Reduction in the sample size from the original sample to the interviewed sample. SIDP-R indicates Structured Interview for DSM-III-R Personality Disorders.
Torgersen S, Kringlen E, Cramer V. The Prevalence of Personality Disorders in a Community Sample. Arch Gen Psychiatry. 2001;58(6):590-596. doi:10.1001/archpsyc.58.6.590
To our knowledge, no previous studies of personality disorders (PDs) in a large representative sample of the common population have been conducted.
A representative sample of 2053 individuals between the ages of 18 and 65 years in Oslo, the capital of Norway, was studied from 1994 to 1997. Information about PDs was obtained by means of the Structured Interview for DSM-III-R Personality Disorders, in conjunction with an interview recording demographic data. The subjects were interviewed primarily at home, but in some instances, also at the clinic.
The prevalence of PDs was 13.4% (SE, 0.7). The prevalence rates (SEs) for specific PDs, irrespective of whether a person had 1 or more PD, were: paranoid, 2.4% (0.3); schizoid, 1.7% (1.6); schizotypal, 0.6% (0.2); antisocial, 0.7% (0.2); sadistic, 0.2% (0.1); borderline, 0.7% (0.2); histrionic, 2.0% (0.3); narcissistic, 0.8; (0.2); avoidant, 5.0% (0.5); dependent, 1.5% (0.3); obsessive-compulsive: 2.0% (0.3); passive-aggressive, 1.7% (0.3); self-defeating, 0.8%, (0.2). The prevalence of PDs was highest among subjects with only a high school education or less, and living without a partner in the center of the city.
Personality disorders were found to be prevalent, with avoidant, schizoid, and paranoid PDs more common, and borderline PD less common than what is usually reported. Personality disorders tend to be more frequent among single individuals from the lower socioeconomic classes in the center of the city. It is impossible to determine what is cause and what is consequence from a cross-sectional study.
IN CONTRAST to symptom disorders, few epidemiological studies of personality disorders (PDs) have been conducted to establish their prevalence. As the structured Diagnostic Interview Schedule (DIS)%1 contains questions pertaining to the antisocial PD, it has been studied in epidemiological studies of symptom disorders.%2- 6 An attempt has also been made to arrive at estimates of the prevalence of obsessive-compulsive%3 and histrionic%7 PDs in such studies. One study has tried to estimate the frequency of borderline PD using the DIS.%8
As to studies of the whole realm of PDs, Table 1 presents an overview of the prevalence of these disorders. The studies show a wide variation of prevalences of all as well as of the specific PDs. The prevalence of any PD varies between 5.9 and 22.5, with a median prevalence of 11.1 and a pooled prevalence of 12.47.
The main limitation of the previous studies is that the samples are not representative of the population at large.
The samples are also small, varying from 133 to 799 individuals, with a median of 302 individuals. Two studies have simply applied clinical interviews,%9,15 thus obtaining the most deviant prevalence. Two studies have applied self-report questionnaires.%11,14 Only 6 of the studies have used structured interviews.%10,12,13,16- 18
The present study was an attempt to examine a relatively representative, large sample from the common population. We sought to establish the prevalence of the specific PDs and study the demographic correlates. From earlier studies, especially clinical samples, our hypotheses were that women are more likely to have a borderline, histrionic, or dependent PD, while men are more likely to have an antisocial, schizoid, or obsessive-compulsive PD. We expected more PDs among younger individuals, especially higher frequencies of antisocial and borderline PDs. As is the case with other psychiatric disorders, we believe that PDs are more common among those with lower socioeconomic status. We chose education as the best measure because low income is so easily a direct consequence of psychiatric disabilities. Our hypothesis was that there are more PDs among those living in the inner city and those living alone. In Norway, as in Scandinavia generally, many couples live together in stable relationships without being married. So we must apply the variable "living alone, living together with a partner," rather than "being married, not being married."
Our point of departure was the National Register of Oslo containing the names and addresses of all citizens in Oslo. The group aged 18 to 65 years encompassed 308 237 individuals. From this population, a sample of 3590 individuals was drawn by chance according to a computer program in the national register. Figure 1 shows the reduction in the sample of 3590 individuals for various reasons.
Four of the individuals were deceased, 345 had moved from Oslo, and 548 were not found at the given address. It is possible that these individuals were not located either because they had provided an incorrect address to the authorities, or because they had moved to another residence inside or outside of Oslo without informing the authorities of their new address. As a result, one fourth of the sample could not be traced.
Of those who were contacted, 51 were either too medically ill, hospitalized in a medical hospital, or too physically disabled or sensory disabled to participate in an interview. Thirty-three persons, 5 of whom were hospitalized in a psychiatric ward, were too emotionally ill to be interviewed. A few others were taken in at a refugee reception center. Eighty-seven others were unable to be interviewed because they were unable to speak a Scandinavian language or English, French, German, or one of the other foreign languages used by the interviewers (including Asian languages). Altogether, close to 7% of those traced could not be interviewed for the above-named reasons.
Of those remaining, 18% refused to be interviewed, or they postponed the interview for too long. All in all, 2066 of a total sample of 3590 were interviewed. However, among these, it was not possible to establish PD diagnoses for 13 because of inadequate information. Consequently, the total sample amounted to 2053 subjects.
In contrast to most studies, we did not select households, but started with a fixed list of potential subjects. Consequently, we know something about those who did not participate. More women than men participated (62.7% and 52.7%, respectively). Those aged 40 years or older were included more frequently than younger individuals (60.8% compared with 55.0%). More subjects living in the town periphery were included than individuals in the center of the city (60.5% and 49.0%, respectively). The center of the city was defined as the 5 regions that meet in the heart of the city, thus constituting the downtown core of the city. All differences are statistically significant with a χ2 test (P< .001). Combined, men aged 40 years or older living in the center of the city participated least often (42.5%), and women aged 40 or older living on the outskirts of the city participated most often (68.7%). The reason for the lower rate of participation in the different demographic groups was not refusal or illness, but incorrect address and relocation without a new correct address.
The Structured Interview for DSM-III-R (SIDP-R) Personality Disorders%19 was applied to assess PDs in the subjects. The interviewers, mainly experienced nurses, but also medical students and experienced interviewers, were trained by using live patient interviews and videos of patient interviews throughout a period of several weeks. All interviews were conducted face-to-face, mostly at home, but some took place at the psychiatric clinic. The SIDP-R consists of 160 questions grouped under 16 thematic sections, such as "relationships," "emotions," and "reactions to stressful situations." At the end of each section is a listing of relevant DSM-III-R criteria rated from 0 to 2, with brief descriptions guiding the ratings. Level 0 corresponds with "not present," and levels 1 and 2, with "present to a moderate degree" and "present to a severe degree," respectively. A rating of 1 or 2 indicates criterion fulfilled.
The instructions for the SIDP-R specify a "five-year rule" which means that behavior typical of the past 5 years was the basis for the ratings. If an individual's personality changed dramatically during the past few years, the personality that dominated most of the time during the last 5 years was considered typical. The PD diagnoses were made without reference to the exclusion criteria, eg, schizophrenia for schizotypal PD.
The reliability was assessed by means of a rater listening to 40 audiotaped interviews. The κ value for any PD was 0.84. The number of persons with a specific PD was too small for making any κ calculation. Instead, intraclass correlations for the scaled PDs were calculated. The intraclass correlation for schizoid PD was 0.78; 0.71 for paranoid; 0.92 for schizotypal; 0.78 for obsessive-compulsive; 0.78 for histrionic; 0.82 for dependent; 0.95 for antisocial; 0.83 for avoidant; 0.89 for borderline; 0.95 for passive-aggressive; 0.87 for sadistic; and 0.85 for self-defeating PD. The median intraclass correlation was thus 0.83.
The prevalence rates were weighted, taking into account the slight, although statistically significant, differences between the interviewed sample and the population at large.
The relationship between PDs and demographic associations was calculated by means of a logistic regression analysis, in which the discrete demographic variables (sex, age group, educational level, living in the center/periphery of the city, living alone/with a partner, married/not married) were independent variables, and each of the PDs were dependent variables. In this way, odds ratios with confidence intervals were calculated taking into account possible correlations between the demographic variables. Multiple regression analysis was also performed. The independent demographic variables were treated as semicontinuous where possible, and the dependent variables were PD scales created by assigning number of criteria fulfilled for the respective PDs.
The α level is 5% if not otherwise stated.
The unweighted prevalence rate of any PD in the present sample was 13.1% in general, 14.6% among women, and 13.7% among men (Table 2). The weighted prevalence is slightly higher at 13.4% (Table 2). Generally, there was very little difference between the unweighted and weighted percentages because our sample was fairly representative of the population at large. The most prevalent PD was avoidant PD (5.0%), followed by paranoid (2.2% and 2.4%, unweighted and weighted, respectively), histrionic, and obsessive-compulsive (both 1.9%-2.0%). The rarest disorders were sadistic (0.2%), schizotypal (0.6%), antisocial (0.6%-0.7%), and borderline PDs (0.7%). The fearful cluster is very common (9.2%-9.4%), while the other 2 clusters are only one third as prevalent. Statistically, antisocial, passive aggressive, and obsessive-compulsive PDs are significantly more common among men. Also, schizoid PD is twice as common among men, but this is not statistically significant. Borderline, histrionic, dependent, and self-defeating PDs are twice as common among women as among men; however, the differences are not statistically significant.
The mean number of PD diagnoses among those with a PD was 1.48. Seventy-one percent of those with PD had only 1, 18.6% had 2, 5.2% had 3, 3.3% had 4, 1.1% had 5, 0.4% had 6, and 0.4% had 7 diagnoses.
Table 3 presents the increased risk of different PDs in various demographic groups. As logistic regression is applied, the possible correlations between the different demographic variables are taken into account.
It is observed that although the prevalence of PDs in general and the prevalence of the different clusters are the same in men and women, there is a sex difference in terms of the specific type of disorders. Men are more often schizoid and passive-aggressive, while women are more often histrionic and dependent.
Personality disorders are diagnosed most frequently in those older than 49 years, which is particularly so for the eccentric cluster.
Those with a high school education or less are more likely to have a PD, especially of the eccentric type. This is in particular the case for paranoid and avoidant PDs. Interestingly, obsessive-compulsive PD is more frequent in subjects with higher education (ie, a college/university education) compared with those with less education.
Living in the center of the city is related to having a PD. Again, this relationship is most often observed for the eccentric cluster, in particular paranoid and schizotypal PDs. Histrionic and passive-aggressive PDs are also relatively more prevalent in the center of the city.
Living without a partner is related to having a PD and to having eccentric and dramatic PDs. In addition, living without a partner is related to the paranoid, schizoid, antisocial, borderline, and self-defeating PDs.
Table 4 presents the statistically significant standardized β weight's from multiple regression analyses in which PD scales are applied by adding the criteria. Age and education are treated as continuous variables. The Enter method was applied. All results from the logistic regression analyses are confirmed, with the exception of the relationship between PDs generally and younger age, more individuals with passive-aggressive PD in the city center, and more women with histrionic PD.
In addition, a number of other relationships appeared in the multiple regression analysis. Antisocial, sadistic, narcissistic, obsessive-compulsive, eccentric, and dramatic traits were related to being male, while self-defeating traits were related to being female. Antisocial, sadistic, borderline, passive-aggressive, and dramatic traits are correlated with younger age. Schizoid, schizotypal, antisocial, sadistic, borderline, dependent, self-defeating, dramatic, and fearful traits are related to less education. Schizoid, borderline, narcissistic, obsessive-compulsive, self-defeating, and dramatic traits are more often observed in the center of the city. Schizoid and obsessive-compulsive traits are more often found among those not married. Schizotypal, histrionic, narcissistic, dependent, passive-aggressive, and fearful traits are more often found among those living without a partner.
Having more than 1 PD diagnosis is common. Our mean number of 1.48 diagnoses among those with a PD is slightly lower than what is observed in clinical populations.%20,21
The observed prevalence of any PDs in the present study is almost exactly the same as the prevalence in the pooled data from earlier studies (Table 1). The prevalence of avoidant, paranoid, and schizoid PDs is considerably higher in this study compared with previous ones (Table 1), whereas the rate of borderline PD is relatively low. Only the Swedish%14 and German%13 studies show a somewhat similar pattern, with a high prevalence of paranoid and avoidant PDs. It is not clear whether these observations reflect an improved sampling by our study, or national character traits that may be shared with other Scandinavians and Northern Europeans. Studies of patient populations in Norway have demonstrated a high prevalence of avoidant and paranoid PDs when compared with, for instance, American studies.%22,23
According to a Norwegian twin study,%24 avoidant, paranoid, and schizoid PDs seem to be less genetically influenced than other PDs. It may be that some cultural factors are responsible for the development of these disorders in Norway, with the consequence that they seem to be both more prevalent and more environmentally determined.
Our finding that schizoid, antisocial, narcissistic and obsessive-compulsive PDs and/or traits are more often observed among men, and histrionic, avoidant, and dependent features are more often found among women has also been reported in earlier studies.%10,25 As in the present study, these researchers did not find that borderline PD was related to the sex of the subject. Most clinical studies do. In fact, we found that borderline PD was twice as frequent among women; however, the low number of subjects with borderline PD made it very difficult to reach statistical significance. Furthermore, it is possible that being female and having borderline features interact to make the person more likely to seek treatment.
Earlier studies%10,25 have also observed that those with antisocial, borderline, and passive-aggressive traits were younger, and those with schizoid features were older. No other study has found that schizotypal, avoidant, and obsessive-compulsive PDs and traits are more frequent in older age.
No other epidemiological study has investigated whether those with PDs or PD traits more often live alone. In Norway, this is significant because many couples live together in stable relationships without being married; however, if one considers marital status (single, married, divorced, separated, widowed), one may make a comparison with previous studies. Earlier studies have shown that those with borderline PD were more often single, those with antisocial PD were more often divorced, those with passive-aggressive PD were more often not married, and those with a PD were more often separated at the time of the interview.%10 The present study showed that those with these PDs and traits more often lived without a partner.
Educational level has seldom been related to the prevalence of PDs. Reich et al%11 found no demographic connection to PDs in general. Nor did Nestad et al%7 find any demographic correlations to histrionic PD.
No other study of PDs has investigated the difference between those living in the city center as opposed to living on the outskirts. Lewis and Booth,%26 however, found that greater London had a higher prevalence of psychiatric morbidity than average. The same authors%27 found the highest General Health Questionnaire scores in those living in built-up urban areas. A somewhat lower score was observed among those in urban areas with "access to gardens or open spaces." The lowest score was found among those living in country areas. Sex, age, marital status, and social class were all controlled for in the study.
The fact that more psychiatric disorders are found in the center of cities has been known since the 1930s.%28 The reason for this is not easy to determine. Two hypotheses have been popular. The drift hypothesis states that those with emotional problems drift to the center, maybe to live anonymously; and the stress hypothesis maintains that there is more emotional hardship in the city center.%29 Which hypothesis is more correct has not yet been decided. Our results cannot be explained by socioeconomic status. The city center has both poorer and more prosperous areas, as has the periphery. Furthermore, we controlled for demographic variables. The population density is higher in the city center. There are 6292 people per square kilometer living in the center of the city, compared with 2912 people per square kilometer living in the periphery. In fact, 5 of 6 of the most densely populated areas in the city (which consists of a total of 25 areas) lie in the center. Anomie as a consequence of population density has been a popular explanation.%29
Our study has several limitations. Although our sample is more representative than earlier studies of PDs, we did not succeed in obtaining exactly the same proportion of the population in all demographic strata. However, we controlled for the sampling bias and obtained slightly higher prevalence rates.
In conclusion, PDs are relatively prevalent. Nearly 1 in 7 individuals in the Oslo catchment area was identified as meeting criteria for such a disorder. Age, education, area of living, and life situation seem to correlate with PDs and traits.
Whether the demographic associations are causes or consequences, and to what extent there are cohort effects taking place, can only be answered by future, prospective longitudinal studies of large community samples.
Accepted for publication January 23, 2000.
This study was supported by a grant from the Norwegian Council for Mental Health, Oslo, and the Foundation for Health and Rehabilitation, Oslo.
Corresponding author and reprints: Svenn Torgersen, PhD, Department of Psychology, Oslo University, PO Box 1039, Blindern, N-0315 Oslo, Norway (e-mail: Svenn.Torgersen@psykologi.uio.no).