, they may be related to social maladjustment in preschizophrenic children, the stress of making new friends, and the fact that preschizophrenic children are more anxious in new environments.^23,24 Life events, such as parental death, parental divorce, or change in social class, are possible but less likely explanations as they are not related to change of the municipality only.

It is extremely unlikely that selective migration or urban-rural differences in the availability of psychiatric services explain the effect of urbanicity at birth; for fixed urbanicity at the 15th birthday, risk increases with increasing degree of urbanization at place of birth, and there was no evidence that age at onset was modified by urbanicity during upbringing, ie, there were no urban-rural differences in the threshold for a psychiatric admission with schizophrenia. Furthermore, services are free and distances are small in Denmark.

The results of the study are based on patients with schizophrenia admitted to a psychiatric hospital or those in outpatient care diagnosed as having schizophrenia. Although not all patients with schizophrenia are admitted to a psychiatric hospital or are in outpatient care during the first episode, many will eventually be admitted or come in for outpatient care and, thus, subsequently will become registered. Furthermore, the classification of degree of urbanization is based on the number of inhabitants in the largest city in the municipality, and is almost certainly a crude proxy variable for the unidentified underlying risk factors and mechanism. If the underlying risk-increasing mechanisms explaining the urban-rural difference can be identified, it is likely that their effects will be much larger than the association with urbanicity during upbringing we demonstrated.

To our knowledge, this is the first study to directly assess the relative importance of urbanicity at birth and during upbringing. Lewis et al⁴ have shown an association between being brought up in urban areas and having schizophrenia later in life, but they did not distinguish this effect from an effect of urban birth. Marcelis et al²⁵ showed that urban residence at onset did not affect schizophrenia risk when controlling for urban birth. However, their results may be biased by the migration of preschizophrenic people before their first hospitalization with schizophrenia. Although our findings must be replicated, they warrant direct tests of the hypothesis that continuous, or repeated, exposures during upbringing that occur more frequently in urbanized areas are responsible for the association between urbanization and schizophrenia risk. Candidate risk factors would include infections, diet, and exposure to pollution.

Accepted for publication May 1, 2001.

This study was supported by the Theodore and Vada Stanley Foundation, Bethesda, Md. The National Centre for Register-Based Research is funded by the Danish National Research Foundation, Copenhagen.

Corresponding author and reprints: Carsten Bøcker Pedersen, MSc, National Centre for Register-Based Research, Aarhus University, Taasingegade 1, 8000 Aarhus C, Denmark (e-mail: cbp@ncrr.au.dk).